158 files changed, 5632 insertions, 734 deletions
diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index a19a5266d5fc..db524e75c4a2 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -6,7 +6,7 @@ config ARM
        select HAVE_MEMBLOCK
        select RTC_LIB
        select SYS_SUPPORTS_APM_EMULATION
-        select GENERIC_ATOMIC64 if (!CPU_32v6K)
+        select GENERIC_ATOMIC64 if (!CPU_32v6K || !AEABI)
        select HAVE_OPROFILE if (HAVE_PERF_EVENTS)
        select HAVE_ARCH_KGDB
        select HAVE_KPROBES if (!XIP_KERNEL)
@@ -646,7 +646,7 @@ config ARCH_S3C2410
        select ARCH_HAS_CPUFREQ
        select HAVE_CLK
        select ARCH_USES_GETTIMEOFFSET
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
        help
          Samsung S3C2410X CPU based systems, such as the Simtec Electronics
          BAST (<http://www.simtec.co.uk/products/EB110ITX/>), the IPAQ 1940 or
@@ -676,8 +676,8 @@ config ARCH_S3C64XX
        select S3C_DEV_NAND
        select USB_ARCH_HAS_OHCI
        select SAMSUNG_GPIOLIB_4BIT
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        help
          Samsung S3C64XX series based systems
@@ -686,10 +686,10 @@ config ARCH_S5P64X0
        select CPU_V6
        select GENERIC_GPIO
        select HAVE_CLK
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        select ARCH_USES_GETTIMEOFFSET
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
-        select HAVE_S3C_RTC
+        select HAVE_S3C_RTC if RTC_CLASS
        help
          Samsung S5P64X0 CPU based systems, such as the Samsung SMDK6440,
          SMDK6450.
@@ -700,7 +700,7 @@ config ARCH_S5P6442
        select GENERIC_GPIO
        select HAVE_CLK
        select ARCH_USES_GETTIMEOFFSET
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        help
          Samsung S5P6442 CPU based systems
@@ -711,9 +711,9 @@ config ARCH_S5PC100
        select CPU_V7
        select ARM_L1_CACHE_SHIFT_6
        select ARCH_USES_GETTIMEOFFSET
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
-        select HAVE_S3C_RTC
+        select HAVE_S3C_RTC if RTC_CLASS
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        help
          Samsung S5PC100 series based systems
@@ -726,9 +726,9 @@ config ARCH_S5PV210
        select ARM_L1_CACHE_SHIFT_6
        select ARCH_HAS_CPUFREQ
        select ARCH_USES_GETTIMEOFFSET
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
-        select HAVE_S3C_RTC
+        select HAVE_S3C_RTC if RTC_CLASS
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        help
          Samsung S5PV210/S5PC110 series based systems
@@ -739,9 +739,9 @@ config ARCH_S5PV310
        select GENERIC_GPIO
        select HAVE_CLK
        select GENERIC_CLOCKEVENTS
-        select HAVE_S3C_RTC
+        select HAVE_S3C_RTC if RTC_CLASS
-        select HAVE_S3C2410_I2C
+        select HAVE_S3C2410_I2C if I2C
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        help
          Samsung S5PV310 series based systems
diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S
index 6825c34646d4..9be21ba648cd 100644
--- a/arch/arm/boot/compressed/head.S
+++ b/arch/arm/boot/compressed/head.S
@@ -1084,6 +1084,6 @@ memdump:	mov	r12, r0
 reloc_end:
                .align
-                .section ".stack", "w"
+                .section ".stack", "aw", %nobits
 user_stack:     .space  4096
 user_stack_end:
diff --git a/arch/arm/boot/compressed/vmlinux.lds.in b/arch/arm/boot/compressed/vmlinux.lds.in
index d08168941bd6..366a924019ac 100644
--- a/arch/arm/boot/compressed/vmlinux.lds.in
+++ b/arch/arm/boot/compressed/vmlinux.lds.in
@@ -57,7 +57,7 @@ SECTIONS
  .bss                  : { *(.bss) }
  _end = .;
-  .stack (NOLOAD)       : { *(.stack) }
+  .stack                : { *(.stack) }
  .stab 0               : { *(.stab) }
  .stabstr 0            : { *(.stabstr) }
diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index ada6359160eb..772f95f1aecd 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -251,15 +251,16 @@ void __init gic_dist_init(unsigned int gic_nr, void __iomem *base,
                writel(cpumask, base + GIC_DIST_TARGET + i * 4 / 4);
        /*
-         * Set priority on all interrupts.
+         * Set priority on all global interrupts.
         */
-        for (i = 0; i < max_irq; i += 4)
+        for (i = 32; i < max_irq; i += 4)
                writel(0xa0a0a0a0, base + GIC_DIST_PRI + i * 4 / 4);
        /*
-         * Disable all interrupts.
+         * Disable all interrupts.  Leave the PPI and SGIs alone
+         * as these enables are banked registers.
         */
-        for (i = 0; i < max_irq; i += 32)
+        for (i = 32; i < max_irq; i += 32)
                writel(0xffffffff, base + GIC_DIST_ENABLE_CLEAR + i * 4 / 32);
        /*
@@ -277,11 +278,30 @@ void __init gic_dist_init(unsigned int gic_nr, void __iomem *base,
 void __cpuinit gic_cpu_init(unsigned int gic_nr, void __iomem *base)
 {
+        void __iomem *dist_base;
+        int i;
        if (gic_nr >= MAX_GIC_NR)
                BUG();
+        dist_base = gic_data[gic_nr].dist_base;
+        BUG_ON(!dist_base);
        gic_data[gic_nr].cpu_base = base;
+        /*
+         * Deal with the banked PPI and SGI interrupts - disable all
+         * PPI interrupts, ensure all SGI interrupts are enabled.
+         */
+        writel(0xffff0000, dist_base + GIC_DIST_ENABLE_CLEAR);
+        writel(0x0000ffff, dist_base + GIC_DIST_ENABLE_SET);
+        /*
+         * Set priority on PPI and SGI interrupts
+         */
+        for (i = 0; i < 32; i += 4)
+                writel(0xa0a0a0a0, dist_base + GIC_DIST_PRI + i * 4 / 4);
        writel(0xf0, base + GIC_CPU_PRIMASK);
        writel(1, base + GIC_CPU_CTRL);
 }
diff --git a/arch/arm/include/asm/assembler.h b/arch/arm/include/asm/assembler.h
index 062b58c029ab..749bb6622404 100644
--- a/arch/arm/include/asm/assembler.h
+++ b/arch/arm/include/asm/assembler.h
@@ -238,7 +238,7 @@
        @ Slightly optimised to avoid incrementing the pointer twice
        usraccoff \instr, \reg, \ptr, \inc, 0, \cond, \abort
        .if     \rept == 2
-        usraccoff \instr, \reg, \ptr, \inc, 4, \cond, \abort
+        usraccoff \instr, \reg, \ptr, \inc, \inc, \cond, \abort
        .endif
        add\cond \ptr, #\rept * \inc
diff --git a/arch/arm/include/asm/hardware/it8152.h b/arch/arm/include/asm/hardware/it8152.h
index 6700c7fc7ebd..21fa272301f8 100644
--- a/arch/arm/include/asm/hardware/it8152.h
+++ b/arch/arm/include/asm/hardware/it8152.h
@@ -75,7 +75,7 @@ extern unsigned long it8152_base_address;
  IT8152_PD_IRQ(1)  USB (USBR)
  IT8152_PD_IRQ(0)  Audio controller (ACR)
 */
-#define IT8152_IRQ(x)   (IRQ_BOARD_END + (x))
+#define IT8152_IRQ(x)   (IRQ_BOARD_START + (x))
 /* IRQ-sources in 3 groups - local devices, LPC (serial), and external PCI */
 #define IT8152_LD_IRQ_COUNT     9
diff --git a/arch/arm/include/asm/mmu.h b/arch/arm/include/asm/mmu.h
index 68870c776671..b4ffe9d5b526 100644
--- a/arch/arm/include/asm/mmu.h
+++ b/arch/arm/include/asm/mmu.h
@@ -13,6 +13,10 @@ typedef struct {
 #ifdef CONFIG_CPU_HAS_ASID
 #define ASID(mm)        ((mm)->context.id & 255)
+/* init_mm.context.id_lock should be initialized. */
+#define INIT_MM_CONTEXT(name)                                                 \
+        .context.id_lock    = __SPIN_LOCK_UNLOCKED(name.context.id_lock),
 #else
 #define ASID(mm)        (0)
 #endif
diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h
index b155414192da..53d1d5deb111 100644
--- a/arch/arm/include/asm/pgtable.h
+++ b/arch/arm/include/asm/pgtable.h
@@ -374,6 +374,9 @@ static inline pte_t *pmd_page_vaddr(pmd_t pmd)
 #define pmd_page(pmd)           pfn_to_page(__phys_to_pfn(pmd_val(pmd)))
+/* we don't need complex calculations here as the pmd is folded into the pgd */
+#define pmd_addr_end(addr,end)  (end)
 /*
 * Conversion functions: convert a page and protection to a page entry,
 * and a page entry and page directory to the page they refer to.
diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
index 54593b0c241b..21e3a4ab3b8c 100644
--- a/arch/arm/kernel/hw_breakpoint.c
+++ b/arch/arm/kernel/hw_breakpoint.c
@@ -748,8 +748,7 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr,
                breakpoint_handler(addr, regs);
                break;
        case ARM_ENTRY_ASYNC_WATCHPOINT:
-                WARN_ON("Asynchronous watchpoint exception taken. "
+                WARN(1, "Asynchronous watchpoint exception taken. Debugging results may be unreliable\n");
-                        "Debugging results may be unreliable");
        case ARM_ENTRY_SYNC_WATCHPOINT:
                watchpoint_handler(addr, regs);
                break;
diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c
index 49643b1467e6..07a50357492a 100644
--- a/arch/arm/kernel/perf_event.c
+++ b/arch/arm/kernel/perf_event.c
@@ -1749,7 +1749,7 @@ static inline int armv7_pmnc_has_overflowed(unsigned long pmnc)
 static inline int armv7_pmnc_counter_has_overflowed(unsigned long pmnc,
                                        enum armv7_counters counter)
 {
-        int ret;
+        int ret = 0;
        if (counter == ARMV7_CYCLE_COUNTER)
                ret = pmnc & ARMV7_FLAG_C;
diff --git a/arch/arm/kernel/stacktrace.c b/arch/arm/kernel/stacktrace.c
index 20b7411e47fd..c2e112e1a05f 100644
--- a/arch/arm/kernel/stacktrace.c
+++ b/arch/arm/kernel/stacktrace.c
@@ -28,7 +28,7 @@ int notrace unwind_frame(struct stackframe *frame)
        /* only go to a higher address on the stack */
        low = frame->sp;
-        high = ALIGN(low, THREAD_SIZE) + THREAD_SIZE;
+        high = ALIGN(low, THREAD_SIZE);
        /* check current frame pointer is within bounds */
        if (fp < (low + 12) || fp + 4 >= high)
diff --git a/arch/arm/kernel/traps.c b/arch/arm/kernel/traps.c
index cda78d59aa31..446aee97436f 100644
--- a/arch/arm/kernel/traps.c
+++ b/arch/arm/kernel/traps.c
@@ -53,10 +53,7 @@ static void dump_mem(const char *, const char *, unsigned long, unsigned long);
 void dump_backtrace_entry(unsigned long where, unsigned long from, unsigned long frame)
 {
 #ifdef CONFIG_KALLSYMS
-        char sym1[KSYM_SYMBOL_LEN], sym2[KSYM_SYMBOL_LEN];
+        printk("[<%08lx>] (%pS) from [<%08lx>] (%pS)\n", where, (void *)where, from, (void *)from);
-        sprint_symbol(sym1, where);
-        sprint_symbol(sym2, from);
-        printk("[<%08lx>] (%s) from [<%08lx>] (%s)\n", where, sym1, from, sym2);
 #else
        printk("Function entered at [<%08lx>] from [<%08lx>]\n", where, from);
 #endif
diff --git a/arch/arm/kernel/unwind.c b/arch/arm/kernel/unwind.c
index 2a161765f6d5..d2cb0b3c9872 100644
--- a/arch/arm/kernel/unwind.c
+++ b/arch/arm/kernel/unwind.c
@@ -279,7 +279,7 @@ int unwind_frame(struct stackframe *frame)
        /* only go to a higher address on the stack */
        low = frame->sp;
-        high = ALIGN(low, THREAD_SIZE) + THREAD_SIZE;
+        high = ALIGN(low, THREAD_SIZE);
        pr_debug("%s(pc = %08lx lr = %08lx sp = %08lx)\n", __func__,
                 frame->pc, frame->lr, frame->sp);
diff --git a/arch/arm/lib/findbit.S b/arch/arm/lib/findbit.S
index 1e4cbd4e7be9..64f6bc1a9132 100644
--- a/arch/arm/lib/findbit.S
+++ b/arch/arm/lib/findbit.S
@@ -174,8 +174,8 @@ ENDPROC(_find_next_bit_be)
 */
 .L_found:
 #if __LINUX_ARM_ARCH__ >= 5
-                rsb     r1, r3, #0
+                rsb     r0, r3, #0
-                and     r3, r3, r1
+                and     r3, r3, r0
                clz     r3, r3
                rsb     r3, r3, #31
                add     r0, r2, r3
@@ -190,5 +190,7 @@ ENDPROC(_find_next_bit_be)
                addeq   r2, r2, #1
                mov     r0, r2
 #endif
+                cmp     r1, r0                  @ Clamp to maxbit
+                movlo   r0, r1
                mov     pc, lr
diff --git a/arch/arm/mach-aaec2000/include/mach/vmalloc.h b/arch/arm/mach-aaec2000/include/mach/vmalloc.h
index cff4e0a996ce..a6299e8321bd 100644
--- a/arch/arm/mach-aaec2000/include/mach/vmalloc.h
+++ b/arch/arm/mach-aaec2000/include/mach/vmalloc.h
@@ -11,6 +11,6 @@
 #ifndef __ASM_ARCH_VMALLOC_H
 #define __ASM_ARCH_VMALLOC_H
-#define VMALLOC_END             0xd0000000
+#define VMALLOC_END             0xd0000000UL
 #endif /* __ASM_ARCH_VMALLOC_H */
diff --git a/arch/arm/mach-bcmring/include/mach/vmalloc.h b/arch/arm/mach-bcmring/include/mach/vmalloc.h
index 3db3a09fd398..7397bd7817d9 100644
--- a/arch/arm/mach-bcmring/include/mach/vmalloc.h
+++ b/arch/arm/mach-bcmring/include/mach/vmalloc.h
@@ -22,4 +22,4 @@
 * 0xe0000000 to 0xefffffff. This gives us 256 MB of vm space and handles
 * larger physical memory designs better.
 */
-#define VMALLOC_END       0xf0000000
+#define VMALLOC_END       0xf0000000UL
diff --git a/arch/arm/mach-clps711x/include/mach/vmalloc.h b/arch/arm/mach-clps711x/include/mach/vmalloc.h
index 30b3a287ed88..467b96137e47 100644
--- a/arch/arm/mach-clps711x/include/mach/vmalloc.h
+++ b/arch/arm/mach-clps711x/include/mach/vmalloc.h
@@ -17,4 +17,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
diff --git a/arch/arm/mach-davinci/dm355.c b/arch/arm/mach-davinci/dm355.c
index 9be261beae7d..2652af124acd 100644
--- a/arch/arm/mach-davinci/dm355.c
+++ b/arch/arm/mach-davinci/dm355.c
@@ -359,8 +359,8 @@ static struct clk_lookup dm355_clks[] = {
        CLK(NULL, "uart1", &uart1_clk),
        CLK(NULL, "uart2", &uart2_clk),
        CLK("i2c_davinci.1", NULL, &i2c_clk),
-        CLK("davinci-asp.0", NULL, &asp0_clk),
+        CLK("davinci-mcbsp.0", NULL, &asp0_clk),
-        CLK("davinci-asp.1", NULL, &asp1_clk),
+        CLK("davinci-mcbsp.1", NULL, &asp1_clk),
        CLK("davinci_mmc.0", NULL, &mmcsd0_clk),
        CLK("davinci_mmc.1", NULL, &mmcsd1_clk),
        CLK("spi_davinci.0", NULL, &spi0_clk),
@@ -664,7 +664,7 @@ static struct resource dm355_asp1_resources[] = {
 };
 static struct platform_device dm355_asp1_device = {
-        .name           = "davinci-asp",
+        .name           = "davinci-mcbsp",
        .id             = 1,
        .num_resources  = ARRAY_SIZE(dm355_asp1_resources),
        .resource       = dm355_asp1_resources,
diff --git a/arch/arm/mach-davinci/dm365.c b/arch/arm/mach-davinci/dm365.c
index a12065e87266..c466d710d3c1 100644
--- a/arch/arm/mach-davinci/dm365.c
+++ b/arch/arm/mach-davinci/dm365.c
@@ -459,7 +459,7 @@ static struct clk_lookup dm365_clks[] = {
        CLK(NULL, "usb", &usb_clk),
        CLK("davinci_emac.1", NULL, &emac_clk),
        CLK("davinci_voicecodec", NULL, &voicecodec_clk),
-        CLK("davinci-asp.0", NULL, &asp0_clk),
+        CLK("davinci-mcbsp", NULL, &asp0_clk),
        CLK(NULL, "rto", &rto_clk),
        CLK(NULL, "mjcp", &mjcp_clk),
        CLK(NULL, NULL, NULL),
@@ -922,8 +922,8 @@ static struct resource dm365_asp_resources[] = {
 };
 static struct platform_device dm365_asp_device = {
-        .name           = "davinci-asp",
+        .name           = "davinci-mcbsp",
-        .id             = 0,
+        .id             = -1,
        .num_resources  = ARRAY_SIZE(dm365_asp_resources),
        .resource       = dm365_asp_resources,
 };
diff --git a/arch/arm/mach-davinci/dm644x.c b/arch/arm/mach-davinci/dm644x.c
index 0608dd776a16..9a2376b3137c 100644
--- a/arch/arm/mach-davinci/dm644x.c
+++ b/arch/arm/mach-davinci/dm644x.c
@@ -302,7 +302,7 @@ static struct clk_lookup dm644x_clks[] = {
        CLK("davinci_emac.1", NULL, &emac_clk),
        CLK("i2c_davinci.1", NULL, &i2c_clk),
        CLK("palm_bk3710", NULL, &ide_clk),
-        CLK("davinci-asp", NULL, &asp_clk),
+        CLK("davinci-mcbsp", NULL, &asp_clk),
        CLK("davinci_mmc.0", NULL, &mmcsd_clk),
        CLK(NULL, "spi", &spi_clk),
        CLK(NULL, "gpio", &gpio_clk),
@@ -580,7 +580,7 @@ static struct resource dm644x_asp_resources[] = {
 };
 static struct platform_device dm644x_asp_device = {
-        .name           = "davinci-asp",
+        .name           = "davinci-mcbsp",
        .id             = -1,
        .num_resources  = ARRAY_SIZE(dm644x_asp_resources),
        .resource       = dm644x_asp_resources,
diff --git a/arch/arm/mach-ebsa110/include/mach/vmalloc.h b/arch/arm/mach-ebsa110/include/mach/vmalloc.h
index 60bde56fba4c..ea141b7a3e03 100644
--- a/arch/arm/mach-ebsa110/include/mach/vmalloc.h
+++ b/arch/arm/mach-ebsa110/include/mach/vmalloc.h
@@ -7,4 +7,4 @@
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */
-#define VMALLOC_END       0xdf000000
+#define VMALLOC_END       0xdf000000UL
diff --git a/arch/arm/mach-ep93xx/include/mach/dma.h b/arch/arm/mach-ep93xx/include/mach/dma.h
index 3a5961d3f3b1..5e31b2b25da9 100644
--- a/arch/arm/mach-ep93xx/include/mach/dma.h
+++ b/arch/arm/mach-ep93xx/include/mach/dma.h
@@ -1,5 +1,13 @@
-/*
+/**
- * arch/arm/mach-ep93xx/include/mach/dma.h
+ * DOC: EP93xx DMA M2P memory to peripheral and peripheral to memory engine
+ *
+ * The EP93xx DMA M2P subsystem handles DMA transfers between memory and
+ * peripherals. DMA M2P channels are available for audio, UARTs and IrDA.
+ * See chapter 10 of the EP93xx users guide for full details on the DMA M2P
+ * engine.
+ *
+ * See sound/soc/ep93xx/ep93xx-pcm.c for an example use of the DMA M2P code.
+ *
 */
 #ifndef __ASM_ARCH_DMA_H
@@ -8,12 +16,34 @@
 #include <linux/list.h>
 #include <linux/types.h>
+/**
+ * struct ep93xx_dma_buffer - Information about a buffer to be transferred
+ * using the DMA M2P engine
+ *
+ * @list: Entry in DMA buffer list
+ * @bus_addr: Physical address of the buffer
+ * @size: Size of the buffer in bytes
+ */
 struct ep93xx_dma_buffer {
        struct list_head        list;
        u32                     bus_addr;
        u16                     size;
 };
+/**
+ * struct ep93xx_dma_m2p_client - Information about a DMA M2P client
+ *
+ * @name: Unique name for this client
+ * @flags: Client flags
+ * @cookie: User data to pass to callback functions
+ * @buffer_started: Non NULL function to call when a transfer is started.
+ *                      The arguments are the user data cookie and the DMA
+ *                      buffer which is starting.
+ * @buffer_finished: Non NULL function to call when a transfer is completed.
+ *                      The arguments are the user data cookie, the DMA buffer
+ *                      which has completed, and a boolean flag indicating if
+ *                      the transfer had an error.
+ */
 struct ep93xx_dma_m2p_client {
        char                    *name;
        u8                      flags;
@@ -24,10 +54,11 @@ struct ep93xx_dma_m2p_client {
                                        struct ep93xx_dma_buffer *buf,
                                        int bytes, int error);
-        /* Internal to the DMA code.  */
+        /* private: Internal use only */
        void                    *channel;
 };
+/* DMA M2P ports */
 #define EP93XX_DMA_M2P_PORT_I2S1        0x00
 #define EP93XX_DMA_M2P_PORT_I2S2        0x01
 #define EP93XX_DMA_M2P_PORT_AAC1        0x02
@@ -39,18 +70,80 @@ struct ep93xx_dma_m2p_client {
 #define EP93XX_DMA_M2P_PORT_UART3       0x08
 #define EP93XX_DMA_M2P_PORT_IRDA        0x09
 #define EP93XX_DMA_M2P_PORT_MASK        0x0f
-#define EP93XX_DMA_M2P_TX               0x00
-#define EP93XX_DMA_M2P_RX               0x10
-#define EP93XX_DMA_M2P_ABORT_ON_ERROR   0x20
-#define EP93XX_DMA_M2P_IGNORE_ERROR     0x40
-#define EP93XX_DMA_M2P_ERROR_MASK       0x60
-int  ep93xx_dma_m2p_client_register(struct ep93xx_dma_m2p_client *m2p);
+/* DMA M2P client flags */
+#define EP93XX_DMA_M2P_TX               0x00    /* Memory to peripheral */
+#define EP93XX_DMA_M2P_RX               0x10    /* Peripheral to memory */
+/*
+ * DMA M2P client error handling flags. See the EP93xx users guide
+ * documentation on the DMA M2P CONTROL register for more details
+ */
+#define EP93XX_DMA_M2P_ABORT_ON_ERROR   0x20    /* Abort on peripheral error */
+#define EP93XX_DMA_M2P_IGNORE_ERROR     0x40    /* Ignore peripheral errors */
+#define EP93XX_DMA_M2P_ERROR_MASK       0x60    /* Mask of error bits */
+/**
+ * ep93xx_dma_m2p_client_register - Register a client with the DMA M2P
+ * subsystem
+ *
+ * @m2p: Client information to register
+ * returns 0 on success
+ *
+ * The DMA M2P subsystem allocates a channel and an interrupt line for the DMA
+ * client
+ */
+int ep93xx_dma_m2p_client_register(struct ep93xx_dma_m2p_client *m2p);
+/**
+ * ep93xx_dma_m2p_client_unregister - Unregister a client from the DMA M2P
+ * subsystem
+ *
+ * @m2p: Client to unregister
+ *
+ * Any transfers currently in progress will be completed in hardware, but
+ * ignored in software.
+ */
 void ep93xx_dma_m2p_client_unregister(struct ep93xx_dma_m2p_client *m2p);
+/**
+ * ep93xx_dma_m2p_submit - Submit a DMA M2P transfer
+ *
+ * @m2p: DMA Client to submit the transfer on
+ * @buf: DMA Buffer to submit
+ *
+ * If the current or next transfer positions are free on the M2P client then
+ * the transfer is started immediately. If not, the transfer is added to the
+ * list of pending transfers. This function must not be called from the
+ * buffer_finished callback for an M2P channel.
+ *
+ */
 void ep93xx_dma_m2p_submit(struct ep93xx_dma_m2p_client *m2p,
                           struct ep93xx_dma_buffer *buf);
+/**
+ * ep93xx_dma_m2p_submit_recursive - Put a DMA transfer on the pending list
+ * for an M2P channel
+ *
+ * @m2p: DMA Client to submit the transfer on
+ * @buf: DMA Buffer to submit
+ *
+ * This function must only be called from the buffer_finished callback for an
+ * M2P channel. It is commonly used to add the next transfer in a chained list
+ * of DMA transfers.
+ */
 void ep93xx_dma_m2p_submit_recursive(struct ep93xx_dma_m2p_client *m2p,
                                     struct ep93xx_dma_buffer *buf);
+/**
+ * ep93xx_dma_m2p_flush - Flush all pending transfers on a DMA M2P client
+ *
+ * @m2p: DMA client to flush transfers on
+ *
+ * Any transfers currently in progress will be completed in hardware, but
+ * ignored in software.
+ *
+ */
 void ep93xx_dma_m2p_flush(struct ep93xx_dma_m2p_client *m2p);
 #endif /* __ASM_ARCH_DMA_H */
diff --git a/arch/arm/mach-footbridge/include/mach/vmalloc.h b/arch/arm/mach-footbridge/include/mach/vmalloc.h
index 0ffbb7c85e59..40ba78e5782b 100644
--- a/arch/arm/mach-footbridge/include/mach/vmalloc.h
+++ b/arch/arm/mach-footbridge/include/mach/vmalloc.h
@@ -7,4 +7,4 @@
 */
-#define VMALLOC_END       0xf0000000
+#define VMALLOC_END       0xf0000000UL
diff --git a/arch/arm/mach-h720x/include/mach/vmalloc.h b/arch/arm/mach-h720x/include/mach/vmalloc.h
index a45915b88756..8520b4a4d4e6 100644
--- a/arch/arm/mach-h720x/include/mach/vmalloc.h
+++ b/arch/arm/mach-h720x/include/mach/vmalloc.h
@@ -5,6 +5,6 @@
 #ifndef __ARCH_ARM_VMALLOC_H
 #define __ARCH_ARM_VMALLOC_H
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
 #endif
diff --git a/arch/arm/mach-imx/eukrea_mbimx27-baseboard.c b/arch/arm/mach-imx/eukrea_mbimx27-baseboard.c
index 026263c665ca..7e1e9dc2c8fc 100644
--- a/arch/arm/mach-imx/eukrea_mbimx27-baseboard.c
+++ b/arch/arm/mach-imx/eukrea_mbimx27-baseboard.c
@@ -250,9 +250,6 @@ static const struct imxuart_platform_data uart_pdata __initconst = {
        .flags = IMXUART_HAVE_RTSCTS,
 };
-#if defined(CONFIG_TOUCHSCREEN_ADS7846) \
-        || defined(CONFIG_TOUCHSCREEN_ADS7846_MODULE)
 #define ADS7846_PENDOWN (GPIO_PORTD | 25)
 static void ads7846_dev_init(void)
@@ -273,9 +270,7 @@ static struct ads7846_platform_data ads7846_config __initdata = {
        .get_pendown_state      = ads7846_get_pendown_state,
        .keep_vref_on           = 1,
 };
-#endif
-#if defined(CONFIG_SPI_IMX) || defined(CONFIG_SPI_IMX_MODULE)
 static struct spi_board_info eukrea_mbimx27_spi_board_info[] __initdata = {
        [0] = {
                .modalias       = "ads7846",
@@ -294,7 +289,6 @@ static const struct spi_imx_master eukrea_mbimx27_spi0_data __initconst = {
        .chipselect     = eukrea_mbimx27_spi_cs,
        .num_chipselect = ARRAY_SIZE(eukrea_mbimx27_spi_cs),
 };
-#endif
 static struct i2c_board_info eukrea_mbimx27_i2c_devices[] = {
        {
diff --git a/arch/arm/mach-integrator/include/mach/vmalloc.h b/arch/arm/mach-integrator/include/mach/vmalloc.h
index e056e7cf5645..2f5a2bafb11f 100644
--- a/arch/arm/mach-integrator/include/mach/vmalloc.h
+++ b/arch/arm/mach-integrator/include/mach/vmalloc.h
@@ -17,4 +17,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
diff --git a/arch/arm/mach-kirkwood/common.c b/arch/arm/mach-kirkwood/common.c
index 51ff23b72d3a..3688123b5ad8 100644
--- a/arch/arm/mach-kirkwood/common.c
+++ b/arch/arm/mach-kirkwood/common.c
@@ -854,10 +854,9 @@ int __init kirkwood_find_tclk(void)
        kirkwood_pcie_id(&dev, &rev);
-        if ((dev == MV88F6281_DEV_ID && (rev == MV88F6281_REV_A0 ||
+        if (dev == MV88F6281_DEV_ID || dev == MV88F6282_DEV_ID)
-                                        rev == MV88F6281_REV_A1)) ||
+                if (((readl(SAMPLE_AT_RESET) >> 21) & 1) == 0)
-            (dev == MV88F6282_DEV_ID))
+                        return 200000000;
-                return 200000000;
        return 166666667;
 }
diff --git a/arch/arm/mach-kirkwood/d2net_v2-setup.c b/arch/arm/mach-kirkwood/d2net_v2-setup.c
index 4aa86e4a152c..a31c9499ab36 100644
--- a/arch/arm/mach-kirkwood/d2net_v2-setup.c
+++ b/arch/arm/mach-kirkwood/d2net_v2-setup.c
@@ -225,5 +225,5 @@ MACHINE_START(D2NET_V2, "LaCie d2 Network v2")
        .init_machine   = d2net_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
diff --git a/arch/arm/mach-kirkwood/lacie_v2-common.c b/arch/arm/mach-kirkwood/lacie_v2-common.c
index d3ea1b6c8a02..285edab776e9 100644
--- a/arch/arm/mach-kirkwood/lacie_v2-common.c
+++ b/arch/arm/mach-kirkwood/lacie_v2-common.c
@@ -111,17 +111,3 @@ void __init lacie_v2_hdd_power_init(int hdd_num)
                        pr_err("Failed to power up HDD%d\n", i + 1);
        }
 }
-/*****************************************************************************
- * Timer
- ****************************************************************************/
-static void lacie_v2_timer_init(void)
-{
-        kirkwood_tclk = 166666667;
-        orion_time_init(IRQ_KIRKWOOD_BRIDGE, kirkwood_tclk);
-}
-struct sys_timer lacie_v2_timer = {
-        .init = lacie_v2_timer_init,
-};
diff --git a/arch/arm/mach-kirkwood/lacie_v2-common.h b/arch/arm/mach-kirkwood/lacie_v2-common.h
index af521315b87b..fc64f578536e 100644
--- a/arch/arm/mach-kirkwood/lacie_v2-common.h
+++ b/arch/arm/mach-kirkwood/lacie_v2-common.h
@@ -13,6 +13,4 @@ void lacie_v2_register_flash(void);
 void lacie_v2_register_i2c_devices(void);
 void lacie_v2_hdd_power_init(int hdd_num);
-extern struct sys_timer lacie_v2_timer;
 #endif
diff --git a/arch/arm/mach-kirkwood/mpp.c b/arch/arm/mach-kirkwood/mpp.c
index 065187d177c6..27901f702feb 100644
--- a/arch/arm/mach-kirkwood/mpp.c
+++ b/arch/arm/mach-kirkwood/mpp.c
@@ -59,7 +59,7 @@ void __init kirkwood_mpp_conf(unsigned int *mpp_list)
        }
        printk("\n");
-        while (*mpp_list) {
+        for ( ; *mpp_list; mpp_list++) {
                unsigned int num = MPP_NUM(*mpp_list);
                unsigned int sel = MPP_SEL(*mpp_list);
                int shift, gpio_mode;
@@ -88,8 +88,6 @@ void __init kirkwood_mpp_conf(unsigned int *mpp_list)
                if (sel != 0)
                        gpio_mode = 0;
                orion_gpio_set_valid(num, gpio_mode);
-                mpp_list++;
        }
        printk(KERN_DEBUG "  final MPP regs:");
diff --git a/arch/arm/mach-kirkwood/netspace_v2-setup.c b/arch/arm/mach-kirkwood/netspace_v2-setup.c
index 5ea66f1f4178..65ee21fd2f3b 100644
--- a/arch/arm/mach-kirkwood/netspace_v2-setup.c
+++ b/arch/arm/mach-kirkwood/netspace_v2-setup.c
@@ -262,7 +262,7 @@ MACHINE_START(NETSPACE_V2, "LaCie Network Space v2")
        .init_machine   = netspace_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
 #endif
@@ -272,7 +272,7 @@ MACHINE_START(INETSPACE_V2, "LaCie Internet Space v2")
        .init_machine   = netspace_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
 #endif
@@ -282,6 +282,6 @@ MACHINE_START(NETSPACE_MAX_V2, "LaCie Network Space Max v2")
        .init_machine   = netspace_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
 #endif
diff --git a/arch/arm/mach-kirkwood/netxbig_v2-setup.c b/arch/arm/mach-kirkwood/netxbig_v2-setup.c
index a1b45d501aef..93afd3c8bfd8 100644
--- a/arch/arm/mach-kirkwood/netxbig_v2-setup.c
+++ b/arch/arm/mach-kirkwood/netxbig_v2-setup.c
@@ -403,7 +403,7 @@ MACHINE_START(NET2BIG_V2, "LaCie 2Big Network v2")
        .init_machine   = netxbig_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
 #endif
@@ -413,6 +413,6 @@ MACHINE_START(NET5BIG_V2, "LaCie 5Big Network v2")
        .init_machine   = netxbig_v2_init,
        .map_io         = kirkwood_map_io,
        .init_irq       = kirkwood_init_irq,
-        .timer          = &lacie_v2_timer,
+        .timer          = &kirkwood_timer,
 MACHINE_END
 #endif
diff --git a/arch/arm/mach-kirkwood/ts41x-setup.c b/arch/arm/mach-kirkwood/ts41x-setup.c
index 8be09a0ce4ac..3587a281d993 100644
--- a/arch/arm/mach-kirkwood/ts41x-setup.c
+++ b/arch/arm/mach-kirkwood/ts41x-setup.c
@@ -27,6 +27,10 @@
 #include "mpp.h"
 #include "tsx1x-common.h"
+/* for the PCIe reset workaround */
+#include <plat/pcie.h>
 #define QNAP_TS41X_JUMPER_JP1   45
 static struct i2c_board_info __initdata qnap_ts41x_i2c_rtc = {
@@ -140,8 +144,16 @@ static void __init qnap_ts41x_init(void)
 static int __init ts41x_pci_init(void)
 {
-        if (machine_is_ts41x())
+        if (machine_is_ts41x()) {
+                /*
+                 * Without this explicit reset, the PCIe SATA controller
+                 * (Marvell 88sx7042/sata_mv) is known to stop working
+                 * after a few minutes.
