aboutsummaryrefslogtreecommitdiffstats
path: root/arch
diff options
context:
space:
mode:
authorAnton Altaparmakov <aia21@cantab.net>2005-10-17 10:00:34 -0400
committerAnton Altaparmakov <aia21@cantab.net>2005-10-17 10:00:34 -0400
commit7946ada30bc45546cefc85809ba0fd07879eff52 (patch)
tree03c382e8e4f9b81f151be953de3b7f61411c9303 /arch
parent98b270362bb9ea6629732e7f5b65b8a6ce4743c7 (diff)
parent3d80636a0d5f056ffc26472d05b6027a7a9f6e1c (diff)
Merge branch 'master' of /usr/src/ntfs-2.6/
Diffstat (limited to 'arch')
-rw-r--r--arch/arm/Makefile2
-rw-r--r--arch/arm/kernel/armksyms.c4
-rw-r--r--arch/arm/kernel/entry-common.S7
-rw-r--r--arch/arm/kernel/vmlinux.lds.S15
-rw-r--r--arch/arm/mach-l7200/core.c20
-rw-r--r--arch/arm/mach-pxa/corgi_lcd.c20
-rw-r--r--arch/arm/mach-pxa/generic.c5
-rw-r--r--arch/arm/mach-pxa/spitz.c4
-rw-r--r--arch/arm/mach-s3c2410/Kconfig1
-rw-r--r--arch/arm/mach-s3c2410/mach-anubis.c2
-rw-r--r--arch/arm/mach-s3c2410/mach-bast.c4
-rw-r--r--arch/arm/mach-s3c2410/mach-vr1000.c2
-rw-r--r--arch/arm/mach-s3c2410/s3c2410.c3
-rw-r--r--arch/arm/mach-s3c2410/s3c2440.c4
-rw-r--r--arch/arm/mach-s3c2410/time.c1
-rw-r--r--arch/arm/mm/alignment.c44
-rw-r--r--arch/arm/nwfpe/fpa11.c5
-rw-r--r--arch/arm/nwfpe/fpa11.h20
-rw-r--r--arch/arm/nwfpe/fpa11_cprt.c3
-rw-r--r--arch/arm/nwfpe/fpopcode.h6
-rw-r--r--arch/arm/nwfpe/softfloat.h3
-rw-r--r--arch/cris/arch-v32/kernel/smp.c2
-rw-r--r--arch/m32r/kernel/entry.S9
-rw-r--r--arch/m32r/kernel/smp.c12
-rw-r--r--arch/m32r/kernel/traps.c33
-rw-r--r--arch/mips/pci/fixup-tb0226.c33
-rw-r--r--arch/ppc/kernel/cputable.c5
-rw-r--r--arch/ppc/kernel/dma-mapping.c4
-rw-r--r--arch/ppc64/kernel/module.c13
-rw-r--r--arch/ppc64/kernel/pSeries_pci.c4
-rw-r--r--arch/ppc64/kernel/pmac_setup.c3
-rw-r--r--arch/sh/kernel/smp.c3
-rw-r--r--arch/sparc/Kconfig4
-rw-r--r--arch/sparc/defconfig1
-rw-r--r--arch/sparc64/kernel/dtlb_base.S14
-rw-r--r--arch/sparc64/kernel/dtlb_prot.S12
-rw-r--r--arch/sparc64/kernel/head.S69
-rw-r--r--arch/sparc64/kernel/itlb_base.S26
-rw-r--r--arch/sparc64/kernel/ktlb.S92
-rw-r--r--arch/sparc64/kernel/pci_iommu.c363
-rw-r--r--arch/sparc64/kernel/pci_psycho.c44
-rw-r--r--arch/sparc64/kernel/pci_sabre.c39
-rw-r--r--arch/sparc64/kernel/pci_schizo.c57
-rw-r--r--arch/sparc64/kernel/smp.c7
-rw-r--r--arch/sparc64/mm/init.c187
-rw-r--r--arch/sparc64/mm/ultra.S16
-rw-r--r--arch/sparc64/prom/misc.c12
-rw-r--r--arch/um/drivers/Makefile2
-rw-r--r--arch/um/drivers/ubd_kern.c556
-rw-r--r--arch/um/drivers/ubd_user.c75
-rw-r--r--arch/um/include/aio.h18
-rw-r--r--arch/um/include/os.h5
-rw-r--r--arch/um/os-Linux/aio.c205
53 files changed, 1020 insertions, 1080 deletions
diff --git a/arch/arm/Makefile b/arch/arm/Makefile
index 7779f2d1acad..299bc0468702 100644
--- a/arch/arm/Makefile
+++ b/arch/arm/Makefile
@@ -53,7 +53,7 @@ tune-$(CONFIG_CPU_ARM926T) :=-mtune=arm9tdmi
53tune-$(CONFIG_CPU_SA110) :=-mtune=strongarm110 53tune-$(CONFIG_CPU_SA110) :=-mtune=strongarm110
54tune-$(CONFIG_CPU_SA1100) :=-mtune=strongarm1100 54tune-$(CONFIG_CPU_SA1100) :=-mtune=strongarm1100
55tune-$(CONFIG_CPU_XSCALE) :=$(call cc-option,-mtune=xscale,-mtune=strongarm110) -Wa,-mcpu=xscale 55tune-$(CONFIG_CPU_XSCALE) :=$(call cc-option,-mtune=xscale,-mtune=strongarm110) -Wa,-mcpu=xscale
56tune-$(CONFIG_CPU_V6) :=-mtune=strongarm 56tune-$(CONFIG_CPU_V6) :=$(call cc-option,-mtune=arm1136j-s,-mtune=strongarm)
57 57
58# Need -Uarm for gcc < 3.x 58# Need -Uarm for gcc < 3.x
59CFLAGS_ABI :=$(call cc-option,-mapcs-32,-mabi=apcs-gnu) $(call cc-option,-mno-thumb-interwork,) 59CFLAGS_ABI :=$(call cc-option,-mapcs-32,-mabi=apcs-gnu) $(call cc-option,-mno-thumb-interwork,)