+                 */
+                orion_pcie_reset((void __iomem *)PCIE_VIRT_BASE);
                kirkwood_pcie_init(KW_PCIE0);
+        }
   return 0;
 }
diff --git a/arch/arm/mach-mmp/include/mach/cputype.h b/arch/arm/mach-mmp/include/mach/cputype.h
index f43a68b213f1..8a3b56dfd35d 100644
--- a/arch/arm/mach-mmp/include/mach/cputype.h
+++ b/arch/arm/mach-mmp/include/mach/cputype.h
@@ -46,7 +46,8 @@ static inline int cpu_is_pxa910(void)
 #ifdef CONFIG_CPU_MMP2
 static inline int cpu_is_mmp2(void)
 {
-        return (((cpu_readid_id() >> 8) & 0xff) == 0x58);
+        return (((read_cpuid_id() >> 8) & 0xff) == 0x58);
+}
 #else
 #define cpu_is_mmp2()   (0)
 #endif
diff --git a/arch/arm/mach-msm/include/mach/vmalloc.h b/arch/arm/mach-msm/include/mach/vmalloc.h
index 31a32ad062dc..d138448eff16 100644
--- a/arch/arm/mach-msm/include/mach/vmalloc.h
+++ b/arch/arm/mach-msm/include/mach/vmalloc.h
@@ -16,7 +16,7 @@
 #ifndef __ASM_ARCH_MSM_VMALLOC_H
 #define __ASM_ARCH_MSM_VMALLOC_H
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
 #endif
diff --git a/arch/arm/mach-mv78xx0/mpp.c b/arch/arm/mach-mv78xx0/mpp.c
index 354ac514eb89..84db2dfc475c 100644
--- a/arch/arm/mach-mv78xx0/mpp.c
+++ b/arch/arm/mach-mv78xx0/mpp.c
@@ -54,7 +54,7 @@ void __init mv78xx0_mpp_conf(unsigned int *mpp_list)
        }
        printk("\n");
-        while (*mpp_list) {
+        for ( ; *mpp_list; mpp_list++) {
                unsigned int num = MPP_NUM(*mpp_list);
                unsigned int sel = MPP_SEL(*mpp_list);
                int shift, gpio_mode;
@@ -83,8 +83,6 @@ void __init mv78xx0_mpp_conf(unsigned int *mpp_list)
                if (sel != 0)
                        gpio_mode = 0;
                orion_gpio_set_valid(num, gpio_mode);
-                mpp_list++;
        }
        printk(KERN_DEBUG "  final MPP regs:");
diff --git a/arch/arm/mach-mx25/devices-imx25.h b/arch/arm/mach-mx25/devices-imx25.h
index 93afa10b13cf..d94d282fa676 100644
--- a/arch/arm/mach-mx25/devices-imx25.h
+++ b/arch/arm/mach-mx25/devices-imx25.h
@@ -42,9 +42,9 @@ extern const struct imx_mxc_nand_data imx25_mxc_nand_data __initconst;
 #define imx25_add_mxc_nand(pdata)       \
        imx_add_mxc_nand(&imx25_mxc_nand_data, pdata)
-extern const struct imx_spi_imx_data imx25_spi_imx_data[] __initconst;
+extern const struct imx_spi_imx_data imx25_cspi_data[] __initconst;
 #define imx25_add_spi_imx(id, pdata)    \
-        imx_add_spi_imx(&imx25_spi_imx_data[id], pdata)
+        imx_add_spi_imx(&imx25_cspi_data[id], pdata)
 #define imx25_add_spi_imx0(pdata)       imx25_add_spi_imx(0, pdata)
 #define imx25_add_spi_imx1(pdata)       imx25_add_spi_imx(1, pdata)
 #define imx25_add_spi_imx2(pdata)       imx25_add_spi_imx(2, pdata)
diff --git a/arch/arm/mach-mx3/mach-pcm037_eet.c b/arch/arm/mach-mx3/mach-pcm037_eet.c
index 99e0894e07db..fda56545d2fd 100644
--- a/arch/arm/mach-mx3/mach-pcm037_eet.c
+++ b/arch/arm/mach-mx3/mach-pcm037_eet.c
@@ -14,6 +14,7 @@
 #include <mach/common.h>
 #include <mach/iomux-mx3.h>
+#include <mach/spi.h>
 #include <asm/mach-types.h>
@@ -59,14 +60,12 @@ static struct spi_board_info pcm037_spi_dev[] = {
 };
 /* Platform Data for MXC CSPI */
-#if defined(CONFIG_SPI_IMX) || defined(CONFIG_SPI_IMX_MODULE)
 static int pcm037_spi1_cs[] = {MXC_SPI_CS(1), IOMUX_TO_GPIO(MX31_PIN_KEY_COL7)};
 static const struct spi_imx_master pcm037_spi1_pdata __initconst = {
        .chipselect = pcm037_spi1_cs,
        .num_chipselect = ARRAY_SIZE(pcm037_spi1_cs),
 };
-#endif
 /* GPIO-keys input device */
 static struct gpio_keys_button pcm037_gpio_keys[] = {
@@ -171,7 +170,7 @@ static struct platform_device pcm037_gpio_keys_device = {
        },
 };
-static int eet_init_devices(void)
+static int __init eet_init_devices(void)
 {
        if (!machine_is_pcm037() || pcm037_variant() != PCM037_EET)
                return 0;
diff --git a/arch/arm/mach-netx/include/mach/vmalloc.h b/arch/arm/mach-netx/include/mach/vmalloc.h
index 7cca3574308f..871f1ef7bff5 100644
--- a/arch/arm/mach-netx/include/mach/vmalloc.h
+++ b/arch/arm/mach-netx/include/mach/vmalloc.h
@@ -16,4 +16,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
diff --git a/arch/arm/mach-omap1/devices.c b/arch/arm/mach-omap1/devices.c
index ea0d80a89da7..e7f9ee63dce5 100644
--- a/arch/arm/mach-omap1/devices.c
+++ b/arch/arm/mach-omap1/devices.c
@@ -321,10 +321,9 @@ static struct platform_device omap_wdt_device = {
 static int __init omap_init_wdt(void)
 {
        if (!cpu_is_omap16xx())
-                return;
+                return -ENODEV;
-        platform_device_register(&omap_wdt_device);
+        return platform_device_register(&omap_wdt_device);
-        return 0;
 }
 subsys_initcall(omap_init_wdt);
 #endif
diff --git a/arch/arm/mach-omap1/include/mach/camera.h b/arch/arm/mach-omap1/include/mach/camera.h
index fd54b452eb22..847d00f0bb0a 100644
--- a/arch/arm/mach-omap1/include/mach/camera.h
+++ b/arch/arm/mach-omap1/include/mach/camera.h
@@ -1,6 +1,8 @@
 #ifndef __ASM_ARCH_CAMERA_H_
 #define __ASM_ARCH_CAMERA_H_
+#include <media/omap1_camera.h>
 void omap1_camera_init(void *);
 static inline void omap1_set_camera_info(struct omap1_cam_platform_data *info)
diff --git a/arch/arm/mach-omap1/include/mach/vmalloc.h b/arch/arm/mach-omap1/include/mach/vmalloc.h
index b001f67d695b..22ec4a479577 100644
--- a/arch/arm/mach-omap1/include/mach/vmalloc.h
+++ b/arch/arm/mach-omap1/include/mach/vmalloc.h
@@ -17,4 +17,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */
-#define VMALLOC_END     0xd8000000
+#define VMALLOC_END     0xd8000000UL
diff --git a/arch/arm/mach-omap2/board-devkit8000.c b/arch/arm/mach-omap2/board-devkit8000.c
index 067f4379c87f..53ac762518bd 100644
--- a/arch/arm/mach-omap2/board-devkit8000.c
+++ b/arch/arm/mach-omap2/board-devkit8000.c
@@ -242,9 +242,6 @@ static int devkit8000_twl_gpio_setup(struct device *dev,
        mmc[0].gpio_cd = gpio + 0;
        omap2_hsmmc_init(mmc);
-        /* link regulators to MMC adapters */
-        devkit8000_vmmc1_supply.dev = mmc[0].dev;
        /* TWL4030_GPIO_MAX + 1 == ledB, PMU_STAT (out, active low LED) */
        gpio_leds[2].gpio = gpio + TWL4030_GPIO_MAX + 1;
diff --git a/arch/arm/mach-omap2/include/mach/vmalloc.h b/arch/arm/mach-omap2/include/mach/vmalloc.h
index 4da31e997efe..866319947760 100644
--- a/arch/arm/mach-omap2/include/mach/vmalloc.h
+++ b/arch/arm/mach-omap2/include/mach/vmalloc.h
@@ -17,4 +17,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */
-#define VMALLOC_END       0xf8000000
+#define VMALLOC_END       0xf8000000UL
diff --git a/arch/arm/mach-orion5x/mpp.c b/arch/arm/mach-orion5x/mpp.c
index bc4c3b9aaf83..db485d3b8144 100644
--- a/arch/arm/mach-orion5x/mpp.c
+++ b/arch/arm/mach-orion5x/mpp.c
@@ -127,7 +127,7 @@ void __init orion5x_mpp_conf(struct orion5x_mpp_mode *mode)
        /* Initialize gpiolib. */
        orion_gpio_init();
-        while (mode->mpp >= 0) {
+        for ( ; mode->mpp >= 0; mode++) {
                u32 *reg;
                int num_type;
                int shift;
@@ -160,8 +160,6 @@ void __init orion5x_mpp_conf(struct orion5x_mpp_mode *mode)
                        orion_gpio_set_unused(mode->mpp);
                orion_gpio_set_valid(mode->mpp, !!(mode->type == MPP_GPIO));
-                mode++;
        }
        writel(mpp_0_7_ctrl, MPP_0_7_CTRL);
diff --git a/arch/arm/mach-orion5x/ts78xx-setup.c b/arch/arm/mach-orion5x/ts78xx-setup.c
index 16f1bd5324be..c1c1cd04bdde 100644
--- a/arch/arm/mach-orion5x/ts78xx-setup.c
+++ b/arch/arm/mach-orion5x/ts78xx-setup.c
@@ -239,7 +239,7 @@ static struct platform_nand_data ts78xx_ts_nand_data = {
 static struct resource ts78xx_ts_nand_resources = {
        .start          = TS_NAND_DATA,
        .end            = TS_NAND_DATA + 4,
-        .flags          = IORESOURCE_IO,
+        .flags          = IORESOURCE_MEM,
 };
 static struct platform_device ts78xx_ts_nand_device = {
diff --git a/arch/arm/mach-pnx4008/include/mach/vmalloc.h b/arch/arm/mach-pnx4008/include/mach/vmalloc.h
index 31b65ee07b0b..184913c71141 100644
--- a/arch/arm/mach-pnx4008/include/mach/vmalloc.h
+++ b/arch/arm/mach-pnx4008/include/mach/vmalloc.h
@@ -17,4 +17,4 @@
 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
 * area for the same reason. ;)
 */
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
diff --git a/arch/arm/mach-pxa/cm-x2xx.c b/arch/arm/mach-pxa/cm-x2xx.c
index ac5598ce9724..d34b99febeb9 100644
--- a/arch/arm/mach-pxa/cm-x2xx.c
+++ b/arch/arm/mach-pxa/cm-x2xx.c
@@ -476,8 +476,6 @@ static void __init cmx2xx_init(void)
 static void __init cmx2xx_init_irq(void)
 {
-        pxa27x_init_irq();
        if (cpu_is_pxa25x()) {
                pxa25x_init_irq();
                cmx2xx_pci_init_irq(CMX255_GPIO_IT8152_IRQ);
diff --git a/arch/arm/mach-pxa/saar.c b/arch/arm/mach-pxa/saar.c
index 4b521e045d75..ffa50e633ee6 100644
--- a/arch/arm/mach-pxa/saar.c
+++ b/arch/arm/mach-pxa/saar.c
@@ -116,7 +116,7 @@ static struct platform_device smc91x_device = {
        },
 };
-#if defined(CONFIG_FB_PXA) || (CONFIG_FB_PXA_MODULE)
+#if defined(CONFIG_FB_PXA) || defined(CONFIG_FB_PXA_MODULE)
 static uint16_t lcd_power_on[] = {
        /* single frame */
        SMART_CMD_NOOP,
diff --git a/arch/arm/mach-rpc/include/mach/vmalloc.h b/arch/arm/mach-rpc/include/mach/vmalloc.h
index 3bcd86fadb81..fb700228637a 100644
--- a/arch/arm/mach-rpc/include/mach/vmalloc.h
+++ b/arch/arm/mach-rpc/include/mach/vmalloc.h
@@ -7,4 +7,4 @@
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */
-#define VMALLOC_END       0xdc000000
+#define VMALLOC_END       0xdc000000UL
diff --git a/arch/arm/mach-s3c64xx/Kconfig b/arch/arm/mach-s3c64xx/Kconfig
index 1ca7bdc6485c..579d2f0f4dd0 100644
--- a/arch/arm/mach-s3c64xx/Kconfig
+++ b/arch/arm/mach-s3c64xx/Kconfig
@@ -143,7 +143,7 @@ config MACH_SMDK6410
        select S3C_DEV_USB_HSOTG
        select S3C_DEV_WDT
        select SAMSUNG_DEV_KEYPAD
-        select HAVE_S3C2410_WATCHDOG
+        select HAVE_S3C2410_WATCHDOG if WATCHDOG
        select S3C64XX_SETUP_SDHCI
        select S3C64XX_SETUP_I2C1
        select S3C64XX_SETUP_IDE
diff --git a/arch/arm/mach-shark/include/mach/vmalloc.h b/arch/arm/mach-shark/include/mach/vmalloc.h
index 8e845b6a7cb5..b10df988526d 100644
--- a/arch/arm/mach-shark/include/mach/vmalloc.h
+++ b/arch/arm/mach-shark/include/mach/vmalloc.h
@@ -1,4 +1,4 @@
 /*
 * arch/arm/mach-shark/include/mach/vmalloc.h
 */
-#define VMALLOC_END       0xd0000000
+#define VMALLOC_END       0xd0000000UL
diff --git a/arch/arm/mach-shmobile/board-ap4evb.c b/arch/arm/mach-shmobile/board-ap4evb.c
index 32d9e2816e56..d3260542b943 100644
--- a/arch/arm/mach-shmobile/board-ap4evb.c
+++ b/arch/arm/mach-shmobile/board-ap4evb.c
@@ -163,11 +163,13 @@ static struct mtd_partition nor_flash_partitions[] = {
                .name           = "loader",
                .offset         = 0x00000000,
                .size           = 512 * 1024,
+                .mask_flags     = MTD_WRITEABLE,
        },
        {
                .name           = "bootenv",
                .offset         = MTDPART_OFS_APPEND,
                .size           = 512 * 1024,
+                .mask_flags     = MTD_WRITEABLE,
        },
        {
                .name           = "kernel_ro",
@@ -581,6 +583,10 @@ static int fsi_set_rate(int is_porta, int rate)
                return -EINVAL;
        switch (rate) {
+        case 44100:
+                clk_set_rate(fsib_clk, clk_round_rate(fsib_clk, 11283000));
+                ret = SH_FSI_ACKMD_256 | SH_FSI_BPFMD_64;
+                break;
        case 48000:
                clk_set_rate(fsib_clk, clk_round_rate(fsib_clk, 85428000));
                clk_set_rate(fdiv_clk, clk_round_rate(fdiv_clk, 12204000));
diff --git a/arch/arm/mach-shmobile/clock-sh7372.c b/arch/arm/mach-shmobile/clock-sh7372.c
index 7db31e6c6bf2..b25ce90a346e 100644
--- a/arch/arm/mach-shmobile/clock-sh7372.c
+++ b/arch/arm/mach-shmobile/clock-sh7372.c
@@ -220,8 +220,7 @@ static void pllc2_disable(struct clk *clk)
        __raw_writel(__raw_readl(PLLC2CR) & ~0x80000000, PLLC2CR);
 }
-static int pllc2_set_rate(struct clk *clk,
+static int pllc2_set_rate(struct clk *clk, unsigned long rate)
-                          unsigned long rate, int algo_id)
 {
        unsigned long value;
        int idx;
@@ -463,8 +462,7 @@ static int fsidiv_enable(struct clk *clk)
        return 0;
 }
-static int fsidiv_set_rate(struct clk *clk,
+static int fsidiv_set_rate(struct clk *clk, unsigned long rate)
-                           unsigned long rate, int algo_id)
 {
        int idx;
diff --git a/arch/arm/mach-shmobile/intc-sh7372.c b/arch/arm/mach-shmobile/intc-sh7372.c
index 4cd3cae38e72..30b2f400666a 100644
--- a/arch/arm/mach-shmobile/intc-sh7372.c
+++ b/arch/arm/mach-shmobile/intc-sh7372.c
@@ -98,7 +98,7 @@ static struct intc_vect intca_vectors[] __initdata = {
        INTC_VECT(IRQ14A, 0x03c0), INTC_VECT(IRQ15A, 0x03e0),
        INTC_VECT(IRQ16A, 0x3200), INTC_VECT(IRQ17A, 0x3220),
        INTC_VECT(IRQ18A, 0x3240), INTC_VECT(IRQ19A, 0x3260),
-        INTC_VECT(IRQ20A, 0x3280), INTC_VECT(IRQ31A, 0x32a0),
+        INTC_VECT(IRQ20A, 0x3280), INTC_VECT(IRQ21A, 0x32a0),
        INTC_VECT(IRQ22A, 0x32c0), INTC_VECT(IRQ23A, 0x32e0),
        INTC_VECT(IRQ24A, 0x3300), INTC_VECT(IRQ25A, 0x3320),
        INTC_VECT(IRQ26A, 0x3340), INTC_VECT(IRQ27A, 0x3360),
diff --git a/arch/arm/mach-ux500/cpu.c b/arch/arm/mach-ux500/cpu.c
index 73fb1a551ec6..608a1372b172 100644
--- a/arch/arm/mach-ux500/cpu.c
+++ b/arch/arm/mach-ux500/cpu.c
@@ -75,14 +75,14 @@ void __init ux500_init_irq(void)
 static inline void ux500_cache_wait(void __iomem *reg, unsigned long mask)
 {
        /* wait for the operation to complete */
-        while (readl(reg) & mask)
+        while (readl_relaxed(reg) & mask)
                ;
 }
 static inline void ux500_cache_sync(void)
 {
        void __iomem *base = __io_address(UX500_L2CC_BASE);
-        writel(0, base + L2X0_CACHE_SYNC);
+        writel_relaxed(0, base + L2X0_CACHE_SYNC);
        ux500_cache_wait(base + L2X0_CACHE_SYNC, 1);
 }
@@ -107,7 +107,7 @@ static void ux500_l2x0_inv_all(void)
        uint32_t l2x0_way_mask = (1<<16) - 1;   /* Bitmask of active ways */
        /* invalidate all ways */
-        writel(l2x0_way_mask, l2x0_base + L2X0_INV_WAY);
+        writel_relaxed(l2x0_way_mask, l2x0_base + L2X0_INV_WAY);
        ux500_cache_wait(l2x0_base + L2X0_INV_WAY, l2x0_way_mask);
        ux500_cache_sync();
 }
diff --git a/arch/arm/mach-versatile/include/mach/vmalloc.h b/arch/arm/mach-versatile/include/mach/vmalloc.h
index ebd8a2543d3b..7d8e069ad51b 100644
--- a/arch/arm/mach-versatile/include/mach/vmalloc.h
+++ b/arch/arm/mach-versatile/include/mach/vmalloc.h
@@ -18,4 +18,4 @@
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
-#define VMALLOC_END             0xd8000000
+#define VMALLOC_END             0xd8000000UL
diff --git a/arch/arm/mach-vexpress/ct-ca9x4.c b/arch/arm/mach-vexpress/ct-ca9x4.c
index c2e405a9e025..fd25ccd7272f 100644
--- a/arch/arm/mach-vexpress/ct-ca9x4.c
+++ b/arch/arm/mach-vexpress/ct-ca9x4.c
@@ -54,7 +54,9 @@ static struct map_desc ct_ca9x4_io_desc[] __initdata = {
 static void __init ct_ca9x4_map_io(void)
 {
+#ifdef CONFIG_LOCAL_TIMERS
        twd_base = MMIO_P2V(A9_MPCORE_TWD);
+#endif
        v2m_map_io(ct_ca9x4_io_desc, ARRAY_SIZE(ct_ca9x4_io_desc));
 }
diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index e4dd0646e859..ac6a36142fcd 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -198,7 +198,7 @@ __dma_alloc_remap(struct page *page, size_t size, gfp_t gfp, pgprot_t prot)
         * fragmentation of the DMA space, and also prevents allocations
         * smaller than a section from crossing a section boundary.
         */
-        bit = fls(size - 1) + 1;
+        bit = fls(size - 1);
        if (bit > SECTION_SHIFT)
                bit = SECTION_SHIFT;
        align = 1 << bit;
diff --git a/arch/arm/mm/ioremap.c b/arch/arm/mm/ioremap.c
index 17e7b0b57e49..55c17a6fb22f 100644
--- a/arch/arm/mm/ioremap.c
+++ b/arch/arm/mm/ioremap.c
@@ -206,8 +206,8 @@ void __iomem * __arm_ioremap_pfn_caller(unsigned long pfn,
         */
        if (pfn_valid(pfn)) {
                printk(KERN_WARNING "BUG: Your driver calls ioremap() on system memory.  This leads\n"
-                       KERN_WARNING "to architecturally unpredictable behaviour on ARMv6+, and ioremap()\n"
+                       "to architecturally unpredictable behaviour on ARMv6+, and ioremap()\n"
-                       KERN_WARNING "will fail in the next kernel release.  Please fix your driver.\n");
+                       "will fail in the next kernel release.  Please fix your driver.\n");
                WARN_ON(1);
        }
diff --git a/arch/arm/plat-mxc/devices/platform-imx-dma.c b/arch/arm/plat-mxc/devices/platform-imx-dma.c
index 02d989018059..3a705c7877dd 100644
--- a/arch/arm/plat-mxc/devices/platform-imx-dma.c
+++ b/arch/arm/plat-mxc/devices/platform-imx-dma.c
@@ -12,15 +12,7 @@
 #include <mach/hardware.h>
 #include <mach/devices-common.h>
-#ifdef SDMA_IS_MERGED
 #include <mach/sdma.h>
-#else
-struct sdma_platform_data {
-        int sdma_version;
-        char *cpu_name;
-        int to_version;
-};
-#endif
 struct imx_imx_sdma_data {
        resource_size_t iobase;
diff --git a/arch/arm/plat-mxc/devices/platform-spi_imx.c b/arch/arm/plat-mxc/devices/platform-spi_imx.c
index e48340ec331e..17f724c9452d 100644
--- a/arch/arm/plat-mxc/devices/platform-spi_imx.c
+++ b/arch/arm/plat-mxc/devices/platform-spi_imx.c
@@ -27,6 +27,7 @@ const struct imx_spi_imx_data imx21_cspi_data[] __initconst = {
        imx_spi_imx_data_entry(MX21, CSPI, "imx21-cspi", _id, _hwid, SZ_4K)
        imx21_cspi_data_entry(0, 1),
        imx21_cspi_data_entry(1, 2),
+};
 #endif
 #ifdef CONFIG_ARCH_MX25
diff --git a/arch/arm/plat-nomadik/timer.c b/arch/arm/plat-nomadik/timer.c
index aedf9c1d645e..63cdc6025bd7 100644
--- a/arch/arm/plat-nomadik/timer.c
+++ b/arch/arm/plat-nomadik/timer.c
@@ -3,6 +3,7 @@
 *
 * Copyright (C) 2008 STMicroelectronics
 * Copyright (C) 2010 Alessandro Rubini
+ * Copyright (C) 2010 Linus Walleij for ST-Ericsson
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2, as
@@ -16,11 +17,13 @@
 #include <linux/clk.h>
 #include <linux/jiffies.h>
 #include <linux/err.h>
+#include <linux/cnt32_to_63.h>
+#include <linux/timer.h>
 #include <asm/mach/time.h>
 #include <plat/mtu.h>
-void __iomem *mtu_base; /* ssigned by machine code */
+void __iomem *mtu_base; /* Assigned by machine code */
 /*
 * Kernel assumes that sched_clock can be called early
@@ -48,16 +51,82 @@ static struct clocksource nmdk_clksrc = {
 /*
 * Override the global weak sched_clock symbol with this
 * local implementation which uses the clocksource to get some
- * better resolution when scheduling the kernel. We accept that
+ * better resolution when scheduling the kernel.
- * this wraps around for now, since it is just a relative time
+ *
- * stamp. (Inspired by OMAP implementation.)
+ * Because the hardware timer period may be quite short
+ * (32.3 secs on the 133 MHz MTU timer selection on ux500)
+ * and because cnt32_to_63() needs to be called at least once per
+ * half period to work properly, a kernel keepwarm() timer is set up
+ * to ensure this requirement is always met.
+ *
+ * Also the sched_clock timer will wrap around at some point,
+ * here we set it to run continously for a year.
 */
+#define SCHED_CLOCK_MIN_WRAP 3600*24*365
+static struct timer_list cnt32_to_63_keepwarm_timer;
+static u32 sched_mult;
+static u32 sched_shift;
 unsigned long long notrace sched_clock(void)
 {
-        return clocksource_cyc2ns(nmdk_clksrc.read(
+        u64 cycles;
-                                  &nmdk_clksrc),
-                                  nmdk_clksrc.mult,
+        if (unlikely(!mtu_base))
-                                  nmdk_clksrc.shift);
+                return 0;
+        cycles = cnt32_to_63(-readl(mtu_base + MTU_VAL(0)));
+        /*
+         * sched_mult is guaranteed to be even so will
+         * shift out bit 63
+         */
+        return (cycles * sched_mult) >> sched_shift;
+}
+/* Just kick sched_clock every so often */
+static void cnt32_to_63_keepwarm(unsigned long data)
+{
+        mod_timer(&cnt32_to_63_keepwarm_timer, round_jiffies(jiffies + data));
+        (void) sched_clock();
+}
+/*
+ * Set up a timer to keep sched_clock():s 32_to_63 algorithm warm
+ * once in half a 32bit timer wrap interval.
+ */
+static void __init nmdk_sched_clock_init(unsigned long rate)
+{
+        u32 v;
+        unsigned long delta;
+        u64 days;
+        /* Find the apropriate mult and shift factors */
+        clocks_calc_mult_shift(&sched_mult, &sched_shift,
+                               rate, NSEC_PER_SEC, SCHED_CLOCK_MIN_WRAP);
+        /* We need to multiply by an even number to get rid of bit 63 */
+        if (sched_mult & 1)
+                sched_mult++;
+        /* Let's see what we get, take max counter and scale it */
+        days = (0xFFFFFFFFFFFFFFFFLLU * sched_mult) >> sched_shift;
+        do_div(days, NSEC_PER_SEC);
+        do_div(days, (3600*24));
+        pr_info("sched_clock: using %d bits @ %lu Hz wrap in %lu days\n",
+                (64 - sched_shift), rate, (unsigned long) days);
+        /*
+         * Program a timer to kick us at half 32bit wraparound
+         * Formula: seconds per wrap = (2^32) / f
+         */
+        v = 0xFFFFFFFFUL / rate;
+        /* We want half of the wrap time to keep cnt32_to_63 warm */
+        v /= 2;
+        pr_debug("sched_clock: prescaled timer rate: %lu Hz, "
+                 "initialize keepwarm timer every %d seconds\n", rate, v);
+        /* Convert seconds to jiffies */
+        delta = msecs_to_jiffies(v*1000);
+        setup_timer(&cnt32_to_63_keepwarm_timer, cnt32_to_63_keepwarm, delta);
+        mod_timer(&cnt32_to_63_keepwarm_timer, round_jiffies(jiffies + delta));
 }
 /* Clockevent device: use one-shot mode */
@@ -161,13 +230,15 @@ void __init nmdk_timer_init(void)
        writel(0, mtu_base + MTU_BGLR(0));
        writel(cr | MTU_CRn_ENA, mtu_base + MTU_CR(0));
-        /* Now the scheduling clock is ready */
+        /* Now the clock source is ready */
        nmdk_clksrc.read = nmdk_read_timer;
        if (clocksource_register(&nmdk_clksrc))
                pr_err("timer: failed to initialize clock source %s\n",
                       nmdk_clksrc.name);
+        nmdk_sched_clock_init(rate);
        /* Timer 1 is used for events */
        clockevents_calc_mult_shift(&nmdk_clkevt, rate, MTU_MIN_RANGE);
diff --git a/arch/arm/plat-omap/devices.c b/arch/arm/plat-omap/devices.c
index 6f42a18b8aa4..fc819120978d 100644
--- a/arch/arm/plat-omap/devices.c
+++ b/arch/arm/plat-omap/devices.c
@@ -284,12 +284,14 @@ void __init omap_dsp_reserve_sdram_memblock(void)
        if (!size)
                return;
-        paddr = __memblock_alloc_base(size, SZ_1M, MEMBLOCK_REAL_LIMIT);
+        paddr = memblock_alloc(size, SZ_1M);
        if (!paddr) {
                pr_err("%s: failed to reserve %x bytes\n",
                                __func__, size);
                return;
        }
+        memblock_free(paddr, size);
+        memblock_remove(paddr, size);
        omap_dsp_phys_mempool_base = paddr;
 }
diff --git a/arch/arm/plat-omap/dma.c b/arch/arm/plat-omap/dma.c
index f5c5b8da9a87..2c2826571d45 100644
--- a/arch/arm/plat-omap/dma.c
+++ b/arch/arm/plat-omap/dma.c
@@ -1983,6 +1983,8 @@ static int omap2_dma_handle_ch(int ch)
        dma_write(OMAP2_DMA_CSR_CLEAR_MASK, CSR(ch));
        dma_write(1 << ch, IRQSTATUS_L0);
+        /* read back the register to flush the write */
+        dma_read(IRQSTATUS_L0);
        /* If the ch is not chained then chain_id will be -1 */
        if (dma_chan[ch].chain_id != -1) {
diff --git a/arch/arm/plat-orion/include/plat/pcie.h b/arch/arm/plat-orion/include/plat/pcie.h
index 3ebfef72b4e7..cc99163e73fd 100644
--- a/arch/arm/plat-orion/include/plat/pcie.h
+++ b/arch/arm/plat-orion/include/plat/pcie.h
@@ -11,12 +11,15 @@
 #ifndef __PLAT_PCIE_H
 #define __PLAT_PCIE_H
+struct pci_bus;
 u32 orion_pcie_dev_id(void __iomem *base);
 u32 orion_pcie_rev(void __iomem *base);
 int orion_pcie_link_up(void __iomem *base);
 int orion_pcie_x4_mode(void __iomem *base);
 int orion_pcie_get_local_bus_nr(void __iomem *base);
 void orion_pcie_set_local_bus_nr(void __iomem *base, int nr);
+void orion_pcie_reset(void __iomem *base);
 void orion_pcie_setup(void __iomem *base,
                      struct mbus_dram_target_info *dram);
 int orion_pcie_rd_conf(void __iomem *base, struct pci_bus *bus,
diff --git a/arch/arm/plat-orion/pcie.c b/arch/arm/plat-orion/pcie.c
index 779553a1595e..af2d733c50b5 100644
--- a/arch/arm/plat-orion/pcie.c
+++ b/arch/arm/plat-orion/pcie.c
@@ -182,11 +182,6 @@ void __init orion_pcie_setup(void __iomem *base,
        u32 mask;
        /*
-         * soft reset PCIe unit
-         */
-        orion_pcie_reset(base);
-        /*
         * Point PCIe unit MBUS decode windows to DRAM space.
         */
        orion_pcie_setup_wins(base, dram);
diff --git a/arch/blackfin/kernel/process.c b/arch/blackfin/kernel/process.c
index cd0c090ebc54..b407bc8ad918 100644
--- a/arch/blackfin/kernel/process.c
+++ b/arch/blackfin/kernel/process.c
@@ -7,7 +7,6 @@
 */
 #include <linux/module.h>
-#include <linux/smp_lock.h>
 #include <linux/unistd.h>
 #include <linux/user.h>
 #include <linux/uaccess.h>
diff --git a/arch/frv/kernel/process.c b/arch/frv/kernel/process.c
index 2b63b0191f52..efad12071c2e 100644
--- a/arch/frv/kernel/process.c
+++ b/arch/frv/kernel/process.c
@@ -16,7 +16,6 @@
 #include <linux/kernel.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
 #include <linux/ptrace.h>
diff --git a/arch/h8300/kernel/process.c b/arch/h8300/kernel/process.c
index 97478138e361..933bd388efb2 100644
--- a/arch/h8300/kernel/process.c
+++ b/arch/h8300/kernel/process.c
@@ -28,7 +28,6 @@
 #include <linux/kernel.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
 #include <linux/ptrace.h>
diff --git a/arch/ia64/hp/sim/simscsi.c b/arch/ia64/hp/sim/simscsi.c
index 3a078ad3aa44..331de723c676 100644
--- a/arch/ia64/hp/sim/simscsi.c
+++ b/arch/ia64/hp/sim/simscsi.c
@@ -202,7 +202,7 @@ simscsi_readwrite10 (struct scsi_cmnd *sc, int mode)
 }
 static int
-simscsi_queuecommand (struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
+simscsi_queuecommand_lck (struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
 {
        unsigned int target_id = sc->device->id;
        char fname[MAX_ROOT_LEN+16];
@@ -326,6 +326,8 @@ simscsi_queuecommand (struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
        return 0;
 }
+static DEF_SCSI_QCMD(simscsi_queuecommand)
 static int
 simscsi_host_reset (struct scsi_cmnd *sc)
 {
diff --git a/arch/m68k/kernel/process.c b/arch/m68k/kernel/process.c
index 18732ab23292..c2a1fc23dd75 100644
--- a/arch/m68k/kernel/process.c
+++ b/arch/m68k/kernel/process.c
@@ -18,7 +18,6 @@
 #include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
 #include <linux/ptrace.h>
diff --git a/arch/m68knommu/kernel/process.c b/arch/m68knommu/kernel/process.c
index 6d3390590e5b..e2a63af5d517 100644
--- a/arch/m68knommu/kernel/process.c
+++ b/arch/m68knommu/kernel/process.c
@@ -19,7 +19,6 @@
 #include <linux/kernel.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
 #include <linux/ptrace.h>
diff --git a/arch/mn10300/kernel/process.c b/arch/mn10300/kernel/process.c
index 0d0f8049a17b..e1b14a6ed544 100644
--- a/arch/mn10300/kernel/process.c
+++ b/arch/mn10300/kernel/process.c
@@ -14,7 +14,6 @@
 #include <linux/kernel.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
 #include <linux/ptrace.h>
diff --git a/arch/parisc/hpux/sys_hpux.c b/arch/parisc/hpux/sys_hpux.c
index ba430a03bc7a..30394081d9b6 100644
--- a/arch/parisc/hpux/sys_hpux.c
+++ b/arch/parisc/hpux/sys_hpux.c
@@ -28,7 +28,6 @@
 #include <linux/namei.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
-#include <linux/smp_lock.h>
 #include <linux/syscalls.h>
 #include <linux/utsname.h>
 #include <linux/vfs.h>
diff --git a/arch/parisc/kernel/sys_parisc32.c b/arch/parisc/kernel/sys_parisc32.c
index 9779ece2b070..88a0ad14a9c9 100644
--- a/arch/parisc/kernel/sys_parisc32.c
+++ b/arch/parisc/kernel/sys_parisc32.c
@@ -20,7 +20,6 @@
 #include <linux/times.h>
 #include <linux/time.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/sem.h>
 #include <linux/msg.h>
 #include <linux/shm.h>
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index b6447190e1a2..e625e9e034ae 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -4,6 +4,10 @@ config PPC32
        bool
        default y if !PPC64
+config 32BIT
+        bool
+        default y if PPC32
 config 64BIT
        bool
        default y if PPC64
diff --git a/arch/powerpc/boot/div64.S b/arch/powerpc/boot/div64.S
index 722f360a32a9..d271ab542673 100644
--- a/arch/powerpc/boot/div64.S
+++ b/arch/powerpc/boot/div64.S
@@ -33,9 +33,10 @@ __div64_32:
        cntlzw  r0,r5           # we are shifting the dividend right
        li      r10,-1          # to make it < 2^32, and shifting
        srw     r10,r10,r0      # the divisor right the same amount,
-        add     r9,r4,r10       # rounding up (so the estimate cannot
+        addc    r9,r4,r10       # rounding up (so the estimate cannot
        andc    r11,r6,r10      # ever be too large, only too small)
        andc    r9,r9,r10
+        addze   r9,r9
        or      r11,r5,r11
        rotlw   r9,r9,r0
        rotlw   r11,r11,r0
diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
index 7a9db64f3f04..42850ee00ada 100644
--- a/arch/powerpc/kernel/kgdb.c
+++ b/arch/powerpc/kernel/kgdb.c
@@ -337,7 +337,7 @@ char *dbg_get_reg(int regno, void *mem, struct pt_regs *regs)
                /* FP registers 32 -> 63 */
 #if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_SPE)
                if (current)
-                        memcpy(mem, current->thread.evr[regno-32],
+                        memcpy(mem, &current->thread.evr[regno-32],
                                        dbg_reg_def[regno].size);
 #else
                /* fp registers not used by kernel, leave zero */
@@ -362,7 +362,7 @@ int dbg_set_reg(int regno, void *mem, struct pt_regs *regs)
        if (regno >= 32 && regno < 64) {
                /* FP registers 32 -> 63 */
 #if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_SPE)
-                memcpy(current->thread.evr[regno-32], mem,
+                memcpy(&current->thread.evr[regno-32], mem,
                                dbg_reg_def[regno].size);
 #else
                /* fp registers not used by kernel, leave zero */
diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index 2a178b0ebcdf..ce6f61c6f871 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -497,9 +497,8 @@ static void __init emergency_stack_init(void)
 }
 /*
- * Called into from start_kernel, after lock_kernel has been called.
+ * Called into from start_kernel this initializes bootmem, which is used
- * Initializes bootmem, which is unsed to manage page allocation until
+ * to manage page allocation until mem_init is called.
- * mem_init is called.
 */
 void __init setup_arch(char **cmdline_p)
 {
diff --git a/arch/powerpc/kernel/sys_ppc32.c b/arch/powerpc/kernel/sys_ppc32.c
index b1b6043a56c4..4e5bf1edc0f2 100644
--- a/arch/powerpc/kernel/sys_ppc32.c
+++ b/arch/powerpc/kernel/sys_ppc32.c
@@ -23,7 +23,6 @@
 #include <linux/resource.h>
 #include <linux/times.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/sem.h>
 #include <linux/msg.h>
 #include <linux/shm.h>
diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
index 83f534d862db..5e9584405c45 100644
--- a/arch/powerpc/mm/hash_utils_64.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -1123,7 +1123,7 @@ void hash_preload(struct mm_struct *mm, unsigned long ea,
        else
 #endif /* CONFIG_PPC_HAS_HASH_64K */
                rc = __hash_page_4K(ea, access, vsid, ptep, trap, local, ssize,
-                                    subpage_protection(pgdir, ea));
+                                    subpage_protection(mm, ea));
        /* Dump some info in case of hash insertion failure, they should
         * never happen so it is really useful to know if/when they do
diff --git a/arch/powerpc/mm/tlb_low_64e.S b/arch/powerpc/mm/tlb_low_64e.S
index 8b04c54e596f..8526bd9d2aa3 100644
--- a/arch/powerpc/mm/tlb_low_64e.S
+++ b/arch/powerpc/mm/tlb_low_64e.S
@@ -138,8 +138,11 @@
        cmpldi  cr0,r15,0                       /* Check for user region */
        std     r14,EX_TLB_ESR(r12)             /* write crazy -1 to frame */
        beq     normal_tlb_miss
+        li      r11,_PAGE_PRESENT|_PAGE_BAP_SX  /* Base perm */
+        oris    r11,r11,_PAGE_ACCESSED@h
        /* XXX replace the RMW cycles with immediate loads + writes */
-1:      mfspr   r10,SPRN_MAS1
+        mfspr   r10,SPRN_MAS1
        cmpldi  cr0,r15,8                       /* Check for vmalloc region */
        rlwinm  r10,r10,0,16,1                  /* Clear TID */
        mtspr   SPRN_MAS1,r10
diff --git a/arch/powerpc/mm/tlb_nohash.c b/arch/powerpc/mm/tlb_nohash.c
index 36c0c449a899..2a030d89bbc6 100644
--- a/arch/powerpc/mm/tlb_nohash.c
+++ b/arch/powerpc/mm/tlb_nohash.c
@@ -585,6 +585,6 @@ void setup_initial_memory_limit(phys_addr_t first_memblock_base,
        ppc64_rma_size = min_t(u64, first_memblock_size, 0x40000000);
        /* Finally limit subsequent allocations */
-        memblock_set_current_limit(ppc64_memblock_base + ppc64_rma_size);
+        memblock_set_current_limit(first_memblock_base + ppc64_rma_size);
 }
 #endif /* CONFIG_PPC64 */
diff --git a/arch/powerpc/platforms/pseries/Kconfig b/arch/powerpc/platforms/pseries/Kconfig
index c667f0f02c34..3139814f6439 100644
--- a/arch/powerpc/platforms/pseries/Kconfig
+++ b/arch/powerpc/platforms/pseries/Kconfig
@@ -47,6 +47,12 @@ config LPARCFG
 config PPC_PSERIES_DEBUG
        depends on PPC_PSERIES && PPC_EARLY_DEBUG
        bool "Enable extra debug logging in platforms/pseries"
+        help
+          Say Y here if you want the pseries core to produce a bunch of
+          debug messages to the system log. Select this if you are having a
+          problem with the pseries core and want to see more of what is
+          going on. This does not enable debugging in lpar.c, which must
+          be manually done due to its verbosity.
        default y
 config PPC_SMLPAR
diff --git a/arch/powerpc/platforms/pseries/eeh.c b/arch/powerpc/platforms/pseries/eeh.c
index 34b7dc12e731..17a11c82e6f8 100644
--- a/arch/powerpc/platforms/pseries/eeh.c
+++ b/arch/powerpc/platforms/pseries/eeh.c
@@ -21,8 +21,6 @@
 * Please address comments and feedback to Linas Vepstas <linas@austin.ibm.com>
 */
-#undef DEBUG
 #include <linux/delay.h>
 #include <linux/init.h>
 #include <linux/list.h>
diff --git a/arch/powerpc/platforms/pseries/pci_dlpar.c b/arch/powerpc/platforms/pseries/pci_dlpar.c
index 4b7a062dee15..5fcc92a12d3e 100644
--- a/arch/powerpc/platforms/pseries/pci_dlpar.c
+++ b/arch/powerpc/platforms/pseries/pci_dlpar.c
@@ -25,8 +25,6 @@
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
-#undef DEBUG
 #include <linux/pci.h>
 #include <asm/pci-bridge.h>
 #include <asm/ppc-pci.h>
diff --git a/arch/s390/Kconfig.debug b/arch/s390/Kconfig.debug
index 45e0c6199f36..05221b13ffb1 100644
--- a/arch/s390/Kconfig.debug
+++ b/arch/s390/Kconfig.debug
@@ -6,6 +6,18 @@ config TRACE_IRQFLAGS_SUPPORT
 source "lib/Kconfig.debug"
+config STRICT_DEVMEM
+        def_bool y
+        prompt "Filter access to /dev/mem"
+        ---help---
+          This option restricts access to /dev/mem.  If this option is
+          disabled, you allow userspace access to all memory, including
+          kernel and userspace memory. Accidental memory access is likely
+          to be disastrous.
+          Memory access is required for experts who want to debug the kernel.
+          If you are unsure, say Y.
 config DEBUG_STRICT_USER_COPY_CHECKS
        bool "Strict user copy size checks"
        ---help---
diff --git a/arch/s390/include/asm/page.h b/arch/s390/include/asm/page.h
index a8729ea7e9ac..3c987e9ec8d6 100644
--- a/arch/s390/include/asm/page.h
+++ b/arch/s390/include/asm/page.h
@@ -130,6 +130,11 @@ struct page;
 void arch_free_page(struct page *page, int order);
 void arch_alloc_page(struct page *page, int order);
+static inline int devmem_is_allowed(unsigned long pfn)
+{
+        return 0;
+}
 #define HAVE_ARCH_FREE_PAGE
 #define HAVE_ARCH_ALLOC_PAGE
diff --git a/arch/s390/kernel/compat_linux.c b/arch/s390/kernel/compat_linux.c
index 1e6449c79ab6..53acaa86dd94 100644
--- a/arch/s390/kernel/compat_linux.c
+++ b/arch/s390/kernel/compat_linux.c
@@ -25,7 +25,6 @@
 #include <linux/resource.h>
 #include <linux/times.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/sem.h>
 #include <linux/msg.h>
 #include <linux/shm.h>
diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c
index d60fc4398516..2564793ec2b6 100644
--- a/arch/s390/kernel/kprobes.c
+++ b/arch/s390/kernel/kprobes.c
@@ -30,6 +30,7 @@
 #include <asm/sections.h>
 #include <linux/module.h>
 #include <linux/slab.h>
+#include <linux/hardirq.h>
 DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
 DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
@@ -212,7 +213,7 @@ static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
        /* Set the PER control regs, turns on single step for this address */
        __ctl_load(kprobe_per_regs, 9, 11);
        regs->psw.mask |= PSW_MASK_PER;
-        regs->psw.mask &= ~(PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK);
+        regs->psw.mask &= ~(PSW_MASK_IO | PSW_MASK_EXT);
 }
 static void __kprobes save_previous_kprobe(struct kprobe_ctlblk *kcb)
@@ -239,7 +240,7 @@ static void __kprobes set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
        __get_cpu_var(current_kprobe) = p;
        /* Save the interrupt and per flags */
        kcb->kprobe_saved_imask = regs->psw.mask &
-            (PSW_MASK_PER | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK);
+                (PSW_MASK_PER | PSW_MASK_IO | PSW_MASK_EXT);
        /* Save the control regs that govern PER */
        __ctl_store(kcb->kprobe_saved_ctl, 9, 11);
 }
@@ -316,8 +317,6 @@ static int __kprobes kprobe_handler(struct pt_regs *regs)
                return 1;
 ss_probe:
-        if (regs->psw.mask & (PSW_MASK_PER | PSW_MASK_IO))
-                local_irq_disable();
        prepare_singlestep(p, regs);
        kcb->kprobe_status = KPROBE_HIT_SS;
        return 1;
@@ -350,6 +349,7 @@ static int __kprobes trampoline_probe_handler(struct kprobe *p,
        struct hlist_node *node, *tmp;
        unsigned long flags, orig_ret_address = 0;
        unsigned long trampoline_address = (unsigned long)&kretprobe_trampoline;
+        kprobe_opcode_t *correct_ret_addr = NULL;
        INIT_HLIST_HEAD(&empty_rp);
        kretprobe_hash_lock(current, &head, &flags);
@@ -372,10 +372,32 @@ static int __kprobes trampoline_probe_handler(struct kprobe *p,
                        /* another task is sharing our hash bucket */
                        continue;
-                if (ri->rp && ri->rp->handler)
+                orig_ret_address = (unsigned long)ri->ret_addr;
-                        ri->rp->handler(ri, regs);
+                if (orig_ret_address != trampoline_address)
+                        /*
+                         * This is the real return address. Any other
+                         * instances associated with this task are for
+                         * other calls deeper on the call stack
+                         */
+                        break;
+        }
+        kretprobe_assert(ri, orig_ret_address, trampoline_address);
+        correct_ret_addr = ri->ret_addr;
+        hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
+                if (ri->task != current)
+                        /* another task is sharing our hash bucket */
+                        continue;
                orig_ret_address = (unsigned long)ri->ret_addr;
+                if (ri->rp && ri->rp->handler) {
+                        ri->ret_addr = correct_ret_addr;
+                        ri->rp->handler(ri, regs);
+                }
                recycle_rp_inst(ri, &empty_rp);
                if (orig_ret_address != trampoline_address) {
@@ -387,7 +409,7 @@ static int __kprobes trampoline_probe_handler(struct kprobe *p,
                        break;
                }
        }
-        kretprobe_assert(ri, orig_ret_address, trampoline_address);
        regs->psw.addr = orig_ret_address | PSW_ADDR_AMODE;
        reset_current_kprobe();
@@ -465,8 +487,6 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs)
                goto out;
        }
        reset_current_kprobe();
-        if (regs->psw.mask & (PSW_MASK_PER | PSW_MASK_IO))
-                local_irq_enable();
 out:
        preempt_enable_no_resched();
@@ -482,7 +502,7 @@ out:
        return 1;
 }
-int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
+static int __kprobes kprobe_trap_handler(struct pt_regs *regs, int trapnr)
 {
        struct kprobe *cur = kprobe_running();
        struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
@@ -508,8 +528,6 @@ int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
                        restore_previous_kprobe(kcb);
                else {
                        reset_current_kprobe();
-                        if (regs->psw.mask & (PSW_MASK_PER | PSW_MASK_IO))
-                                local_irq_enable();
                }
                preempt_enable_no_resched();
                break;
@@ -553,6 +571,18 @@ int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
        return 0;
 }
+int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
+{
+        int ret;
+        if (regs->psw.mask & (PSW_MASK_IO | PSW_MASK_EXT))
+                local_irq_disable();
+        ret = kprobe_trap_handler(regs, trapnr);
+        if (regs->psw.mask & (PSW_MASK_IO | PSW_MASK_EXT))
+                local_irq_restore(regs->psw.mask & ~PSW_MASK_PER);
+        return ret;
+}
 /*
 * Wrapper routine to for handling exceptions.