diff --git a/arch/arm/kernel/armksyms.c b/arch/arm/kernel/armksyms.c
index 835d450797a1..7b17a87a3311 100644
--- a/arch/arm/kernel/armksyms.c
+++ b/arch/arm/kernel/armksyms.c
@@ -45,8 +45,8 @@ extern void fp_enter(void);
45 45
46#define EXPORT_SYMBOL_ALIAS(sym,orig) \ 46#define EXPORT_SYMBOL_ALIAS(sym,orig) \
47 EXPORT_CRC_ALIAS(sym) \ 47 EXPORT_CRC_ALIAS(sym) \
48 const struct kernel_symbol __ksymtab_##sym \ 48 static const struct kernel_symbol __ksymtab_##sym \
49 __attribute__((section("__ksymtab"))) = \ 49 __attribute_used__ __attribute__((section("__ksymtab"))) = \
50 { (unsigned long)&orig, #sym }; 50 { (unsigned long)&orig, #sym };
51 51
52/* 52/*
diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S
index 81d450ac3fab..066597f4345a 100644
--- a/arch/arm/kernel/entry-common.S
+++ b/arch/arm/kernel/entry-common.S
@@ -106,15 +106,10 @@ ENTRY(ret_from_fork)
106 .endm 106 .endm
107 107
108.Larm700bug: 108.Larm700bug:
109 ldr r0, [sp, #S_PSR] @ Get calling cpsr
110 sub lr, lr, #4
111 str lr, [r8]
112 msr spsr_cxsf, r0
113 ldmia sp, {r0 - lr}^ @ Get calling r0 - lr 109 ldmia sp, {r0 - lr}^ @ Get calling r0 - lr
114 mov r0, r0 110 mov r0, r0
115 ldr lr, [sp, #S_PC] @ Get PC
116 add sp, sp, #S_FRAME_SIZE 111 add sp, sp, #S_FRAME_SIZE
117 movs pc, lr 112 subs pc, lr, #4
118#else 113#else
119 .macro arm710_bug_check, instr, temp 114 .macro arm710_bug_check, instr, temp
120 .endm 115 .endm
diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index 08e58ecd44be..0d5db5279c5c 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -89,13 +89,6 @@ SECTIONS
89 *(.got) /* Global offset table */ 89 *(.got) /* Global offset table */
90 } 90 }
91 91
92 . = ALIGN(16);
93 __ex_table : { /* Exception table */
94 __start___ex_table = .;
95 *(__ex_table)
96 __stop___ex_table = .;
97 }
98
99 RODATA 92 RODATA
100 93
101 _etext = .; /* End of text and rodata section */ 94 _etext = .; /* End of text and rodata section */
@@ -138,6 +131,14 @@ SECTIONS
138 *(.data.cacheline_aligned) 131 *(.data.cacheline_aligned)
139 132
140 /* 133 /*
134 * The exception fixup table (might need resorting at runtime)
135 */
136 . = ALIGN(32);
137 __start___ex_table = .;
138 *(__ex_table)
139 __stop___ex_table = .;
140
141 /*
141 * and the usual data section 142 * and the usual data section
142 */ 143 */
143 *(.data) 144 *(.data)
diff --git a/arch/arm/mach-l7200/core.c b/arch/arm/mach-l7200/core.c
index 5fd8c9f97f9a..03ed742ae2be 100644
--- a/arch/arm/mach-l7200/core.c
+++ b/arch/arm/mach-l7200/core.c
@@ -7,11 +7,17 @@
7 */ 7 */
8#include <linux/kernel.h> 8#include <linux/kernel.h>
9#include <linux/init.h> 9#include <linux/init.h>
10#include <linux/device.h>
10 11
12#include <asm/types.h>
13#include <asm/irq.h>
14#include <asm/mach-types.h>
11#include <asm/hardware.h> 15#include <asm/hardware.h>
12#include <asm/page.h> 16#include <asm/page.h>
13 17
18#include <asm/mach/arch.h>
14#include <asm/mach/map.h> 19#include <asm/mach/map.h>
20#include <asm/mach/irq.h>
15 21
16/* 22/*
17 * IRQ base register 23 * IRQ base register
@@ -47,6 +53,12 @@ static void l7200_unmask_irq(unsigned int irq)
47{ 53{
48 IRQ_ENABLE = 1 << irq; 54 IRQ_ENABLE = 1 << irq;
49} 55}
56
57static struct irqchip l7200_irq_chip = {
58 .ack = l7200_mask_irq,
59 .mask = l7200_mask_irq,
60 .unmask = l7200_unmask_irq
61};
50 62
51static void __init l7200_init_irq(void) 63static void __init l7200_init_irq(void)
52{ 64{
@@ -56,11 +68,9 @@ static void __init l7200_init_irq(void)