 */
@@ -560,8 +590,12 @@ int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
                                       unsigned long val, void *data)
 {
        struct die_args *args = (struct die_args *)data;
+        struct pt_regs *regs = args->regs;
        int ret = NOTIFY_DONE;
+        if (regs->psw.mask & (PSW_MASK_IO | PSW_MASK_EXT))
+                local_irq_disable();
        switch (val) {
        case DIE_BPT:
                if (kprobe_handler(args->regs))
@@ -572,16 +606,17 @@ int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
                        ret = NOTIFY_STOP;
                break;
        case DIE_TRAP:
-                /* kprobe_running() needs smp_processor_id() */
+                if (!preemptible() && kprobe_running() &&
-                preempt_disable();
+                    kprobe_trap_handler(args->regs, args->trapnr))
-                if (kprobe_running() &&
-                    kprobe_fault_handler(args->regs, args->trapnr))
                        ret = NOTIFY_STOP;
-                preempt_enable();
                break;
        default:
                break;
        }
+        if (regs->psw.mask & (PSW_MASK_IO | PSW_MASK_EXT))
+                local_irq_restore(regs->psw.mask & ~PSW_MASK_PER);
        return ret;
 }
@@ -595,6 +630,7 @@ int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
        /* setup return addr to the jprobe handler routine */
        regs->psw.addr = (unsigned long)(jp->entry) | PSW_ADDR_AMODE;
+        regs->psw.mask &= ~(PSW_MASK_IO | PSW_MASK_EXT);
        /* r14 is the function return address */
        kcb->jprobe_saved_r14 = (unsigned long)regs->gprs[14];
diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c
index 38e641cdd977..45b405ca2567 100644
--- a/arch/s390/mm/gup.c
+++ b/arch/s390/mm/gup.c
@@ -20,18 +20,17 @@
 static inline int gup_pte_range(pmd_t *pmdp, pmd_t pmd, unsigned long addr,
                unsigned long end, int write, struct page **pages, int *nr)
 {
-        unsigned long mask, result;
+        unsigned long mask;
        pte_t *ptep, pte;
        struct page *page;
-        result = write ? 0 : _PAGE_RO;
+        mask = (write ? _PAGE_RO : 0) | _PAGE_INVALID | _PAGE_SPECIAL;
-        mask = result | _PAGE_INVALID | _PAGE_SPECIAL;
        ptep = ((pte_t *) pmd_deref(pmd)) + pte_index(addr);
        do {
                pte = *ptep;
                barrier();
-                if ((pte_val(pte) & mask) != result)
+                if ((pte_val(pte) & mask) != 0)
                        return 0;
                VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
                page = pte_page(pte);
diff --git a/arch/sh/boards/mach-ecovec24/setup.c b/arch/sh/boards/mach-ecovec24/setup.c
index 2eaeb9e59585..f48c492a68d3 100644
--- a/arch/sh/boards/mach-ecovec24/setup.c
+++ b/arch/sh/boards/mach-ecovec24/setup.c
@@ -720,32 +720,6 @@ static struct platform_device camera_devices[] = {
 };
 /* FSI */
-/*
- * FSI-B use external clock which came from da7210.
- * So, we should change parent of fsi
- */
-#define FCLKBCR         0xa415000c
-static void fsimck_init(struct clk *clk)
-{
-        u32 status = __raw_readl(clk->enable_reg);
-        /* use external clock */
-        status &= ~0x000000ff;
-        status |= 0x00000080;
-        __raw_writel(status, clk->enable_reg);
-}
-static struct clk_ops fsimck_clk_ops = {
-        .init = fsimck_init,
-};
-static struct clk fsimckb_clk = {
-        .ops            = &fsimck_clk_ops,
-        .enable_reg     = (void __iomem *)FCLKBCR,
-        .rate           = 0, /* unknown */
-};
 static struct sh_fsi_platform_info fsi_info = {
        .portb_flags = SH_FSI_BRS_INV |
                       SH_FSI_OUT_SLAVE_MODE |
@@ -1264,10 +1238,10 @@ static int __init arch_setup(void)
        /* change parent of FSI B */
        clk = clk_get(NULL, "fsib_clk");
        if (!IS_ERR(clk)) {
-                clk_register(&fsimckb_clk);
+                /* 48kHz dummy clock was used to make sure 1/1 divide */
-                clk_set_parent(clk, &fsimckb_clk);
+                clk_set_rate(&sh7724_fsimckb_clk, 48000);
-                clk_set_rate(clk, 11000);
+                clk_set_parent(clk, &sh7724_fsimckb_clk);
-                clk_set_rate(&fsimckb_clk, 11000);
+                clk_set_rate(clk, 48000);
                clk_put(clk);
        }
diff --git a/arch/sh/boards/mach-se/7724/setup.c b/arch/sh/boards/mach-se/7724/setup.c
index c31d228fdfc6..673530500e27 100644
--- a/arch/sh/boards/mach-se/7724/setup.c
+++ b/arch/sh/boards/mach-se/7724/setup.c
@@ -871,14 +871,14 @@ static int __init devices_setup(void)
        /* set SPU2 clock to 83.4 MHz */
        clk = clk_get(NULL, "spu_clk");
-        if (clk) {
+        if (!IS_ERR(clk)) {
                clk_set_rate(clk, clk_round_rate(clk, 83333333));
                clk_put(clk);
        }
        /* change parent of FSI A */
        clk = clk_get(NULL, "fsia_clk");
-        if (clk) {
+        if (!IS_ERR(clk)) {
                clk_register(&fsimcka_clk);
                clk_set_parent(clk, &fsimcka_clk);
                clk_set_rate(clk, 11000);
diff --git a/arch/sh/include/asm/processor_32.h b/arch/sh/include/asm/processor_32.h
index 46d5179c9f49..e3c73cdd8c90 100644
--- a/arch/sh/include/asm/processor_32.h
+++ b/arch/sh/include/asm/processor_32.h
@@ -199,10 +199,13 @@ extern unsigned long get_wchan(struct task_struct *p);
 #define ARCH_HAS_PREFETCHW
 static inline void prefetch(void *x)
 {
-        __asm__ __volatile__ ("pref @%0\n\t" : : "r" (x) : "memory");
+        __builtin_prefetch(x, 0, 3);
 }
-#define prefetchw(x)    prefetch(x)
+static inline void prefetchw(void *x)
+{
+        __builtin_prefetch(x, 1, 3);
+}
 #endif
 #endif /* __KERNEL__ */
diff --git a/arch/sh/include/cpu-sh4/cpu/sh7724.h b/arch/sh/include/cpu-sh4/cpu/sh7724.h
index 4c27b68789b3..7eb435999426 100644
--- a/arch/sh/include/cpu-sh4/cpu/sh7724.h
+++ b/arch/sh/include/cpu-sh4/cpu/sh7724.h
@@ -303,4 +303,7 @@ enum {
        SHDMA_SLAVE_SDHI1_RX,
 };
+extern struct clk sh7724_fsimcka_clk;
+extern struct clk sh7724_fsimckb_clk;
 #endif /* __ASM_SH7724_H__ */
diff --git a/arch/sh/kernel/cpu/sh4/clock-sh4-202.c b/arch/sh/kernel/cpu/sh4/clock-sh4-202.c
index 4eabc68cd753..b601fa3978d1 100644
--- a/arch/sh/kernel/cpu/sh4/clock-sh4-202.c
+++ b/arch/sh/kernel/cpu/sh4/clock-sh4-202.c
@@ -110,7 +110,7 @@ static int shoc_clk_verify_rate(struct clk *clk, unsigned long rate)
        return 0;
 }
-static int shoc_clk_set_rate(struct clk *clk, unsigned long rate, int algo_id)
+static int shoc_clk_set_rate(struct clk *clk, unsigned long rate)
 {
        unsigned long frqcr3;
        unsigned int tmp;
diff --git a/arch/sh/kernel/cpu/sh4a/clock-sh7724.c b/arch/sh/kernel/cpu/sh4a/clock-sh7724.c
index 0fe2e9329cb2..271c0b325a9a 100644
--- a/arch/sh/kernel/cpu/sh4a/clock-sh7724.c
+++ b/arch/sh/kernel/cpu/sh4a/clock-sh7724.c
@@ -111,12 +111,21 @@ static struct clk div3_clk = {
        .parent         = &pll_clk,
 };
+/* External input clock (pin name: FSIMCKA/FSIMCKB ) */
+struct clk sh7724_fsimcka_clk = {
+};
+struct clk sh7724_fsimckb_clk = {
+};
 static struct clk *main_clks[] = {
        &r_clk,
        &extal_clk,
        &fll_clk,
        &pll_clk,
        &div3_clk,
+        &sh7724_fsimcka_clk,
+        &sh7724_fsimckb_clk,
 };
 static void div4_kick(struct clk *clk)
@@ -154,16 +163,38 @@ struct clk div4_clks[DIV4_NR] = {
        [DIV4_M1] = DIV4(FRQCRB, 4, 0x2f7c, CLK_ENABLE_ON_INIT),
 };
-enum { DIV6_V, DIV6_FA, DIV6_FB, DIV6_I, DIV6_S, DIV6_NR };
+enum { DIV6_V, DIV6_I, DIV6_S, DIV6_NR };
 static struct clk div6_clks[DIV6_NR] = {
        [DIV6_V] = SH_CLK_DIV6(&div3_clk, VCLKCR, 0),
-        [DIV6_FA] = SH_CLK_DIV6(&div3_clk, FCLKACR, 0),
-        [DIV6_FB] = SH_CLK_DIV6(&div3_clk, FCLKBCR, 0),
        [DIV6_I] = SH_CLK_DIV6(&div3_clk, IRDACLKCR, 0),
        [DIV6_S] = SH_CLK_DIV6(&div3_clk, SPUCLKCR, CLK_ENABLE_ON_INIT),
 };
+enum { DIV6_FA, DIV6_FB, DIV6_REPARENT_NR };
+/* Indices are important - they are the actual src selecting values */
+static struct clk *fclkacr_parent[] = {
+        [0] = &div3_clk,
+        [1] = NULL,
+        [2] = &sh7724_fsimcka_clk,
+        [3] = NULL,
+};
+static struct clk *fclkbcr_parent[] = {
+        [0] = &div3_clk,
+        [1] = NULL,
+        [2] = &sh7724_fsimckb_clk,
+        [3] = NULL,
+};
+static struct clk div6_reparent_clks[DIV6_REPARENT_NR] = {
+        [DIV6_FA] = SH_CLK_DIV6_EXT(&div3_clk, FCLKACR, 0,
+                                      fclkacr_parent, ARRAY_SIZE(fclkacr_parent), 6, 2),
+        [DIV6_FB] = SH_CLK_DIV6_EXT(&div3_clk, FCLKBCR, 0,
+                                      fclkbcr_parent, ARRAY_SIZE(fclkbcr_parent), 6, 2),
+};
 static struct clk mstp_clks[HWBLK_NR] = {
        SH_HWBLK_CLK(HWBLK_TLB, &div4_clks[DIV4_I], CLK_ENABLE_ON_INIT),
        SH_HWBLK_CLK(HWBLK_IC, &div4_clks[DIV4_I], CLK_ENABLE_ON_INIT),
@@ -240,8 +271,8 @@ static struct clk_lookup lookups[] = {
        /* DIV6 clocks */
        CLKDEV_CON_ID("video_clk", &div6_clks[DIV6_V]),
-        CLKDEV_CON_ID("fsia_clk", &div6_clks[DIV6_FA]),
+        CLKDEV_CON_ID("fsia_clk", &div6_reparent_clks[DIV6_FA]),
-        CLKDEV_CON_ID("fsib_clk", &div6_clks[DIV6_FB]),
+        CLKDEV_CON_ID("fsib_clk", &div6_reparent_clks[DIV6_FB]),
        CLKDEV_CON_ID("irda_clk", &div6_clks[DIV6_I]),
        CLKDEV_CON_ID("spu_clk", &div6_clks[DIV6_S]),
@@ -376,6 +407,9 @@ int __init arch_clk_init(void)
                ret = sh_clk_div6_register(div6_clks, DIV6_NR);
        if (!ret)
+                ret = sh_clk_div6_reparent_register(div6_reparent_clks, DIV6_REPARENT_NR);
+        if (!ret)
                ret = sh_hwblk_clk_register(mstp_clks, HWBLK_NR);
        return ret;
diff --git a/arch/sh/kernel/sys_sh.c b/arch/sh/kernel/sys_sh.c
index 81f58371613d..8c6a350df751 100644
--- a/arch/sh/kernel/sys_sh.c
+++ b/arch/sh/kernel/sys_sh.c
@@ -88,7 +88,7 @@ asmlinkage int sys_cacheflush(unsigned long addr, unsigned long len, int op)
        }
        if (op & CACHEFLUSH_I)
-                flush_cache_all();
+                flush_icache_range(addr, addr+len);
        up_read(&current->mm->mmap_sem);
        return 0;
diff --git a/arch/sh/kernel/vsyscall/vsyscall-trapa.S b/arch/sh/kernel/vsyscall/vsyscall-trapa.S
index 3b6eb34c43fa..3e70f851cdc6 100644
--- a/arch/sh/kernel/vsyscall/vsyscall-trapa.S
+++ b/arch/sh/kernel/vsyscall/vsyscall-trapa.S
@@ -8,9 +8,9 @@ __kernel_vsyscall:
         * fill out .eh_frame -- PFM. */
 .LEND_vsyscall:
        .size __kernel_vsyscall,.-.LSTART_vsyscall
-        .previous
        .section .eh_frame,"a",@progbits
+        .previous
 .LCIE:
        .ualong .LCIE_end - .LCIE_start
 .LCIE_start:
diff --git a/arch/sparc/kernel/leon_smp.c b/arch/sparc/kernel/leon_smp.c
index 7524689b03d2..16582d85368a 100644
--- a/arch/sparc/kernel/leon_smp.c
+++ b/arch/sparc/kernel/leon_smp.c
@@ -12,7 +12,6 @@
 #include <linux/sched.h>
 #include <linux/threads.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/interrupt.h>
 #include <linux/kernel_stat.h>
 #include <linux/init.h>
diff --git a/arch/sparc/kernel/sys_sparc32.c b/arch/sparc/kernel/sys_sparc32.c
index e6375a750d9a..6db18c6927fb 100644
--- a/arch/sparc/kernel/sys_sparc32.c
+++ b/arch/sparc/kernel/sys_sparc32.c
@@ -17,7 +17,6 @@
 #include <linux/resource.h>
 #include <linux/times.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/sem.h>
 #include <linux/msg.h>
 #include <linux/shm.h>
diff --git a/arch/sparc/kernel/sys_sparc_32.c b/arch/sparc/kernel/sys_sparc_32.c
index 675c9e11ada5..42b282fa6112 100644
--- a/arch/sparc/kernel/sys_sparc_32.c
+++ b/arch/sparc/kernel/sys_sparc_32.c
@@ -19,7 +19,6 @@
 #include <linux/mman.h>
 #include <linux/utsname.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/ipc.h>
 #include <asm/uaccess.h>
diff --git a/arch/sparc/kernel/unaligned_32.c b/arch/sparc/kernel/unaligned_32.c
index 12b9f352595f..4491f4cb2695 100644
--- a/arch/sparc/kernel/unaligned_32.c
+++ b/arch/sparc/kernel/unaligned_32.c
@@ -16,7 +16,6 @@
 #include <asm/system.h>
 #include <asm/uaccess.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/perf_event.h>
 enum direction {
diff --git a/arch/sparc/kernel/windows.c b/arch/sparc/kernel/windows.c
index b351770cbdd6..3107381e576d 100644
--- a/arch/sparc/kernel/windows.c
+++ b/arch/sparc/kernel/windows.c
@@ -9,7 +9,6 @@
 #include <linux/string.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <asm/uaccess.h>
diff --git a/arch/tile/Kconfig b/arch/tile/Kconfig
index 07ec8a865c1d..e11b5fcb70eb 100644
--- a/arch/tile/Kconfig
+++ b/arch/tile/Kconfig
@@ -329,6 +329,18 @@ endmenu  # Tilera-specific configuration
 menu "Bus options"
+config PCI
+        bool "PCI support"
+        default y
+        select PCI_DOMAINS
+        ---help---
+          Enable PCI root complex support, so PCIe endpoint devices can
+          be attached to the Tile chip.  Many, but not all, PCI devices
+          are supported under Tilera's root complex driver.
+config PCI_DOMAINS
+        bool
 config NO_IOMEM
        def_bool !PCI
diff --git a/arch/tile/include/asm/cacheflush.h b/arch/tile/include/asm/cacheflush.h
index c5741da4eeac..14a3f8556ace 100644
--- a/arch/tile/include/asm/cacheflush.h
+++ b/arch/tile/include/asm/cacheflush.h
@@ -137,4 +137,56 @@ static inline void finv_buffer(void *buffer, size_t size)
        mb_incoherent();
 }
+/*
+ * Flush & invalidate a VA range that is homed remotely on a single core,
+ * waiting until the memory controller holds the flushed values.
+ */
+static inline void finv_buffer_remote(void *buffer, size_t size)
+{
+        char *p;
+        int i;
+        /*
+         * Flush and invalidate the buffer out of the local L1/L2
+         * and request the home cache to flush and invalidate as well.
+         */
+        __finv_buffer(buffer, size);
+        /*
+         * Wait for the home cache to acknowledge that it has processed
+         * all the flush-and-invalidate requests.  This does not mean
+         * that the flushed data has reached the memory controller yet,
+         * but it does mean the home cache is processing the flushes.
+         */
+        __insn_mf();
+        /*
+         * Issue a load to the last cache line, which can't complete
+         * until all the previously-issued flushes to the same memory
+         * controller have also completed.  If we weren't striping
+         * memory, that one load would be sufficient, but since we may
+         * be, we also need to back up to the last load issued to
+         * another memory controller, which would be the point where
+         * we crossed an 8KB boundary (the granularity of striping
+         * across memory controllers).  Keep backing up and doing this
+         * until we are before the beginning of the buffer, or have
+         * hit all the controllers.
+         */
+        for (i = 0, p = (char *)buffer + size - 1;
+             i < (1 << CHIP_LOG_NUM_MSHIMS()) && p >= (char *)buffer;
+             ++i) {
+                const unsigned long STRIPE_WIDTH = 8192;
+                /* Force a load instruction to issue. */
+                *(volatile char *)p;
+                /* Jump to end of previous stripe. */
+                p -= STRIPE_WIDTH;
+                p = (char *)((unsigned long)p | (STRIPE_WIDTH - 1));
+        }
+        /* Wait for the loads (and thus flushes) to have completed. */
+        __insn_mf();
+}
 #endif /* _ASM_TILE_CACHEFLUSH_H */
diff --git a/arch/tile/include/asm/io.h b/arch/tile/include/asm/io.h
index ee43328713ab..d3cbb9b14cbe 100644
--- a/arch/tile/include/asm/io.h
+++ b/arch/tile/include/asm/io.h
@@ -55,9 +55,6 @@ extern void iounmap(volatile void __iomem *addr);
 #define ioremap_writethrough(physaddr, size)    ioremap(physaddr, size)
 #define ioremap_fullcache(physaddr, size)       ioremap(physaddr, size)
-void __iomem *ioport_map(unsigned long port, unsigned int len);
-extern inline void ioport_unmap(void __iomem *addr) {}
 #define mmiowb()
 /* Conversion between virtual and physical mappings.  */
@@ -189,12 +186,22 @@ static inline void memcpy_toio(volatile void __iomem *dst, const void *src,
 * we never run, uses them unconditionally.
 */
-static inline int ioport_panic(void)
+static inline long ioport_panic(void)
 {
        panic("inb/outb and friends do not exist on tile");
        return 0;
 }
+static inline void __iomem *ioport_map(unsigned long port, unsigned int len)
+{
+        return (void __iomem *) ioport_panic();
+}
+static inline void ioport_unmap(void __iomem *addr)
+{
+        ioport_panic();
+}
 static inline u8 inb(unsigned long addr)
 {
        return ioport_panic();
diff --git a/arch/tile/include/asm/pci-bridge.h b/arch/tile/include/asm/pci-bridge.h
deleted file mode 100644
index e853b0e2793b..000000000000
--- a/arch/tile/include/asm/pci-bridge.h
+++ /dev/null
@@ -1,117 +0,0 @@
-/*
- * Copyright 2010 Tilera Corporation. All Rights Reserved.
- *
- *   This program is free software; you can redistribute it and/or
- *   modify it under the terms of the GNU General Public License
- *   as published by the Free Software Foundation, version 2.
- *
- *   This program is distributed in the hope that it will be useful, but
- *   WITHOUT ANY WARRANTY; without even the implied warranty of
- *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
- *   NON INFRINGEMENT.  See the GNU General Public License for
- *   more details.
- */
-#ifndef _ASM_TILE_PCI_BRIDGE_H
-#define _ASM_TILE_PCI_BRIDGE_H
-#include <linux/ioport.h>
-#include <linux/pci.h>
-struct device_node;
-struct pci_controller;
-/*
- * pci_io_base returns the memory address at which you can access
- * the I/O space for PCI bus number `bus' (or NULL on error).
- */
-extern void __iomem *pci_bus_io_base(unsigned int bus);
-extern unsigned long pci_bus_io_base_phys(unsigned int bus);
-extern unsigned long pci_bus_mem_base_phys(unsigned int bus);
-/* Allocate a new PCI host bridge structure */
-extern struct pci_controller *pcibios_alloc_controller(void);
-/* Helper function for setting up resources */
-extern void pci_init_resource(struct resource *res, unsigned long start,
-                              unsigned long end, int flags, char *name);
-/* Get the PCI host controller for a bus */
-extern struct pci_controller *pci_bus_to_hose(int bus);
-/*
- * Structure of a PCI controller (host bridge)
- */
-struct pci_controller {
-        int index;              /* PCI domain number */
-        struct pci_bus *root_bus;
-        int first_busno;
-        int last_busno;
-        int hv_cfg_fd[2];       /* config{0,1} fds for this PCIe controller */
-        int hv_mem_fd;          /* fd to Hypervisor for MMIO operations */
-        struct pci_ops *ops;
-        int irq_base;           /* Base IRQ from the Hypervisor */
-        int plx_gen1;           /* flag for PLX Gen 1 configuration */
-        /* Address ranges that are routed to this controller/bridge. */
-        struct resource mem_resources[3];
-};
-static inline struct pci_controller *pci_bus_to_host(struct pci_bus *bus)
-{
-        return bus->sysdata;
-}
-extern void setup_indirect_pci_nomap(struct pci_controller *hose,
-                               void __iomem *cfg_addr, void __iomem *cfg_data);
-extern void setup_indirect_pci(struct pci_controller *hose,
-                               u32 cfg_addr, u32 cfg_data);
-extern void setup_grackle(struct pci_controller *hose);
-extern unsigned char common_swizzle(struct pci_dev *, unsigned char *);
-/*
- *   The following code swizzles for exactly one bridge.  The routine
- *   common_swizzle below handles multiple bridges.  But there are a
- *   some boards that don't follow the PCI spec's suggestion so we
- *   break this piece out separately.
- */
-static inline unsigned char bridge_swizzle(unsigned char pin,
-                unsigned char idsel)
-{
-        return (((pin-1) + idsel) % 4) + 1;
-}
-/*
- * The following macro is used to lookup irqs in a standard table
- * format for those PPC systems that do not already have PCI
- * interrupts properly routed.
- */
-/* FIXME - double check this */
-#define PCI_IRQ_TABLE_LOOKUP ({ \
-        long _ctl_ = -1; \
-        if (idsel >= min_idsel && idsel <= max_idsel && pin <= irqs_per_slot) \
-                _ctl_ = pci_irq_table[idsel - min_idsel][pin-1]; \
-        _ctl_; \
-})
-/*
- * Scan the buses below a given PCI host bridge and assign suitable
- * resources to all devices found.
- */
-extern int pciauto_bus_scan(struct pci_controller *, int);
-#ifdef CONFIG_PCI
-extern unsigned long pci_address_to_pio(phys_addr_t address);
-#else
-static inline unsigned long pci_address_to_pio(phys_addr_t address)
-{
-        return (unsigned long)-1;
-}
-#endif
-#endif /* _ASM_TILE_PCI_BRIDGE_H */
diff --git a/arch/tile/include/asm/pci.h b/arch/tile/include/asm/pci.h
index b0c15da2d5d5..c3fc458a0d32 100644
--- a/arch/tile/include/asm/pci.h
+++ b/arch/tile/include/asm/pci.h
@@ -15,7 +15,29 @@
 #ifndef _ASM_TILE_PCI_H
 #define _ASM_TILE_PCI_H
-#include <asm/pci-bridge.h>
+#include <linux/pci.h>
+/*
+ * Structure of a PCI controller (host bridge)
+ */
+struct pci_controller {
+        int index;              /* PCI domain number */
+        struct pci_bus *root_bus;
+        int first_busno;
+        int last_busno;
+        int hv_cfg_fd[2];       /* config{0,1} fds for this PCIe controller */
+        int hv_mem_fd;          /* fd to Hypervisor for MMIO operations */
+        struct pci_ops *ops;
+        int irq_base;           /* Base IRQ from the Hypervisor */
+        int plx_gen1;           /* flag for PLX Gen 1 configuration */
+        /* Address ranges that are routed to this controller/bridge. */
+        struct resource mem_resources[3];
+};
 /*
 * The hypervisor maps the entirety of CPA-space as bus addresses, so
@@ -24,56 +46,12 @@
 */
 #define PCI_DMA_BUS_IS_PHYS     1
-struct pci_controller *pci_bus_to_hose(int bus);
-unsigned char __init common_swizzle(struct pci_dev *dev, unsigned char *pinp);
 int __init tile_pci_init(void);
-void pci_iounmap(struct pci_dev *dev, void __iomem *addr);
-void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max);
-void __devinit pcibios_fixup_bus(struct pci_bus *bus);
-int __devinit _tile_cfg_read(struct pci_controller *hose,
+void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max);
-                                    int bus,
+static inline void pci_iounmap(struct pci_dev *dev, void __iomem *addr) {}
-                                    int slot,
-                                    int function,
-                                    int offset,
-                                    int size,
-                                    u32 *val);
-int __devinit _tile_cfg_write(struct pci_controller *hose,
-                                     int bus,
-                                     int slot,
-                                     int function,
-                                     int offset,
-                                     int size,
-                                     u32 val);
-/*
+void __devinit pcibios_fixup_bus(struct pci_bus *bus);
- * These are used to to config reads and writes in the early stages of
- * setup before the driver infrastructure has been set up enough to be
- * able to do config reads and writes.
- */
-#define early_cfg_read(where, size, value) \
-        _tile_cfg_read(controller, \
-                       current_bus, \
-                       pci_slot, \
-                       pci_fn, \
-                       where, \
-                       size, \
-                       value)
-#define early_cfg_write(where, size, value) \
-        _tile_cfg_write(controller, \
-                       current_bus, \
-                       pci_slot, \
-                       pci_fn, \
-                       where, \
-                       size, \
-                       value)
-#define PCICFG_BYTE     1
-#define PCICFG_WORD     2
-#define PCICFG_DWORD    4
 #define TILE_NUM_PCIE   2
@@ -88,33 +66,33 @@ static inline int pci_proc_domain(struct pci_bus *bus)
 }
 /*
- * I/O space is currently not supported.
+ * pcibios_assign_all_busses() tells whether or not the bus numbers
+ * should be reassigned, in case the BIOS didn't do it correctly, or
+ * in case we don't have a BIOS and we want to let Linux do it.
 */
+static inline int pcibios_assign_all_busses(void)
+{
+        return 1;
+}
-#define TILE_PCIE_LOWER_IO              0x0
+/*
-#define TILE_PCIE_UPPER_IO              0x10000
+ * No special bus mastering setup handling.
-#define TILE_PCIE_PCIE_IO_SIZE          0x0000FFFF
+ */
-#define _PAGE_NO_CACHE          0
-#define _PAGE_GUARDED           0
-#define pcibios_assign_all_busses()    pci_assign_all_buses
-extern int pci_assign_all_buses;
 static inline void pcibios_set_master(struct pci_dev *dev)
 {
-        /* No special bus mastering setup handling */
 }
 #define PCIBIOS_MIN_MEM         0
-#define PCIBIOS_MIN_IO          TILE_PCIE_LOWER_IO
+#define PCIBIOS_MIN_IO          0
 /*
 * This flag tells if the platform is TILEmpower that needs
 * special configuration for the PLX switch chip.
 */
-extern int blade_pci;
+extern int tile_plx_gen1;
+/* Use any cpu for PCI. */
+#define cpumask_of_pcibus(bus) cpu_online_mask
 /* implement the pci_ DMA API in terms of the generic device dma_ one */
 #include <asm-generic/pci-dma-compat.h>
@@ -122,7 +100,4 @@ extern int blade_pci;
 /* generic pci stuff */
 #include <asm-generic/pci.h>
-/* Use any cpu for PCI. */
-#define cpumask_of_pcibus(bus) cpu_online_mask
 #endif /* _ASM_TILE_PCI_H */
diff --git a/arch/tile/include/asm/processor.h b/arch/tile/include/asm/processor.h
index 1747ff3946b2..a9e7c8760334 100644
--- a/arch/tile/include/asm/processor.h
+++ b/arch/tile/include/asm/processor.h
@@ -292,8 +292,18 @@ extern int kstack_hash;
 /* Are we using huge pages in the TLB for kernel data? */
 extern int kdata_huge;
+/* Support standard Linux prefetching. */
+#define ARCH_HAS_PREFETCH
+#define prefetch(x) __builtin_prefetch(x)
 #define PREFETCH_STRIDE CHIP_L2_LINE_SIZE()
+/* Bring a value into the L1D, faulting the TLB if necessary. */
+#ifdef __tilegx__
+#define prefetch_L1(x) __insn_prefetch_l1_fault((void *)(x))
+#else
+#define prefetch_L1(x) __insn_prefetch_L1((void *)(x))
+#endif
 #else /* __ASSEMBLY__ */
 /* Do some slow action (e.g. read a slow SPR). */
diff --git a/arch/tile/include/hv/drv_xgbe_impl.h b/arch/tile/include/hv/drv_xgbe_impl.h
new file mode 100644
index 000000000000..3a73b2b44913
--- /dev/null
+++ b/arch/tile/include/hv/drv_xgbe_impl.h
@@ -0,0 +1,300 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+/**
+ * @file drivers/xgbe/impl.h
+ * Implementation details for the NetIO library.
+ */
+#ifndef __DRV_XGBE_IMPL_H__
+#define __DRV_XGBE_IMPL_H__
+#include <hv/netio_errors.h>
+#include <hv/netio_intf.h>
+#include <hv/drv_xgbe_intf.h>
+/** How many groups we have (log2). */
+#define LOG2_NUM_GROUPS (12)
+/** How many groups we have. */
+#define NUM_GROUPS (1 << LOG2_NUM_GROUPS)
+/** Number of output requests we'll buffer per tile. */
+#define EPP_REQS_PER_TILE (32)
+/** Words used in an eDMA command without checksum acceleration. */
+#define EDMA_WDS_NO_CSUM      8
+/** Words used in an eDMA command with checksum acceleration. */
+#define EDMA_WDS_CSUM        10
+/** Total available words in the eDMA command FIFO. */
+#define EDMA_WDS_TOTAL      128
+/*
+ * FIXME: These definitions are internal and should have underscores!
+ * NOTE: The actual numeric values here are intentional and allow us to
+ * optimize the concept "if small ... else if large ... else ...", by
+ * checking for the low bit being set, and then for non-zero.
+ * These are used as array indices, so they must have the values (0, 1, 2)
+ * in some order.
+ */
+#define SIZE_SMALL (1)       /**< Small packet queue. */
+#define SIZE_LARGE (2)       /**< Large packet queue. */
+#define SIZE_JUMBO (0)       /**< Jumbo packet queue. */
+/** The number of "SIZE_xxx" values. */
+#define NETIO_NUM_SIZES 3
+/*
+ * Default numbers of packets for IPP drivers.  These values are chosen
+ * such that CIPP1 will not overflow its L2 cache.
+ */
+/** The default number of small packets. */
+#define NETIO_DEFAULT_SMALL_PACKETS 2750
+/** The default number of large packets. */
+#define NETIO_DEFAULT_LARGE_PACKETS 2500
+/** The default number of jumbo packets. */
+#define NETIO_DEFAULT_JUMBO_PACKETS 250
+/** Log2 of the size of a memory arena. */
+#define NETIO_ARENA_SHIFT      24      /* 16 MB */
+/** Size of a memory arena. */
+#define NETIO_ARENA_SIZE       (1 << NETIO_ARENA_SHIFT)
+/** A queue of packets.
+ *
+ * This structure partially defines a queue of packets waiting to be
+ * processed.  The queue as a whole is written to by an interrupt handler and
+ * read by non-interrupt code; this data structure is what's touched by the
+ * interrupt handler.  The other part of the queue state, the read offset, is
+ * kept in user space, not in hypervisor space, so it is in a separate data
+ * structure.
+ *
+ * The read offset (__packet_receive_read in the user part of the queue
+ * structure) points to the next packet to be read. When the read offset is
+ * equal to the write offset, the queue is empty; therefore the queue must
+ * contain one more slot than the required maximum queue size.
+ *
+ * Here's an example of all 3 state variables and what they mean.  All
+ * pointers move left to right.
+ *
+ * @code
+ *   I   I   V   V   V   V   I   I   I   I
+ *   0   1   2   3   4   5   6   7   8   9  10
+ *           ^       ^       ^               ^
+ *           |               |               |
+ *           |               |               __last_packet_plus_one
+ *           |               __buffer_write
+ *           __packet_receive_read
+ * @endcode
+ *
+ * This queue has 10 slots, and thus can hold 9 packets (_last_packet_plus_one
+ * = 10).  The read pointer is at 2, and the write pointer is at 6; thus,
+ * there are valid, unread packets in slots 2, 3, 4, and 5.  The remaining
+ * slots are invalid (do not contain a packet).
+ */
+typedef struct {
+  /** Byte offset of the next notify packet to be written: zero for the first
+   *  packet on the queue, sizeof (netio_pkt_t) for the second packet on the
+   *  queue, etc. */
+  volatile uint32_t __packet_write;
+  /** Offset of the packet after the last valid packet (i.e., when any
+   *  pointer is incremented to this value, it wraps back to zero). */
+  uint32_t __last_packet_plus_one;
+}
+__netio_packet_queue_t;
+/** A queue of buffers.
+ *
+ * This structure partially defines a queue of empty buffers which have been
+ * obtained via requests to the IPP.  (The elements of the queue are packet
+ * handles, which are transformed into a full netio_pkt_t when the buffer is
+ * retrieved.)  The queue as a whole is written to by an interrupt handler and
+ * read by non-interrupt code; this data structure is what's touched by the
+ * interrupt handler.  The other parts of the queue state, the read offset and
+ * requested write offset, are kept in user space, not in hypervisor space, so
+ * they are in a separate data structure.
+ *
+ * The read offset (__buffer_read in the user part of the queue structure)
+ * points to the next buffer to be read. When the read offset is equal to the
+ * write offset, the queue is empty; therefore the queue must contain one more
+ * slot than the required maximum queue size.
+ *
+ * The requested write offset (__buffer_requested_write in the user part of
+ * the queue structure) points to the slot which will hold the next buffer we
+ * request from the IPP, once we get around to sending such a request.  When
+ * the requested write offset is equal to the write offset, no requests for
+ * new buffers are outstanding; when the requested write offset is one greater
+ * than the read offset, no more requests may be sent.
+ *
+ * Note that, unlike the packet_queue, the buffer_queue places incoming
+ * buffers at decreasing addresses.  This makes the check for "is it time to
+ * wrap the buffer pointer" cheaper in the assembly code which receives new
+ * buffers, and means that the value which defines the queue size,
+ * __last_buffer, is different than in the packet queue.  Also, the offset
+ * used in the packet_queue is already scaled by the size of a packet; here we
+ * use unscaled slot indices for the offsets.  (These differences are
+ * historical, and in the future it's possible that the packet_queue will look
+ * more like this queue.)
+ *
+ * @code
+ * Here's an example of all 4 state variables and what they mean.  Remember:
+ * all pointers move right to left.
+ *
+ *   V   V   V   I   I   R   R   V   V   V
+ *   0   1   2   3   4   5   6   7   8   9
+ *           ^       ^       ^           ^
+ *           |       |       |           |
+ *           |       |       |           __last_buffer
+ *           |       |       __buffer_write
+ *           |       __buffer_requested_write
+ *           __buffer_read
+ * @endcode
+ *
+ * This queue has 10 slots, and thus can hold 9 buffers (_last_buffer = 9).
+ * The read pointer is at 2, and the write pointer is at 6; thus, there are
+ * valid, unread buffers in slots 2, 1, 0, 9, 8, and 7.  The requested write
+ * pointer is at 4; thus, requests have been made to the IPP for buffers which
+ * will be placed in slots 6 and 5 when they arrive.  Finally, the remaining
+ * slots are invalid (do not contain a buffer).
+ */
+typedef struct
+{
+  /** Ordinal number of the next buffer to be written: 0 for the first slot in
+   *  the queue, 1 for the second slot in the queue, etc. */
+  volatile uint32_t __buffer_write;
+  /** Ordinal number of the last buffer (i.e., when any pointer is decremented
+   *  below zero, it is reloaded with this value). */
+  uint32_t __last_buffer;
+}
+__netio_buffer_queue_t;
+/**
+ * An object for providing Ethernet packets to a process.
+ */
+typedef struct __netio_queue_impl_t
+{
+  /** The queue of packets waiting to be received. */
+  __netio_packet_queue_t __packet_receive_queue;
+  /** The intr bit mask that IDs this device. */
+  unsigned int __intr_id;
+  /** Offset to queues of empty buffers, one per size. */
+  uint32_t __buffer_queue[NETIO_NUM_SIZES];
+  /** The address of the first EPP tile, or -1 if no EPP. */
+  /* ISSUE: Actually this is always "0" or "~0". */
+  uint32_t __epp_location;
+  /** The queue ID that this queue represents. */
+  unsigned int __queue_id;
+  /** Number of acknowledgements received. */
+  volatile uint32_t __acks_received;
+  /** Last completion number received for packet_sendv. */
+  volatile uint32_t __last_completion_rcv;
+  /** Number of packets allowed to be outstanding. */
+  uint32_t __max_outstanding;
+  /** First VA available for packets. */
+  void* __va_0;
+  /** First VA in second range available for packets. */
+  void* __va_1;
+  /** Padding to align the "__packets" field to the size of a netio_pkt_t. */
+  uint32_t __padding[3];
+  /** The packets themselves. */
+  netio_pkt_t __packets[0];
+}
+netio_queue_impl_t;
+/**
+ * An object for managing the user end of a NetIO queue.
+ */
+typedef struct __netio_queue_user_impl_t
+{
+  /** The next incoming packet to be read. */
+  uint32_t __packet_receive_read;
+  /** The next empty buffers to be read, one index per size. */
+  uint8_t __buffer_read[NETIO_NUM_SIZES];
+  /** Where the empty buffer we next request from the IPP will go, one index
+   * per size. */
+  uint8_t __buffer_requested_write[NETIO_NUM_SIZES];
+  /** PCIe interface flag. */
+  uint8_t __pcie;
+  /** Number of packets left to be received before we send a credit update. */
+  uint32_t __receive_credit_remaining;
+  /** Value placed in __receive_credit_remaining when it reaches zero. */
+  uint32_t __receive_credit_interval;
+  /** First fast I/O routine index. */
+  uint32_t __fastio_index;
+  /** Number of acknowledgements expected. */
+  uint32_t __acks_outstanding;
+  /** Last completion number requested. */
+  uint32_t __last_completion_req;
+  /** File descriptor for driver. */
+  int __fd;
+}
+netio_queue_user_impl_t;
+#define NETIO_GROUP_CHUNK_SIZE   64   /**< Max # groups in one IPP request */
+#define NETIO_BUCKET_CHUNK_SIZE  64   /**< Max # buckets in one IPP request */
+/** Internal structure used to convey packet send information to the
+ * hypervisor.  FIXME: Actually, it's not used for that anymore, but
+ * netio_packet_send() still uses it internally.
+ */
+typedef struct
+{
+  uint16_t flags;              /**< Packet flags (__NETIO_SEND_FLG_xxx) */
+  uint16_t transfer_size;      /**< Size of packet */
+  uint32_t va;                 /**< VA of start of packet */
+  __netio_pkt_handle_t handle; /**< Packet handle */
+  uint32_t csum0;              /**< First checksum word */
+  uint32_t csum1;              /**< Second checksum word */
+}
+__netio_send_cmd_t;
+/** Flags used in two contexts:
+ *  - As the "flags" member in the __netio_send_cmd_t, above; used only
+ *    for netio_pkt_send_{prepare,commit}.
+ *  - As part of the flags passed to the various send packet fast I/O calls.
+ */
+/** Need acknowledgement on this packet.  Note that some code in the
+ *  normal send_pkt fast I/O handler assumes that this is equal to 1. */
+#define __NETIO_SEND_FLG_ACK    0x1
+/** Do checksum on this packet.  (Only used with the __netio_send_cmd_t;
+ *  normal packet sends use a special fast I/O index to denote checksumming,
+ *  and multi-segment sends test the checksum descriptor.) */
+#define __NETIO_SEND_FLG_CSUM   0x2
+/** Get a completion on this packet.  Only used with multi-segment sends.  */
+#define __NETIO_SEND_FLG_COMPLETION 0x4
+/** Position of the number-of-extra-segments value in the flags word.
+    Only used with multi-segment sends. */
+#define __NETIO_SEND_FLG_XSEG_SHIFT 3
+/** Width of the number-of-extra-segments value in the flags word. */
+#define __NETIO_SEND_FLG_XSEG_WIDTH 2
+#endif /* __DRV_XGBE_IMPL_H__ */
diff --git a/arch/tile/include/hv/drv_xgbe_intf.h b/arch/tile/include/hv/drv_xgbe_intf.h
new file mode 100644
index 000000000000..146e47d5334b
--- /dev/null
+++ b/arch/tile/include/hv/drv_xgbe_intf.h
@@ -0,0 +1,615 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+/**
+ * @file drv_xgbe_intf.h
+ * Interface to the hypervisor XGBE driver.
+ */
+#ifndef __DRV_XGBE_INTF_H__
+#define __DRV_XGBE_INTF_H__
+/**
+ * An object for forwarding VAs and PAs to the hypervisor.
+ * @ingroup types
+ *
+ * This allows the supervisor to specify a number of areas of memory to
+ * store packet buffers.
+ */
+typedef struct
+{
+  /** The physical address of the memory. */
+  HV_PhysAddr pa;
+  /** Page table entry for the memory.  This is only used to derive the
+   *  memory's caching mode; the PA bits are ignored. */
+  HV_PTE pte;
+  /** The virtual address of the memory. */
+  HV_VirtAddr va;
+  /** Size (in bytes) of the memory area. */
+  int size;
+}
+netio_ipp_address_t;
+/** The various pread/pwrite offsets into the hypervisor-level driver.
+ * @ingroup types
+ */
+typedef enum
+{
+  /** Inform the Linux driver of the address of the NetIO arena memory.
+   *  This offset is actually only used to convey information from netio
+   *  to the Linux driver; it never makes it from there to the hypervisor.
+   *  Write-only; takes a uint32_t specifying the VA address. */
+  NETIO_FIXED_ADDR               = 0x5000000000000000ULL,
+  /** Inform the Linux driver of the size of the NetIO arena memory.
+   *  This offset is actually only used to convey information from netio
+   *  to the Linux driver; it never makes it from there to the hypervisor.
+   *  Write-only; takes a uint32_t specifying the VA size. */
+  NETIO_FIXED_SIZE               = 0x5100000000000000ULL,
+  /** Register current tile with IPP.  Write then read: write, takes a
+   *  netio_input_config_t, read returns a pointer to a netio_queue_impl_t. */
+  NETIO_IPP_INPUT_REGISTER_OFF   = 0x6000000000000000ULL,
+  /** Unregister current tile from IPP.  Write-only, takes a dummy argument. */
+  NETIO_IPP_INPUT_UNREGISTER_OFF = 0x6100000000000000ULL,
+  /** Start packets flowing.  Write-only, takes a dummy argument. */
+  NETIO_IPP_INPUT_INIT_OFF       = 0x6200000000000000ULL,
+  /** Stop packets flowing.  Write-only, takes a dummy argument. */
+  NETIO_IPP_INPUT_UNINIT_OFF     = 0x6300000000000000ULL,
+  /** Configure group (typically we group on VLAN).  Write-only: takes an
+   *  array of netio_group_t's, low 24 bits of the offset is the base group
+   *  number times the size of a netio_group_t. */
+  NETIO_IPP_INPUT_GROUP_CFG_OFF  = 0x6400000000000000ULL,
+  /** Configure bucket.  Write-only: takes an array of netio_bucket_t's, low
+   *  24 bits of the offset is the base bucket number times the size of a
+   *  netio_bucket_t. */
+  NETIO_IPP_INPUT_BUCKET_CFG_OFF = 0x6500000000000000ULL,
+  /** Get/set a parameter.  Read or write: read or write data is the parameter
+   *  value, low 32 bits of the offset is a __netio_getset_offset_t. */
+  NETIO_IPP_PARAM_OFF            = 0x6600000000000000ULL,
+  /** Get fast I/O index.  Read-only; returns a 4-byte base index value. */
+  NETIO_IPP_GET_FASTIO_OFF       = 0x6700000000000000ULL,
+  /** Configure hijack IP address.  Packets with this IPv4 dest address
+   *  go to bucket NETIO_NUM_BUCKETS - 1.  Write-only: takes an IP address
+   *  in some standard form.  FIXME: Define the form! */
+  NETIO_IPP_INPUT_HIJACK_CFG_OFF  = 0x6800000000000000ULL,
+  /**
+   * Offsets beyond this point are reserved for the supervisor (although that
+   * enforcement must be done by the supervisor driver itself).
+   */
+  NETIO_IPP_USER_MAX_OFF         = 0x6FFFFFFFFFFFFFFFULL,
+  /** Register I/O memory.  Write-only, takes a netio_ipp_address_t. */
+  NETIO_IPP_IOMEM_REGISTER_OFF   = 0x7000000000000000ULL,
+  /** Unregister I/O memory.  Write-only, takes a netio_ipp_address_t. */
+  NETIO_IPP_IOMEM_UNREGISTER_OFF = 0x7100000000000000ULL,
+  /* Offsets greater than 0x7FFFFFFF can't be used directly from Linux
+   * userspace code due to limitations in the pread/pwrite syscalls. */
+  /** Drain LIPP buffers. */
+  NETIO_IPP_DRAIN_OFF              = 0xFA00000000000000ULL,
+  /** Supply a netio_ipp_address_t to be used as shared memory for the
+   *  LEPP command queue. */
+  NETIO_EPP_SHM_OFF              = 0xFB00000000000000ULL,
+  /* 0xFC... is currently unused. */
+  /** Stop IPP/EPP tiles.  Write-only, takes a dummy argument.  */
+  NETIO_IPP_STOP_SHIM_OFF        = 0xFD00000000000000ULL,
+  /** Start IPP/EPP tiles.  Write-only, takes a dummy argument.  */
+  NETIO_IPP_START_SHIM_OFF       = 0xFE00000000000000ULL,
+  /** Supply packet arena.  Write-only, takes an array of
+    * netio_ipp_address_t values. */
+  NETIO_IPP_ADDRESS_OFF          = 0xFF00000000000000ULL,
+} netio_hv_offset_t;
+/** Extract the base offset from an offset */
+#define NETIO_BASE_OFFSET(off)    ((off) & 0xFF00000000000000ULL)
+/** Extract the local offset from an offset */
+#define NETIO_LOCAL_OFFSET(off)   ((off) & 0x00FFFFFFFFFFFFFFULL)
+/**
+ * Get/set offset.
+ */
+typedef union
+{
+  struct
+  {
+    uint64_t addr:48;        /**< Class-specific address */
+    unsigned int class:8;    /**< Class (e.g., NETIO_PARAM) */
+    unsigned int opcode:8;   /**< High 8 bits of NETIO_IPP_PARAM_OFF */
+  }
+  bits;                      /**< Bitfields */
+  uint64_t word;             /**< Aggregated value to use as the offset */
+}
+__netio_getset_offset_t;
+/**
+ * Fast I/O index offsets (must be contiguous).
+ */
+typedef enum
+{
+  NETIO_FASTIO_ALLOCATE         = 0, /**< Get empty packet buffer */
+  NETIO_FASTIO_FREE_BUFFER      = 1, /**< Give buffer back to IPP */
+  NETIO_FASTIO_RETURN_CREDITS   = 2, /**< Give credits to IPP */
+  NETIO_FASTIO_SEND_PKT_NOCK    = 3, /**< Send a packet, no checksum */
+  NETIO_FASTIO_SEND_PKT_CK      = 4, /**< Send a packet, with checksum */
+  NETIO_FASTIO_SEND_PKT_VEC     = 5, /**< Send a vector of packets */
+  NETIO_FASTIO_SENDV_PKT        = 6, /**< Sendv one packet */
+  NETIO_FASTIO_NUM_INDEX        = 7, /**< Total number of fast I/O indices */
+} netio_fastio_index_t;
+/** 3-word return type for Fast I/O call. */
+typedef struct
+{
+  int err;            /**< Error code. */
+  uint32_t val0;      /**< Value.  Meaning depends upon the specific call. */
+  uint32_t val1;      /**< Value.  Meaning depends upon the specific call. */
+} netio_fastio_rv3_t;
+/** 0-argument fast I/O call */
+int __netio_fastio0(uint32_t fastio_index);
+/** 1-argument fast I/O call */
+int __netio_fastio1(uint32_t fastio_index, uint32_t arg0);
+/** 3-argument fast I/O call, 2-word return value */
+netio_fastio_rv3_t __netio_fastio3_rv3(uint32_t fastio_index, uint32_t arg0,
+                                       uint32_t arg1, uint32_t arg2);
+/** 4-argument fast I/O call */
+int __netio_fastio4(uint32_t fastio_index, uint32_t arg0, uint32_t arg1,
+                    uint32_t arg2, uint32_t arg3);
+/** 6-argument fast I/O call */
+int __netio_fastio6(uint32_t fastio_index, uint32_t arg0, uint32_t arg1,
+                    uint32_t arg2, uint32_t arg3, uint32_t arg4, uint32_t arg5);
+/** 9-argument fast I/O call */
+int __netio_fastio9(uint32_t fastio_index, uint32_t arg0, uint32_t arg1,
+                    uint32_t arg2, uint32_t arg3, uint32_t arg4, uint32_t arg5,
+                    uint32_t arg6, uint32_t arg7, uint32_t arg8);
+/** Allocate an empty packet.
+ * @param fastio_index Fast I/O index.
+ * @param size Size of the packet to allocate.
+ */
+#define __netio_fastio_allocate(fastio_index, size) \
+  __netio_fastio1((fastio_index) + NETIO_FASTIO_ALLOCATE, size)
+/** Free a buffer.
+ * @param fastio_index Fast I/O index.
+ * @param handle Handle for the packet to free.
+ */
+#define __netio_fastio_free_buffer(fastio_index, handle) \
+  __netio_fastio1((fastio_index) + NETIO_FASTIO_FREE_BUFFER, handle)
+/** Increment our receive credits.
+ * @param fastio_index Fast I/O index.
+ * @param credits Number of credits to add.
+ */
+#define __netio_fastio_return_credits(fastio_index, credits) \
+  __netio_fastio1((fastio_index) + NETIO_FASTIO_RETURN_CREDITS, credits)
+/** Send packet, no checksum.
+ * @param fastio_index Fast I/O index.
+ * @param ackflag Nonzero if we want an ack.
+ * @param size Size of the packet.
+ * @param va Virtual address of start of packet.
+ * @param handle Packet handle.
+ */
+#define __netio_fastio_send_pkt_nock(fastio_index, ackflag, size, va, handle) \
+  __netio_fastio4((fastio_index) + NETIO_FASTIO_SEND_PKT_NOCK, ackflag, \
+                  size, va, handle)
+/** Send packet, calculate checksum.
+ * @param fastio_index Fast I/O index.
+ * @param ackflag Nonzero if we want an ack.
+ * @param size Size of the packet.
+ * @param va Virtual address of start of packet.
+ * @param handle Packet handle.
+ * @param csum0 Shim checksum header.
+ * @param csum1 Checksum seed.
+ */
+#define __netio_fastio_send_pkt_ck(fastio_index, ackflag, size, va, handle, \
+                                   csum0, csum1) \
+  __netio_fastio6((fastio_index) + NETIO_FASTIO_SEND_PKT_CK, ackflag, \
+                  size, va, handle, csum0, csum1)
+/** Format for the "csum0" argument to the __netio_fastio_send routines
+ * and LEPP.  Note that this is currently exactly identical to the
+ * ShimProtocolOffloadHeader.
+ */
+typedef union
+{
+  struct
+  {
+    unsigned int start_byte:7;       /**< The first byte to be checksummed */
+    unsigned int count:14;           /**< Number of bytes to be checksummed. */
+    unsigned int destination_byte:7; /**< The byte to write the checksum to. */
+    unsigned int reserved:4;         /**< Reserved. */
+  } bits;                            /**< Decomposed method of access. */
+  unsigned int word;                 /**< To send out the IDN. */
+} __netio_checksum_header_t;
+/** Sendv packet with 1 or 2 segments.
+ * @param fastio_index Fast I/O index.
+ * @param flags Ack/csum/notify flags in low 3 bits; number of segments minus
+ *        1 in next 2 bits; expected checksum in high 16 bits.
+ * @param confno Confirmation number to request, if notify flag set.
+ * @param csum0 Checksum descriptor; if zero, no checksum.
+ * @param va_F Virtual address of first segment.
+ * @param va_L Virtual address of last segment, if 2 segments.
+ * @param len_F_L Length of first segment in low 16 bits; length of last
+ *        segment, if 2 segments, in high 16 bits.
+ */
+#define __netio_fastio_sendv_pkt_1_2(fastio_index, flags, confno, csum0, \
+                                     va_F, va_L, len_F_L) \
+  __netio_fastio6((fastio_index) + NETIO_FASTIO_SENDV_PKT, flags, confno, \
+                  csum0, va_F, va_L, len_F_L)
+/** Send packet on PCIe interface.
+ * @param fastio_index Fast I/O index.
+ * @param flags Ack/csum/notify flags in low 3 bits.
+ * @param confno Confirmation number to request, if notify flag set.
+ * @param csum0 Checksum descriptor; Hard wired 0, not needed for PCIe.
+ * @param va_F Virtual address of the packet buffer.
+ * @param va_L Virtual address of last segment, if 2 segments. Hard wired 0.
+ * @param len_F_L Length of the packet buffer in low 16 bits.
+ */
+#define __netio_fastio_send_pcie_pkt(fastio_index, flags, confno, csum0, \
+                                     va_F, va_L, len_F_L) \
+  __netio_fastio6((fastio_index) + PCIE_FASTIO_SENDV_PKT, flags, confno, \
+                  csum0, va_F, va_L, len_F_L)
+/** Sendv packet with 3 or 4 segments.
+ * @param fastio_index Fast I/O index.
+ * @param flags Ack/csum/notify flags in low 3 bits; number of segments minus
+ *        1 in next 2 bits; expected checksum in high 16 bits.
+ * @param confno Confirmation number to request, if notify flag set.
+ * @param csum0 Checksum descriptor; if zero, no checksum.
+ * @param va_F Virtual address of first segment.
+ * @param va_L Virtual address of last segment (third segment if 3 segments,
+ *        fourth segment if 4 segments).
+ * @param len_F_L Length of first segment in low 16 bits; length of last
+ *        segment in high 16 bits.
+ * @param va_M0 Virtual address of "middle 0" segment; this segment is sent
+ *        second when there are three segments, and third if there are four.
+ * @param va_M1 Virtual address of "middle 1" segment; this segment is sent
+ *        second when there are four segments.
+ * @param len_M0_M1 Length of middle 0 segment in low 16 bits; length of middle
+ *        1 segment, if 4 segments, in high 16 bits.
+ */
+#define __netio_fastio_sendv_pkt_3_4(fastio_index, flags, confno, csum0, va_F, \
+                                     va_L, len_F_L, va_M0, va_M1, len_M0_M1) \
+  __netio_fastio9((fastio_index) + NETIO_FASTIO_SENDV_PKT, flags, confno, \
+                  csum0, va_F, va_L, len_F_L, va_M0, va_M1, len_M0_M1)
+/** Send vector of packets.
+ * @param fastio_index Fast I/O index.
+ * @param seqno Number of packets transmitted so far on this interface;
+ *        used to decide which packets should be acknowledged.
+ * @param nentries Number of entries in vector.
+ * @param va Virtual address of start of vector entry array.
+ * @return 3-word netio_fastio_rv3_t structure.  The structure's err member
+ *         is an error code, or zero if no error.  The val0 member is the
+ *         updated value of seqno; it has been incremented by 1 for each
+ *         packet sent.  That increment may be less than nentries if an
+ *         error occured, or if some of the entries in the vector contain
+ *         handles equal to NETIO_PKT_HANDLE_NONE.  The val1 member is the
+ *         updated value of nentries; it has been decremented by 1 for each
+ *         vector entry processed.  Again, that decrement may be less than
+ *         nentries (leaving the returned value positive) if an error
+ *         occurred.
+ */
+#define __netio_fastio_send_pkt_vec(fastio_index, seqno, nentries, va) \
+  __netio_fastio3_rv3((fastio_index) + NETIO_FASTIO_SEND_PKT_VEC, seqno, \
+                      nentries, va)
+/** An egress DMA command for LEPP. */
+typedef struct
+{
+  /** Is this a TSO transfer?
+   *
+   * NOTE: This field is always 0, to distinguish it from
+   * lepp_tso_cmd_t.  It must come first!
+   */
+  uint8_t tso               : 1;
+  /** Unused padding bits. */
+  uint8_t _unused           : 3;
+  /** Should this packet be sent directly from caches instead of DRAM,
+   * using hash-for-home to locate the packet data?
+   */
+  uint8_t hash_for_home     : 1;
+  /** Should we compute a checksum? */
+  uint8_t compute_checksum  : 1;
+  /** Is this the final buffer for this packet?
+   *
+   * A single packet can be split over several input buffers (a "gather"
+   * operation).  This flag indicates that this is the last buffer
+   * in a packet.
+   */
+  uint8_t end_of_packet     : 1;
+  /** Should LEPP advance 'comp_busy' when this DMA is fully finished? */
+  uint8_t send_completion   : 1;
+  /** High bits of Client Physical Address of the start of the buffer
+   *  to be egressed.
+   *
+   *  NOTE: Only 6 bits are actually needed here, as CPAs are
+   *  currently 38 bits.  So two bits could be scavenged from this.
+   */
+  uint8_t cpa_hi;
+  /** The number of bytes to be egressed. */
+  uint16_t length;
+  /** Low 32 bits of Client Physical Address of the start of the buffer
+   *  to be egressed.
+   */
+  uint32_t cpa_lo;
+  /** Checksum information (only used if 'compute_checksum'). */
+  __netio_checksum_header_t checksum_data;
+} lepp_cmd_t;
+/** A chunk of physical memory for a TSO egress. */
+typedef struct
+{
+  /** The low bits of the CPA. */
+  uint32_t cpa_lo;
+  /** The high bits of the CPA. */
+  uint16_t cpa_hi               : 15;
+  /** Should this packet be sent directly from caches instead of DRAM,
+   *  using hash-for-home to locate the packet data?
+   */
+  uint16_t hash_for_home        : 1;
+  /** The length in bytes. */
+  uint16_t length;
+} lepp_frag_t;
+/** An LEPP command that handles TSO. */
+typedef struct
+{
+  /** Is this a TSO transfer?
+   *
+   *  NOTE: This field is always 1, to distinguish it from
+   *  lepp_cmd_t.  It must come first!
+   */
+  uint8_t tso             : 1;
+  /** Unused padding bits. */
+  uint8_t _unused         : 7;
+  /** Size of the header[] array in bytes.  It must be in the range
+   *  [40, 127], which are the smallest header for a TCP packet over
+   *  Ethernet and the maximum possible prepend size supported by
+   *  hardware, respectively.  Note that the array storage must be
+   *  padded out to a multiple of four bytes so that the following
+   *  LEPP command is aligned properly.
+   */
+  uint8_t header_size;
+  /** Byte offset of the IP header in header[]. */
+  uint8_t ip_offset;
+  /** Byte offset of the TCP header in header[]. */
+  uint8_t tcp_offset;
+  /** The number of bytes to use for the payload of each packet,
+   *  except of course the last one, which may not have enough bytes.
+   *  This means that each Ethernet packet except the last will have a
+   *  size of header_size + payload_size.
+   */
+  uint16_t payload_size;
+  /** The length of the 'frags' array that follows this struct. */
+  uint16_t num_frags;
+  /** The actual frags. */
+  lepp_frag_t frags[0 /* Variable-sized; num_frags entries. */];
+  /*
+   * The packet header template logically follows frags[],
+   * but you can't declare that in C.
+   *
+   * uint32_t header[header_size_in_words_rounded_up];
+   */
+} lepp_tso_cmd_t;
+/** An LEPP completion ring entry. */
+typedef void* lepp_comp_t;
+/** Maximum number of frags for one TSO command.  This is adapted from
+ *  linux's "MAX_SKB_FRAGS", and presumably over-estimates by one, for
+ *  our page size of exactly 65536.  We add one for a "body" fragment.
+ */
+#define LEPP_MAX_FRAGS (65536 / HV_PAGE_SIZE_SMALL + 2 + 1)
+/** Total number of bytes needed for an lepp_tso_cmd_t. */
+#define LEPP_TSO_CMD_SIZE(num_frags, header_size) \
+  (sizeof(lepp_tso_cmd_t) + \
+   (num_frags) * sizeof(lepp_frag_t) + \
+   (((header_size) + 3) & -4))
+/** The size of the lepp "cmd" queue. */
+#define LEPP_CMD_QUEUE_BYTES \
+ (((CHIP_L2_CACHE_SIZE() - 2 * CHIP_L2_LINE_SIZE()) / \
+  (sizeof(lepp_cmd_t) + sizeof(lepp_comp_t))) * sizeof(lepp_cmd_t))
+/** The largest possible command that can go in lepp_queue_t::cmds[]. */
+#define LEPP_MAX_CMD_SIZE LEPP_TSO_CMD_SIZE(LEPP_MAX_FRAGS, 128)
+/** The largest possible value of lepp_queue_t::cmd_{head, tail} (inclusive).
+ */
+#define LEPP_CMD_LIMIT \
+  (LEPP_CMD_QUEUE_BYTES - LEPP_MAX_CMD_SIZE)
+/** The maximum number of completions in an LEPP queue. */
+#define LEPP_COMP_QUEUE_SIZE \
+  ((LEPP_CMD_LIMIT + sizeof(lepp_cmd_t) - 1) / sizeof(lepp_cmd_t))
+/** Increment an index modulo the queue size. */
+#define LEPP_QINC(var) \
+  (var = __insn_mnz(var - (LEPP_COMP_QUEUE_SIZE - 1), var + 1))
+/** A queue used to convey egress commands from the client to LEPP. */
+typedef struct
+{
+  /** Index of first completion not yet processed by user code.
+   *  If this is equal to comp_busy, there are no such completions.
+   *
+   *  NOTE: This is only read/written by the user.
+   */
+  unsigned int comp_head;
+  /** Index of first completion record not yet completed.
+   *  If this is equal to comp_tail, there are no such completions.
+   *  This index gets advanced (modulo LEPP_QUEUE_SIZE) whenever
+   *  a command with the 'completion' bit set is finished.
+   *
+   *  NOTE: This is only written by LEPP, only read by the user.
+   */
+  volatile unsigned int comp_busy;
+  /** Index of the first empty slot in the completion ring.
+   *  Entries from this up to but not including comp_head (in ring order)
+   *  can be filled in with completion data.
+   *
+   *  NOTE: This is only read/written by the user.
+   */
+  unsigned int comp_tail;
+  /** Byte index of first command enqueued for LEPP but not yet processed.
+   *
+   *  This is always divisible by sizeof(void*) and always <= LEPP_CMD_LIMIT.
+   *
+   *  NOTE: LEPP advances this counter as soon as it no longer needs
+   *  the cmds[] storage for this entry, but the transfer is not actually
+   *  complete (i.e. the buffer pointed to by the command is no longer
+   *  needed) until comp_busy advances.
+   *
+   *  If this is equal to cmd_tail, the ring is empty.
+   *
+   *  NOTE: This is only written by LEPP, only read by the user.
+   */
+  volatile unsigned int cmd_head;
+  /** Byte index of first empty slot in the command ring.  This field can
+   *  be incremented up to but not equal to cmd_head (because that would
+   *  mean the ring is empty).
+   *
+   *  This is always divisible by sizeof(void*) and always <= LEPP_CMD_LIMIT.
+   *
+   *  NOTE: This is read/written by the user, only read by LEPP.
+   */
+  volatile unsigned int cmd_tail;
+  /** A ring of variable-sized egress DMA commands.
+   *
+   *  NOTE: Only written by the user, only read by LEPP.
+   */
+  char cmds[LEPP_CMD_QUEUE_BYTES]
+    __attribute__((aligned(CHIP_L2_LINE_SIZE())));
+  /** A ring of user completion data.
+   *  NOTE: Only read/written by the user.
+   */
+  lepp_comp_t comps[LEPP_COMP_QUEUE_SIZE]
+    __attribute__((aligned(CHIP_L2_LINE_SIZE())));
+} lepp_queue_t;
+/** An internal helper function for determining the number of entries
+ *  available in a ring buffer, given that there is one sentinel.
+ */
+static inline unsigned int
+_lepp_num_free_slots(unsigned int head, unsigned int tail)
+{
+  /*
+   * One entry is reserved for use as a sentinel, to distinguish
+   * "empty" from "full".  So we compute
+   * (head - tail - 1) % LEPP_QUEUE_SIZE, but without using a slow % operation.
+   */
+  return (head - tail - 1) + ((head <= tail) ? LEPP_COMP_QUEUE_SIZE : 0);
+}
+/** Returns how many new comp entries can be enqueued. */
+static inline unsigned int
+lepp_num_free_comp_slots(const lepp_queue_t* q)
+{
+  return _lepp_num_free_slots(q->comp_head, q->comp_tail);
+}
+static inline int
+lepp_qsub(int v1, int v2)
+{
+  int delta = v1 - v2;
+  return delta + ((delta >> 31) & LEPP_COMP_QUEUE_SIZE);
+}
+/** FIXME: Check this from linux, via a new "pwrite()" call. */
+#define LIPP_VERSION 1
+/** We use exactly two bytes of alignment padding. */
+#define LIPP_PACKET_PADDING 2
+/** The minimum size of a "small" buffer (including the padding). */
+#define LIPP_SMALL_PACKET_SIZE 128
+/*
+ * NOTE: The following two values should total to less than around
+ * 13582, to keep the total size used for "lipp_state_t" below 64K.
+ */
+/** The maximum number of "small" buffers.
+ *  This is enough for 53 network cpus with 128 credits.  Note that
+ *  if these are exhausted, we will fall back to using large buffers.
+ */
+#define LIPP_SMALL_BUFFERS 6785
+/** The maximum number of "large" buffers.
+ *  This is enough for 53 network cpus with 128 credits.
+ */
+#define LIPP_LARGE_BUFFERS 6785
+#endif /* __DRV_XGBE_INTF_H__ */
diff --git a/arch/tile/include/hv/netio_errors.h b/arch/tile/include/hv/netio_errors.h
new file mode 100644
index 000000000000..e1591bff61b5
--- /dev/null
+++ b/arch/tile/include/hv/netio_errors.h
@@ -0,0 +1,122 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+/**
+ * Error codes returned from NetIO routines.
+ */
+#ifndef __NETIO_ERRORS_H__
+#define __NETIO_ERRORS_H__
+/**
+ * @addtogroup error
+ *
+ * @brief The error codes returned by NetIO functions.
+ *
+ * NetIO functions return 0 (defined as ::NETIO_NO_ERROR) on success, and
+ * a negative value if an error occurs.
+ *
+ * In cases where a NetIO function failed due to a error reported by
+ * system libraries, the error code will be the negation of the
+ * system errno at the time of failure.  The @ref netio_strerror()
+ * function will deliver error strings for both NetIO and system error
+ * codes.
+ *
+ * @{
+ */
+/** The set of all NetIO errors. */
+typedef enum
+{
+  /** Operation successfully completed. */
+  NETIO_NO_ERROR        = 0,
+  /** A packet was successfully retrieved from an input queue. */
+  NETIO_PKT             = 0,
+  /** Largest NetIO error number. */
+  NETIO_ERR_MAX         = -701,
+  /** The tile is not registered with the IPP. */
+  NETIO_NOT_REGISTERED  = -701,
+  /** No packet was available to retrieve from the input queue. */
+  NETIO_NOPKT           = -702,
+  /** The requested function is not implemented. */
+  NETIO_NOT_IMPLEMENTED = -703,
+  /** On a registration operation, the target queue already has the maximum
+   *  number of tiles registered for it, and no more may be added.  On a
+   *  packet send operation, the output queue is full and nothing more can
+   *  be queued until some of the queued packets are actually transmitted. */
+  NETIO_QUEUE_FULL      = -704,
+  /** The calling process or thread is not bound to exactly one CPU. */
+  NETIO_BAD_AFFINITY    = -705,
+  /** Cannot allocate memory on requested controllers. */
+  NETIO_CANNOT_HOME     = -706,
+  /** On a registration operation, the IPP specified is not configured
+   *  to support the options requested; for instance, the application
+   *  wants a specific type of tagged headers which the configured IPP
+   *  doesn't support.  Or, the supplied configuration information is
+   *  not self-consistent, or is out of range; for instance, specifying
+   *  both NETIO_RECV and NETIO_NO_RECV, or asking for more than
+   *  NETIO_MAX_SEND_BUFFERS to be preallocated.  On a VLAN or bucket
+   *  configure operation, the number of items, or the base item, was
+   *  out of range.
+   */
+  NETIO_BAD_CONFIG      = -707,
+  /** Too many tiles have registered to transmit packets. */
+  NETIO_TOOMANY_XMIT    = -708,
+  /** Packet transmission was attempted on a queue which was registered
+      with transmit disabled. */
+  NETIO_UNREG_XMIT      = -709,
+  /** This tile is already registered with the IPP. */
+  NETIO_ALREADY_REGISTERED = -710,
+  /** The Ethernet link is down. The application should try again later. */
+  NETIO_LINK_DOWN       = -711,
+  /** An invalid memory buffer has been specified.  This may be an unmapped
+   * virtual address, or one which does not meet alignment requirements.
+   * For netio_input_register(), this error may be returned when multiple
+   * processes specify different memory regions to be used for NetIO
+   * buffers.  That can happen if these processes specify explicit memory
+   * regions with the ::NETIO_FIXED_BUFFER_VA flag, or if tmc_cmem_init()
+   * has not been called by a common ancestor of the processes.
+   */
+  NETIO_FAULT           = -712,
+  /** Cannot combine user-managed shared memory and cache coherence. */
+  NETIO_BAD_CACHE_CONFIG = -713,
+  /** Smallest NetIO error number. */
+  NETIO_ERR_MIN         = -713,
+#ifndef __DOXYGEN__
+  /** Used internally to mean that no response is needed; never returned to
+   *  an application. */
+  NETIO_NO_RESPONSE     = 1
+#endif
+} netio_error_t;
+/** @} */
+#endif /* __NETIO_ERRORS_H__ */
diff --git a/arch/tile/include/hv/netio_intf.h b/arch/tile/include/hv/netio_intf.h
new file mode 100644
index 000000000000..8d20972aba2c
--- /dev/null
+++ b/arch/tile/include/hv/netio_intf.h
@@ -0,0 +1,2975 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+/**
+ * NetIO interface structures and macros.
+ */
+#ifndef __NETIO_INTF_H__
+#define __NETIO_INTF_H__
+#include <hv/netio_errors.h>
+#ifdef __KERNEL__
+#include <linux/types.h>
+#else
+#include <stdint.h>
+#endif
+#if !defined(__HV__) && !defined(__BOGUX__) && !defined(__KERNEL__)
+#include <assert.h>
+#define netio_assert assert  /**< Enable assertions from macros */
+#else
+#define netio_assert(...) ((void)(0))  /**< Disable assertions from macros */
+#endif
+/*
+ * If none of these symbols are defined, we're building libnetio in an
+ * environment where we have pthreads, so we'll enable locking.
+ */
+#if !defined(__HV__) && !defined(__BOGUX__) && !defined(__KERNEL__) && \
+    !defined(__NEWLIB__)
+#define _NETIO_PTHREAD       /**< Include a mutex in netio_queue_t below */
+/*
+ * If NETIO_UNLOCKED is defined, we don't do use per-cpu locks on
+ * per-packet NetIO operations.  We still do pthread locking on things
+ * like netio_input_register, though.  This is used for building
+ * libnetio_unlocked.
+ */
+#ifndef NETIO_UNLOCKED
+/* Avoid PLT overhead by using our own inlined per-cpu lock. */
+#include <sched.h>
+typedef int _netio_percpu_mutex_t;
+static __inline int
+_netio_percpu_mutex_init(_netio_percpu_mutex_t* lock)
+{
+  *lock = 0;
+  return 0;
+}
+static __inline int
+_netio_percpu_mutex_lock(_netio_percpu_mutex_t* lock)
+{
+  while (__builtin_expect(__insn_tns(lock), 0))
+    sched_yield();
+  return 0;
+}
+static __inline int
+_netio_percpu_mutex_unlock(_netio_percpu_mutex_t* lock)
+{
+  *lock = 0;
+  return 0;
+}
+#else /* NETIO_UNLOCKED */
+/* Don't do any locking for per-packet NetIO operations. */
+typedef int _netio_percpu_mutex_t;
+#define _netio_percpu_mutex_init(L)
+#define _netio_percpu_mutex_lock(L)
+#define _netio_percpu_mutex_unlock(L)
+#endif /* NETIO_UNLOCKED */
+#endif /* !__HV__, !__BOGUX, !__KERNEL__, !__NEWLIB__ */
+/** How many tiles can register for a given queue.
+ *  @ingroup setup */
+#define NETIO_MAX_TILES_PER_QUEUE  64
+/** Largest permissible queue identifier.
+ *  @ingroup setup  */
+#define NETIO_MAX_QUEUE_ID        255
+#ifndef __DOXYGEN__
+/* Metadata packet checksum/ethertype flags. */
+/** The L4 checksum has not been calculated. */
+#define _NETIO_PKT_NO_L4_CSUM_SHIFT           0
+#define _NETIO_PKT_NO_L4_CSUM_RMASK           1
+#define _NETIO_PKT_NO_L4_CSUM_MASK \
+         (_NETIO_PKT_NO_L4_CSUM_RMASK << _NETIO_PKT_NO_L4_CSUM_SHIFT)
+/** The L3 checksum has not been calculated. */
+#define _NETIO_PKT_NO_L3_CSUM_SHIFT           1
+#define _NETIO_PKT_NO_L3_CSUM_RMASK           1
+#define _NETIO_PKT_NO_L3_CSUM_MASK \
+         (_NETIO_PKT_NO_L3_CSUM_RMASK << _NETIO_PKT_NO_L3_CSUM_SHIFT)
+/** The L3 checksum is incorrect (or perhaps has not been calculated). */
+#define _NETIO_PKT_BAD_L3_CSUM_SHIFT          2
+#define _NETIO_PKT_BAD_L3_CSUM_RMASK          1
+#define _NETIO_PKT_BAD_L3_CSUM_MASK \
+         (_NETIO_PKT_BAD_L3_CSUM_RMASK << _NETIO_PKT_BAD_L3_CSUM_SHIFT)
+/** The Ethernet packet type is unrecognized. */
+#define _NETIO_PKT_TYPE_UNRECOGNIZED_SHIFT    3
+#define _NETIO_PKT_TYPE_UNRECOGNIZED_RMASK    1
+#define _NETIO_PKT_TYPE_UNRECOGNIZED_MASK \
+         (_NETIO_PKT_TYPE_UNRECOGNIZED_RMASK << \
+          _NETIO_PKT_TYPE_UNRECOGNIZED_SHIFT)
+/* Metadata packet type flags. */
+/** Where the packet type bits are; this field is the index into
+ *  _netio_pkt_info. */
+#define _NETIO_PKT_TYPE_SHIFT        4
+#define _NETIO_PKT_TYPE_RMASK        0x3F
+/** How many VLAN tags the packet has, and, if we have two, which one we
+ *  actually grouped on.  A VLAN within a proprietary (Marvell or Broadcom)
+ *  tag is counted here. */
+#define _NETIO_PKT_VLAN_SHIFT        4
+#define _NETIO_PKT_VLAN_RMASK        0x3
+#define _NETIO_PKT_VLAN_MASK \
+         (_NETIO_PKT_VLAN_RMASK << _NETIO_PKT_VLAN_SHIFT)
+#define _NETIO_PKT_VLAN_NONE         0   /* No VLAN tag. */
+#define _NETIO_PKT_VLAN_ONE          1   /* One VLAN tag. */
+#define _NETIO_PKT_VLAN_TWO_OUTER    2   /* Two VLAN tags, outer one used. */
+#define _NETIO_PKT_VLAN_TWO_INNER    3   /* Two VLAN tags, inner one used. */
+/** Which proprietary tags the packet has. */
+#define _NETIO_PKT_TAG_SHIFT         6
+#define _NETIO_PKT_TAG_RMASK         0x3
+#define _NETIO_PKT_TAG_MASK \
+          (_NETIO_PKT_TAG_RMASK << _NETIO_PKT_TAG_SHIFT)
+#define _NETIO_PKT_TAG_NONE          0   /* No proprietary tags. */
+#define _NETIO_PKT_TAG_MRVL          1   /* Marvell HyperG.Stack tags. */
+#define _NETIO_PKT_TAG_MRVL_EXT      2   /* HyperG.Stack extended tags. */
+#define _NETIO_PKT_TAG_BRCM          3   /* Broadcom HiGig tags. */
+/** Whether a packet has an LLC + SNAP header. */
+#define _NETIO_PKT_SNAP_SHIFT        8
+#define _NETIO_PKT_SNAP_RMASK        0x1
+#define _NETIO_PKT_SNAP_MASK \
+          (_NETIO_PKT_SNAP_RMASK << _NETIO_PKT_SNAP_SHIFT)
+/* NOTE: Bits 9 and 10 are unused. */
+/** Length of any custom data before the L2 header, in words. */
+#define _NETIO_PKT_CUSTOM_LEN_SHIFT  11
+#define _NETIO_PKT_CUSTOM_LEN_RMASK  0x1F
+#define _NETIO_PKT_CUSTOM_LEN_MASK \
+          (_NETIO_PKT_CUSTOM_LEN_RMASK << _NETIO_PKT_CUSTOM_LEN_SHIFT)
+/** The L4 checksum is incorrect (or perhaps has not been calculated). */
+#define _NETIO_PKT_BAD_L4_CSUM_SHIFT 16
+#define _NETIO_PKT_BAD_L4_CSUM_RMASK 0x1
+#define _NETIO_PKT_BAD_L4_CSUM_MASK \
+          (_NETIO_PKT_BAD_L4_CSUM_RMASK << _NETIO_PKT_BAD_L4_CSUM_SHIFT)
+/** Length of the L2 header, in words. */
+#define _NETIO_PKT_L2_LEN_SHIFT  17
+#define _NETIO_PKT_L2_LEN_RMASK  0x1F
+#define _NETIO_PKT_L2_LEN_MASK \
+          (_NETIO_PKT_L2_LEN_RMASK << _NETIO_PKT_L2_LEN_SHIFT)
+/* Flags in minimal packet metadata. */
+/** We need an eDMA checksum on this packet. */
+#define _NETIO_PKT_NEED_EDMA_CSUM_SHIFT            0
+#define _NETIO_PKT_NEED_EDMA_CSUM_RMASK            1
+#define _NETIO_PKT_NEED_EDMA_CSUM_MASK \
+         (_NETIO_PKT_NEED_EDMA_CSUM_RMASK << _NETIO_PKT_NEED_EDMA_CSUM_SHIFT)
+/* Data within the packet information table. */
+/* Note that, for efficiency, code which uses these fields assumes that none
+ * of the shift values below are zero.  See uses below for an explanation. */
+/** Offset within the L2 header of the innermost ethertype (in halfwords). */
+#define _NETIO_PKT_INFO_ETYPE_SHIFT       6
+#define _NETIO_PKT_INFO_ETYPE_RMASK    0x1F
+/** Offset within the L2 header of the VLAN tag (in halfwords). */
+#define _NETIO_PKT_INFO_VLAN_SHIFT       11
+#define _NETIO_PKT_INFO_VLAN_RMASK     0x1F
+#endif
+/** The size of a memory buffer representing a small packet.