56 FIQ_ENABLECLEAR = 0xffffffff; /* clear all fast interrupt enables */ 68 FIQ_ENABLECLEAR = 0xffffffff; /* clear all fast interrupt enables */
57 69
58 for (irq = 0; irq < NR_IRQS; irq++) { 70 for (irq = 0; irq < NR_IRQS; irq++) {
59 irq_desc[irq].valid = 1; 71 set_irq_chip(irq, &l7200_irq_chip);
60 irq_desc[irq].probe_ok = 1; 72 set_irq_flags(irq, IRQF_VALID);
61 irq_desc[irq].mask_ack = l7200_mask_irq; 73 set_irq_handler(irq, do_level_IRQ);
62 irq_desc[irq].mask = l7200_mask_irq;
63 irq_desc[irq].unmask = l7200_unmask_irq;
64 } 74 }
65 75
66 init_FIQ(); 76 init_FIQ();
diff --git a/arch/arm/mach-pxa/corgi_lcd.c b/arch/arm/mach-pxa/corgi_lcd.c
index c02ef7c0f7ef..850538fadece 100644
--- a/arch/arm/mach-pxa/corgi_lcd.c
+++ b/arch/arm/mach-pxa/corgi_lcd.c
@@ -467,6 +467,7 @@ void corgi_put_hsync(void)
467{ 467{
468 if (get_hsync_time) 468 if (get_hsync_time)
469 symbol_put(w100fb_get_hsynclen); 469 symbol_put(w100fb_get_hsynclen);
470 get_hsync_time = NULL;
470} 471}
471 472
472void corgi_wait_hsync(void) 473void corgi_wait_hsync(void)
@@ -476,20 +477,37 @@ void corgi_wait_hsync(void)
476#endif 477#endif
477 478
478#ifdef CONFIG_PXA_SHARP_Cxx00 479#ifdef CONFIG_PXA_SHARP_Cxx00
480static struct device *spitz_pxafb_dev;
481
482static int is_pxafb_device(struct device * dev, void * data)
483{
484 struct platform_device *pdev = container_of(dev, struct platform_device, dev);
485
486 return (strncmp(pdev->name, "pxa2xx-fb", 9) == 0);
487}
488
479unsigned long spitz_get_hsync_len(void) 489unsigned long spitz_get_hsync_len(void)
480{ 490{
491 if (!spitz_pxafb_dev) {
492 spitz_pxafb_dev = bus_find_device(&platform_bus_type, NULL, NULL, is_pxafb_device);
493 if (!spitz_pxafb_dev)
494 return 0;
495 }
481 if (!get_hsync_time) 496 if (!get_hsync_time)
482 get_hsync_time = symbol_get(pxafb_get_hsync_time); 497 get_hsync_time = symbol_get(pxafb_get_hsync_time);
483 if (!get_hsync_time) 498 if (!get_hsync_time)
484 return 0; 499 return 0;
485 500
486 return pxafb_get_hsync_time(&pxafb_device.dev); 501 return pxafb_get_hsync_time(spitz_pxafb_dev);
487} 502}
488 503
489void spitz_put_hsync(void) 504void spitz_put_hsync(void)
490{ 505{
506 put_device(spitz_pxafb_dev);
491 if (get_hsync_time) 507 if (get_hsync_time)
492 symbol_put(pxafb_get_hsync_time); 508 symbol_put(pxafb_get_hsync_time);
509 spitz_pxafb_dev = NULL;
510 get_hsync_time = NULL;
493} 511}
494 512
495void spitz_wait_hsync(void) 513void spitz_wait_hsync(void)
diff --git a/arch/arm/mach-pxa/generic.c b/arch/arm/mach-pxa/generic.c
index d0660a8c4b70..d327c127eddb 100644
--- a/arch/arm/mach-pxa/generic.c
+++ b/arch/arm/mach-pxa/generic.c
@@ -208,6 +208,11 @@ static struct platform_device pxafb_device = {
208 .resource = pxafb_resources, 208 .resource = pxafb_resources,
209}; 209};
210 210
211void __init set_pxa_fb_parent(struct device *parent_dev)
212{
213 pxafb_device.dev.parent = parent_dev;
214}
215
211static struct platform_device ffuart_device = { 216static struct platform_device ffuart_device = {
212 .name = "pxa2xx-uart", 217 .name = "pxa2xx-uart",
213 .id = 0, 218 .id = 0,
diff --git a/arch/arm/mach-pxa/spitz.c b/arch/arm/mach-pxa/spitz.c
index 568afe3d6e1a..d0ab428c2d7d 100644
--- a/arch/arm/mach-pxa/spitz.c
+++ b/arch/arm/mach-pxa/spitz.c
@@ -36,7 +36,6 @@
36#include <asm/arch/irq.h> 36#include <asm/arch/irq.h>
37#include <asm/arch/mmc.h> 37#include <asm/arch/mmc.h>
38#include <asm/arch/udc.h> 38#include <asm/arch/udc.h>
39#include <asm/arch/ohci.h>
40#include <asm/arch/pxafb.h> 39#include <asm/arch/pxafb.h>
41#include <asm/arch/akita.h> 40#include <asm/arch/akita.h>
42#include <asm/arch/spitz.h> 41#include <asm/arch/spitz.h>