+ *  @ingroup egress */
+#define SMALL_PACKET_SIZE 256
+/** The size of a memory buffer representing a large packet.
+ *  @ingroup egress */
+#define LARGE_PACKET_SIZE 2048
+/** The size of a memory buffer representing a jumbo packet.
+ *  @ingroup egress */
+#define JUMBO_PACKET_SIZE (12 * 1024)
+/* Common ethertypes.
+ * @ingroup ingress */
+/** @{ */
+/** The ethertype of IPv4. */
+#define ETHERTYPE_IPv4 (0x0800)
+/** The ethertype of ARP. */
+#define ETHERTYPE_ARP (0x0806)
+/** The ethertype of VLANs. */
+#define ETHERTYPE_VLAN (0x8100)
+/** The ethertype of a Q-in-Q header. */
+#define ETHERTYPE_Q_IN_Q (0x9100)
+/** The ethertype of IPv6. */
+#define ETHERTYPE_IPv6 (0x86DD)
+/** The ethertype of MPLS. */
+#define ETHERTYPE_MPLS (0x8847)
+/** @} */
+/** The possible return values of NETIO_PKT_STATUS.
+ * @ingroup ingress
+ */
+typedef enum
+{
+  /** No problems were detected with this packet. */
+  NETIO_PKT_STATUS_OK,
+  /** The packet is undersized; this is expected behavior if the packet's
+    * ethertype is unrecognized, but otherwise the packet is likely corrupt. */
+  NETIO_PKT_STATUS_UNDERSIZE,
+  /** The packet is oversized and some trailing bytes have been discarded.
+      This is expected behavior for short packets, since it's impossible to
+      precisely determine the amount of padding which may have been added to
+      them to make them meet the minimum Ethernet packet size. */
+  NETIO_PKT_STATUS_OVERSIZE,
+  /** The packet was judged to be corrupt by hardware (for instance, it had
+      a bad CRC, or part of it was discarded due to lack of buffer space in
+      the I/O shim) and should be discarded. */
+  NETIO_PKT_STATUS_BAD
+} netio_pkt_status_t;
+/** Log2 of how many buckets we have. */
+#define NETIO_LOG2_NUM_BUCKETS (10)
+/** How many buckets we have.
+ * @ingroup ingress */
+#define NETIO_NUM_BUCKETS (1 << NETIO_LOG2_NUM_BUCKETS)
+/**
+ * @brief A group-to-bucket identifier.
+ *
+ * @ingroup setup
+ *
+ * This tells us what to do with a given group.
+ */
+typedef union {
+  /** The header broken down into bits. */
+  struct {
+    /** Whether we should balance on L4, if available */
+    unsigned int __balance_on_l4:1;
+    /** Whether we should balance on L3, if available */
+    unsigned int __balance_on_l3:1;
+    /** Whether we should balance on L2, if available */
+    unsigned int __balance_on_l2:1;
+    /** Reserved for future use */
+    unsigned int __reserved:1;
+    /** The base bucket to use to send traffic */
+    unsigned int __bucket_base:NETIO_LOG2_NUM_BUCKETS;
+    /** The mask to apply to the balancing value. This must be one less
+     * than a power of two, e.g. 0x3 or 0xFF.
+     */
+    unsigned int __bucket_mask:NETIO_LOG2_NUM_BUCKETS;
+    /** Pad to 32 bits */
+    unsigned int __padding:(32 - 4 - 2 * NETIO_LOG2_NUM_BUCKETS);
+  } bits;
+  /** To send out the IDN. */
+  unsigned int word;
+}
+netio_group_t;
+/**
+ * @brief A VLAN-to-bucket identifier.
+ *
+ * @ingroup setup
+ *
+ * This tells us what to do with a given VLAN.
+ */
+typedef netio_group_t netio_vlan_t;
+/**
+ * A bucket-to-queue mapping.
+ * @ingroup setup
+ */
+typedef unsigned char netio_bucket_t;
+/**
+ * A packet size can always fit in a netio_size_t.
+ * @ingroup setup
+ */
+typedef unsigned int netio_size_t;
+/**
+ * @brief Ethernet standard (ingress) packet metadata.
+ *
+ * @ingroup ingress
+ *
+ * This is additional data associated with each packet.
+ * This structure is opaque and accessed through the @ref ingress.
+ *
+ * Also, the buffer population operation currently assumes that standard
+ * metadata is at least as large as minimal metadata, and will need to be
+ * modified if that is no longer the case.
+ */
+typedef struct
+{
+#ifdef __DOXYGEN__
+  /** This structure is opaque. */
+  unsigned char opaque[24];
+#else
+  /** The overall ordinal of the packet */
+  unsigned int __packet_ordinal;
+  /** The ordinal of the packet within the group */
+  unsigned int __group_ordinal;
+  /** The best flow hash IPP could compute. */
+  unsigned int __flow_hash;
+  /** Flags pertaining to checksum calculation, packet type, etc. */
+  unsigned int __flags;
+  /** The first word of "user data". */
+  unsigned int __user_data_0;
+  /** The second word of "user data". */
+  unsigned int __user_data_1;
+#endif
+}
+netio_pkt_metadata_t;
+/** To ensure that the L3 header is aligned mod 4, the L2 header should be
+ * aligned mod 4 plus 2, since every supported L2 header is 4n + 2 bytes
+ * long.  The standard way to do this is to simply add 2 bytes of padding
+ * before the L2 header.
+ */
+#define NETIO_PACKET_PADDING 2
+/**
+ * @brief Ethernet minimal (egress) packet metadata.
+ *
+ * @ingroup egress
+ *
+ * This structure represents information about packets which have
+ * been processed by @ref netio_populate_buffer() or
+ * @ref netio_populate_prepend_buffer().  This structure is opaque
+ * and accessed through the @ref egress.
+ *
+ * @internal This structure is actually copied into the memory used by
+ * standard metadata, which is assumed to be large enough.
+ */
+typedef struct
+{
+#ifdef __DOXYGEN__
+  /** This structure is opaque. */
+  unsigned char opaque[14];
+#else
+  /** The offset of the L2 header from the start of the packet data. */
+  unsigned short l2_offset;
+  /** The offset of the L3 header from the start of the packet data. */
+  unsigned short l3_offset;
+  /** Where to write the checksum. */
+  unsigned char csum_location;
+  /** Where to start checksumming from. */
+  unsigned char csum_start;
+  /** Flags pertaining to checksum calculation etc. */
+  unsigned short flags;
+  /** The L2 length of the packet. */
+  unsigned short l2_length;
+  /** The checksum with which to seed the checksum generator. */
+  unsigned short csum_seed;
+  /** How much to checksum. */
+  unsigned short csum_length;
+#endif
+}
+netio_pkt_minimal_metadata_t;
+#ifndef __DOXYGEN__
+/**
+ * @brief An I/O notification header.
+ *
+ * This is the first word of data received from an I/O shim in a notification
+ * packet. It contains framing and status information.
+ */
+typedef union
+{
+  unsigned int word; /**< The whole word. */
+  /** The various fields. */
+  struct
+  {
+    unsigned int __channel:7;    /**< Resource channel. */
+    unsigned int __type:4;       /**< Type. */
+    unsigned int __ack:1;        /**< Whether an acknowledgement is needed. */
+    unsigned int __reserved:1;   /**< Reserved. */
+    unsigned int __protocol:1;   /**< A protocol-specific word is added. */
+    unsigned int __status:2;     /**< Status of the transfer. */
+    unsigned int __framing:2;    /**< Framing of the transfer. */
+    unsigned int __transfer_size:14; /**< Transfer size in bytes (total). */
+  } bits;
+}
+__netio_pkt_notif_t;
+/**
+ * Returns the base address of the packet.
+ */
+#define _NETIO_PKT_HANDLE_BASE(p) \
+  ((unsigned char*)((p).word & 0xFFFFFFC0))
+/**
+ * Returns the base address of the packet.
+ */
+#define _NETIO_PKT_BASE(p) \
+  _NETIO_PKT_HANDLE_BASE(p->__packet)
+/**
+ * @brief An I/O notification packet (second word)
+ *
+ * This is the second word of data received from an I/O shim in a notification
+ * packet.  This is the virtual address of the packet buffer, plus some flag
+ * bits.  (The virtual address of the packet is always 256-byte aligned so we
+ * have room for 8 bits' worth of flags in the low 8 bits.)
+ *
+ * @internal
+ * NOTE: The low two bits must contain "__queue", so the "packet size"
+ * (SIZE_SMALL, SIZE_LARGE, or SIZE_JUMBO) can be determined quickly.
+ *
+ * If __addr or __offset are moved, _NETIO_PKT_BASE
+ * (defined right below this) must be changed.
+ */
+typedef union
+{
+  unsigned int word; /**< The whole word. */
+  /** The various fields. */
+  struct
+  {
+    /** Which queue the packet will be returned to once it is sent back to
+        the IPP.  This is one of the SIZE_xxx values. */
+    unsigned int __queue:2;
+    /** The IPP handle of the sending IPP. */
+    unsigned int __ipp_handle:2;
+    /** Reserved for future use. */
+    unsigned int __reserved:1;
+    /** If 1, this packet has minimal (egress) metadata; otherwise, it
+        has standard (ingress) metadata. */
+    unsigned int __minimal:1;
+    /** Offset of the metadata within the packet.  This value is multiplied
+     *  by 64 and added to the base packet address to get the metadata
+     *  address.  Note that this field is aligned within the word such that
+     *  you can easily extract the metadata address with a 26-bit mask. */
+    unsigned int __offset:2;
+    /** The top 24 bits of the packet's virtual address. */
+    unsigned int __addr:24;
+  } bits;
+}
+__netio_pkt_handle_t;
+#endif /* !__DOXYGEN__ */
+/**
+ * @brief A handle for an I/O packet's storage.
+ * @ingroup ingress
+ *
+ * netio_pkt_handle_t encodes the concept of a ::netio_pkt_t with its
+ * packet metadata removed.  It is a much smaller type that exists to
+ * facilitate applications where the full ::netio_pkt_t type is too
+ * large, such as those that cache enormous numbers of packets or wish
+ * to transmit packet descriptors over the UDN.
+ *
+ * Because there is no metadata, most ::netio_pkt_t operations cannot be
+ * performed on a netio_pkt_handle_t.  It supports only
+ * netio_free_handle() (to free the buffer) and
+ * NETIO_PKT_CUSTOM_DATA_H() (to access a pointer to its contents).
+ * The application must acquire any additional metadata it wants from the
+ * original ::netio_pkt_t and record it separately.
+ *
+ * A netio_pkt_handle_t can be extracted from a ::netio_pkt_t by calling
+ * NETIO_PKT_HANDLE().  An invalid handle (analogous to NULL) can be
+ * created by assigning the value ::NETIO_PKT_HANDLE_NONE. A handle can
+ * be tested for validity with NETIO_PKT_HANDLE_IS_VALID().
+ */
+typedef struct
+{
+  unsigned int word; /**< Opaque bits. */
+} netio_pkt_handle_t;
+/**
+ * @brief A packet descriptor.
+ *
+ * @ingroup ingress
+ * @ingroup egress
+ *
+ * This data structure represents a packet.  The structure is manipulated
+ * through the @ref ingress and the @ref egress.
+ *
+ * While the contents of a netio_pkt_t are opaque, the structure itself is
+ * portable.  This means that it may be shared between all tiles which have
+ * done a netio_input_register() call for the interface on which the pkt_t
+ * was initially received (via netio_get_packet()) or retrieved (via
+ * netio_get_buffer()).  The contents of a netio_pkt_t can be transmitted to
+ * another tile via shared memory, or via a UDN message, or by other means.
+ * The destination tile may then use the pkt_t as if it had originally been
+ * received locally; it may read or write the packet's data, read its
+ * metadata, free the packet, send the packet, transfer the netio_pkt_t to
+ * yet another tile, and so forth.
+ *
+ * Once a netio_pkt_t has been transferred to a second tile, the first tile
+ * should not reference the original copy; in particular, if more than one
+ * tile frees or sends the same netio_pkt_t, the IPP's packet free lists will
+ * become corrupted.  Note also that each tile which reads or modifies
+ * packet data must obey the memory coherency rules outlined in @ref input.
+ */
+typedef struct
+{
+#ifdef __DOXYGEN__
+  /** This structure is opaque. */
+  unsigned char opaque[32];
+#else
+  /** For an ingress packet (one with standard metadata), this is the
+   *  notification header we got from the I/O shim.  For an egress packet
+   *  (one with minimal metadata), this word is zero if the packet has not
+   *  been populated, and nonzero if it has. */
+  __netio_pkt_notif_t __notif_header;
+  /** Virtual address of the packet buffer, plus state flags. */
+  __netio_pkt_handle_t __packet;
+  /** Metadata associated with the packet. */
+  netio_pkt_metadata_t __metadata;
+#endif
+}
+netio_pkt_t;
+#ifndef __DOXYGEN__
+#define __NETIO_PKT_NOTIF_HEADER(pkt) ((pkt)->__notif_header)
+#define __NETIO_PKT_IPP_HANDLE(pkt) ((pkt)->__packet.bits.__ipp_handle)
+#define __NETIO_PKT_QUEUE(pkt) ((pkt)->__packet.bits.__queue)
+#define __NETIO_PKT_NOTIF_HEADER_M(mda, pkt) ((pkt)->__notif_header)
+#define __NETIO_PKT_IPP_HANDLE_M(mda, pkt) ((pkt)->__packet.bits.__ipp_handle)
+#define __NETIO_PKT_MINIMAL(pkt) ((pkt)->__packet.bits.__minimal)
+#define __NETIO_PKT_QUEUE_M(mda, pkt) ((pkt)->__packet.bits.__queue)
+#define __NETIO_PKT_FLAGS_M(mda, pkt) ((mda)->__flags)
+/* Packet information table, used by the attribute access functions below. */
+extern const uint16_t _netio_pkt_info[];
+#endif /* __DOXYGEN__ */
+#ifndef __DOXYGEN__
+/* These macros are deprecated and will disappear in a future MDE release. */
+#define NETIO_PKT_GOOD_CHECKSUM(pkt) \
+  NETIO_PKT_L4_CSUM_CORRECT(pkt)
+#define NETIO_PKT_GOOD_CHECKSUM_M(mda, pkt) \
+  NETIO_PKT_L4_CSUM_CORRECT_M(mda, pkt)
+#endif /* __DOXYGEN__ */
+/* Packet attribute access functions. */
+/** Return a pointer to the metadata for a packet.
+ * @ingroup ingress
+ *
+ * Calling this function once and passing the result to other retrieval
+ * functions with a "_M" suffix usually improves performance.  This
+ * function must be called on an 'ingress' packet (i.e. one retrieved
+ * by @ref netio_get_packet(), on which @ref netio_populate_buffer() or
+ * @ref netio_populate_prepend_buffer have not been called). Use of this
+ * function on an 'egress' packet will cause an assertion failure.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to the packet's standard metadata.
+ */
+static __inline netio_pkt_metadata_t*
+NETIO_PKT_METADATA(netio_pkt_t* pkt)
+{
+  netio_assert(!pkt->__packet.bits.__minimal);
+  return &pkt->__metadata;
+}
+/** Return a pointer to the minimal metadata for a packet.
+ * @ingroup egress
+ *
+ * Calling this function once and passing the result to other retrieval
+ * functions with a "_MM" suffix usually improves performance.  This
+ * function must be called on an 'egress' packet (i.e. one on which
+ * @ref netio_populate_buffer() or @ref netio_populate_prepend_buffer()
+ * have been called, or one retrieved by @ref netio_get_buffer()). Use of
+ * this function on an 'ingress' packet will cause an assertion failure.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to the packet's standard metadata.
+ */
+static __inline netio_pkt_minimal_metadata_t*
+NETIO_PKT_MINIMAL_METADATA(netio_pkt_t* pkt)
+{
+  netio_assert(pkt->__packet.bits.__minimal);
+  return (netio_pkt_minimal_metadata_t*) &pkt->__metadata;
+}
+/** Determine whether a packet has 'minimal' metadata.
+ * @ingroup pktfuncs
+ *
+ * This function will return nonzero if the packet is an 'egress'
+ * packet (i.e. one on which @ref netio_populate_buffer() or
+ * @ref netio_populate_prepend_buffer() have been called, or one
+ * retrieved by @ref netio_get_buffer()), and zero if the packet
+ * is an 'ingress' packet (i.e. one retrieved by @ref netio_get_packet(),
+ * which has not been converted into an 'egress' packet).
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the packet has minimal metadata.
+ */
+static __inline unsigned int
+NETIO_PKT_IS_MINIMAL(netio_pkt_t* pkt)
+{
+  return pkt->__packet.bits.__minimal;
+}
+/** Return a handle for a packet's storage.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A handle for the packet's storage.
+ */
+static __inline netio_pkt_handle_t
+NETIO_PKT_HANDLE(netio_pkt_t* pkt)
+{
+  netio_pkt_handle_t h;
+  h.word = pkt->__packet.word;
+  return h;
+}
+/** A special reserved value indicating the absence of a packet handle.
+ *
+ * @ingroup pktfuncs
+ */
+#define NETIO_PKT_HANDLE_NONE ((netio_pkt_handle_t) { 0 })
+/** Test whether a packet handle is valid.
+ *
+ * Applications may wish to use the reserved value NETIO_PKT_HANDLE_NONE
+ * to indicate no packet at all.  This function tests to see if a packet
+ * handle is a real handle, not this special reserved value.
+ *
+ * @ingroup pktfuncs
+ *
+ * @param[in] handle Handle on which to operate.
+ * @return One if the packet handle is valid, else zero.
+ */
+static __inline unsigned int
+NETIO_PKT_HANDLE_IS_VALID(netio_pkt_handle_t handle)
+{
+  return handle.word != 0;
+}
+/** Return a pointer to the start of the packet's custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup ingress
+ *
+ * @param[in] handle Handle on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_CUSTOM_DATA_H(netio_pkt_handle_t handle)
+{
+  return _NETIO_PKT_HANDLE_BASE(handle) + NETIO_PACKET_PADDING;
+}
+/** Return the length of the packet's custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ *
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet's custom header, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_CUSTOM_HEADER_LENGTH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  /*
+   * Note that we effectively need to extract a quantity from the flags word
+   * which is measured in words, and then turn it into bytes by shifting
+   * it left by 2.  We do this all at once by just shifting right two less
+   * bits, and shifting the mask up two bits.
+   */
+  return ((mda->__flags >> (_NETIO_PKT_CUSTOM_LEN_SHIFT - 2)) &
+          (_NETIO_PKT_CUSTOM_LEN_RMASK << 2));
+}
+/** Return the length of the packet, starting with the custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_CUSTOM_LENGTH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (__NETIO_PKT_NOTIF_HEADER(pkt).bits.__transfer_size -
+          NETIO_PACKET_PADDING);
+}
+/** Return a pointer to the start of the packet's custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_CUSTOM_DATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return NETIO_PKT_CUSTOM_DATA_H(NETIO_PKT_HANDLE(pkt));
+}
+/** Return the length of the packet's L2 (Ethernet plus VLAN or SNAP) header.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet's L2 header, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_HEADER_LENGTH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  /*
+   * Note that we effectively need to extract a quantity from the flags word
+   * which is measured in words, and then turn it into bytes by shifting
+   * it left by 2.  We do this all at once by just shifting right two less
+   * bits, and shifting the mask up two bits.  We then add two bytes.
+   */
+  return ((mda->__flags >> (_NETIO_PKT_L2_LEN_SHIFT - 2)) &
+          (_NETIO_PKT_L2_LEN_RMASK << 2)) + 2;
+}
+/** Return the length of the packet, starting with the L2 (Ethernet) header.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_LENGTH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (NETIO_PKT_CUSTOM_LENGTH_M(mda, pkt) -
+          NETIO_PKT_CUSTOM_HEADER_LENGTH_M(mda,pkt));
+}
+/** Return a pointer to the start of the packet's L2 (Ethernet) header.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_L2_DATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (NETIO_PKT_CUSTOM_DATA_M(mda, pkt) +
+          NETIO_PKT_CUSTOM_HEADER_LENGTH_M(mda, pkt));
+}
+/** Retrieve the length of the packet, starting with the L3 (generally,
+ *  the IP) header.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Length of the packet's L3 header and data, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L3_LENGTH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (NETIO_PKT_L2_LENGTH_M(mda, pkt) -
+          NETIO_PKT_L2_HEADER_LENGTH_M(mda,pkt));
+}
+/** Return a pointer to the packet's L3 (generally, the IP) header.
+ * @ingroup ingress
+ *
+ * Note that we guarantee word alignment of the L3 header.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to the packet's L3 header.
+ */
+static __inline unsigned char*
+NETIO_PKT_L3_DATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (NETIO_PKT_L2_DATA_M(mda, pkt) +
+          NETIO_PKT_L2_HEADER_LENGTH_M(mda, pkt));
+}
+/** Return the ordinal of the packet.
+ * @ingroup ingress
+ *
+ * Each packet is given an ordinal number when it is delivered by the IPP.
+ * In the medium term, the ordinal is unique and monotonically increasing,
+ * being incremented by 1 for each packet; the ordinal of the first packet
+ * delivered after the IPP starts is zero.  (Since the ordinal is of finite
+ * size, given enough input packets, it will eventually wrap around to zero;
+ * in the long term, therefore, ordinals are not unique.)  The ordinals
+ * handed out by different IPPs are not disjoint, so two packets from
+ * different IPPs may have identical ordinals.  Packets dropped by the
+ * IPP or by the I/O shim are not assigned ordinals.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's per-IPP packet ordinal.
+ */
+static __inline unsigned int
+NETIO_PKT_ORDINAL_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return mda->__packet_ordinal;
+}
+/** Return the per-group ordinal of the packet.
+ * @ingroup ingress
+ *
+ * Each packet is given a per-group ordinal number when it is
+ * delivered by the IPP. By default, the group is the packet's VLAN,
+ * although IPP can be recompiled to use different values.  In
+ * the medium term, the ordinal is unique and monotonically
+ * increasing, being incremented by 1 for each packet; the ordinal of
+ * the first packet distributed to a particular group is zero.
+ * (Since the ordinal is of finite size, given enough input packets,
+ * it will eventually wrap around to zero; in the long term,
+ * therefore, ordinals are not unique.)  The ordinals handed out by
+ * different IPPs are not disjoint, so two packets from different IPPs
+ * may have identical ordinals; similarly, packets distributed to
+ * different groups may have identical ordinals.  Packets dropped by
+ * the IPP or by the I/O shim are not assigned ordinals.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's per-IPP, per-group ordinal.
+ */
+static __inline unsigned int
+NETIO_PKT_GROUP_ORDINAL_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return mda->__group_ordinal;
+}
+/** Return the VLAN ID assigned to the packet.
+ * @ingroup ingress
+ *
+ * This value is usually contained within the packet header.
+ *
+ * This value will be zero if the packet does not have a VLAN tag, or if
+ * this value was not extracted from the packet.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's VLAN ID.
+ */
+static __inline unsigned short
+NETIO_PKT_VLAN_ID_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  int vl = (mda->__flags >> _NETIO_PKT_VLAN_SHIFT) & _NETIO_PKT_VLAN_RMASK;
+  unsigned short* pkt_p;
+  int index;
+  unsigned short val;
+  if (vl == _NETIO_PKT_VLAN_NONE)
+    return 0;
+  pkt_p = (unsigned short*) NETIO_PKT_L2_DATA_M(mda, pkt);
+  index = (mda->__flags >> _NETIO_PKT_TYPE_SHIFT) & _NETIO_PKT_TYPE_RMASK;
+  val = pkt_p[(_netio_pkt_info[index] >> _NETIO_PKT_INFO_VLAN_SHIFT) &
+              _NETIO_PKT_INFO_VLAN_RMASK];
+#ifdef __TILECC__
+  return (__insn_bytex(val) >> 16) & 0xFFF;
+#else
+  return (__builtin_bswap32(val) >> 16) & 0xFFF;
+#endif
+}
+/** Return the ethertype of the packet.
+ * @ingroup ingress
+ *
+ * This value is usually contained within the packet header.
+ *
+ * This value is reliable if @ref NETIO_PKT_ETHERTYPE_RECOGNIZED_M()
+ * returns true, and otherwise, may not be well defined.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's ethertype.
+ */
+static __inline unsigned short
+NETIO_PKT_ETHERTYPE_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  unsigned short* pkt_p = (unsigned short*) NETIO_PKT_L2_DATA_M(mda, pkt);
+  int index = (mda->__flags >> _NETIO_PKT_TYPE_SHIFT) & _NETIO_PKT_TYPE_RMASK;
+  unsigned short val =
+    pkt_p[(_netio_pkt_info[index] >> _NETIO_PKT_INFO_ETYPE_SHIFT) &
+          _NETIO_PKT_INFO_ETYPE_RMASK];
+  return __builtin_bswap32(val) >> 16;
+}
+/** Return the flow hash computed on the packet.
+ * @ingroup ingress
+ *
+ * For TCP and UDP packets, this hash is calculated by hashing together
+ * the "5-tuple" values, specifically the source IP address, destination
+ * IP address, protocol type, source port and destination port.
+ * The hash value is intended to be helpful for millions of distinct
+ * flows.
+ *
+ * For IPv4 or IPv6 packets which are neither TCP nor UDP, the flow hash is
+ * derived by hashing together the source and destination IP addresses.
+ *
+ * For MPLS-encapsulated packets, the flow hash is derived by hashing
+ * the first MPLS label.
+ *
+ * For all other packets the flow hash is computed from the source
+ * and destination Ethernet addresses.
+ *
+ * The hash is symmetric, meaning it produces the same value if the
+ * source and destination are swapped. The only exceptions are
+ * tunneling protocols 0x04 (IP in IP Encapsulation), 0x29 (Simple
+ * Internet Protocol), 0x2F (General Routing Encapsulation) and 0x32
+ * (Encap Security Payload), which use only the destination address
+ * since the source address is not meaningful.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's 32-bit flow hash.
+ */
+static __inline unsigned int
+NETIO_PKT_FLOW_HASH_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return mda->__flow_hash;
+}
+/** Return the first word of "user data" for the packet.
+ *
+ * The contents of the user data words depend on the IPP.
+ *
+ * When using the standard ipp1, ipp2, or ipp4 sub-drivers, the first
+ * word of user data contains the least significant bits of the 64-bit
+ * arrival cycle count (see @c get_cycle_count_low()).
+ *
+ * See the <em>System Programmer's Guide</em> for details.
+ *
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's first word of "user data".
+ */
+static __inline unsigned int
+NETIO_PKT_USER_DATA_0_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return mda->__user_data_0;
+}
+/** Return the second word of "user data" for the packet.
+ *
+ * The contents of the user data words depend on the IPP.
+ *
+ * When using the standard ipp1, ipp2, or ipp4 sub-drivers, the second
+ * word of user data contains the most significant bits of the 64-bit
+ * arrival cycle count (see @c get_cycle_count_high()).
+ *
+ * See the <em>System Programmer's Guide</em> for details.
+ *
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's second word of "user data".
+ */
+static __inline unsigned int
+NETIO_PKT_USER_DATA_1_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return mda->__user_data_1;
+}
+/** Determine whether the L4 (TCP/UDP) checksum was calculated.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the L4 checksum was calculated.
+ */
+static __inline unsigned int
+NETIO_PKT_L4_CSUM_CALCULATED_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return !(mda->__flags & _NETIO_PKT_NO_L4_CSUM_MASK);
+}
+/** Determine whether the L4 (TCP/UDP) checksum was calculated and found to
+ *  be correct.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the checksum was calculated and is correct.
+ */
+static __inline unsigned int
+NETIO_PKT_L4_CSUM_CORRECT_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return !(mda->__flags &
+           (_NETIO_PKT_BAD_L4_CSUM_MASK | _NETIO_PKT_NO_L4_CSUM_MASK));
+}
+/** Determine whether the L3 (IP) checksum was calculated.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the L3 (IP) checksum was calculated.
+*/
+static __inline unsigned int
+NETIO_PKT_L3_CSUM_CALCULATED_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return !(mda->__flags & _NETIO_PKT_NO_L3_CSUM_MASK);
+}
+/** Determine whether the L3 (IP) checksum was calculated and found to be
+ *  correct.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the checksum was calculated and is correct.
+ */
+static __inline unsigned int
+NETIO_PKT_L3_CSUM_CORRECT_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return !(mda->__flags &
+           (_NETIO_PKT_BAD_L3_CSUM_MASK | _NETIO_PKT_NO_L3_CSUM_MASK));
+}
+/** Determine whether the ethertype was recognized and L3 packet data was
+ *  processed.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the ethertype was recognized and L3 packet data was
+ *   processed.
+ */
+static __inline unsigned int
+NETIO_PKT_ETHERTYPE_RECOGNIZED_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return !(mda->__flags & _NETIO_PKT_TYPE_UNRECOGNIZED_MASK);
+}
+/** Retrieve the status of a packet and any errors that may have occurred
+ * during ingress processing (length mismatches, CRC errors, etc.).
+ * @ingroup ingress
+ *
+ * Note that packets for which @ref NETIO_PKT_ETHERTYPE_RECOGNIZED()
+ * returns zero are always reported as underlength, as there is no a priori
+ * means to determine their length.  Normally, applications should use
+ * @ref NETIO_PKT_BAD_M() instead of explicitly checking status with this
+ * function.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's status.
+ */
+static __inline netio_pkt_status_t
+NETIO_PKT_STATUS_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (netio_pkt_status_t) __NETIO_PKT_NOTIF_HEADER(pkt).bits.__status;
+}
+/** Report whether a packet is bad (i.e., was shorter than expected based on
+ *  its headers, or had a bad CRC).
+ * @ingroup ingress
+ *
+ * Note that this function does not verify L3 or L4 checksums.
+ *
+ * @param[in] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the packet is bad and should be discarded.
+ */
+static __inline unsigned int
+NETIO_PKT_BAD_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return ((NETIO_PKT_STATUS_M(mda, pkt) & 1) &&
+          (NETIO_PKT_ETHERTYPE_RECOGNIZED_M(mda, pkt) ||
+           NETIO_PKT_STATUS_M(mda, pkt) == NETIO_PKT_STATUS_BAD));
+}
+/** Return the length of the packet, starting with the L2 (Ethernet) header.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_LENGTH_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt)
+{
+  return mmd->l2_length;
+}
+/** Return the length of the L2 (Ethernet) header.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet's L2 header, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_HEADER_LENGTH_MM(netio_pkt_minimal_metadata_t* mmd,
+                              netio_pkt_t* pkt)
+{
+  return mmd->l3_offset - mmd->l2_offset;
+}
+/** Return the length of the packet, starting with the L3 (IP) header.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return Length of the packet's L3 header and data, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L3_LENGTH_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt)
+{
+  return (NETIO_PKT_L2_LENGTH_MM(mmd, pkt) -
+          NETIO_PKT_L2_HEADER_LENGTH_MM(mmd, pkt));
+}
+/** Return a pointer to the packet's L3 (generally, the IP) header.
+ * @ingroup egress
+ *
+ * Note that we guarantee word alignment of the L3 header.
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to the packet's L3 header.
+ */
+static __inline unsigned char*
+NETIO_PKT_L3_DATA_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt)
+{
+  return _NETIO_PKT_BASE(pkt) + mmd->l3_offset;
+}
+/** Return a pointer to the packet's L2 (Ethernet) header.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_L2_DATA_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt)
+{
+  return _NETIO_PKT_BASE(pkt) + mmd->l2_offset;
+}
+/** Retrieve the status of a packet and any errors that may have occurred
+ * during ingress processing (length mismatches, CRC errors, etc.).
+ * @ingroup ingress
+ *
+ * Note that packets for which @ref NETIO_PKT_ETHERTYPE_RECOGNIZED()
+ * returns zero are always reported as underlength, as there is no a priori
+ * means to determine their length.  Normally, applications should use
+ * @ref NETIO_PKT_BAD() instead of explicitly checking status with this
+ * function.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's status.
+ */
+static __inline netio_pkt_status_t
+NETIO_PKT_STATUS(netio_pkt_t* pkt)
+{
+  netio_assert(!pkt->__packet.bits.__minimal);
+  return (netio_pkt_status_t) __NETIO_PKT_NOTIF_HEADER(pkt).bits.__status;
+}
+/** Report whether a packet is bad (i.e., was shorter than expected based on
+ *  its headers, or had a bad CRC).
+ * @ingroup ingress
+ *
+ * Note that this function does not verify L3 or L4 checksums.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the packet is bad and should be discarded.
+ */
+static __inline unsigned int
+NETIO_PKT_BAD(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_BAD_M(mda, pkt);
+}
+/** Return the length of the packet's custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet's custom header, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_CUSTOM_HEADER_LENGTH(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_CUSTOM_HEADER_LENGTH_M(mda, pkt);
+}
+/** Return the length of the packet, starting with the custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return  The length of the packet, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_CUSTOM_LENGTH(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_CUSTOM_LENGTH_M(mda, pkt);
+}
+/** Return a pointer to the packet's custom header.
+ *  A custom header may or may not be present, depending upon the IPP; its
+ *  contents and alignment are also IPP-dependent.  Currently, none of the
+ *  standard IPPs supplied by Tilera produce a custom header.  If present,
+ *  the custom header precedes the L2 header in the packet buffer.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_CUSTOM_DATA(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_CUSTOM_DATA_M(mda, pkt);
+}
+/** Return the length of the packet's L2 (Ethernet plus VLAN or SNAP) header.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The length of the packet's L2 header, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_HEADER_LENGTH(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_L2_HEADER_LENGTH_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_L2_HEADER_LENGTH_M(mda, pkt);
+  }
+}
+/** Return the length of the packet, starting with the L2 (Ethernet) header.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return  The length of the packet, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L2_LENGTH(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_L2_LENGTH_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_L2_LENGTH_M(mda, pkt);
+  }
+}
+/** Return a pointer to the packet's L2 (Ethernet) header.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to start of the packet.
+ */
+static __inline unsigned char*
+NETIO_PKT_L2_DATA(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_L2_DATA_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_L2_DATA_M(mda, pkt);
+  }
+}
+/** Retrieve the length of the packet, starting with the L3 (generally, the IP)
+ * header.
+ * @ingroup pktfuncs
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Length of the packet's L3 header and data, in bytes.
+ */
+static __inline netio_size_t
+NETIO_PKT_L3_LENGTH(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_L3_LENGTH_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_L3_LENGTH_M(mda, pkt);
+  }
+}
+/** Return a pointer to the packet's L3 (generally, the IP) header.
+ * @ingroup pktfuncs
+ *
+ * Note that we guarantee word alignment of the L3 header.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return A pointer to the packet's L3 header.
+ */
+static __inline unsigned char*
+NETIO_PKT_L3_DATA(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_L3_DATA_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_L3_DATA_M(mda, pkt);
+  }
+}
+/** Return the ordinal of the packet.
+ * @ingroup ingress
+ *
+ * Each packet is given an ordinal number when it is delivered by the IPP.
+ * In the medium term, the ordinal is unique and monotonically increasing,
+ * being incremented by 1 for each packet; the ordinal of the first packet
+ * delivered after the IPP starts is zero.  (Since the ordinal is of finite
+ * size, given enough input packets, it will eventually wrap around to zero;
+ * in the long term, therefore, ordinals are not unique.)  The ordinals
+ * handed out by different IPPs are not disjoint, so two packets from
+ * different IPPs may have identical ordinals.  Packets dropped by the
+ * IPP or by the I/O shim are not assigned ordinals.
+ *
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's per-IPP packet ordinal.
+ */
+static __inline unsigned int
+NETIO_PKT_ORDINAL(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_ORDINAL_M(mda, pkt);
+}
+/** Return the per-group ordinal of the packet.
+ * @ingroup ingress
+ *
+ * Each packet is given a per-group ordinal number when it is
+ * delivered by the IPP. By default, the group is the packet's VLAN,
+ * although IPP can be recompiled to use different values.  In
+ * the medium term, the ordinal is unique and monotonically
+ * increasing, being incremented by 1 for each packet; the ordinal of
+ * the first packet distributed to a particular group is zero.
+ * (Since the ordinal is of finite size, given enough input packets,
+ * it will eventually wrap around to zero; in the long term,
+ * therefore, ordinals are not unique.)  The ordinals handed out by
+ * different IPPs are not disjoint, so two packets from different IPPs
+ * may have identical ordinals; similarly, packets distributed to
+ * different groups may have identical ordinals.  Packets dropped by
+ * the IPP or by the I/O shim are not assigned ordinals.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's per-IPP, per-group ordinal.
+ */
+static __inline unsigned int
+NETIO_PKT_GROUP_ORDINAL(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_GROUP_ORDINAL_M(mda, pkt);
+}
+/** Return the VLAN ID assigned to the packet.
+ * @ingroup ingress
+ *
+ * This is usually also contained within the packet header.  If the packet
+ * does not have a VLAN tag, the VLAN ID returned by this function is zero.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's VLAN ID.
+ */
+static __inline unsigned short
+NETIO_PKT_VLAN_ID(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_VLAN_ID_M(mda, pkt);
+}
+/** Return the ethertype of the packet.
+ * @ingroup ingress
+ *
+ * This value is reliable if @ref NETIO_PKT_ETHERTYPE_RECOGNIZED()
+ * returns true, and otherwise, may not be well defined.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's ethertype.
+ */
+static __inline unsigned short
+NETIO_PKT_ETHERTYPE(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_ETHERTYPE_M(mda, pkt);
+}
+/** Return the flow hash computed on the packet.
+ * @ingroup ingress
+ *
+ * For TCP and UDP packets, this hash is calculated by hashing together
+ * the "5-tuple" values, specifically the source IP address, destination
+ * IP address, protocol type, source port and destination port.
+ * The hash value is intended to be helpful for millions of distinct
+ * flows.
+ *
+ * For IPv4 or IPv6 packets which are neither TCP nor UDP, the flow hash is
+ * derived by hashing together the source and destination IP addresses.
+ *
+ * For MPLS-encapsulated packets, the flow hash is derived by hashing
+ * the first MPLS label.
+ *
+ * For all other packets the flow hash is computed from the source
+ * and destination Ethernet addresses.
+ *
+ * The hash is symmetric, meaning it produces the same value if the
+ * source and destination are swapped. The only exceptions are
+ * tunneling protocols 0x04 (IP in IP Encapsulation), 0x29 (Simple
+ * Internet Protocol), 0x2F (General Routing Encapsulation) and 0x32
+ * (Encap Security Payload), which use only the destination address
+ * since the source address is not meaningful.
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's 32-bit flow hash.
+ */
+static __inline unsigned int
+NETIO_PKT_FLOW_HASH(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_FLOW_HASH_M(mda, pkt);
+}
+/** Return the first word of "user data" for the packet.
+ *
+ * The contents of the user data words depend on the IPP.
+ *
+ * When using the standard ipp1, ipp2, or ipp4 sub-drivers, the first
+ * word of user data contains the least significant bits of the 64-bit
+ * arrival cycle count (see @c get_cycle_count_low()).
+ *
+ * See the <em>System Programmer's Guide</em> for details.
+ *
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's first word of "user data".
+ */
+static __inline unsigned int
+NETIO_PKT_USER_DATA_0(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_USER_DATA_0_M(mda, pkt);
+}
+/** Return the second word of "user data" for the packet.
+ *
+ * The contents of the user data words depend on the IPP.
+ *
+ * When using the standard ipp1, ipp2, or ipp4 sub-drivers, the second
+ * word of user data contains the most significant bits of the 64-bit
+ * arrival cycle count (see @c get_cycle_count_high()).
+ *
+ * See the <em>System Programmer's Guide</em> for details.
+ *
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return The packet's second word of "user data".
+ */
+static __inline unsigned int
+NETIO_PKT_USER_DATA_1(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_USER_DATA_1_M(mda, pkt);
+}
+/** Determine whether the L4 (TCP/UDP) checksum was calculated.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the L4 checksum was calculated.
+ */
+static __inline unsigned int
+NETIO_PKT_L4_CSUM_CALCULATED(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_L4_CSUM_CALCULATED_M(mda, pkt);
+}
+/** Determine whether the L4 (TCP/UDP) checksum was calculated and found to
+ *  be correct.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the checksum was calculated and is correct.
+ */
+static __inline unsigned int
+NETIO_PKT_L4_CSUM_CORRECT(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_L4_CSUM_CORRECT_M(mda, pkt);
+}
+/** Determine whether the L3 (IP) checksum was calculated.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the L3 (IP) checksum was calculated.
+*/
+static __inline unsigned int
+NETIO_PKT_L3_CSUM_CALCULATED(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_L3_CSUM_CALCULATED_M(mda, pkt);
+}
+/** Determine whether the L3 (IP) checksum was calculated and found to be
+ *  correct.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the checksum was calculated and is correct.
+ */
+static __inline unsigned int
+NETIO_PKT_L3_CSUM_CORRECT(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_L3_CSUM_CORRECT_M(mda, pkt);
+}
+/** Determine whether the Ethertype was recognized and L3 packet data was
+ *  processed.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ * @return Nonzero if the Ethertype was recognized and L3 packet data was
+ *   processed.
+ */
+static __inline unsigned int
+NETIO_PKT_ETHERTYPE_RECOGNIZED(netio_pkt_t* pkt)
+{
+  netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+  return NETIO_PKT_ETHERTYPE_RECOGNIZED_M(mda, pkt);
+}
+/** Set an egress packet's L2 length, using a metadata pointer to speed the
+ * computation.
+ * @ingroup egress
+ *
+ * @param[in,out] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @param[in] len Packet L2 length, in bytes.
+ */
+static __inline void
+NETIO_PKT_SET_L2_LENGTH_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt,
+                           int len)
+{
+  mmd->l2_length = len;
+}
+/** Set an egress packet's L2 length.
+ * @ingroup egress
+ *
+ * @param[in,out] pkt Packet on which to operate.
+ * @param[in] len Packet L2 length, in bytes.
+ */
+static __inline void
+NETIO_PKT_SET_L2_LENGTH(netio_pkt_t* pkt, int len)
+{
+  netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+  NETIO_PKT_SET_L2_LENGTH_MM(mmd, pkt, len);
+}
+/** Set an egress packet's L2 header length, using a metadata pointer to
+ *  speed the computation.
+ * @ingroup egress
+ *
+ * It is not normally necessary to call this routine; only the L2 length,
+ * not the header length, is needed to transmit a packet.  It may be useful if
+ * the egress packet will later be processed by code which expects to use
+ * functions like @ref NETIO_PKT_L3_DATA() to get a pointer to the L3 payload.
+ *
+ * @param[in,out] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @param[in] len Packet L2 header length, in bytes.
+ */
+static __inline void
+NETIO_PKT_SET_L2_HEADER_LENGTH_MM(netio_pkt_minimal_metadata_t* mmd,
+                                  netio_pkt_t* pkt, int len)
+{
+  mmd->l3_offset = mmd->l2_offset + len;
+}
+/** Set an egress packet's L2 header length.
+ * @ingroup egress
+ *
+ * It is not normally necessary to call this routine; only the L2 length,
+ * not the header length, is needed to transmit a packet.  It may be useful if
+ * the egress packet will later be processed by code which expects to use
+ * functions like @ref NETIO_PKT_L3_DATA() to get a pointer to the L3 payload.
+ *
+ * @param[in,out] pkt Packet on which to operate.
+ * @param[in] len Packet L2 header length, in bytes.
+ */
+static __inline void
+NETIO_PKT_SET_L2_HEADER_LENGTH(netio_pkt_t* pkt, int len)
+{
+  netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+  NETIO_PKT_SET_L2_HEADER_LENGTH_MM(mmd, pkt, len);
+}
+/** Set up an egress packet for hardware checksum computation, using a
+ *  metadata pointer to speed the operation.
+ * @ingroup egress
+ *
+ *  NetIO provides the ability to automatically calculate a standard
+ *  16-bit Internet checksum on transmitted packets.  The application
+ *  may specify the point in the packet where the checksum starts, the
+ *  number of bytes to be checksummed, and the two bytes in the packet
+ *  which will be replaced with the completed checksum.  (If the range
+ *  of bytes to be checksummed includes the bytes to be replaced, the
+ *  initial values of those bytes will be included in the checksum.)
+ *
+ *  For some protocols, the packet checksum covers data which is not present
+ *  in the packet, or is at least not contiguous to the main data payload.
+ *  For instance, the TCP checksum includes a "pseudo-header" which includes
+ *  the source and destination IP addresses of the packet.  To accommodate
+ *  this, the checksum engine may be "seeded" with an initial value, which
+ *  the application would need to compute based on the specific protocol's
+ *  requirements.  Note that the seed is given in host byte order (little-
+ *  endian), not network byte order (big-endian); code written to compute a
+ *  pseudo-header checksum in network byte order will need to byte-swap it
+ *  before use as the seed.
+ *
+ *  Note that the checksum is computed as part of the transmission process,
+ *  so it will not be present in the packet upon completion of this routine.
+ *
+ * @param[in,out] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ * @param[in] start Offset within L2 packet of the first byte to include in
+ *   the checksum.
+ * @param[in] length Number of bytes to include in the checksum.
+ *   the checksum.
+ * @param[in] location Offset within L2 packet of the first of the two bytes
+ *   to be replaced with the calculated checksum.
+ * @param[in] seed Initial value of the running checksum before any of the
+ *   packet data is added.
+ */
+static __inline void
+NETIO_PKT_DO_EGRESS_CSUM_MM(netio_pkt_minimal_metadata_t* mmd,
+                            netio_pkt_t* pkt, int start, int length,
+                            int location, uint16_t seed)
+{
+  mmd->csum_start = start;
+  mmd->csum_length = length;
+  mmd->csum_location = location;
+  mmd->csum_seed = seed;
+  mmd->flags |= _NETIO_PKT_NEED_EDMA_CSUM_MASK;
+}
+/** Set up an egress packet for hardware checksum computation.
+ * @ingroup egress
+ *
+ *  NetIO provides the ability to automatically calculate a standard
+ *  16-bit Internet checksum on transmitted packets.  The application
+ *  may specify the point in the packet where the checksum starts, the
+ *  number of bytes to be checksummed, and the two bytes in the packet
+ *  which will be replaced with the completed checksum.  (If the range
+ *  of bytes to be checksummed includes the bytes to be replaced, the
+ *  initial values of those bytes will be included in the checksum.)
+ *
+ *  For some protocols, the packet checksum covers data which is not present
+ *  in the packet, or is at least not contiguous to the main data payload.
+ *  For instance, the TCP checksum includes a "pseudo-header" which includes
+ *  the source and destination IP addresses of the packet.  To accommodate
+ *  this, the checksum engine may be "seeded" with an initial value, which
+ *  the application would need to compute based on the specific protocol's
+ *  requirements.  Note that the seed is given in host byte order (little-
+ *  endian), not network byte order (big-endian); code written to compute a
+ *  pseudo-header checksum in network byte order will need to byte-swap it
+ *  before use as the seed.
+ *
+ *  Note that the checksum is computed as part of the transmission process,
+ *  so it will not be present in the packet upon completion of this routine.
+ *
+ * @param[in,out] pkt Packet on which to operate.
+ * @param[in] start Offset within L2 packet of the first byte to include in
+ *   the checksum.
+ * @param[in] length Number of bytes to include in the checksum.
+ *   the checksum.
+ * @param[in] location Offset within L2 packet of the first of the two bytes
+ *   to be replaced with the calculated checksum.
+ * @param[in] seed Initial value of the running checksum before any of the
+ *   packet data is added.
+ */
+static __inline void
+NETIO_PKT_DO_EGRESS_CSUM(netio_pkt_t* pkt, int start, int length,
+                         int location, uint16_t seed)
+{
+  netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+  NETIO_PKT_DO_EGRESS_CSUM_MM(mmd, pkt, start, length, location, seed);
+}
+/** Return the number of bytes which could be prepended to a packet, using a
+ *  metadata pointer to speed the operation.
+ *  See @ref netio_populate_prepend_buffer() to get a full description of
+ *  prepending.
+ *
+ * @param[in,out] mda Pointer to packet's standard metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline int
+NETIO_PKT_PREPEND_AVAIL_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+  return (pkt->__packet.bits.__offset << 6) +
+         NETIO_PKT_CUSTOM_HEADER_LENGTH_M(mda, pkt);
+}
+/** Return the number of bytes which could be prepended to a packet, using a
+ *  metadata pointer to speed the operation.
+ *  See @ref netio_populate_prepend_buffer() to get a full description of
+ *  prepending.
+ * @ingroup egress
+ *
+ * @param[in,out] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline int
+NETIO_PKT_PREPEND_AVAIL_MM(netio_pkt_minimal_metadata_t* mmd, netio_pkt_t* pkt)
+{
+  return (pkt->__packet.bits.__offset << 6) + mmd->l2_offset;
+}
+/** Return the number of bytes which could be prepended to a packet.
+ *  See @ref netio_populate_prepend_buffer() to get a full description of
+ *  prepending.
+ * @ingroup egress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline int
+NETIO_PKT_PREPEND_AVAIL(netio_pkt_t* pkt)
+{
+  if (NETIO_PKT_IS_MINIMAL(pkt))
+  {
+    netio_pkt_minimal_metadata_t* mmd = NETIO_PKT_MINIMAL_METADATA(pkt);
+    return NETIO_PKT_PREPEND_AVAIL_MM(mmd, pkt);
+  }
+  else
+  {
+    netio_pkt_metadata_t* mda = NETIO_PKT_METADATA(pkt);
+    return NETIO_PKT_PREPEND_AVAIL_M(mda, pkt);
+  }
+}
+/** Flush a packet's minimal metadata from the cache, using a metadata pointer
+ *  to speed the operation.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_MINIMAL_METADATA_MM(netio_pkt_minimal_metadata_t* mmd,
+                                    netio_pkt_t* pkt)
+{
+}
+/** Invalidate a packet's minimal metadata from the cache, using a metadata
+ *  pointer to speed the operation.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_INV_MINIMAL_METADATA_MM(netio_pkt_minimal_metadata_t* mmd,
+                                  netio_pkt_t* pkt)
+{
+}
+/** Flush and then invalidate a packet's minimal metadata from the cache,
+ *  using a metadata pointer to speed the operation.
+ * @ingroup egress
+ *
+ * @param[in] mmd Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_INV_MINIMAL_METADATA_MM(netio_pkt_minimal_metadata_t* mmd,
+                                        netio_pkt_t* pkt)
+{
+}
+/** Flush a packet's metadata from the cache, using a metadata pointer
+ *  to speed the operation.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's minimal metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_METADATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+}
+/** Invalidate a packet's metadata from the cache, using a metadata
+ *  pointer to speed the operation.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_INV_METADATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+}
+/** Flush and then invalidate a packet's metadata from the cache,
+ *  using a metadata pointer to speed the operation.
+ * @ingroup ingress
+ *
+ * @param[in] mda Pointer to packet's metadata.
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_INV_METADATA_M(netio_pkt_metadata_t* mda, netio_pkt_t* pkt)
+{
+}
+/** Flush a packet's minimal metadata from the cache.
+ * @ingroup egress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_MINIMAL_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Invalidate a packet's minimal metadata from the cache.
+ * @ingroup egress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_INV_MINIMAL_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Flush and then invalidate a packet's minimal metadata from the cache.
+ * @ingroup egress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_INV_MINIMAL_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Flush a packet's metadata from the cache.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Invalidate a packet's metadata from the cache.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_INV_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Flush and then invalidate a packet's metadata from the cache.
+ * @ingroup ingress
+ *
+ * @param[in] pkt Packet on which to operate.
+ */
+static __inline void
+NETIO_PKT_FLUSH_INV_METADATA(netio_pkt_t* pkt)
+{
+}
+/** Number of NUMA nodes we can distribute buffers to.
+ * @ingroup setup */
+#define NETIO_NUM_NODE_WEIGHTS  16
+/**
+ * @brief An object for specifying the characteristics of NetIO communication
+ * endpoint.
+ *
+ * @ingroup setup
+ *
+ * The @ref netio_input_register() function uses this structure to define
+ * how an application tile will communicate with an IPP.
+ *
+ *
+ * Future updates to NetIO may add new members to this structure,
+ * which can affect the success of the registration operation.  Thus,
+ * if dynamically initializing the structure, applications are urged to
+ * zero it out first, for example:
+ *
+ * @code
+ * netio_input_config_t config;
+ * memset(&config, 0, sizeof (config));
+ * config.flags = NETIO_RECV | NETIO_XMIT_CSUM | NETIO_TAG_NONE;
+ * config.num_receive_packets = NETIO_MAX_RECEIVE_PKTS;
+ * config.queue_id = 0;
+ *     .
+ *     .
+ *     .
+ * @endcode
+ *
+ * since that guarantees that any unused structure members, including
+ * members which did not exist when the application was first developed,
+ * will not have unexpected values.
+ *
+ * If statically initializing the structure, we strongly recommend use of
+ * C99-style named initializers, for example:
+ *
+ * @code
+ * netio_input_config_t config = {
+ *    .flags = NETIO_RECV | NETIO_XMIT_CSUM | NETIO_TAG_NONE,
+ *    .num_receive_packets = NETIO_MAX_RECEIVE_PKTS,
+ *    .queue_id = 0,
+ * },
+ * @endcode
+ *
+ * instead of the old-style structure initialization:
+ *
+ * @code
+ * // Bad example! Currently equivalent to the above, but don't do this.
+ * netio_input_config_t config = {
+ *    NETIO_RECV | NETIO_XMIT_CSUM | NETIO_TAG_NONE, NETIO_MAX_RECEIVE_PKTS, 0
+ * },
+ * @endcode
+ *
+ * since the C99 style requires no changes to the code if elements of the
+ * config structure are rearranged.  (It also makes the initialization much
+ * easier to understand.)
+ *
+ * Except for items which address a particular tile's transmit or receive
+ * characteristics, such as the ::NETIO_RECV flag, applications are advised
+ * to specify the same set of configuration data on all registrations.
+ * This prevents differing results if multiple tiles happen to do their
+ * registration operations in a different order on different invocations of
+ * the application.  This is particularly important for things like link
+ * management flags, and buffer size and homing specifications.
+ *
+ * Unless the ::NETIO_FIXED_BUFFER_VA flag is specified in flags, the NetIO
+ * buffer pool is automatically created and mapped into the application's
+ * virtual address space at an address chosen by the operating system,
+ * using the common memory (cmem) facility in the Tilera Multicore
+ * Components library.  The cmem facility allows multiple processes to gain
+ * access to shared memory which is mapped into each process at an
+ * identical virtual address.  In order for this to work, the processes
+ * must have a common ancestor, which must create the common memory using
+ * tmc_cmem_init().
+ *
+ * In programs using the iLib process creation API, or in programs which use
+ * only one process (which include programs using the pthreads library),
+ * tmc_cmem_init() is called automatically.  All other applications
+ * must call it explicitly, before any child processes which might call
+ * netio_input_register() are created.
+ */
+typedef struct
+{
+  /** Registration characteristics.
+      This value determines several characteristics of the registration;
+      flags for different types of behavior are ORed together to make the
+      final flag value.  Generally applications should specify exactly
+      one flag from each of the following categories:
+      - Whether the application will be receiving packets on this queue
+        (::NETIO_RECV or ::NETIO_NO_RECV).
+      - Whether the application will be transmitting packets on this queue,
+        and if so, whether it will request egress checksum calculation
+        (::NETIO_XMIT, ::NETIO_XMIT_CSUM, or ::NETIO_NO_XMIT).  It is
+        legal to call netio_get_buffer() without one of the XMIT flags,
+        as long as ::NETIO_RECV is specified; in this case, the retrieved
+        buffers must be passed to another tile for transmission.
+      - Whether the application expects any vendor-specific tags in
+        its packets' L2 headers (::NETIO_TAG_NONE, ::NETIO_TAG_BRCM,
+        or ::NETIO_TAG_MRVL).  This must match the configuration of the
+        target IPP.
+      To accommodate applications written to previous versions of the NetIO
+      interface, none of the flags above are currently required; if omitted,
+      NetIO behaves more or less as if ::NETIO_RECV | ::NETIO_XMIT_CSUM |
+      ::NETIO_TAG_NONE were used.  However, explicit specification of
+      the relevant flags allows NetIO to do a better job of resource
+      allocation, allows earlier detection of certain configuration errors,
+      and may enable advanced features or higher performance in the future,
+      so their use is strongly recommended.
+      Note that specifying ::NETIO_NO_RECV along with ::NETIO_NO_XMIT
+      is a special case, intended primarily for use by programs which
+      retrieve network statistics or do link management operations.
+      When these flags are both specified, the resulting queue may not
+      be used with NetIO routines other than netio_get(), netio_set(),
+      and netio_input_unregister().  See @ref link for more information
+      on link management.
+      Other flags are optional; their use is described below.
+  */
+  int flags;
+  /** Interface name.  This is a string which identifies the specific
+      Ethernet controller hardware to be used.  The format of the string
+      is a device type and a device index, separated by a slash; so,
+      the first 10 Gigabit Ethernet controller is named "xgbe/0", while
+      the second 10/100/1000 Megabit Ethernet controller is named "gbe/1".
+   */
+  const char* interface;
+  /** Receive packet queue size.  This specifies the maximum number
+      of ingress packets that can be received on this queue without
+      being retrieved by @ref netio_get_packet().  If the IPP's distribution
+      algorithm calls for a packet to be sent to this queue, and this
+      number of packets are already pending there, the new packet
+      will either be discarded, or sent to another tile registered
+      for the same queue_id (see @ref drops).  This value must
+      be at least ::NETIO_MIN_RECEIVE_PKTS, can always be at least
+      ::NETIO_MAX_RECEIVE_PKTS, and may be larger than that on certain
+      interfaces.
+   */
+  int num_receive_packets;
+  /** The queue ID being requested.  Legal values for this range from 0
+      to ::NETIO_MAX_QUEUE_ID, inclusive.  ::NETIO_MAX_QUEUE_ID is always
+      greater than or equal to the number of tiles; this allows one queue
+      for each tile, plus at least one additional queue.  Some applications
+      may wish to use the additional queue as a destination for unwanted
+      packets, since packets delivered to queues for which no tiles have
+      registered are discarded.
+   */
+  unsigned int queue_id;
+  /** Maximum number of small send buffers to be held in the local empty
+      buffer cache.  This specifies the size of the area which holds
+      empty small egress buffers requested from the IPP but not yet
+      retrieved via @ref netio_get_buffer().  This value must be greater
+      than zero if the application will ever use @ref netio_get_buffer()
+      to allocate empty small egress buffers; it may be no larger than
+      ::NETIO_MAX_SEND_BUFFERS.  See @ref epp for more details on empty
+      buffer caching.
+   */
+  int num_send_buffers_small_total;
+  /** Number of small send buffers to be preallocated at registration.
+      If this value is nonzero, the specified number of empty small egress
+      buffers will be requested from the IPP during the netio_input_register
+      operation; this may speed the execution of @ref netio_get_buffer().
+      This may be no larger than @ref num_send_buffers_small_total.  See @ref
+      epp for more details on empty buffer caching.
+   */
+  int num_send_buffers_small_prealloc;
+  /** Maximum number of large send buffers to be held in the local empty
+      buffer cache.  This specifies the size of the area which holds empty
+      large egress buffers requested from the IPP but not yet retrieved via
+      @ref netio_get_buffer().  This value must be greater than zero if the
+      application will ever use @ref netio_get_buffer() to allocate empty
+      large egress buffers; it may be no larger than ::NETIO_MAX_SEND_BUFFERS.
+      See @ref epp for more details on empty buffer caching.
+   */
+  int num_send_buffers_large_total;
+  /** Number of large send buffers to be preallocated at registration.
+      If this value is nonzero, the specified number of empty large egress
+      buffers will be requested from the IPP during the netio_input_register
+      operation; this may speed the execution of @ref netio_get_buffer().
+      This may be no larger than @ref num_send_buffers_large_total.  See @ref
+      epp for more details on empty buffer caching.
+   */
+  int num_send_buffers_large_prealloc;
+  /** Maximum number of jumbo send buffers to be held in the local empty
+      buffer cache.  This specifies the size of the area which holds empty
+      jumbo egress buffers requested from the IPP but not yet retrieved via
+      @ref netio_get_buffer().  This value must be greater than zero if the
+      application will ever use @ref netio_get_buffer() to allocate empty
+      jumbo egress buffers; it may be no larger than ::NETIO_MAX_SEND_BUFFERS.
+      See @ref epp for more details on empty buffer caching.
+   */
+  int num_send_buffers_jumbo_total;
+  /** Number of jumbo send buffers to be preallocated at registration.
+      If this value is nonzero, the specified number of empty jumbo egress
+      buffers will be requested from the IPP during the netio_input_register
+      operation; this may speed the execution of @ref netio_get_buffer().
+      This may be no larger than @ref num_send_buffers_jumbo_total.  See @ref
+      epp for more details on empty buffer caching.
+   */
+  int num_send_buffers_jumbo_prealloc;
+  /** Total packet buffer size.  This determines the total size, in bytes,
+      of the NetIO buffer pool.  Note that the maximum number of available
+      buffers of each size is determined during hypervisor configuration
+      (see the <em>System Programmer's Guide</em> for details); this just
+      influences how much host memory is allocated for those buffers.
+      The buffer pool is allocated from common memory, which will be
+      automatically initialized if needed.  If your buffer pool is larger
+      than 240 MB, you might need to explicitly call @c tmc_cmem_init(),
+      as described in the Application Libraries Reference Manual (UG227).
+      Packet buffers are currently allocated in chunks of 16 MB; this
+      value will be rounded up to the next larger multiple of 16 MB.
+      If this value is zero, a default of 32 MB will be used; this was
+      the value used by previous versions of NetIO.  Note that taking this
+      default also affects the placement of buffers on Linux NUMA nodes.
+      See @ref buffer_node_weights for an explanation of buffer placement.
+      In order to successfully allocate packet buffers, Linux must have
+      available huge pages on the relevant Linux NUMA nodes.  See the
+      <em>System Programmer's Guide</em> for information on configuring
+      huge page support in Linux.
+   */
+  uint64_t total_buffer_size;
+  /** Buffer placement weighting factors.
+      This array specifies the relative amount of buffering to place
+      on each of the available Linux NUMA nodes.  This array is
+      indexed by the NUMA node, and the values in the array are
+      proportional to the amount of buffer space to allocate on that
+      node.
+      If memory striping is enabled in the Hypervisor, then there is
+      only one logical NUMA node (node 0). In that case, NetIO will by
+      default ignore the suggested buffer node weights, and buffers
+      will be striped across the physical memory controllers. See
+      UG209 System Programmer's Guide for a description of the
+      hypervisor option that controls memory striping.
+      If memory striping is disabled, then there are up to four NUMA
+      nodes, corresponding to the four DDRAM controllers in the TILE
+      processor architecture.  See UG100 Tile Processor Architecture
+      Overview for a diagram showing the location of each of the DDRAM
+      controllers relative to the tile array.
+      For instance, if memory striping is disabled, the following
+      configuration strucure:
+      @code
+      netio_input_config_t config = {
+            .
+            .
+            .
+        .total_buffer_size = 4 * 16 * 1024 * 1024;
+        .buffer_node_weights = { 1, 0, 1, 0 },
+      },
+      @endcode
+      would result in 32 MB of buffers being placed on controller 0, and
+      32 MB on controller 2.  (Since buffers are allocated in units of
+      16 MB, some sets of weights will not be able to be matched exactly.)
+      For the weights to be effective, @ref total_buffer_size must be
+      nonzero.  If @ref total_buffer_size is zero, causing the default
+      32 MB of buffer space to be used, then any specified weights will
+      be ignored, and buffers will positioned as they were in previous
+      versions of NetIO:
+      - For xgbe/0 and gbe/0, 16 MB of buffers will be placed on controller 1,
+        and the other 16 MB will be placed on controller 2.
+      - For xgbe/1 and gbe/1, 16 MB of buffers will be placed on controller 2,
+        and the other 16 MB will be placed on controller 3.
+      If @ref total_buffer_size is nonzero, but all weights are zero,
+      then all buffer space will be allocated on Linux NUMA node zero.
+      By default, the specified buffer placement is treated as a hint;
+      if sufficient free memory is not available on the specified
+      controllers, the buffers will be allocated elsewhere.  However,
+      if the ::NETIO_STRICT_HOMING flag is specified in @ref flags, then a
+      failure to allocate buffer space exactly as requested will cause the
+      registration operation to fail with an error of ::NETIO_CANNOT_HOME.
+      Note that maximal network performance cannot be achieved with
+      only one memory controller.
+   */
+  uint8_t buffer_node_weights[NETIO_NUM_NODE_WEIGHTS];
+  /** Fixed virtual address for packet buffers.  Only valid when
+      ::NETIO_FIXED_BUFFER_VA is specified in @ref flags; see the
+      description of that flag for details.
+   */
+  void* fixed_buffer_va;
+  /**
+      Maximum number of outstanding send packet requests.  This value is
+      only relevant when an EPP is in use; it determines the number of
+      slots in the EPP's outgoing packet queue which this tile is allowed
+      to consume, and thus the number of packets which may be sent before
+      the sending tile must wait for an acknowledgment from the EPP.
+      Modifying this value is generally only helpful when using @ref
+      netio_send_packet_vector(), where it can help improve performance by
+      allowing a single vector send operation to process more packets.
+      Typically it is not specified, and the default, which divides the
+      outgoing packet slots evenly between all tiles on the chip, is used.
+      If a registration asks for more outgoing packet queue slots than are
+      available, ::NETIO_TOOMANY_XMIT will be returned.  The total number
+      of packet queue slots which are available for all tiles for each EPP
+      is subject to change, but is currently ::NETIO_TOTAL_SENDS_OUTSTANDING.
+      This value is ignored if ::NETIO_XMIT is not specified in flags.
+      If you want to specify a large value here for a specific tile, you are
+      advised to specify NETIO_NO_XMIT on other, non-transmitting tiles so
+      that they do not consume a default number of packet slots.  Any tile
+      transmitting is required to have at least ::NETIO_MIN_SENDS_OUTSTANDING
+      slots allocated to it; values less than that will be silently
+      increased by the NetIO library.
+   */
+  int num_sends_outstanding;
+}
+netio_input_config_t;
+/** Registration flags; used in the @ref netio_input_config_t structure.
+ * @addtogroup setup
+ */
+/** @{ */
+/** Fail a registration request if we can't put packet buffers
+    on the specified memory controllers. */
+#define NETIO_STRICT_HOMING   0x00000002
+/** This application expects no tags on its L2 headers. */
+#define NETIO_TAG_NONE        0x00000004
+/** This application expects Marvell extended tags on its L2 headers. */
+#define NETIO_TAG_MRVL        0x00000008
+/** This application expects Broadcom tags on its L2 headers. */
+#define NETIO_TAG_BRCM        0x00000010
+/** This registration may call routines which receive packets. */
+#define NETIO_RECV            0x00000020
+/** This registration may not call routines which receive packets. */
+#define NETIO_NO_RECV         0x00000040
+/** This registration may call routines which transmit packets. */
+#define NETIO_XMIT            0x00000080
+/** This registration may call routines which transmit packets with
+    checksum acceleration. */
+#define NETIO_XMIT_CSUM       0x00000100
+/** This registration may not call routines which transmit packets. */
+#define NETIO_NO_XMIT         0x00000200
+/** This registration wants NetIO buffers mapped at an application-specified
+    virtual address.
+    NetIO buffers are by default created by the TMC common memory facility,
+    which must be configured by a common ancestor of all processes sharing
+    a network interface.  When this flag is specified, NetIO buffers are
+    instead mapped at an address chosen by the application (and specified
+    in @ref netio_input_config_t::fixed_buffer_va).  This allows multiple
+    unrelated but cooperating processes to share a NetIO interface.
+    All processes sharing the same interface must specify this flag,
+    and all must specify the same fixed virtual address.
+    @ref netio_input_config_t::fixed_buffer_va must be a
+    multiple of 16 MB, and the packet buffers will occupy @ref
+    netio_input_config_t::total_buffer_size bytes of virtual address
+    space, beginning at that address.  If any of those virtual addresses
+    are currently occupied by other memory objects, like application or
+    shared library code or data, @ref netio_input_register() will return
+    ::NETIO_FAULT.  While it is impossible to provide a fixed_buffer_va
+    which will work for all applications, a good first guess might be to
+    use 0xb0000000 minus @ref netio_input_config_t::total_buffer_size.
+    If that fails, it might be helpful to consult the running application's
+    virtual address description file (/proc/<em>pid</em>/maps) to see
+    which regions of virtual address space are available.
+ */
+#define NETIO_FIXED_BUFFER_VA 0x00000400
+/** This registration call will not complete unless the network link
+    is up.  The process will wait several seconds for this to happen (the
+    precise interval is link-dependent), but if the link does not come up,
+    ::NETIO_LINK_DOWN will be returned.  This flag is the default if
+    ::NETIO_NOREQUIRE_LINK_UP is not specified.  Note that this flag by
+    itself does not request that the link be brought up; that can be done
+    with the ::NETIO_AUTO_LINK_UPDN or ::NETIO_AUTO_LINK_UP flags (the
+    latter is the default if no NETIO_AUTO_LINK_xxx flags are specified),
+    or by explicitly setting the link's desired state via netio_set().
+    If the link is not brought up by one of those methods, and this flag
+    is specified, the registration operation will return ::NETIO_LINK_DOWN.
+    This flag is ignored if it is specified along with ::NETIO_NO_XMIT and
+    ::NETIO_NO_RECV.  See @ref link for more information on link
+    management.
+ */
+#define NETIO_REQUIRE_LINK_UP    0x00000800
+/** This registration call will complete even if the network link is not up.
+    Whenever the link is not up, packets will not be sent or received:
+    netio_get_packet() will return ::NETIO_NOPKT once all queued packets
+    have been drained, and netio_send_packet() and similar routines will
+    return NETIO_QUEUE_FULL once the outgoing packet queue in the EPP
+    or the I/O shim is full.  See @ref link for more information on link
+    management.
+ */
+#define NETIO_NOREQUIRE_LINK_UP  0x00001000
+#ifndef __DOXYGEN__
+/*
+ * These are part of the implementation of the NETIO_AUTO_LINK_xxx flags,
+ * but should not be used directly by applications, and are thus not
+ * documented.
+ */
+#define _NETIO_AUTO_UP        0x00002000
+#define _NETIO_AUTO_DN        0x00004000
+#define _NETIO_AUTO_PRESENT   0x00008000
+#endif
+/** Set the desired state of the link to up, allowing any speeds which are
+    supported by the link hardware, as part of this registration operation.
+    Do not take down the link automatically.  This is the default if
+    no other NETIO_AUTO_LINK_xxx flags are specified.  This flag is ignored
+    if it is specified along with ::NETIO_NO_XMIT and ::NETIO_NO_RECV.
+    See @ref link for more information on link management.
+ */
+#define NETIO_AUTO_LINK_UP     (_NETIO_AUTO_PRESENT | _NETIO_AUTO_UP)
+/** Set the desired state of the link to up, allowing any speeds which are
+    supported by the link hardware, as part of this registration operation.
+    Set the desired state of the link to down the next time no tiles are
+    registered for packet reception or transmission.  This flag is ignored
+    if it is specified along with ::NETIO_NO_XMIT and ::NETIO_NO_RECV.
+    See @ref link for more information on link management.
+ */
+#define NETIO_AUTO_LINK_UPDN   (_NETIO_AUTO_PRESENT | _NETIO_AUTO_UP | \
+                                _NETIO_AUTO_DN)
+/** Set the desired state of the link to down the next time no tiles are
+    registered for packet reception or transmission.  This flag is ignored
+    if it is specified along with ::NETIO_NO_XMIT and ::NETIO_NO_RECV.
+    See @ref link for more information on link management.
+ */
+#define NETIO_AUTO_LINK_DN     (_NETIO_AUTO_PRESENT | _NETIO_AUTO_DN)
+/** Do not bring up the link automatically as part of this registration
+    operation.  Do not take down the link automatically.  This flag
+    is ignored if it is specified along with ::NETIO_NO_XMIT and
+    ::NETIO_NO_RECV.  See @ref link for more information on link management.
+  */
+#define NETIO_AUTO_LINK_NONE   _NETIO_AUTO_PRESENT
+/** Minimum number of receive packets. */
+#define NETIO_MIN_RECEIVE_PKTS            16
+/** Lower bound on the maximum number of receive packets; may be higher
+    than this on some interfaces. */
+#define NETIO_MAX_RECEIVE_PKTS           128
+/** Maximum number of send buffers, per packet size. */
+#define NETIO_MAX_SEND_BUFFERS            16
+/** Number of EPP queue slots, and thus outstanding sends, per EPP. */
+#define NETIO_TOTAL_SENDS_OUTSTANDING   2015
+/** Minimum number of EPP queue slots, and thus outstanding sends, per
+ *  transmitting tile. */
+#define NETIO_MIN_SENDS_OUTSTANDING       16
+/**@}*/
+#ifndef __DOXYGEN__
+/**
+ * An object for providing Ethernet packets to a process.
+ */
+struct __netio_queue_impl_t;
+/**
+ * An object for managing the user end of a NetIO queue.
+ */
+struct __netio_queue_user_impl_t;
+#endif /* !__DOXYGEN__ */
+/** A netio_queue_t describes a NetIO communications endpoint.
+ * @ingroup setup
+ */
+typedef struct
+{
+#ifdef __DOXYGEN__
+  uint8_t opaque[8];                 /**< This is an opaque structure. */
+#else
+  struct __netio_queue_impl_t* __system_part;    /**< The system part. */
+  struct __netio_queue_user_impl_t* __user_part; /**< The user part. */
+#ifdef _NETIO_PTHREAD
+  _netio_percpu_mutex_t lock;                    /**< Queue lock. */
+#endif
+#endif
+}
+netio_queue_t;
+/**
+ * @brief Packet send context.
+ *
+ * @ingroup egress
+ *
+ * Packet send context for use with netio_send_packet_prepare and _commit.
+ */
+typedef struct
+{
+#ifdef __DOXYGEN__
+  uint8_t opaque[44];   /**< This is an opaque structure. */
+#else
+  uint8_t flags;        /**< Defined below */
+  uint8_t datalen;      /**< Number of valid words pointed to by data. */
+  uint32_t request[9];  /**< Request to be sent to the EPP or shim.  Note
+                             that this is smaller than the 11-word maximum
+                             request size, since some constant values are
+                             not saved in the context. */
+  uint32_t *data;       /**< Data to be sent to the EPP or shim via IDN. */
+#endif
+}
+netio_send_pkt_context_t;
+#ifndef __DOXYGEN__
+#define SEND_PKT_CTX_USE_EPP   1  /**< We're sending to an EPP. */
+#define SEND_PKT_CTX_SEND_CSUM 2  /**< Request includes a checksum. */
+#endif
+/**
+ * @brief Packet vector entry.
+ *
+ * @ingroup egress
+ *
+ * This data structure is used with netio_send_packet_vector() to send multiple
+ * packets with one NetIO call.  The structure should be initialized by
+ * calling netio_pkt_vector_set(), rather than by setting the fields
+ * directly.
+ *
+ * This structure is guaranteed to be a power of two in size, no
+ * bigger than one L2 cache line, and to be aligned modulo its size.
+ */
+typedef struct
+#ifndef __DOXYGEN__
+__attribute__((aligned(8)))
+#endif
+{
+  /** Reserved for use by the user application.  When initialized with
+   *  the netio_set_pkt_vector_entry() function, this field is guaranteed
+   *  to be visible to readers only after all other fields are already
+   *  visible.  This way it can be used as a valid flag or generation
+   *  counter. */
+  uint8_t user_data;
+  /* Structure members below this point should not be accessed directly by
+   * applications, as they may change in the future. */
+  /** Low 8 bits of the packet address to send.  The high bits are
+   *  acquired from the 'handle' field. */
+  uint8_t buffer_address_low;
+  /** Number of bytes to transmit. */
+  uint16_t size;
+  /** The raw handle from a netio_pkt_t.  If this is NETIO_PKT_HANDLE_NONE,
+   *  this vector entry will be skipped and no packet will be transmitted. */
+  netio_pkt_handle_t handle;
+}
+netio_pkt_vector_entry_t;
+/**
+ * @brief Initialize fields in a packet vector entry.
+ *
+ * @ingroup egress
+ *
+ * @param[out] v Pointer to the vector entry to be initialized.
+ * @param[in] pkt Packet to be transmitted when the vector entry is passed to
+ *        netio_send_packet_vector().  Note that the packet's attributes
+ *        (e.g., its L2 offset and length) are captured at the time this
+ *        routine is called; subsequent changes in those attributes will not
+ *        be reflected in the packet which is actually transmitted.
+ *        Changes in the packet's contents, however, will be so reflected.
+ *        If this is NULL, no packet will be transmitted.
+ * @param[in] user_data User data to be set in the vector entry.
+ *        This function guarantees that the "user_data" field will become
+ *        visible to a reader only after all other fields have become visible.
+ *        This allows a structure in a ring buffer to be written and read
+ *        by a polling reader without any locks or other synchronization.
+ */
+static __inline void
+netio_pkt_vector_set(volatile netio_pkt_vector_entry_t* v, netio_pkt_t* pkt,
+                     uint8_t user_data)
+{
+  if (pkt)
+  {
+    if (NETIO_PKT_IS_MINIMAL(pkt))
+    {
+      netio_pkt_minimal_metadata_t* mmd =
+        (netio_pkt_minimal_metadata_t*) &pkt->__metadata;
+      v->buffer_address_low = (uintptr_t) NETIO_PKT_L2_DATA_MM(mmd, pkt) & 0xFF;
+      v->size = NETIO_PKT_L2_LENGTH_MM(mmd, pkt);
+    }
+    else
+    {
+      netio_pkt_metadata_t* mda = &pkt->__metadata;
+      v->buffer_address_low = (uintptr_t) NETIO_PKT_L2_DATA_M(mda, pkt) & 0xFF;
+      v->size = NETIO_PKT_L2_LENGTH_M(mda, pkt);
+    }
+    v->handle.word = pkt->__packet.word;
+  }
+  else
+  {
+    v->handle.word = 0;   /* Set handle to NETIO_PKT_HANDLE_NONE. */
+  }
+  __asm__("" : : : "memory");
+  v->user_data = user_data;
+}
+/**
+ * Flags and structures for @ref netio_get() and @ref netio_set().
+ * @ingroup config
+ */
+/** @{ */
+/** Parameter class; addr is a NETIO_PARAM_xxx value. */
+#define NETIO_PARAM       0
+/** Interface MAC address. This address is only valid with @ref netio_get().
+ *  The value is a 6-byte MAC address.  Depending upon the overall system
+ *  design, a MAC address may or may not be available for each interface. */
+#define NETIO_PARAM_MAC        0
+/** Determine whether to suspend output on the receipt of pause frames.
+ *  If the value is nonzero, the I/O shim will suspend output when a pause
+ *  frame is received.  If the value is zero, pause frames will be ignored. */
+#define NETIO_PARAM_PAUSE_IN   1
+/** Determine whether to send pause frames if the I/O shim packet FIFOs are
+ *  nearly full.  If the value is zero, pause frames are not sent.  If
+ *  the value is nonzero, it is the delay value which will be sent in any
+ *  pause frames which are output, in units of 512 bit times. */
+#define NETIO_PARAM_PAUSE_OUT  2
+/** Jumbo frame support.  The value is a 4-byte integer.  If the value is
+ *  nonzero, the MAC will accept frames of up to 10240 bytes.  If the value
+ *  is zero, the MAC will only accept frames of up to 1544 bytes. */
+#define NETIO_PARAM_JUMBO      3
+/** I/O shim's overflow statistics register.  The value is two 16-bit integers.
+ *  The first 16-bit value (or the low 16 bits, if the value is treated as a
+ *  32-bit number) is the count of packets which were completely dropped and
+ *  not delivered by the shim.  The second 16-bit value (or the high 16 bits,
+ *  if the value is treated as a 32-bit number) is the count of packets
+ *  which were truncated and thus only partially delivered by the shim.  This
+ *  register is automatically reset to zero after it has been read.
+ */
+#define NETIO_PARAM_OVERFLOW   4
+/** IPP statistics.  This address is only valid with @ref netio_get().  The
+ *  value is a netio_stat_t structure.  Unlike the I/O shim statistics, the
+ *  IPP statistics are not all reset to zero on read; see the description
+ *  of the netio_stat_t for details. */
+#define NETIO_PARAM_STAT 5
+/** Possible link state.  The value is a combination of "NETIO_LINK_xxx"
+ *  flags.  With @ref netio_get(), this will indicate which flags are
+ *  actually supported by the hardware.
+ *
+ *  For historical reasons, specifying this value to netio_set() will have
+ *  the same behavior as using ::NETIO_PARAM_LINK_CONFIG, but this usage is
+ *  discouraged.
+ */
+#define NETIO_PARAM_LINK_POSSIBLE_STATE 6
+/** Link configuration. The value is a combination of "NETIO_LINK_xxx" flags.
+ *  With @ref netio_set(), this will attempt to immediately bring up the
+ *  link using whichever of the requested flags are supported by the
+ *  hardware, or take down the link if the flags are zero; if this is
+ *  not possible, an error will be returned.  Many programs will want
+ *  to use ::NETIO_PARAM_LINK_DESIRED_STATE instead.
+ *
+ *  For historical reasons, specifying this value to netio_get() will
+ *  have the same behavior as using ::NETIO_PARAM_LINK_POSSIBLE_STATE,
+ *  but this usage is discouraged.
+ */
+#define NETIO_PARAM_LINK_CONFIG NETIO_PARAM_LINK_POSSIBLE_STATE
+/** Current link state. This address is only valid with @ref netio_get().
+ *  The value is zero or more of the "NETIO_LINK_xxx" flags, ORed together.
+ *  If the link is down, the value ANDed with NETIO_LINK_SPEED will be
+ *  zero; if the link is up, the value ANDed with NETIO_LINK_SPEED will
+ *  result in exactly one of the NETIO_LINK_xxx values, indicating the
+ *  current speed. */
+#define NETIO_PARAM_LINK_CURRENT_STATE 7
+/** Variant symbol for current state, retained for compatibility with
+ *  pre-MDE-2.1 programs. */
+#define NETIO_PARAM_LINK_STATUS NETIO_PARAM_LINK_CURRENT_STATE
+/** Packet Coherence protocol. This address is only valid with @ref netio_get().