@@ -304,7 +303,6 @@ static struct platform_device *devices[] __initdata = {
304 &spitzkbd_device, 303 &spitzkbd_device,
305 &spitzts_device, 304 &spitzts_device,
306 &spitzbl_device, 305 &spitzbl_device,
307 &spitzbattery_device,
308}; 306};
309 307
310static void __init common_init(void) 308static void __init common_init(void)
@@ -328,7 +326,7 @@ static void __init common_init(void)
328 326
329 platform_add_devices(devices, ARRAY_SIZE(devices)); 327 platform_add_devices(devices, ARRAY_SIZE(devices));
330 pxa_set_mci_info(&spitz_mci_platform_data); 328 pxa_set_mci_info(&spitz_mci_platform_data);
331 pxafb_device.dev.parent = &spitzssp_device.dev; 329 set_pxa_fb_parent(&spitzssp_device.dev);
332 set_pxa_fb_info(&spitz_pxafb_info); 330 set_pxa_fb_info(&spitz_pxafb_info);
333} 331}
334 332
diff --git a/arch/arm/mach-s3c2410/Kconfig b/arch/arm/mach-s3c2410/Kconfig
index 06807c6ee68a..c796bcdd6158 100644
--- a/arch/arm/mach-s3c2410/Kconfig
+++ b/arch/arm/mach-s3c2410/Kconfig
@@ -12,6 +12,7 @@ config MACH_ANUBIS
12config ARCH_BAST 12config ARCH_BAST
13 bool "Simtec Electronics BAST (EB2410ITX)" 13 bool "Simtec Electronics BAST (EB2410ITX)"
14 select CPU_S3C2410 14 select CPU_S3C2410
15 select ISA
15 help 16 help
16 Say Y here if you are using the Simtec Electronics EB2410ITX 17 Say Y here if you are using the Simtec Electronics EB2410ITX
17 development board (also known as BAST) 18 development board (also known as BAST)
diff --git a/arch/arm/mach-s3c2410/mach-anubis.c b/arch/arm/mach-s3c2410/mach-anubis.c
index 7c05f27fe1d6..5ae80f4e3e67 100644
--- a/arch/arm/mach-s3c2410/mach-anubis.c
+++ b/arch/arm/mach-s3c2410/mach-anubis.c
@@ -125,7 +125,7 @@ static int external_map[] = { 2 };
125static int chip0_map[] = { 0 }; 125static int chip0_map[] = { 0 };
126static int chip1_map[] = { 1 }; 126static int chip1_map[] = { 1 };
127 127
128struct mtd_partition anubis_default_nand_part[] = { 128static struct mtd_partition anubis_default_nand_part[] = {
129 [0] = { 129 [0] = {
130 .name = "Boot Agent", 130 .name = "Boot Agent",
131 .size = SZ_16K, 131 .size = SZ_16K,
diff --git a/arch/arm/mach-s3c2410/mach-bast.c b/arch/arm/mach-s3c2410/mach-bast.c
index ed1f07d7252f..8ca955984645 100644
--- a/arch/arm/mach-s3c2410/mach-bast.c
+++ b/arch/arm/mach-s3c2410/mach-bast.c
@@ -230,7 +230,7 @@ static int chip0_map[] = { 1 };
230static int chip1_map[] = { 2 }; 230static int chip1_map[] = { 2 };
231static int chip2_map[] = { 3 }; 231static int chip2_map[] = { 3 };
232 232
233struct mtd_partition bast_default_nand_part[] = { 233static struct mtd_partition bast_default_nand_part[] = {
234 [0] = { 234 [0] = {
235 .name = "Boot Agent", 235 .name = "Boot Agent",
236 .size = SZ_16K, 236 .size = SZ_16K,
@@ -340,7 +340,7 @@ static struct resource bast_dm9k_resource[] = {
340 * better IO routines can be written and tested 340 * better IO routines can be written and tested
341*/ 341*/
342 342
343struct dm9000_plat_data bast_dm9k_platdata = { 343static struct dm9000_plat_data bast_dm9k_platdata = {
344 .flags = DM9000_PLATF_16BITONLY 344 .flags = DM9000_PLATF_16BITONLY
345}; 345};
346 346
diff --git a/arch/arm/mach-s3c2410/mach-vr1000.c b/arch/arm/mach-s3c2410/mach-vr1000.c
index 663a7f98fc0b..46b259673c18 100644
--- a/arch/arm/mach-s3c2410/mach-vr1000.c
+++ b/arch/arm/mach-s3c2410/mach-vr1000.c
@@ -288,7 +288,7 @@ static struct resource vr1000_dm9k1_resource[] = {
288 * better IO routines can be written and tested 288 * better IO routines can be written and tested
289*/ 289*/
290 290
291struct dm9000_plat_data vr1000_dm9k_platdata = { 291static struct dm9000_plat_data vr1000_dm9k_platdata = {
292 .flags = DM9000_PLATF_16BITONLY, 292 .flags = DM9000_PLATF_16BITONLY,
293}; 293};
294 294
diff --git a/arch/arm/mach-s3c2410/s3c2410.c b/arch/arm/mach-s3c2410/s3c2410.c
index 0b88993dfd27..a8bf5ec82602 100644
--- a/arch/arm/mach-s3c2410/s3c2410.c
+++ b/arch/arm/mach-s3c2410/s3c2410.c
@@ -125,9 +125,6 @@ static struct platform_device *uart_devices[] __initdata = {
125 &s3c_uart2 125 &s3c_uart2
126}; 126};
127 127
128/* store our uart devices for the serial driver console */
129struct platform_device *s3c2410_uart_devices[3];
130
131static int s3c2410_uart_count = 0; 128static int s3c2410_uart_count = 0;
132 129
133/* uart registration process */ 130/* uart registration process */
diff --git a/arch/arm/mach-s3c2410/s3c2440.c b/arch/arm/mach-s3c2410/s3c2440.c
index d4c8281b55f6..833fa36bce05 100644
--- a/arch/arm/mach-s3c2410/s3c2440.c
+++ b/arch/arm/mach-s3c2410/s3c2440.c
@@ -151,7 +151,7 @@ void __init s3c2440_init_uarts(struct s3c2410_uartcfg *cfg, int no)
151 151
152#ifdef CONFIG_PM 152#ifdef CONFIG_PM
153 153
154struct sleep_save s3c2440_sleep[] = { 154static struct sleep_save s3c2440_sleep[] = {
155 SAVE_ITEM(S3C2440_DSC0), 155 SAVE_ITEM(S3C2440_DSC0),
156 SAVE_ITEM(S3C2440_DSC1), 156 SAVE_ITEM(S3C2440_DSC1),
157 SAVE_ITEM(S3C2440_GPJDAT), 157 SAVE_ITEM(S3C2440_GPJDAT),
@@ -260,7 +260,7 @@ void __init s3c2440_init_clocks(int xtal)
260 * as a driver which may support both 2410 and 2440 may try and use it. 260 * as a driver which may support both 2410 and 2440 may try and use it.
261*/ 261*/
262 262
263int __init s3c2440_core_init(void) 263static int __init s3c2440_core_init(void)
264{ 264{
265 return sysdev_class_register(&s3c2440_sysclass); 265 return sysdev_class_register(&s3c2440_sysclass);
266} 266}
diff --git a/arch/arm/mach-s3c2410/time.c b/arch/arm/mach-s3c2410/time.c
index c0acfb2ad790..8a00e3c3cd08 100644
--- a/arch/arm/mach-s3c2410/time.c
+++ b/arch/arm/mach-s3c2410/time.c
@@ -38,6 +38,7 @@
38#include <asm/hardware/clock.h> 38#include <asm/hardware/clock.h>
39 39
40#include "clock.h" 40#include "clock.h"
41#include "cpu.h"
41 42
42static unsigned long timer_startval; 43static unsigned long timer_startval;
43static unsigned long timer_usec_ticks; 44static unsigned long timer_usec_ticks;
diff --git a/arch/arm/mm/alignment.c b/arch/arm/mm/alignment.c
index f35e69e9c65c..705c98921c37 100644
--- a/arch/arm/mm/alignment.c
+++ b/arch/arm/mm/alignment.c
@@ -111,7 +111,7 @@ proc_alignment_read(char *page, char **start, off_t off, int count, int *eof,
111} 111}
112 112
113static int proc_alignment_write(struct file *file, const char __user *buffer, 113static int proc_alignment_write(struct file *file, const char __user *buffer,
114 unsigned long count, void *data) 114 unsigned long count, void *data)
115{ 115{
116 char mode; 116 char mode;
117 117
@@ -119,7 +119,7 @@ static int proc_alignment_write(struct file *file, const char __user *buffer,
119 if (get_user(mode, buffer)) 119 if (get_user(mode, buffer))
120 return -EFAULT; 120 return -EFAULT;
121 if (mode >= '0' && mode <= '5') 121 if (mode >= '0' && mode <= '5')
122 ai_usermode = mode - '0'; 122 ai_usermode = mode - '0';
123 } 123 }
124 return count; 124 return count;
125} 125}
@@ -262,7 +262,7 @@ union offset_union {
262 goto fault; \ 262 goto fault; \
263 } while (0) 263 } while (0)
264 264
265#define put32_unaligned_check(val,addr) \ 265#define put32_unaligned_check(val,addr) \
266 __put32_unaligned_check("strb", val, addr) 266 __put32_unaligned_check("strb", val, addr)
267 267