+ *  The value is nonzero if the interface is configured for cache-coherent DMA.
+ */
+#define NETIO_PARAM_COHERENT 8
+/** Desired link state. The value is a conbination of "NETIO_LINK_xxx"
+ *  flags, which specify the desired state for the link.  With @ref
+ *  netio_set(), this will, in the background, attempt to bring up the link
+ *  using whichever of the requested flags are reasonable, or take down the
+ *  link if the flags are zero.  The actual link up or down operation may
+ *  happen after this call completes.  If the link state changes in the
+ *  future, the system will continue to try to get back to the desired link
+ *  state; for instance, if the link is brought up successfully, and then
+ *  the network cable is disconnected, the link will go down.  However, the
+ *  desired state of the link is still up, so if the cable is reconnected,
+ *  the link will be brought up again.
+ *
+ *  With @ref netio_get(), this will indicate the desired state for the
+ *  link, as set with a previous netio_set() call, or implicitly by a
+ *  netio_input_register() or netio_input_unregister() operation.  This may
+ *  not reflect the current state of the link; to get that, use
+ *  ::NETIO_PARAM_LINK_CURRENT_STATE. */
+#define NETIO_PARAM_LINK_DESIRED_STATE 9
+/** NetIO statistics structure.  Retrieved using the ::NETIO_PARAM_STAT
+ *  address passed to @ref netio_get(). */
+typedef struct
+{
+  /** Number of packets which have been received by the IPP and forwarded
+   *  to a tile's receive queue for processing.  This value wraps at its
+   *  maximum, and is not cleared upon read. */
+  uint32_t packets_received;
+  /** Number of packets which have been dropped by the IPP, because they could
+   *  not be received, or could not be forwarded to a tile.  The former happens
+   *  when the IPP does not have a free packet buffer of suitable size for an
+   *  incoming frame.  The latter happens when all potential destination tiles
+   *  for a packet, as defined by the group, bucket, and queue configuration,
+   *  have full receive queues.   This value wraps at its maximum, and is not
+   *  cleared upon read. */
+  uint32_t packets_dropped;
+  /*
+   * Note: the #defines after each of the following four one-byte values
+   * denote their location within the third word of the netio_stat_t.  They
+   * are intended for use only by the IPP implementation and are thus omitted
+   * from the Doxygen output.
+   */
+  /** Number of packets dropped because no worker was able to accept a new
+   *  packet.  This value saturates at its maximum, and is cleared upon
+   *  read. */
+  uint8_t drops_no_worker;
+#ifndef __DOXYGEN__
+#define NETIO_STAT_DROPS_NO_WORKER   0
+#endif
+  /** Number of packets dropped because no small buffers were available.
+   *  This value saturates at its maximum, and is cleared upon read. */
+  uint8_t drops_no_smallbuf;
+#ifndef __DOXYGEN__
+#define NETIO_STAT_DROPS_NO_SMALLBUF 1
+#endif
+  /** Number of packets dropped because no large buffers were available.
+   *  This value saturates at its maximum, and is cleared upon read. */
+  uint8_t drops_no_largebuf;
+#ifndef __DOXYGEN__
+#define NETIO_STAT_DROPS_NO_LARGEBUF 2
+#endif
+  /** Number of packets dropped because no jumbo buffers were available.
+   *  This value saturates at its maximum, and is cleared upon read. */
+  uint8_t drops_no_jumbobuf;
+#ifndef __DOXYGEN__
+#define NETIO_STAT_DROPS_NO_JUMBOBUF 3
+#endif
+}
+netio_stat_t;
+/** Link can run, should run, or is running at 10 Mbps. */
+#define NETIO_LINK_10M         0x01
+/** Link can run, should run, or is running at 100 Mbps. */
+#define NETIO_LINK_100M        0x02
+/** Link can run, should run, or is running at 1 Gbps. */
+#define NETIO_LINK_1G          0x04
+/** Link can run, should run, or is running at 10 Gbps. */
+#define NETIO_LINK_10G         0x08
+/** Link should run at the highest speed supported by the link and by
+ *  the device connected to the link.  Only usable as a value for
+ *  the link's desired state; never returned as a value for the current
+ *  or possible states. */
+#define NETIO_LINK_ANYSPEED    0x10
+/** All legal link speeds. */
+#define NETIO_LINK_SPEED  (NETIO_LINK_10M  | \
+                           NETIO_LINK_100M | \
+                           NETIO_LINK_1G   | \
+                           NETIO_LINK_10G  | \
+                           NETIO_LINK_ANYSPEED)
+/** MAC register class.  Addr is a register offset within the MAC.
+ *  Registers within the XGbE and GbE MACs are documented in the Tile
+ *  Processor I/O Device Guide (UG104). MAC registers start at address
+ *  0x4000, and do not include the MAC_INTERFACE registers. */
+#define NETIO_MAC             1
+/** MDIO register class (IEEE 802.3 clause 22 format).  Addr is the "addr"
+ *  member of a netio_mdio_addr_t structure. */
+#define NETIO_MDIO            2
+/** MDIO register class (IEEE 802.3 clause 45 format).  Addr is the "addr"
+ *  member of a netio_mdio_addr_t structure. */
+#define NETIO_MDIO_CLAUSE45   3
+/** NetIO MDIO address type.  Retrieved or provided using the ::NETIO_MDIO
+ *  address passed to @ref netio_get() or @ref netio_set(). */
+typedef union
+{
+  struct
+  {
+    unsigned int reg:16;  /**< MDIO register offset.  For clause 22 access,
+                               must be less than 32. */
+    unsigned int phy:5;   /**< Which MDIO PHY to access. */
+    unsigned int dev:5;   /**< Which MDIO device to access within that PHY.
+                               Applicable for clause 45 access only; ignored
+                               for clause 22 access. */
+  }
+  bits;                   /**< Container for bitfields. */
+  uint64_t addr;          /**< Value to pass to @ref netio_get() or
+                           *   @ref netio_set(). */
+}
+netio_mdio_addr_t;
+/** @} */
+#endif /* __NETIO_INTF_H__ */
diff --git a/arch/tile/kernel/Makefile b/arch/tile/kernel/Makefile
index 112b1e248f05..b4c8e8ec45dc 100644
--- a/arch/tile/kernel/Makefile
+++ b/arch/tile/kernel/Makefile
@@ -15,3 +15,4 @@ obj-$(CONFIG_SMP)		+= smpboot.o smp.o tlb.o
 obj-$(CONFIG_MODULES)           += module.o
 obj-$(CONFIG_EARLY_PRINTK)      += early_printk.o
 obj-$(CONFIG_KEXEC)             += machine_kexec.o relocate_kernel.o
+obj-$(CONFIG_PCI)               += pci.o
diff --git a/arch/tile/kernel/compat.c b/arch/tile/kernel/compat.c
index 67617a05e602..dbc213adf5e1 100644
--- a/arch/tile/kernel/compat.c
+++ b/arch/tile/kernel/compat.c
@@ -21,7 +21,6 @@
 #include <linux/kdev_t.h>
 #include <linux/fs.h>
 #include <linux/fcntl.h>
-#include <linux/smp_lock.h>
 #include <linux/uaccess.h>
 #include <linux/signal.h>
 #include <asm/syscalls.h>
diff --git a/arch/tile/kernel/compat_signal.c b/arch/tile/kernel/compat_signal.c
index fb64b99959d4..543d6a33aa26 100644
--- a/arch/tile/kernel/compat_signal.c
+++ b/arch/tile/kernel/compat_signal.c
@@ -15,7 +15,6 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/kernel.h>
 #include <linux/signal.h>
 #include <linux/errno.h>
diff --git a/arch/tile/kernel/pci.c b/arch/tile/kernel/pci.c
new file mode 100644
index 000000000000..a1ee25be9ad9
--- /dev/null
+++ b/arch/tile/kernel/pci.c
@@ -0,0 +1,621 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/capability.h>
+#include <linux/sched.h>
+#include <linux/errno.h>
+#include <linux/bootmem.h>
+#include <linux/irq.h>
+#include <linux/io.h>
+#include <linux/uaccess.h>
+#include <asm/processor.h>
+#include <asm/sections.h>
+#include <asm/byteorder.h>
+#include <asm/hv_driver.h>
+#include <hv/drv_pcie_rc_intf.h>
+/*
+ * Initialization flow and process
+ * -------------------------------
+ *
+ * This files containes the routines to search for PCI buses,
+ * enumerate the buses, and configure any attached devices.
+ *
+ * There are two entry points here:
+ * 1) tile_pci_init
+ *    This sets up the pci_controller structs, and opens the
+ *    FDs to the hypervisor.  This is called from setup_arch() early
+ *    in the boot process.
+ * 2) pcibios_init
+ *    This probes the PCI bus(es) for any attached hardware.  It's
+ *    called by subsys_initcall.  All of the real work is done by the
+ *    generic Linux PCI layer.
+ *
+ */
+/*
+ * This flag tells if the platform is TILEmpower that needs
+ * special configuration for the PLX switch chip.
+ */
+int __write_once tile_plx_gen1;
+static struct pci_controller controllers[TILE_NUM_PCIE];
+static int num_controllers;
+static struct pci_ops tile_cfg_ops;
+/*
+ * We don't need to worry about the alignment of resources.
+ */
+resource_size_t pcibios_align_resource(void *data, const struct resource *res,
+                            resource_size_t size, resource_size_t align)
+{
+        return res->start;
+}
+EXPORT_SYMBOL(pcibios_align_resource);
+/*
+ * Open a FD to the hypervisor PCI device.
+ *
+ * controller_id is the controller number, config type is 0 or 1 for
+ * config0 or config1 operations.
+ */
+static int __init tile_pcie_open(int controller_id, int config_type)
+{
+        char filename[32];
+        int fd;
+        sprintf(filename, "pcie/%d/config%d", controller_id, config_type);
+        fd = hv_dev_open((HV_VirtAddr)filename, 0);
+        return fd;
+}
+/*
+ * Get the IRQ numbers from the HV and set up the handlers for them.
+ */
+static int __init tile_init_irqs(int controller_id,
+                                 struct pci_controller *controller)
+{
+        char filename[32];
+        int fd;
+        int ret;
+        int x;
+        struct pcie_rc_config rc_config;
+        sprintf(filename, "pcie/%d/ctl", controller_id);
+        fd = hv_dev_open((HV_VirtAddr)filename, 0);
+        if (fd < 0) {
+                pr_err("PCI: hv_dev_open(%s) failed\n", filename);
+                return -1;
+        }
+        ret = hv_dev_pread(fd, 0, (HV_VirtAddr)(&rc_config),
+                           sizeof(rc_config), PCIE_RC_CONFIG_MASK_OFF);
+        hv_dev_close(fd);
+        if (ret != sizeof(rc_config)) {
+                pr_err("PCI: wanted %zd bytes, got %d\n",
+                       sizeof(rc_config), ret);
+                return -1;
+        }
+        /* Record irq_base so that we can map INTx to IRQ # later. */
+        controller->irq_base = rc_config.intr;
+        for (x = 0; x < 4; x++)
+                tile_irq_activate(rc_config.intr + x,
+                                  TILE_IRQ_HW_CLEAR);
+        if (rc_config.plx_gen1)
+                controller->plx_gen1 = 1;
+        return 0;
+}
+/*
+ * First initialization entry point, called from setup_arch().
+ *
+ * Find valid controllers and fill in pci_controller structs for each
+ * of them.
+ *
+ * Returns the number of controllers discovered.
+ */
+int __init tile_pci_init(void)
+{
+        int i;
+        pr_info("PCI: Searching for controllers...\n");
+        /* Do any configuration we need before using the PCIe */
+        for (i = 0; i < TILE_NUM_PCIE; i++) {
+                int hv_cfg_fd0 = -1;
+                int hv_cfg_fd1 = -1;
+                int hv_mem_fd = -1;
+                char name[32];
+                struct pci_controller *controller;
+                /*
+                 * Open the fd to the HV.  If it fails then this
+                 * device doesn't exist.
+                 */
+                hv_cfg_fd0 = tile_pcie_open(i, 0);
+                if (hv_cfg_fd0 < 0)
+                        continue;
+                hv_cfg_fd1 = tile_pcie_open(i, 1);
+                if (hv_cfg_fd1 < 0) {
+                        pr_err("PCI: Couldn't open config fd to HV "
+                            "for controller %d\n", i);
+                        goto err_cont;
+                }
+                sprintf(name, "pcie/%d/mem", i);
+                hv_mem_fd = hv_dev_open((HV_VirtAddr)name, 0);
+                if (hv_mem_fd < 0) {
+                        pr_err("PCI: Could not open mem fd to HV!\n");
+                        goto err_cont;
+                }
+                pr_info("PCI: Found PCI controller #%d\n", i);
+                controller = &controllers[num_controllers];
+                if (tile_init_irqs(i, controller)) {
+                        pr_err("PCI: Could not initialize "
+                               "IRQs, aborting.\n");
+                        goto err_cont;
+                }
+                controller->index = num_controllers;
+                controller->hv_cfg_fd[0] = hv_cfg_fd0;
+                controller->hv_cfg_fd[1] = hv_cfg_fd1;
+                controller->hv_mem_fd = hv_mem_fd;
+                controller->first_busno = 0;
+                controller->last_busno = 0xff;
+                controller->ops = &tile_cfg_ops;
+                num_controllers++;
+                continue;
+err_cont:
+                if (hv_cfg_fd0 >= 0)
+                        hv_dev_close(hv_cfg_fd0);
+                if (hv_cfg_fd1 >= 0)
+                        hv_dev_close(hv_cfg_fd1);
+                if (hv_mem_fd >= 0)
+                        hv_dev_close(hv_mem_fd);
+                continue;
+        }
+        /*
+         * Before using the PCIe, see if we need to do any platform-specific
+         * configuration, such as the PLX switch Gen 1 issue on TILEmpower.
+         */
+        for (i = 0; i < num_controllers; i++) {
+                struct pci_controller *controller = &controllers[i];
+                if (controller->plx_gen1)
+                        tile_plx_gen1 = 1;
+        }
+        return num_controllers;
+}
+/*
+ * (pin - 1) converts from the PCI standard's [1:4] convention to
+ * a normal [0:3] range.
+ */
+static int tile_map_irq(struct pci_dev *dev, u8 slot, u8 pin)
+{
+        struct pci_controller *controller =
+                (struct pci_controller *)dev->sysdata;
+        return (pin - 1) + controller->irq_base;
+}
+static void __init fixup_read_and_payload_sizes(void)
+{
+        struct pci_dev *dev = NULL;
+        int smallest_max_payload = 0x1; /* Tile maxes out at 256 bytes. */
+        int max_read_size = 0x2; /* Limit to 512 byte reads. */
+        u16 new_values;
+        /* Scan for the smallest maximum payload size. */
+        while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+                int pcie_caps_offset;
+                u32 devcap;
+                int max_payload;
+                pcie_caps_offset = pci_find_capability(dev, PCI_CAP_ID_EXP);
+                if (pcie_caps_offset == 0)
+                        continue;
+                pci_read_config_dword(dev, pcie_caps_offset + PCI_EXP_DEVCAP,
+                                      &devcap);
+                max_payload = devcap & PCI_EXP_DEVCAP_PAYLOAD;
+                if (max_payload < smallest_max_payload)
+                        smallest_max_payload = max_payload;
+        }
+        /* Now, set the max_payload_size for all devices to that value. */
+        new_values = (max_read_size << 12) | (smallest_max_payload << 5);
+        while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+                int pcie_caps_offset;
+                u16 devctl;
+                pcie_caps_offset = pci_find_capability(dev, PCI_CAP_ID_EXP);
+                if (pcie_caps_offset == 0)
+                        continue;
+                pci_read_config_word(dev, pcie_caps_offset + PCI_EXP_DEVCTL,
+                                     &devctl);
+                devctl &= ~(PCI_EXP_DEVCTL_PAYLOAD | PCI_EXP_DEVCTL_READRQ);
+                devctl |= new_values;
+                pci_write_config_word(dev, pcie_caps_offset + PCI_EXP_DEVCTL,
+                                      devctl);
+        }
+}
+/*
+ * Second PCI initialization entry point, called by subsys_initcall.
+ *
+ * The controllers have been set up by the time we get here, by a call to
+ * tile_pci_init.
+ */
+static int __init pcibios_init(void)
+{
+        int i;
+        pr_info("PCI: Probing PCI hardware\n");
+        /*
+         * Delay a bit in case devices aren't ready.  Some devices are
+         * known to require at least 20ms here, but we use a more
+         * conservative value.
+         */
+        mdelay(250);
+        /* Scan all of the recorded PCI controllers.  */
+        for (i = 0; i < num_controllers; i++) {
+                struct pci_controller *controller = &controllers[i];
+                struct pci_bus *bus;
+                pr_info("PCI: initializing controller #%d\n", i);
+                /*
+                 * This comes from the generic Linux PCI driver.
+                 *
+                 * It reads the PCI tree for this bus into the Linux
+                 * data structures.
+                 *
+                 * This is inlined in linux/pci.h and calls into
+                 * pci_scan_bus_parented() in probe.c.
+                 */
+                bus = pci_scan_bus(0, controller->ops, controller);
+                controller->root_bus = bus;
+                controller->last_busno = bus->subordinate;
+        }
+        /* Do machine dependent PCI interrupt routing */
+        pci_fixup_irqs(pci_common_swizzle, tile_map_irq);
+        /*
+         * This comes from the generic Linux PCI driver.
+         *
+         * It allocates all of the resources (I/O memory, etc)
+         * associated with the devices read in above.
+         */
+        pci_assign_unassigned_resources();
+        /* Configure the max_read_size and max_payload_size values. */
+        fixup_read_and_payload_sizes();
+        /* Record the I/O resources in the PCI controller structure. */
+        for (i = 0; i < num_controllers; i++) {
+                struct pci_bus *root_bus = controllers[i].root_bus;
+                struct pci_bus *next_bus;
+                struct pci_dev *dev;
+                list_for_each_entry(dev, &root_bus->devices, bus_list) {
+                        /* Find the PCI host controller, ie. the 1st bridge. */
+                        if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI &&
+                                (PCI_SLOT(dev->devfn) == 0)) {
+                                next_bus = dev->subordinate;
+                                controllers[i].mem_resources[0] =
+                                        *next_bus->resource[0];
+                                controllers[i].mem_resources[1] =
+                                         *next_bus->resource[1];
+                                controllers[i].mem_resources[2] =
+                                         *next_bus->resource[2];
+                                break;
+                        }
+                }
+        }
+        return 0;
+}
+subsys_initcall(pcibios_init);
+/*
+ * No bus fixups needed.
+ */
+void __devinit pcibios_fixup_bus(struct pci_bus *bus)
+{
+        /* Nothing needs to be done. */
+}
+/*
+ * This can be called from the generic PCI layer, but doesn't need to
+ * do anything.
+ */
+char __devinit *pcibios_setup(char *str)
+{
+        /* Nothing needs to be done. */
+        return str;
+}
+/*
+ * This is called from the generic Linux layer.
+ */
+void __init pcibios_update_irq(struct pci_dev *dev, int irq)
+{
+        pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
+}
+/*
+ * Enable memory and/or address decoding, as appropriate, for the
+ * device described by the 'dev' struct.
+ *
+ * This is called from the generic PCI layer, and can be called
+ * for bridges or endpoints.
+ */
+int pcibios_enable_device(struct pci_dev *dev, int mask)
+{
+        u16 cmd, old_cmd;
+        u8 header_type;
+        int i;
+        struct resource *r;
+        pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
+        pci_read_config_word(dev, PCI_COMMAND, &cmd);
+        old_cmd = cmd;
+        if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE) {
+                /*
+                 * For bridges, we enable both memory and I/O decoding
+                 * in call cases.
+                 */
+                cmd |= PCI_COMMAND_IO;
+                cmd |= PCI_COMMAND_MEMORY;
+        } else {
+                /*
+                 * For endpoints, we enable memory and/or I/O decoding
+                 * only if they have a memory resource of that type.
+                 */
+                for (i = 0; i < 6; i++) {
+                        r = &dev->resource[i];
+                        if (r->flags & IORESOURCE_UNSET) {
+                                pr_err("PCI: Device %s not available "
+                                       "because of resource collisions\n",
+                                       pci_name(dev));
+                                return -EINVAL;
+                        }
+                        if (r->flags & IORESOURCE_IO)
+                                cmd |= PCI_COMMAND_IO;
+                        if (r->flags & IORESOURCE_MEM)
+                                cmd |= PCI_COMMAND_MEMORY;
+                }
+        }
+        /*
+         * We only write the command if it changed.
+         */
+        if (cmd != old_cmd)
+                pci_write_config_word(dev, PCI_COMMAND, cmd);
+        return 0;
+}
+void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max)
+{
+        unsigned long start = pci_resource_start(dev, bar);
+        unsigned long len = pci_resource_len(dev, bar);
+        unsigned long flags = pci_resource_flags(dev, bar);
+        if (!len)
+                return NULL;
+        if (max && len > max)
+                len = max;
+        if (!(flags & IORESOURCE_MEM)) {
+                pr_info("PCI: Trying to map invalid resource %#lx\n", flags);
+                start = 0;
+        }
+        return (void __iomem *)start;
+}
+EXPORT_SYMBOL(pci_iomap);
+/****************************************************************
+ *
+ * Tile PCI config space read/write routines
+ *
+ ****************************************************************/
+/*
+ * These are the normal read and write ops
+ * These are expanded with macros from  pci_bus_read_config_byte() etc.
+ *
+ * devfn is the combined PCI slot & function.
+ *
+ * offset is in bytes, from the start of config space for the
+ * specified bus & slot.
+ */
+static int __devinit tile_cfg_read(struct pci_bus *bus,
+                                   unsigned int devfn,
+                                   int offset,
+                                   int size,
+                                   u32 *val)
+{
+        struct pci_controller *controller = bus->sysdata;
+        int busnum = bus->number & 0xff;
+        int slot = (devfn >> 3) & 0x1f;
+        int function = devfn & 0x7;
+        u32 addr;
+        int config_mode = 1;
+        /*
+         * There is no bridge between the Tile and bus 0, so we
+         * use config0 to talk to bus 0.
+         *
+         * If we're talking to a bus other than zero then we
+         * must have found a bridge.
+         */
+        if (busnum == 0) {
+                /*
+                 * We fake an empty slot for (busnum == 0) && (slot > 0),
+                 * since there is only one slot on bus 0.
+                 */
+                if (slot) {
+                        *val = 0xFFFFFFFF;
+                        return 0;
+                }
+                config_mode = 0;
+        }
+        addr = busnum << 20;            /* Bus in 27:20 */
+        addr |= slot << 15;             /* Slot (device) in 19:15 */
+        addr |= function << 12;         /* Function is in 14:12 */
+        addr |= (offset & 0xFFF);       /* byte address in 0:11 */
+        return hv_dev_pread(controller->hv_cfg_fd[config_mode], 0,
+                            (HV_VirtAddr)(val), size, addr);
+}
+/*
+ * See tile_cfg_read() for relevent comments.
+ * Note that "val" is the value to write, not a pointer to that value.
+ */
+static int __devinit tile_cfg_write(struct pci_bus *bus,
+                                    unsigned int devfn,
+                                    int offset,
+                                    int size,
+                                    u32 val)
+{
+        struct pci_controller *controller = bus->sysdata;
+        int busnum = bus->number & 0xff;
+        int slot = (devfn >> 3) & 0x1f;
+        int function = devfn & 0x7;
+        u32 addr;
+        int config_mode = 1;
+        HV_VirtAddr valp = (HV_VirtAddr)&val;
+        /*
+         * For bus 0 slot 0 we use config 0 accesses.
+         */
+        if (busnum == 0) {
+                /*
+                 * We fake an empty slot for (busnum == 0) && (slot > 0),
+                 * since there is only one slot on bus 0.
+                 */
+                if (slot)
+                        return 0;
+                config_mode = 0;
+        }
+        addr = busnum << 20;            /* Bus in 27:20 */
+        addr |= slot << 15;             /* Slot (device) in 19:15 */
+        addr |= function << 12;         /* Function is in 14:12 */
+        addr |= (offset & 0xFFF);       /* byte address in 0:11 */
+#ifdef __BIG_ENDIAN
+        /* Point to the correct part of the 32-bit "val". */
+        valp += 4 - size;
+#endif
+        return hv_dev_pwrite(controller->hv_cfg_fd[config_mode], 0,
+                             valp, size, addr);
+}
+static struct pci_ops tile_cfg_ops = {
+        .read =         tile_cfg_read,
+        .write =        tile_cfg_write,
+};
+/*
+ * In the following, each PCI controller's mem_resources[1]
+ * represents its (non-prefetchable) PCI memory resource.
+ * mem_resources[0] and mem_resources[2] refer to its PCI I/O and
+ * prefetchable PCI memory resources, respectively.
+ * For more details, see pci_setup_bridge() in setup-bus.c.
+ * By comparing the target PCI memory address against the
+ * end address of controller 0, we can determine the controller
+ * that should accept the PCI memory access.
+ */
+#define TILE_READ(size, type)                                           \
+type _tile_read##size(unsigned long addr)                               \
+{                                                                       \
+        type val;                                                       \
+        int idx = 0;                                                    \
+        if (addr > controllers[0].mem_resources[1].end &&               \
+            addr > controllers[0].mem_resources[2].end)                 \
+                idx = 1;                                                \
+        if (hv_dev_pread(controllers[idx].hv_mem_fd, 0,                 \
+                         (HV_VirtAddr)(&val), sizeof(type), addr))      \
+                pr_err("PCI: read %zd bytes at 0x%lX failed\n",         \
+                       sizeof(type), addr);                             \
+        return val;                                                     \
+}                                                                       \
+EXPORT_SYMBOL(_tile_read##size)
+TILE_READ(b, u8);
+TILE_READ(w, u16);
+TILE_READ(l, u32);
+TILE_READ(q, u64);
+#define TILE_WRITE(size, type)                                          \
+void _tile_write##size(type val, unsigned long addr)                    \
+{                                                                       \
+        int idx = 0;                                                    \
+        if (addr > controllers[0].mem_resources[1].end &&               \
+            addr > controllers[0].mem_resources[2].end)                 \
+                idx = 1;                                                \
+        if (hv_dev_pwrite(controllers[idx].hv_mem_fd, 0,                \
+                          (HV_VirtAddr)(&val), sizeof(type), addr))     \
+                pr_err("PCI: write %zd bytes at 0x%lX failed\n",        \
+                       sizeof(type), addr);                             \
+}                                                                       \
+EXPORT_SYMBOL(_tile_write##size)
+TILE_WRITE(b, u8);
+TILE_WRITE(w, u16);
+TILE_WRITE(l, u32);
+TILE_WRITE(q, u64);
diff --git a/arch/tile/kernel/setup.c b/arch/tile/kernel/setup.c
index fb0b3cbeae14..f18573643ed1 100644
--- a/arch/tile/kernel/setup.c
+++ b/arch/tile/kernel/setup.c
@@ -840,7 +840,7 @@ static int __init topology_init(void)
        for_each_online_node(i)
                register_one_node(i);
-        for_each_present_cpu(i)
+        for (i = 0; i < smp_height * smp_width; ++i)
                register_cpu(&cpu_devices[i], i);
        return 0;
diff --git a/arch/tile/kernel/signal.c b/arch/tile/kernel/signal.c
index 687719d4abd1..757407e36696 100644
--- a/arch/tile/kernel/signal.c
+++ b/arch/tile/kernel/signal.c
@@ -16,7 +16,6 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/kernel.h>
 #include <linux/signal.h>
 #include <linux/errno.h>
diff --git a/arch/tile/kernel/smpboot.c b/arch/tile/kernel/smpboot.c
index 74d62d098edf..b949edcec200 100644
--- a/arch/tile/kernel/smpboot.c
+++ b/arch/tile/kernel/smpboot.c
@@ -18,7 +18,6 @@
 #include <linux/mm.h>
 #include <linux/sched.h>
 #include <linux/kernel_stat.h>
-#include <linux/smp_lock.h>
 #include <linux/bootmem.h>
 #include <linux/notifier.h>
 #include <linux/cpu.h>
diff --git a/arch/tile/kernel/sys.c b/arch/tile/kernel/sys.c
index 7e764669a022..e2187d24a9b4 100644
--- a/arch/tile/kernel/sys.c
+++ b/arch/tile/kernel/sys.c
@@ -20,7 +20,6 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/syscalls.h>
 #include <linux/mman.h>
 #include <linux/file.h>
diff --git a/arch/tile/lib/memchr_32.c b/arch/tile/lib/memchr_32.c
index 6235283b4859..cc3d9badf030 100644
--- a/arch/tile/lib/memchr_32.c
+++ b/arch/tile/lib/memchr_32.c
@@ -18,12 +18,24 @@
 void *memchr(const void *s, int c, size_t n)
 {
+        const uint32_t *last_word_ptr;
+        const uint32_t *p;
+        const char *last_byte_ptr;
+        uintptr_t s_int;
+        uint32_t goal, before_mask, v, bits;
+        char *ret;
+        if (__builtin_expect(n == 0, 0)) {
+                /* Don't dereference any memory if the array is empty. */
+                return NULL;
+        }
        /* Get an aligned pointer. */
-        const uintptr_t s_int = (uintptr_t) s;
+        s_int = (uintptr_t) s;
-        const uint32_t *p = (const uint32_t *)(s_int & -4);
+        p = (const uint32_t *)(s_int & -4);
        /* Create four copies of the byte for which we are looking. */
-        const uint32_t goal = 0x01010101 * (uint8_t) c;
+        goal = 0x01010101 * (uint8_t) c;
        /* Read the first word, but munge it so that bytes before the array
         * will not match goal.
@@ -31,23 +43,14 @@ void *memchr(const void *s, int c, size_t n)
         * Note that this shift count expression works because we know
         * shift counts are taken mod 32.
         */
-        const uint32_t before_mask = (1 << (s_int << 3)) - 1;
+        before_mask = (1 << (s_int << 3)) - 1;
-        uint32_t v = (*p | before_mask) ^ (goal & before_mask);
+        v = (*p | before_mask) ^ (goal & before_mask);
        /* Compute the address of the last byte. */
-        const char *const last_byte_ptr = (const char *)s + n - 1;
+        last_byte_ptr = (const char *)s + n - 1;
        /* Compute the address of the word containing the last byte. */
-        const uint32_t *const last_word_ptr =
+        last_word_ptr = (const uint32_t *)((uintptr_t) last_byte_ptr & -4);
-            (const uint32_t *)((uintptr_t) last_byte_ptr & -4);
-        uint32_t bits;
-        char *ret;
-        if (__builtin_expect(n == 0, 0)) {
-                /* Don't dereference any memory if the array is empty. */
-                return NULL;
-        }
        while ((bits = __insn_seqb(v, goal)) == 0) {
                if (__builtin_expect(p == last_word_ptr, 0)) {
diff --git a/arch/tile/lib/spinlock_32.c b/arch/tile/lib/spinlock_32.c
index 485e24d62c6b..5cd1c4004eca 100644
--- a/arch/tile/lib/spinlock_32.c
+++ b/arch/tile/lib/spinlock_32.c
@@ -167,23 +167,30 @@ void arch_write_lock_slow(arch_rwlock_t *rwlock, u32 val)
         * when we compare them.
         */
        u32 my_ticket_;
+        u32 iterations = 0;
-        /* Take out the next ticket; this will also stop would-be readers. */
+        /*
-        if (val & 1)
+         * Wait until there are no readers, then bump up the next
-                val = get_rwlock(rwlock);
+         * field and capture the ticket value.
-        rwlock->lock = __insn_addb(val, 1 << WR_NEXT_SHIFT);
+         */
+        for (;;) {
+                if (!(val & 1)) {
+                        if ((val >> RD_COUNT_SHIFT) == 0)
+                                break;
+                        rwlock->lock = val;
+                }
+                delay_backoff(iterations++);
+                val = __insn_tns((int *)&rwlock->lock);
+        }
-        /* Extract my ticket value from the original word. */
+        /* Take out the next ticket and extract my ticket value. */
+        rwlock->lock = __insn_addb(val, 1 << WR_NEXT_SHIFT);
        my_ticket_ = val >> WR_NEXT_SHIFT;
-        /*
+        /* Wait until the "current" field matches our ticket. */
-         * Wait until the "current" field matches our ticket, and
-         * there are no remaining readers.
-         */
        for (;;) {
                u32 curr_ = val >> WR_CURR_SHIFT;
-                u32 readers = val >> RD_COUNT_SHIFT;
+                u32 delta = ((my_ticket_ - curr_) & WR_MASK);
-                u32 delta = ((my_ticket_ - curr_) & WR_MASK) + !!readers;
                if (likely(delta == 0))
                        break;
diff --git a/arch/tile/mm/fault.c b/arch/tile/mm/fault.c
index f295b4ac941d..dcebfc831cd6 100644
--- a/arch/tile/mm/fault.c
+++ b/arch/tile/mm/fault.c
@@ -24,7 +24,6 @@
 #include <linux/mman.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/interrupt.h>
 #include <linux/init.h>
 #include <linux/tty.h>
diff --git a/arch/tile/mm/hugetlbpage.c b/arch/tile/mm/hugetlbpage.c
index 24688b697a8d..201a582c4137 100644
--- a/arch/tile/mm/hugetlbpage.c
+++ b/arch/tile/mm/hugetlbpage.c
@@ -21,7 +21,6 @@
 #include <linux/mm.h>
 #include <linux/hugetlb.h>
 #include <linux/pagemap.h>
-#include <linux/smp_lock.h>
 #include <linux/slab.h>
 #include <linux/err.h>
 #include <linux/sysctl.h>
diff --git a/arch/um/drivers/line.c b/arch/um/drivers/line.c
index 7f7338c90784..1664cce7b0ac 100644
--- a/arch/um/drivers/line.c
+++ b/arch/um/drivers/line.c
@@ -727,6 +727,9 @@ struct winch {
 static void free_winch(struct winch *winch, int free_irq_ok)
 {
+        if (free_irq_ok)
+                free_irq(WINCH_IRQ, winch);
        list_del(&winch->list);
        if (winch->pid != -1)
@@ -735,8 +738,6 @@ static void free_winch(struct winch *winch, int free_irq_ok)
                os_close_file(winch->fd);
        if (winch->stack != 0)
                free_stack(winch->stack, 0);
-        if (free_irq_ok)
-                free_irq(WINCH_IRQ, winch);
        kfree(winch);
 }
diff --git a/arch/um/include/asm/ptrace-generic.h b/arch/um/include/asm/ptrace-generic.h
index 2cd899f75a3c..b7c5bab9bd77 100644
--- a/arch/um/include/asm/ptrace-generic.h
+++ b/arch/um/include/asm/ptrace-generic.h
@@ -38,8 +38,8 @@ struct pt_regs {
 struct task_struct;
-extern long subarch_ptrace(struct task_struct *child, long request, long addr,
+extern long subarch_ptrace(struct task_struct *child, long request,
-                           long data);
+        unsigned long addr, unsigned long data);
 extern unsigned long getreg(struct task_struct *child, int regno);
 extern int putreg(struct task_struct *child, int regno, unsigned long value);
 extern int get_fpregs(struct user_i387_struct __user *buf,
diff --git a/arch/um/kernel/exec.c b/arch/um/kernel/exec.c
index 340268be00b5..09bd7b585726 100644
--- a/arch/um/kernel/exec.c
+++ b/arch/um/kernel/exec.c
@@ -5,7 +5,6 @@
 #include "linux/stddef.h"
 #include "linux/fs.h"
-#include "linux/smp_lock.h"
 #include "linux/ptrace.h"
 #include "linux/sched.h"
 #include "linux/slab.h"
diff --git a/arch/um/kernel/ptrace.c b/arch/um/kernel/ptrace.c
index a5e33f29bbeb..701b672c1122 100644
--- a/arch/um/kernel/ptrace.c
+++ b/arch/um/kernel/ptrace.c
@@ -122,7 +122,7 @@ long arch_ptrace(struct task_struct *child, long request,
                break;
        case PTRACE_SET_THREAD_AREA:
-                ret = ptrace_set_thread_area(child, addr, datavp);
+                ret = ptrace_set_thread_area(child, addr, vp);
                break;
        case PTRACE_FAULTINFO: {
diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
index 849813f398e7..5852519b2d0f 100644
--- a/arch/x86/ia32/sys_ia32.c
+++ b/arch/x86/ia32/sys_ia32.c
@@ -28,7 +28,6 @@
 #include <linux/syscalls.h>
 #include <linux/times.h>
 #include <linux/utsname.h>
-#include <linux/smp_lock.h>
 #include <linux/mm.h>
 #include <linux/uio.h>
 #include <linux/poll.h>
diff --git a/arch/x86/include/asm/apic.h b/arch/x86/include/asm/apic.h
index 286de34b0ed6..f6ce0bda3b98 100644
--- a/arch/x86/include/asm/apic.h
+++ b/arch/x86/include/asm/apic.h
@@ -141,13 +141,13 @@ static inline void native_apic_msr_write(u32 reg, u32 v)
 static inline u32 native_apic_msr_read(u32 reg)
 {
-        u32 low, high;
+        u64 msr;
        if (reg == APIC_DFR)
                return -1;
-        rdmsr(APIC_BASE_MSR + (reg >> 4), low, high);
+        rdmsrl(APIC_BASE_MSR + (reg >> 4), msr);
-        return low;
+        return (u32)msr;
 }
 static inline void native_x2apic_wait_icr_idle(void)
@@ -181,12 +181,12 @@ extern void enable_x2apic(void);
 extern void x2apic_icr_write(u32 low, u32 id);
 static inline int x2apic_enabled(void)
 {
-        int msr, msr2;
+        u64 msr;
        if (!cpu_has_x2apic)
                return 0;
-        rdmsr(MSR_IA32_APICBASE, msr, msr2);
+        rdmsrl(MSR_IA32_APICBASE, msr);
        if (msr & X2APIC_ENABLE)
                return 1;
        return 0;
diff --git a/arch/x86/include/asm/fixmap.h b/arch/x86/include/asm/fixmap.h
index 4d293dced62f..9479a037419f 100644
--- a/arch/x86/include/asm/fixmap.h
+++ b/arch/x86/include/asm/fixmap.h
@@ -216,8 +216,8 @@ static inline unsigned long virt_to_fix(const unsigned long vaddr)
 }
 /* Return an pointer with offset calculated */
-static inline unsigned long __set_fixmap_offset(enum fixed_addresses idx,
+static __always_inline unsigned long
-                                phys_addr_t phys, pgprot_t flags)
+__set_fixmap_offset(enum fixed_addresses idx, phys_addr_t phys, pgprot_t flags)
 {
        __set_fixmap(idx, phys, flags);
        return fix_to_virt(idx) + (phys & (PAGE_SIZE - 1));
diff --git a/arch/x86/include/asm/uv/uv_mmrs.h b/arch/x86/include/asm/uv/uv_mmrs.h
index b2f2d2e05cec..6d90adf4428a 100644
--- a/arch/x86/include/asm/uv/uv_mmrs.h
+++ b/arch/x86/include/asm/uv/uv_mmrs.h
@@ -806,6 +806,78 @@ union uvh_node_present_table_u {
 };
 /* ========================================================================= */
+/*                 UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR                  */
+/* ========================================================================= */
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR 0x16000c8UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_BASE_SHFT 24
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_BASE_MASK 0x00000000ff000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_M_ALIAS_SHFT 48
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_M_ALIAS_MASK 0x001f000000000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_ENABLE_SHFT 63
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR_ENABLE_MASK 0x8000000000000000UL
+union uvh_rh_gam_alias210_overlay_config_0_mmr_u {
+    unsigned long       v;
+    struct uvh_rh_gam_alias210_overlay_config_0_mmr_s {
+        unsigned long   rsvd_0_23: 24;  /*    */
+        unsigned long   base    :  8;  /* RW */
+        unsigned long   rsvd_32_47: 16;  /*    */
+        unsigned long   m_alias :  5;  /* RW */
+        unsigned long   rsvd_53_62: 10;  /*    */
+        unsigned long   enable  :  1;  /* RW */
+    } s;
+};
+/* ========================================================================= */
+/*                 UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR                  */
+/* ========================================================================= */
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR 0x16000d8UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_BASE_SHFT 24
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_BASE_MASK 0x00000000ff000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_M_ALIAS_SHFT 48
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_M_ALIAS_MASK 0x001f000000000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_ENABLE_SHFT 63
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR_ENABLE_MASK 0x8000000000000000UL
+union uvh_rh_gam_alias210_overlay_config_1_mmr_u {
+    unsigned long       v;
+    struct uvh_rh_gam_alias210_overlay_config_1_mmr_s {
+        unsigned long   rsvd_0_23: 24;  /*    */
+        unsigned long   base    :  8;  /* RW */
+        unsigned long   rsvd_32_47: 16;  /*    */
+        unsigned long   m_alias :  5;  /* RW */
+        unsigned long   rsvd_53_62: 10;  /*    */
+        unsigned long   enable  :  1;  /* RW */
+    } s;
+};
+/* ========================================================================= */
+/*                 UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR                  */
+/* ========================================================================= */
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR 0x16000e8UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_BASE_SHFT 24
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_BASE_MASK 0x00000000ff000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_M_ALIAS_SHFT 48
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_M_ALIAS_MASK 0x001f000000000000UL
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_ENABLE_SHFT 63
+#define UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR_ENABLE_MASK 0x8000000000000000UL
+union uvh_rh_gam_alias210_overlay_config_2_mmr_u {
+    unsigned long       v;
+    struct uvh_rh_gam_alias210_overlay_config_2_mmr_s {
+        unsigned long   rsvd_0_23: 24;  /*    */
+        unsigned long   base    :  8;  /* RW */
+        unsigned long   rsvd_32_47: 16;  /*    */
+        unsigned long   m_alias :  5;  /* RW */
+        unsigned long   rsvd_53_62: 10;  /*    */
+        unsigned long   enable  :  1;  /* RW */
+    } s;
+};
+/* ========================================================================= */
 /*                UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_0_MMR                  */
 /* ========================================================================= */
 #define UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_0_MMR 0x16000d0UL
@@ -857,6 +929,29 @@ union uvh_rh_gam_alias210_redirect_config_2_mmr_u {
 };
 /* ========================================================================= */
+/*                          UVH_RH_GAM_CONFIG_MMR                            */
+/* ========================================================================= */
+#define UVH_RH_GAM_CONFIG_MMR 0x1600000UL
+#define UVH_RH_GAM_CONFIG_MMR_M_SKT_SHFT 0
+#define UVH_RH_GAM_CONFIG_MMR_M_SKT_MASK 0x000000000000003fUL
+#define UVH_RH_GAM_CONFIG_MMR_N_SKT_SHFT 6
+#define UVH_RH_GAM_CONFIG_MMR_N_SKT_MASK 0x00000000000003c0UL
+#define UVH_RH_GAM_CONFIG_MMR_MMIOL_CFG_SHFT 12
+#define UVH_RH_GAM_CONFIG_MMR_MMIOL_CFG_MASK 0x0000000000001000UL
+union uvh_rh_gam_config_mmr_u {
+    unsigned long       v;
+    struct uvh_rh_gam_config_mmr_s {
+        unsigned long   m_skt     :  6;  /* RW */
+        unsigned long   n_skt     :  4;  /* RW */
+        unsigned long   rsvd_10_11:  2;  /*    */
+        unsigned long   mmiol_cfg :  1;  /* RW */
+        unsigned long   rsvd_13_63: 51;  /*    */
+    } s;
+};
+/* ========================================================================= */
 /*                    UVH_RH_GAM_GRU_OVERLAY_CONFIG_MMR                      */
 /* ========================================================================= */
 #define UVH_RH_GAM_GRU_OVERLAY_CONFIG_MMR 0x1600010UL
@@ -987,97 +1082,5 @@ union uvh_rtc1_int_config_u {
    } s;
 };
-/* ========================================================================= */
-/*                          UVH_SI_ADDR_MAP_CONFIG                           */
-/* ========================================================================= */
-#define UVH_SI_ADDR_MAP_CONFIG 0xc80000UL
-#define UVH_SI_ADDR_MAP_CONFIG_M_SKT_SHFT 0
-#define UVH_SI_ADDR_MAP_CONFIG_M_SKT_MASK 0x000000000000003fUL
-#define UVH_SI_ADDR_MAP_CONFIG_N_SKT_SHFT 8
-#define UVH_SI_ADDR_MAP_CONFIG_N_SKT_MASK 0x0000000000000f00UL
-union uvh_si_addr_map_config_u {
-    unsigned long       v;
-    struct uvh_si_addr_map_config_s {
-        unsigned long   m_skt :  6;  /* RW */
-        unsigned long   rsvd_6_7:  2;  /*    */
-        unsigned long   n_skt :  4;  /* RW */
-        unsigned long   rsvd_12_63: 52;  /*    */
-    } s;
-};
-/* ========================================================================= */
-/*                       UVH_SI_ALIAS0_OVERLAY_CONFIG                        */
-/* ========================================================================= */
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG 0xc80008UL
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_BASE_SHFT 24
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_BASE_MASK 0x00000000ff000000UL
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_M_ALIAS_SHFT 48
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_M_ALIAS_MASK 0x001f000000000000UL
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_ENABLE_SHFT 63
-#define UVH_SI_ALIAS0_OVERLAY_CONFIG_ENABLE_MASK 0x8000000000000000UL
-union uvh_si_alias0_overlay_config_u {
-    unsigned long       v;
-    struct uvh_si_alias0_overlay_config_s {
-        unsigned long   rsvd_0_23: 24;  /*    */
-        unsigned long   base    :  8;  /* RW */
-        unsigned long   rsvd_32_47: 16;  /*    */
-        unsigned long   m_alias :  5;  /* RW */
-        unsigned long   rsvd_53_62: 10;  /*    */
-        unsigned long   enable  :  1;  /* RW */
-    } s;
-};
-/* ========================================================================= */
-/*                       UVH_SI_ALIAS1_OVERLAY_CONFIG                        */
-/* ========================================================================= */
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG 0xc80010UL
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_BASE_SHFT 24
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_BASE_MASK 0x00000000ff000000UL
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_M_ALIAS_SHFT 48
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_M_ALIAS_MASK 0x001f000000000000UL
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_ENABLE_SHFT 63
-#define UVH_SI_ALIAS1_OVERLAY_CONFIG_ENABLE_MASK 0x8000000000000000UL
-union uvh_si_alias1_overlay_config_u {
-    unsigned long       v;
-    struct uvh_si_alias1_overlay_config_s {
-        unsigned long   rsvd_0_23: 24;  /*    */
-        unsigned long   base    :  8;  /* RW */
-        unsigned long   rsvd_32_47: 16;  /*    */
-        unsigned long   m_alias :  5;  /* RW */
-        unsigned long   rsvd_53_62: 10;  /*    */
-        unsigned long   enable  :  1;  /* RW */
-    } s;
-};
-/* ========================================================================= */
-/*                       UVH_SI_ALIAS2_OVERLAY_CONFIG                        */
-/* ========================================================================= */
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG 0xc80018UL
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_BASE_SHFT 24
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_BASE_MASK 0x00000000ff000000UL
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_M_ALIAS_SHFT 48
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_M_ALIAS_MASK 0x001f000000000000UL
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_ENABLE_SHFT 63
-#define UVH_SI_ALIAS2_OVERLAY_CONFIG_ENABLE_MASK 0x8000000000000000UL
-union uvh_si_alias2_overlay_config_u {
-    unsigned long       v;
-    struct uvh_si_alias2_overlay_config_s {
-        unsigned long   rsvd_0_23: 24;  /*    */
-        unsigned long   base    :  8;  /* RW */
-        unsigned long   rsvd_32_47: 16;  /*    */
-        unsigned long   m_alias :  5;  /* RW */
-        unsigned long   rsvd_53_62: 10;  /*    */
-        unsigned long   enable  :  1;  /* RW */
-    } s;
-};
-#endif /* _ASM_X86_UV_UV_MMRS_H */
+#endif /* __ASM_UV_MMRS_X86_H__ */
diff --git a/arch/x86/include/asm/xen/interface.h b/arch/x86/include/asm/xen/interface.h
index e8506c1f0c55..1c10c88ee4e1 100644
--- a/arch/x86/include/asm/xen/interface.h
+++ b/arch/x86/include/asm/xen/interface.h
@@ -61,9 +61,9 @@ DEFINE_GUEST_HANDLE(void);
 #define HYPERVISOR_VIRT_START mk_unsigned_long(__HYPERVISOR_VIRT_START)
 #endif
-#ifndef machine_to_phys_mapping
+#define MACH2PHYS_VIRT_START  mk_unsigned_long(__MACH2PHYS_VIRT_START)
-#define machine_to_phys_mapping ((unsigned long *)HYPERVISOR_VIRT_START)
+#define MACH2PHYS_VIRT_END    mk_unsigned_long(__MACH2PHYS_VIRT_END)
-#endif
+#define MACH2PHYS_NR_ENTRIES  ((MACH2PHYS_VIRT_END-MACH2PHYS_VIRT_START)>>__MACH2PHYS_SHIFT)
 /* Maximum number of virtual CPUs in multi-processor guests. */
 #define MAX_VIRT_CPUS 32
diff --git a/arch/x86/include/asm/xen/interface_32.h b/arch/x86/include/asm/xen/interface_32.h
index 42a7e004ae5c..8413688b2571 100644
--- a/arch/x86/include/asm/xen/interface_32.h
+++ b/arch/x86/include/asm/xen/interface_32.h
@@ -32,6 +32,11 @@
 /* And the trap vector is... */
 #define TRAP_INSTR "int $0x82"
+#define __MACH2PHYS_VIRT_START 0xF5800000
+#define __MACH2PHYS_VIRT_END   0xF6800000
+#define __MACH2PHYS_SHIFT      2
 /*
 * Virtual addresses beyond this are not modifiable by guest OSes. The
 * machine->physical mapping table starts at this address, read-only.