268#define put32t_unaligned_check(val,addr) \ 268#define put32t_unaligned_check(val,addr) \
@@ -306,19 +306,19 @@ do_alignment_ldrhstrh(unsigned long addr, unsigned long instr, struct pt_regs *r
306 return TYPE_LDST; 306 return TYPE_LDST;
307 307
308 user: 308 user:
309 if (LDST_L_BIT(instr)) { 309 if (LDST_L_BIT(instr)) {
310 unsigned long val; 310 unsigned long val;
311 get16t_unaligned_check(val, addr); 311 get16t_unaligned_check(val, addr);
312 312
313 /* signed half-word? */ 313 /* signed half-word? */
314 if (instr & 0x40) 314 if (instr & 0x40)
315 val = (signed long)((signed short) val); 315 val = (signed long)((signed short) val);
316 316
317 regs->uregs[rd] = val; 317 regs->uregs[rd] = val;
318 } else 318 } else
319 put16t_unaligned_check(regs->uregs[rd], addr); 319 put16t_unaligned_check(regs->uregs[rd], addr);
320 320
321 return TYPE_LDST; 321 return TYPE_LDST;
322 322
323 fault: 323 fault:
324 return TYPE_FAULT; 324 return TYPE_FAULT;
@@ -342,11 +342,11 @@ do_alignment_ldrdstrd(unsigned long addr, unsigned long instr,
342 unsigned long val; 342 unsigned long val;
343 get32_unaligned_check(val, addr); 343 get32_unaligned_check(val, addr);
344 regs->uregs[rd] = val; 344 regs->uregs[rd] = val;
345 get32_unaligned_check(val, addr+4); 345 get32_unaligned_check(val, addr + 4);
346 regs->uregs[rd+1] = val; 346 regs->uregs[rd + 1] = val;
347 } else { 347 } else {
348 put32_unaligned_check(regs->uregs[rd], addr); 348 put32_unaligned_check(regs->uregs[rd], addr);
349 put32_unaligned_check(regs->uregs[rd+1], addr+4); 349 put32_unaligned_check(regs->uregs[rd + 1], addr + 4);
350 } 350 }
351 351
352 return TYPE_LDST; 352 return TYPE_LDST;
@@ -356,11 +356,11 @@ do_alignment_ldrdstrd(unsigned long addr, unsigned long instr,
356 unsigned long val; 356 unsigned long val;
357 get32t_unaligned_check(val, addr); 357 get32t_unaligned_check(val, addr);
358 regs->uregs[rd] = val; 358 regs->uregs[rd] = val;
359 get32t_unaligned_check(val, addr+4); 359 get32t_unaligned_check(val, addr + 4);
360 regs->uregs[rd+1] = val; 360 regs->uregs[rd + 1] = val;
361 } else { 361 } else {
362 put32t_unaligned_check(regs->uregs[rd], addr); 362 put32t_unaligned_check(regs->uregs[rd], addr);
363 put32t_unaligned_check(regs->uregs[rd+1], addr+4); 363 put32t_unaligned_check(regs->uregs[rd + 1], addr + 4);
364 } 364 }
365 365
366 return TYPE_LDST; 366 return TYPE_LDST;
@@ -443,7 +443,7 @@ do_alignment_ldmstm(unsigned long addr, unsigned long instr, struct pt_regs *reg
443 if (LDST_P_EQ_U(instr)) /* U = P */ 443 if (LDST_P_EQ_U(instr)) /* U = P */
444 eaddr += 4; 444 eaddr += 4;
445 445
446 /* 446 /*
447 * For alignment faults on the ARM922T/ARM920T the MMU makes 447 * For alignment faults on the ARM922T/ARM920T the MMU makes
448 * the FSR (and hence addr) equal to the updated base address 448 * the FSR (and hence addr) equal to the updated base address
449 * of the multiple access rather than the restored value. 449 * of the multiple access rather than the restored value.
@@ -570,7 +570,7 @@ thumb2arm(u16 tinstr)
570 /* 6.5.1 Format 3: */ 570 /* 6.5.1 Format 3: */
571 case 0x4800 >> 11: /* 7.1.28 LDR(3) */ 571 case 0x4800 >> 11: /* 7.1.28 LDR(3) */
572 /* NOTE: This case is not technically possible. We're 572 /* NOTE: This case is not technically possible. We're
573 * loading 32-bit memory data via PC relative 573 * loading 32-bit memory data via PC relative
574 * addressing mode. So we can and should eliminate 574 * addressing mode. So we can and should eliminate
575 * this case. But I'll leave it here for now. 575 * this case. But I'll leave it here for now.