diff --git a/arch/x86/include/asm/xen/interface_64.h b/arch/x86/include/asm/xen/interface_64.h
index 100d2662b97c..839a4811cf98 100644
--- a/arch/x86/include/asm/xen/interface_64.h
+++ b/arch/x86/include/asm/xen/interface_64.h
@@ -39,18 +39,7 @@
 #define __HYPERVISOR_VIRT_END   0xFFFF880000000000
 #define __MACH2PHYS_VIRT_START  0xFFFF800000000000
 #define __MACH2PHYS_VIRT_END    0xFFFF804000000000
+#define __MACH2PHYS_SHIFT       3
-#ifndef HYPERVISOR_VIRT_START
-#define HYPERVISOR_VIRT_START mk_unsigned_long(__HYPERVISOR_VIRT_START)
-#define HYPERVISOR_VIRT_END   mk_unsigned_long(__HYPERVISOR_VIRT_END)
-#endif
-#define MACH2PHYS_VIRT_START  mk_unsigned_long(__MACH2PHYS_VIRT_START)
-#define MACH2PHYS_VIRT_END    mk_unsigned_long(__MACH2PHYS_VIRT_END)
-#define MACH2PHYS_NR_ENTRIES  ((MACH2PHYS_VIRT_END-MACH2PHYS_VIRT_START)>>3)
-#ifndef machine_to_phys_mapping
-#define machine_to_phys_mapping ((unsigned long *)HYPERVISOR_VIRT_START)
-#endif
 /*
 * int HYPERVISOR_set_segment_base(unsigned int which, unsigned long base)
diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index dd8c1414b3d5..8760cc60a21c 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -5,6 +5,7 @@
 #include <linux/types.h>
 #include <linux/spinlock.h>
 #include <linux/pfn.h>
+#include <linux/mm.h>
 #include <asm/uaccess.h>
 #include <asm/page.h>
@@ -35,6 +36,8 @@ typedef struct xpaddr {
 #define MAX_DOMAIN_PAGES                                                \
    ((unsigned long)((u64)CONFIG_XEN_MAX_DOMAIN_MEMORY * 1024 * 1024 * 1024 / PAGE_SIZE))
+extern unsigned long *machine_to_phys_mapping;
+extern unsigned int   machine_to_phys_order;
 extern unsigned long get_phys_to_machine(unsigned long pfn);
 extern bool set_phys_to_machine(unsigned long pfn, unsigned long mfn);
@@ -69,10 +72,8 @@ static inline unsigned long mfn_to_pfn(unsigned long mfn)
        if (xen_feature(XENFEAT_auto_translated_physmap))
                return mfn;
-#if 0
        if (unlikely((mfn >> machine_to_phys_order) != 0))
-                return max_mapnr;
+                return ~0;
-#endif
        pfn = 0;
        /*
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 850657d1b0ed..3f838d537392 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -52,7 +52,6 @@
 #include <asm/mce.h>
 #include <asm/kvm_para.h>
 #include <asm/tsc.h>
-#include <asm/atomic.h>
 unsigned int num_processors;
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index ed4118de249e..194539aea175 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -379,14 +379,14 @@ struct redir_addr {
 #define DEST_SHIFT UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_0_MMR_DEST_BASE_SHFT
 static __initdata struct redir_addr redir_addrs[] = {
-        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_0_MMR, UVH_SI_ALIAS0_OVERLAY_CONFIG},
+        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_0_MMR, UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_0_MMR},
-        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_1_MMR, UVH_SI_ALIAS1_OVERLAY_CONFIG},
+        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_1_MMR, UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_1_MMR},
-        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_2_MMR, UVH_SI_ALIAS2_OVERLAY_CONFIG},
+        {UVH_RH_GAM_ALIAS210_REDIRECT_CONFIG_2_MMR, UVH_RH_GAM_ALIAS210_OVERLAY_CONFIG_2_MMR},
 };
 static __init void get_lowmem_redirect(unsigned long *base, unsigned long *size)
 {
-        union uvh_si_alias0_overlay_config_u alias;
+        union uvh_rh_gam_alias210_overlay_config_2_mmr_u alias;
        union uvh_rh_gam_alias210_redirect_config_2_mmr_u redirect;
        int i;
@@ -660,7 +660,7 @@ void uv_nmi_init(void)
 void __init uv_system_init(void)
 {
-        union uvh_si_addr_map_config_u m_n_config;
+        union uvh_rh_gam_config_mmr_u  m_n_config;
        union uvh_node_id_u node_id;
        unsigned long gnode_upper, lowmem_redir_base, lowmem_redir_size;
        int bytes, nid, cpu, lcpu, pnode, blade, i, j, m_val, n_val;
@@ -670,7 +670,7 @@ void __init uv_system_init(void)
        map_low_mmrs();
-        m_n_config.v = uv_read_local_mmr(UVH_SI_ADDR_MAP_CONFIG);
+        m_n_config.v = uv_read_local_mmr(UVH_RH_GAM_CONFIG_MMR );
        m_val = m_n_config.s.m_skt;
        n_val = m_n_config.s.n_skt;
        mmr_base =
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 46d58448c3af..e421b8cd6944 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -280,11 +280,11 @@ static struct amd_nb *amd_alloc_nb(int cpu, int nb_id)
        struct amd_nb *nb;
        int i;
-        nb = kmalloc(sizeof(struct amd_nb), GFP_KERNEL);
+        nb = kmalloc_node(sizeof(struct amd_nb), GFP_KERNEL | __GFP_ZERO,
+                          cpu_to_node(cpu));
        if (!nb)
                return NULL;
-        memset(nb, 0, sizeof(*nb));
        nb->nb_id = nb_id;
        /*
diff --git a/arch/x86/kernel/cpuid.c b/arch/x86/kernel/cpuid.c
index 1b7b31ab7d86..212a6a42527c 100644
--- a/arch/x86/kernel/cpuid.c
+++ b/arch/x86/kernel/cpuid.c
@@ -33,7 +33,6 @@
 #include <linux/init.h>
 #include <linux/poll.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/major.h>
 #include <linux/fs.h>
 #include <linux/device.h>
diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index ec592caac4b4..cd21b654dec6 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -315,14 +315,18 @@ static void kgdb_remove_all_hw_break(void)
                if (!breakinfo[i].enabled)
                        continue;
                bp = *per_cpu_ptr(breakinfo[i].pev, cpu);
-                if (bp->attr.disabled == 1)
+                if (!bp->attr.disabled) {
+                        arch_uninstall_hw_breakpoint(bp);
+                        bp->attr.disabled = 1;
                        continue;
+                }
                if (dbg_is_early)
                        early_dr7 &= ~encode_dr7(i, breakinfo[i].len,
                                                 breakinfo[i].type);
-                else
+                else if (hw_break_release_slot(i))
-                        arch_uninstall_hw_breakpoint(bp);
+                        printk(KERN_ERR "KGDB: hw bpt remove failed %lx\n",
-                bp->attr.disabled = 1;
+                               breakinfo[i].addr);
+                breakinfo[i].enabled = 0;
        }
 }
diff --git a/arch/x86/kernel/microcode_amd.c b/arch/x86/kernel/microcode_amd.c
index e1af7c055c7d..ce0cb4721c9a 100644
--- a/arch/x86/kernel/microcode_amd.c
+++ b/arch/x86/kernel/microcode_amd.c
@@ -212,7 +212,7 @@ static int install_equiv_cpu_table(const u8 *buf)
                return 0;
        }
-        equiv_cpu_table = (struct equiv_cpu_entry *) vmalloc(size);
+        equiv_cpu_table = vmalloc(size);
        if (!equiv_cpu_table) {
                pr_err("failed to allocate equivalent CPU table\n");
                return 0;
diff --git a/arch/x86/kernel/mmconf-fam10h_64.c b/arch/x86/kernel/mmconf-fam10h_64.c
index 71825806cd44..6da143c2a6b8 100644
--- a/arch/x86/kernel/mmconf-fam10h_64.c
+++ b/arch/x86/kernel/mmconf-fam10h_64.c
@@ -217,13 +217,13 @@ void __cpuinit fam10h_check_enable_mmcfg(void)
        wrmsrl(address, val);
 }
-static int __devinit set_check_enable_amd_mmconf(const struct dmi_system_id *d)
+static int __init set_check_enable_amd_mmconf(const struct dmi_system_id *d)
 {
        pci_probe |= PCI_CHECK_ENABLE_AMD_MMCONF;
        return 0;
 }
-static const struct dmi_system_id __cpuinitconst mmconf_dmi_table[] = {
+static const struct dmi_system_id __initconst mmconf_dmi_table[] = {
        {
                .callback = set_check_enable_amd_mmconf,
                .ident = "Sun Microsystems Machine",
@@ -234,7 +234,8 @@ static const struct dmi_system_id __cpuinitconst mmconf_dmi_table[] = {
        {}
 };
-void __cpuinit check_enable_amd_mmconf_dmi(void)
+/* Called from a __cpuinit function, but only on the BSP. */
+void __ref check_enable_amd_mmconf_dmi(void)
 {
        dmi_check_system(mmconf_dmi_table);
 }
diff --git a/arch/x86/kernel/msr.c b/arch/x86/kernel/msr.c
index 7bf2dc4c8f70..12fcbe2c143e 100644
--- a/arch/x86/kernel/msr.c
+++ b/arch/x86/kernel/msr.c
@@ -30,7 +30,6 @@
 #include <linux/init.h>
 #include <linux/poll.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/major.h>
 #include <linux/fs.h>
 #include <linux/device.h>
diff --git a/arch/x86/kernel/pvclock.c b/arch/x86/kernel/pvclock.c
index bab3b9e6f66d..008b91eefa18 100644
--- a/arch/x86/kernel/pvclock.c
+++ b/arch/x86/kernel/pvclock.c
@@ -41,44 +41,6 @@ void pvclock_set_flags(u8 flags)
        valid_flags = flags;
 }
-/*
- * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction,
- * yielding a 64-bit result.
- */
-static inline u64 scale_delta(u64 delta, u32 mul_frac, int shift)
-{
-        u64 product;
-#ifdef __i386__
-        u32 tmp1, tmp2;
-#endif
-        if (shift < 0)
-                delta >>= -shift;
-        else
-                delta <<= shift;
-#ifdef __i386__
-        __asm__ (
-                "mul  %5       ; "
-                "mov  %4,%%eax ; "
-                "mov  %%edx,%4 ; "
-                "mul  %5       ; "
-                "xor  %5,%5    ; "
-                "add  %4,%%eax ; "
-                "adc  %5,%%edx ; "
-                : "=A" (product), "=r" (tmp1), "=r" (tmp2)
-                : "a" ((u32)delta), "1" ((u32)(delta >> 32)), "2" (mul_frac) );
-#elif defined(__x86_64__)
-        __asm__ (
-                "mul %%rdx ; shrd $32,%%rdx,%%rax"
-                : "=a" (product) : "0" (delta), "d" ((u64)mul_frac) );
-#else
-#error implement me!
-#endif
-        return product;
-}
 static u64 pvclock_get_nsec_offset(struct pvclock_shadow_time *shadow)
 {
        u64 delta = native_read_tsc() - shadow->tsc_timestamp;
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 82e144a4e514..1ca12298ffc7 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -3395,6 +3395,7 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
        vcpu->arch.regs[VCPU_REGS_RIP] = svm->vmcb->save.rip;
        load_host_msrs(vcpu);
+        kvm_load_ldt(ldt_selector);
        loadsegment(fs, fs_selector);
 #ifdef CONFIG_X86_64
        load_gs_index(gs_selector);
@@ -3402,7 +3403,6 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
 #else
        loadsegment(gs, gs_selector);
 #endif
-        kvm_load_ldt(ldt_selector);
        reload_tss(vcpu);
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 8da0e45ff7c9..ff21fdda0c53 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -821,10 +821,9 @@ static void vmx_save_host_state(struct kvm_vcpu *vcpu)
 #endif
 #ifdef CONFIG_X86_64
-        if (is_long_mode(&vmx->vcpu)) {
+        rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base);
-                rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base);
+        if (is_long_mode(&vmx->vcpu))
                wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base);
-        }
 #endif
        for (i = 0; i < vmx->save_nmsrs; ++i)
                kvm_set_shared_msr(vmx->guest_msrs[i].index,
@@ -839,23 +838,23 @@ static void __vmx_load_host_state(struct vcpu_vmx *vmx)
        ++vmx->vcpu.stat.host_state_reload;
        vmx->host_state.loaded = 0;
-        if (vmx->host_state.fs_reload_needed)
+#ifdef CONFIG_X86_64
-                loadsegment(fs, vmx->host_state.fs_sel);
+        if (is_long_mode(&vmx->vcpu))
+                rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base);
+#endif
        if (vmx->host_state.gs_ldt_reload_needed) {
                kvm_load_ldt(vmx->host_state.ldt_sel);
 #ifdef CONFIG_X86_64
                load_gs_index(vmx->host_state.gs_sel);
-                wrmsrl(MSR_KERNEL_GS_BASE, current->thread.gs);
 #else
                loadsegment(gs, vmx->host_state.gs_sel);
 #endif
        }
+        if (vmx->host_state.fs_reload_needed)
+                loadsegment(fs, vmx->host_state.fs_sel);
        reload_tss();
 #ifdef CONFIG_X86_64
-        if (is_long_mode(&vmx->vcpu)) {
+        wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base);
-                rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base);
-                wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base);
-        }
 #endif
        if (current_thread_info()->status & TS_USEDFPU)
                clts();
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index 49358481c733..12cdbb17ad18 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -251,7 +251,7 @@ static void __cpuinit calculate_tlb_offset(void)
        }
 }
-static int tlb_cpuhp_notify(struct notifier_block *n,
+static int __cpuinit tlb_cpuhp_notify(struct notifier_block *n,
                unsigned long action, void *hcpu)
 {
        switch (action & 0xf) {
diff --git a/arch/x86/pci/acpi.c b/arch/x86/pci/acpi.c
index 15466c096ba5..0972315c3860 100644
--- a/arch/x86/pci/acpi.c
+++ b/arch/x86/pci/acpi.c
@@ -138,7 +138,6 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
        struct acpi_resource_address64 addr;
        acpi_status status;
        unsigned long flags;
-        struct resource *root, *conflict;
        u64 start, end;
        status = resource_to_addr(acpi_res, &addr);
@@ -146,12 +145,10 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
                return AE_OK;
        if (addr.resource_type == ACPI_MEMORY_RANGE) {
-                root = &iomem_resource;
                flags = IORESOURCE_MEM;
                if (addr.info.mem.caching == ACPI_PREFETCHABLE_MEMORY)
                        flags |= IORESOURCE_PREFETCH;
        } else if (addr.resource_type == ACPI_IO_RANGE) {
-                root = &ioport_resource;
                flags = IORESOURCE_IO;
        } else
                return AE_OK;
@@ -172,25 +169,90 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
                return AE_OK;
        }
-        conflict = insert_resource_conflict(root, res);
+        info->res_num++;
-        if (conflict) {
+        if (addr.translation_offset)
-                dev_err(&info->bridge->dev,
+                dev_info(&info->bridge->dev, "host bridge window %pR "
-                        "address space collision: host bridge window %pR "
+                         "(PCI address [%#llx-%#llx])\n",
-                        "conflicts with %s %pR\n",
+                         res, res->start - addr.translation_offset,
-                        res, conflict->name, conflict);
+                         res->end - addr.translation_offset);
-        } else {
+        else
-                pci_bus_add_resource(info->bus, res, 0);
+                dev_info(&info->bridge->dev, "host bridge window %pR\n", res);
-                info->res_num++;
-                if (addr.translation_offset)
+        return AE_OK;
-                        dev_info(&info->bridge->dev, "host bridge window %pR "
+}
-                                 "(PCI address [%#llx-%#llx])\n",
-                                 res, res->start - addr.translation_offset,
+static bool resource_contains(struct resource *res, resource_size_t point)
-                                 res->end - addr.translation_offset);
+{
+        if (res->start <= point && point <= res->end)
+                return true;
+        return false;
+}
+static void coalesce_windows(struct pci_root_info *info, int type)
+{
+        int i, j;
+        struct resource *res1, *res2;
+        for (i = 0; i < info->res_num; i++) {
+                res1 = &info->res[i];
+                if (!(res1->flags & type))
+                        continue;
+                for (j = i + 1; j < info->res_num; j++) {
+                        res2 = &info->res[j];
+                        if (!(res2->flags & type))
+                                continue;
+                        /*
+                         * I don't like throwing away windows because then
+                         * our resources no longer match the ACPI _CRS, but
+                         * the kernel resource tree doesn't allow overlaps.
+                         */
+                        if (resource_contains(res1, res2->start) ||
+                            resource_contains(res1, res2->end) ||
+                            resource_contains(res2, res1->start) ||
+                            resource_contains(res2, res1->end)) {
+                                res1->start = min(res1->start, res2->start);
+                                res1->end = max(res1->end, res2->end);
+                                dev_info(&info->bridge->dev,
+                                         "host bridge window expanded to %pR; %pR ignored\n",
+                                         res1, res2);
+                                res2->flags = 0;
+                        }
+                }
+        }
+}
+static void add_resources(struct pci_root_info *info)
+{
+        int i;
+        struct resource *res, *root, *conflict;
+        if (!pci_use_crs)
+                return;
+        coalesce_windows(info, IORESOURCE_MEM);
+        coalesce_windows(info, IORESOURCE_IO);
+        for (i = 0; i < info->res_num; i++) {
+                res = &info->res[i];
+                if (res->flags & IORESOURCE_MEM)
+                        root = &iomem_resource;
+                else if (res->flags & IORESOURCE_IO)
+                        root = &ioport_resource;
                else
-                        dev_info(&info->bridge->dev,
+                        continue;
-                                 "host bridge window %pR\n", res);
+                conflict = insert_resource_conflict(root, res);
+                if (conflict)
+                        dev_err(&info->bridge->dev,
+                                "address space collision: host bridge window %pR "
+                                "conflicts with %s %pR\n",
+                                res, conflict->name, conflict);
+                else
+                        pci_bus_add_resource(info->bus, res, 0);
        }
-        return AE_OK;
 }
 static void
@@ -224,6 +286,7 @@ get_current_resources(struct acpi_device *device, int busnum,
        acpi_walk_resources(device->handle, METHOD_NAME__CRS, setup_resource,
                                &info);
+        add_resources(&info);
        return;
 name_alloc_fail:
diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c
index 117f5b8daf75..d7b5109f7a9c 100644
--- a/arch/x86/pci/xen.c
+++ b/arch/x86/pci/xen.c
@@ -147,8 +147,10 @@ static int xen_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
                irq = xen_allocate_pirq(v[i], 0, /* not sharable */
                        (type == PCI_CAP_ID_MSIX) ?
                        "pcifront-msi-x" : "pcifront-msi");
-                if (irq < 0)
+                if (irq < 0) {
-                        return -1;
+                        ret = -1;
+                        goto free;
+                }
                ret = set_irq_msi(irq, msidesc);
                if (ret)
@@ -164,7 +166,7 @@ error:
        if (ret == -ENODEV)
                dev_err(&dev->dev, "Xen PCI frontend has not registered" \
                        " MSI/MSI-X support!\n");
+free:
        kfree(v);
        return ret;
 }
diff --git a/arch/x86/platform/uv/tlb_uv.c b/arch/x86/platform/uv/tlb_uv.c
index 20ea20a39e2a..a318194002b5 100644
--- a/arch/x86/platform/uv/tlb_uv.c
+++ b/arch/x86/platform/uv/tlb_uv.c
@@ -1343,8 +1343,8 @@ uv_activation_descriptor_init(int node, int pnode)
         * each bau_desc is 64 bytes; there are 8 (UV_ITEMS_PER_DESCRIPTOR)
         * per cpu; and up to 32 (UV_ADP_SIZE) cpu's per uvhub
         */
-        bau_desc = (struct bau_desc *)kmalloc_node(sizeof(struct bau_desc)*
+        bau_desc = kmalloc_node(sizeof(struct bau_desc) * UV_ADP_SIZE
-                UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR, GFP_KERNEL, node);
+                                * UV_ITEMS_PER_DESCRIPTOR, GFP_KERNEL, node);
        BUG_ON(!bau_desc);
        pa = uv_gpa(bau_desc); /* need the real nasid*/
@@ -1402,9 +1402,9 @@ uv_payload_queue_init(int node, int pnode)
        struct bau_payload_queue_entry *pqp_malloc;
        struct bau_control *bcp;
-        pqp = (struct bau_payload_queue_entry *) kmalloc_node(
+        pqp = kmalloc_node((DEST_Q_SIZE + 1)
-                (DEST_Q_SIZE + 1) * sizeof(struct bau_payload_queue_entry),
+                           * sizeof(struct bau_payload_queue_entry),
-                GFP_KERNEL, node);
+                           GFP_KERNEL, node);
        BUG_ON(!pqp);
        pqp_malloc = pqp;
@@ -1520,8 +1520,7 @@ static void __init uv_init_per_cpu(int nuvhubs)
        timeout_us = calculate_destination_timeout();
-        uvhub_descs = (struct uvhub_desc *)
+        uvhub_descs = kmalloc(nuvhubs * sizeof(struct uvhub_desc), GFP_KERNEL);
-                kmalloc(nuvhubs * sizeof(struct uvhub_desc), GFP_KERNEL);
        memset(uvhub_descs, 0, nuvhubs * sizeof(struct uvhub_desc));
        uvhub_mask = kzalloc((nuvhubs+7)/8, GFP_KERNEL);
        for_each_present_cpu(cpu) {
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 235c0f4d3861..02c710bebf7a 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -75,6 +75,11 @@ DEFINE_PER_CPU(struct vcpu_info, xen_vcpu_info);
 enum xen_domain_type xen_domain_type = XEN_NATIVE;
 EXPORT_SYMBOL_GPL(xen_domain_type);
+unsigned long *machine_to_phys_mapping = (void *)MACH2PHYS_VIRT_START;
+EXPORT_SYMBOL(machine_to_phys_mapping);
+unsigned int   machine_to_phys_order;
+EXPORT_SYMBOL(machine_to_phys_order);
 struct start_info *xen_start_info;
 EXPORT_SYMBOL_GPL(xen_start_info);
@@ -1090,6 +1095,8 @@ static void __init xen_setup_stackprotector(void)
 /* First C function to be called on Xen boot */
 asmlinkage void __init xen_start_kernel(void)
 {
+        struct physdev_set_iopl set_iopl;
+        int rc;
        pgd_t *pgd;
        if (!xen_start_info)
@@ -1097,6 +1104,8 @@ asmlinkage void __init xen_start_kernel(void)
        xen_domain_type = XEN_PV_DOMAIN;
+        xen_setup_machphys_mapping();
        /* Install Xen paravirt ops */
        pv_info = xen_info;
        pv_init_ops = xen_init_ops;
@@ -1191,8 +1200,6 @@ asmlinkage void __init xen_start_kernel(void)
        /* Allocate and initialize top and mid mfn levels for p2m structure */
        xen_build_mfn_list_list();
-        init_mm.pgd = pgd;
        /* keep using Xen gdt for now; no urgent need to change it */
 #ifdef CONFIG_X86_32
@@ -1202,10 +1209,18 @@ asmlinkage void __init xen_start_kernel(void)
 #else
        pv_info.kernel_rpl = 0;
 #endif
        /* set the limit of our address space */
        xen_reserve_top();
+        /* We used to do this in xen_arch_setup, but that is too late on AMD
+         * were early_cpu_init (run before ->arch_setup()) calls early_amd_init
+         * which pokes 0xcf8 port.
+         */
+        set_iopl.iopl = 1;
+        rc = HYPERVISOR_physdev_op(PHYSDEVOP_set_iopl, &set_iopl);
+        if (rc != 0)
+                xen_raw_printk("physdev_op failed %d\n", rc);
 #ifdef CONFIG_X86_32
        /* set up basic CPUID stuff */
        cpu_detect(&new_cpu_data);
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index c237b810b03f..a1feff9e59b6 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -2034,6 +2034,20 @@ static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn)
        set_page_prot(pmd, PAGE_KERNEL_RO);
 }
+void __init xen_setup_machphys_mapping(void)
+{
+        struct xen_machphys_mapping mapping;
+        unsigned long machine_to_phys_nr_ents;
+        if (HYPERVISOR_memory_op(XENMEM_machphys_mapping, &mapping) == 0) {
+                machine_to_phys_mapping = (unsigned long *)mapping.v_start;
+                machine_to_phys_nr_ents = mapping.max_mfn + 1;
+        } else {
+                machine_to_phys_nr_ents = MACH2PHYS_NR_ENTRIES;
+        }
+        machine_to_phys_order = fls(machine_to_phys_nr_ents - 1);
+}
 #ifdef CONFIG_X86_64
 static void convert_pfn_mfn(void *v)
 {
@@ -2119,44 +2133,83 @@ __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
        return pgd;
 }
 #else   /* !CONFIG_X86_64 */
-static RESERVE_BRK_ARRAY(pmd_t, level2_kernel_pgt, PTRS_PER_PMD);
+static RESERVE_BRK_ARRAY(pmd_t, initial_kernel_pmd, PTRS_PER_PMD);
+static RESERVE_BRK_ARRAY(pmd_t, swapper_kernel_pmd, PTRS_PER_PMD);
+static __init void xen_write_cr3_init(unsigned long cr3)
+{
+        unsigned long pfn = PFN_DOWN(__pa(swapper_pg_dir));
+        BUG_ON(read_cr3() != __pa(initial_page_table));
+        BUG_ON(cr3 != __pa(swapper_pg_dir));
+        /*
+         * We are switching to swapper_pg_dir for the first time (from
+         * initial_page_table) and therefore need to mark that page
+         * read-only and then pin it.
+         *
+         * Xen disallows sharing of kernel PMDs for PAE
+         * guests. Therefore we must copy the kernel PMD from
+         * initial_page_table into a new kernel PMD to be used in
+         * swapper_pg_dir.
+         */
+        swapper_kernel_pmd =
+                extend_brk(sizeof(pmd_t) * PTRS_PER_PMD, PAGE_SIZE);
+        memcpy(swapper_kernel_pmd, initial_kernel_pmd,
+               sizeof(pmd_t) * PTRS_PER_PMD);
+        swapper_pg_dir[KERNEL_PGD_BOUNDARY] =
+                __pgd(__pa(swapper_kernel_pmd) | _PAGE_PRESENT);
+        set_page_prot(swapper_kernel_pmd, PAGE_KERNEL_RO);
+        set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
+        xen_write_cr3(cr3);
+        pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, pfn);
+        pin_pagetable_pfn(MMUEXT_UNPIN_TABLE,
+                          PFN_DOWN(__pa(initial_page_table)));
+        set_page_prot(initial_page_table, PAGE_KERNEL);
+        set_page_prot(initial_kernel_pmd, PAGE_KERNEL);
+        pv_mmu_ops.write_cr3 = &xen_write_cr3;
+}
 __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
                                         unsigned long max_pfn)
 {
        pmd_t *kernel_pmd;
-        level2_kernel_pgt = extend_brk(sizeof(pmd_t *) * PTRS_PER_PMD, PAGE_SIZE);
+        initial_kernel_pmd =
+                extend_brk(sizeof(pmd_t) * PTRS_PER_PMD, PAGE_SIZE);
        max_pfn_mapped = PFN_DOWN(__pa(xen_start_info->pt_base) +
                                  xen_start_info->nr_pt_frames * PAGE_SIZE +
                                  512*1024);
        kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd);
-        memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD);
+        memcpy(initial_kernel_pmd, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD);
-        xen_map_identity_early(level2_kernel_pgt, max_pfn);
+        xen_map_identity_early(initial_kernel_pmd, max_pfn);
-        memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD);
+        memcpy(initial_page_table, pgd, sizeof(pgd_t) * PTRS_PER_PGD);
-        set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY],
+        initial_page_table[KERNEL_PGD_BOUNDARY] =
-                        __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT));
+                __pgd(__pa(initial_kernel_pmd) | _PAGE_PRESENT);
-        set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
+        set_page_prot(initial_kernel_pmd, PAGE_KERNEL_RO);
-        set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
+        set_page_prot(initial_page_table, PAGE_KERNEL_RO);
        set_page_prot(empty_zero_page, PAGE_KERNEL_RO);
        pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
-        xen_write_cr3(__pa(swapper_pg_dir));
+        pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE,
+                          PFN_DOWN(__pa(initial_page_table)));
-        pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir)));
+        xen_write_cr3(__pa(initial_page_table));
        memblock_x86_reserve_range(__pa(xen_start_info->pt_base),
                      __pa(xen_start_info->pt_base +
                           xen_start_info->nr_pt_frames * PAGE_SIZE),
                      "XEN PAGETABLES");
-        return swapper_pg_dir;
+        return initial_page_table;
 }
 #endif  /* CONFIG_X86_64 */
@@ -2290,7 +2343,11 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = {
        .write_cr2 = xen_write_cr2,
        .read_cr3 = xen_read_cr3,
+#ifdef CONFIG_X86_32
+        .write_cr3 = xen_write_cr3_init,
+#else
        .write_cr3 = xen_write_cr3,
+#endif
        .flush_tlb_user = xen_flush_tlb,
        .flush_tlb_kernel = xen_flush_tlb,
@@ -2627,7 +2684,8 @@ int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
        prot = __pgprot(pgprot_val(prot) | _PAGE_IOMAP);
-        vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP;
+        BUG_ON(!((vma->vm_flags & (VM_PFNMAP | VM_RESERVED | VM_IO)) ==
+                                (VM_PFNMAP | VM_RESERVED | VM_IO)));
        rmd.mfn = mfn;
        rmd.prot = prot;
diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c
index b1dbdaa23ecc..01afd8a94607 100644
--- a/arch/x86/xen/setup.c
+++ b/arch/x86/xen/setup.c
@@ -23,7 +23,6 @@
 #include <xen/interface/callback.h>
 #include <xen/interface/memory.h>
 #include <xen/interface/physdev.h>
-#include <xen/interface/memory.h>
 #include <xen/features.h>
 #include "xen-ops.h"
@@ -118,16 +117,18 @@ static unsigned long __init xen_return_unused_memory(unsigned long max_pfn,
                                                     const struct e820map *e820)
 {
        phys_addr_t max_addr = PFN_PHYS(max_pfn);
-        phys_addr_t last_end = 0;
+        phys_addr_t last_end = ISA_END_ADDRESS;
        unsigned long released = 0;
        int i;
+        /* Free any unused memory above the low 1Mbyte. */
        for (i = 0; i < e820->nr_map && last_end < max_addr; i++) {
                phys_addr_t end = e820->map[i].addr;
                end = min(max_addr, end);
-                released += xen_release_chunk(last_end, end);
+                if (last_end < end)
-                last_end = e820->map[i].addr + e820->map[i].size;
+                        released += xen_release_chunk(last_end, end);
+                last_end = max(last_end, e820->map[i].addr + e820->map[i].size);
        }
        if (last_end < max_addr)
@@ -164,6 +165,7 @@ char * __init xen_memory_setup(void)
                XENMEM_memory_map;
        rc = HYPERVISOR_memory_op(op, &memmap);
        if (rc == -ENOSYS) {
+                BUG_ON(xen_initial_domain());
                memmap.nr_entries = 1;
                map[0].addr = 0ULL;
                map[0].size = mem_end;
@@ -201,12 +203,13 @@ char * __init xen_memory_setup(void)
        }
        /*
-         * Even though this is normal, usable memory under Xen, reserve
+         * In domU, the ISA region is normal, usable memory, but we
-         * ISA memory anyway because too many things think they can poke
+         * reserve ISA memory anyway because too many things poke
         * about in there.
         *
-         * In a dom0 kernel, this region is identity mapped with the
+         * In Dom0, the host E820 information can leave gaps in the
-         * hardware ISA area, so it really is out of bounds.
+         * ISA range, which would cause us to release those pages.  To
+         * avoid this, we unconditionally reserve them here.
         */
        e820_add_region(ISA_START_ADDRESS, ISA_END_ADDRESS - ISA_START_ADDRESS,
                        E820_RESERVED);
@@ -244,8 +247,7 @@ char * __init xen_memory_setup(void)
        else
                extra_pages = 0;
-        if (!xen_initial_domain())
+        xen_add_extra_mem(extra_pages);
-                xen_add_extra_mem(extra_pages);
        return "Xen";
 }
@@ -333,9 +335,6 @@ void __cpuinit xen_enable_syscall(void)
 void __init xen_arch_setup(void)
 {
-        struct physdev_set_iopl set_iopl;
-        int rc;
        xen_panic_handler_init();
        HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments);
@@ -352,11 +351,6 @@ void __init xen_arch_setup(void)
        xen_enable_sysenter();
        xen_enable_syscall();
-        set_iopl.iopl = 1;
-        rc = HYPERVISOR_physdev_op(PHYSDEVOP_set_iopl, &set_iopl);
-        if (rc != 0)
-                printk(KERN_INFO "physdev_op failed %d\n", rc);
 #ifdef CONFIG_ACPI
        if (!(xen_start_info->flags & SIF_INITDOMAIN)) {
                printk(KERN_INFO "ACPI in unprivileged domain disabled\n");