576 */ 576 */
@@ -642,7 +642,7 @@ do_alignment(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
642 642
643 if (fault) { 643 if (fault) {
644 type = TYPE_FAULT; 644 type = TYPE_FAULT;
645 goto bad_or_fault; 645 goto bad_or_fault;
646 } 646 }
647 647
648 if (user_mode(regs)) 648 if (user_mode(regs))
diff --git a/arch/arm/nwfpe/fpa11.c b/arch/arm/nwfpe/fpa11.c
index 7690f731ee87..7b3d74d73c80 100644
--- a/arch/arm/nwfpe/fpa11.c
+++ b/arch/arm/nwfpe/fpa11.c
@@ -31,11 +31,6 @@
31#include <linux/string.h> 31#include <linux/string.h>
32#include <asm/system.h> 32#include <asm/system.h>
33 33
34/* forward declarations */
35unsigned int EmulateCPDO(const unsigned int);
36unsigned int EmulateCPDT(const unsigned int);
37unsigned int EmulateCPRT(const unsigned int);
38
39/* Reset the FPA11 chip. Called to initialize and reset the emulator. */ 34/* Reset the FPA11 chip. Called to initialize and reset the emulator. */
40static void resetFPA11(void) 35static void resetFPA11(void)
41{ 36{
diff --git a/arch/arm/nwfpe/fpa11.h b/arch/arm/nwfpe/fpa11.h
index 93523ae4b7a1..9677ae8448e8 100644
--- a/arch/arm/nwfpe/fpa11.h
+++ b/arch/arm/nwfpe/fpa11.h
@@ -95,4 +95,24 @@ extern int8 SetRoundingMode(const unsigned int);
95extern int8 SetRoundingPrecision(const unsigned int); 95extern int8 SetRoundingPrecision(const unsigned int);
96extern void nwfpe_init_fpa(union fp_state *fp); 96extern void nwfpe_init_fpa(union fp_state *fp);
97 97
98extern unsigned int EmulateAll(unsigned int opcode);
99
100extern unsigned int EmulateCPDT(const unsigned int opcode);
101extern unsigned int EmulateCPDO(const unsigned int opcode);
102extern unsigned int EmulateCPRT(const unsigned int opcode);
103
104/* fpa11_cpdt.c */
105extern unsigned int PerformLDF(const unsigned int opcode);
106extern unsigned int PerformSTF(const unsigned int opcode);
107extern unsigned int PerformLFM(const unsigned int opcode);
108extern unsigned int PerformSFM(const unsigned int opcode);
109
110/* single_cpdo.c */
111
112extern unsigned int SingleCPDO(struct roundingData *roundData,
113 const unsigned int opcode, FPREG * rFd);
114/* double_cpdo.c */
115extern unsigned int DoubleCPDO(struct roundingData *roundData,
116 const unsigned int opcode, FPREG * rFd);
117
98#endif 118#endif
diff --git a/arch/arm/nwfpe/fpa11_cprt.c b/arch/arm/nwfpe/fpa11_cprt.c
index adf8d3000540..7c67023655e4 100644
--- a/arch/arm/nwfpe/fpa11_cprt.c
+++ b/arch/arm/nwfpe/fpa11_cprt.c
@@ -26,12 +26,11 @@
26#include "fpa11.inl" 26#include "fpa11.inl"
27#include "fpmodule.h" 27#include "fpmodule.h"
28#include "fpmodule.inl" 28#include "fpmodule.inl"
29#include "softfloat.h"
29 30
30#ifdef CONFIG_FPE_NWFPE_XP 31#ifdef CONFIG_FPE_NWFPE_XP
31extern flag floatx80_is_nan(floatx80); 32extern flag floatx80_is_nan(floatx80);
32#endif 33#endif
33extern flag float64_is_nan(float64);
34extern flag float32_is_nan(float32);
35 34
36unsigned int PerformFLT(const unsigned int opcode); 35unsigned int PerformFLT(const unsigned int opcode);
37unsigned int PerformFIX(const unsigned int opcode); 36unsigned int PerformFIX(const unsigned int opcode);
diff --git a/arch/arm/nwfpe/fpopcode.h b/arch/arm/nwfpe/fpopcode.h
index 1777e92a88e6..6528e081c83f 100644
--- a/arch/arm/nwfpe/fpopcode.h
+++ b/arch/arm/nwfpe/fpopcode.h
@@ -476,4 +476,10 @@ static inline unsigned int getDestinationSize(const unsigned int opcode)
476 return (nRc); 476 return (nRc);
477} 477}
478 478
479extern unsigned int checkCondition(const unsigned int opcode,
480 const unsigned int ccodes);
481
482extern const float64 float64Constant[];
483extern const float32 float32Constant[];
484
479#endif 485#endif
diff --git a/arch/arm/nwfpe/softfloat.h b/arch/arm/nwfpe/softfloat.h
index 1c8799b9ee4d..14151700b6b2 100644
--- a/arch/arm/nwfpe/softfloat.h
+++ b/arch/arm/nwfpe/softfloat.h
@@ -265,4 +265,7 @@ static inline flag float64_lt_nocheck(float64 a, float64 b)
265 return (a != b) && (aSign ^ (a < b)); 265 return (a != b) && (aSign ^ (a < b));
266} 266}
267 267
268extern flag float32_is_nan( float32 a );
269extern flag float64_is_nan( float64 a );
270
268#endif 271#endif
diff --git a/arch/cris/arch-v32/kernel/smp.c b/arch/cris/arch-v32/kernel/smp.c
index 2c5cae04a95c..957f551ba5ce 100644
--- a/arch/cris/arch-v32/kernel/smp.c
+++ b/arch/cris/arch-v32/kernel/smp.c
@@ -15,6 +15,7 @@
15#include <linux/kernel.h> 15#include <linux/kernel.h>
16#include <linux/cpumask.h> 16#include <linux/cpumask.h>
17#include <linux/interrupt.h> 17#include <linux/interrupt.h>
18#include <linux/module.h>
18 19
19#define IPI_SCHEDULE 1 20#define IPI_SCHEDULE 1
20#define IPI_CALL 2 21#define IPI_CALL 2
@@ -28,6 +29,7 @@ spinlock_t cris_atomic_locks[] = { [0 ... LOCK_COUNT - 1] = SPIN_LOCK_UNLOCKED};
28/* CPU masks */ 29/* CPU masks */
29cpumask_t cpu_online_map = CPU_MASK_NONE; 30cpumask_t cpu_online_map = CPU_MASK_NONE;
30cpumask_t phys_cpu_present_map = CPU_MASK_NONE; 31cpumask_t phys_cpu_present_map = CPU_MASK_NONE;
32EXPORT_SYMBOL(phys_cpu_present_map);
31 33
32/* Variables used during SMP boot */ 34/* Variables used during SMP boot */
33volatile int cpu_now_booting = 0; 35volatile int cpu_now_booting = 0;
diff --git a/arch/m32r/kernel/entry.S b/arch/m32r/kernel/entry.S
index dddbf6b5ed2c..85920fb8d08c 100644
--- a/arch/m32r/kernel/entry.S
+++ b/arch/m32r/kernel/entry.S
@@ -681,6 +681,15 @@ ENTRY(debug_trap)
681 bl do_debug_trap 681 bl do_debug_trap
682 bra error_code 682 bra error_code
683 683
684ENTRY(ill_trap)
685 /* void ill_trap(void) */
686 SWITCH_TO_KERNEL_STACK
687 SAVE_ALL
688 ldi r1, #0 ; error_code ; FIXME
689 mv r0, sp ; pt_regs
690 bl do_ill_trap
691 bra error_code
692
684 693
685/* Cache flushing handler */ 694/* Cache flushing handler */
686ENTRY(cache_flushing_handler) 695ENTRY(cache_flushing_handler)
diff --git a/arch/m32r/kernel/smp.c b/arch/m32r/kernel/smp.c
index a4576ac7e870..8b1f6eb76870 100644
--- a/arch/m32r/kernel/smp.c
+++ b/arch/m32r/kernel/smp.c
@@ -275,12 +275,14 @@ static void flush_tlb_all_ipi(void *info)
275 *==========================================================================*/ 275 *==========================================================================*/
276void smp_flush_tlb_mm(struct mm_struct *mm) 276void smp_flush_tlb_mm(struct mm_struct *mm)
277{ 277{
278 int cpu_id = smp_processor_id(); 278 int cpu_id;
279 cpumask_t cpu_mask; 279 cpumask_t cpu_mask;
280 unsigned long *mmc = &mm->context[cpu_id]; 280 unsigned long *mmc;
281 unsigned long flags; 281 unsigned long flags;
282 282
283 preempt_disable(); 283 preempt_disable();
284 cpu_id = smp_processor_id();
285 mmc = &mm->context[cpu_id];
284 cpu_mask = mm->cpu_vm_mask; 286 cpu_mask = mm->cpu_vm_mask;
285 cpu_clear(cpu_id, cpu_mask); 287 cpu_clear(cpu_id, cpu_mask);
286 288
@@ -343,12 +345,14 @@ void smp_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
343void smp_flush_tlb_page(struct vm_area_struct *vma, unsigned long va) 345void smp_flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
344{ 346{
345 struct mm_struct *mm = vma->vm_mm; 347 struct mm_struct *mm = vma->vm_mm;
346 int cpu_id = smp_processor_id(); 348 int cpu_id;
347 cpumask_t cpu_mask; 349 cpumask_t cpu_mask;
348 unsigned long *mmc = &mm->context[cpu_id]; 350 unsigned long *mmc;
349 unsigned long flags; 351 unsigned long flags;
350 352
351 preempt_disable(); 353 preempt_disable();
354 cpu_id = smp_processor_id();
355 mmc = &mm->context[cpu_id];
352 cpu_mask = mm->cpu_vm_mask; 356 cpu_mask = mm->cpu_vm_mask;
353 cpu_clear(cpu_id, cpu_mask); 357 cpu_clear(cpu_id, cpu_mask);
354 358
diff --git a/arch/m32r/kernel/traps.c b/arch/m32r/kernel/traps.c
index 01922271d17e..5fe8ed6d62dc 100644
--- a/arch/m32r/kernel/traps.c
+++ b/arch/m32r/kernel/traps.c
@@ -5,8 +5,6 @@
5 * Hitoshi Yamamoto 5 * Hitoshi Yamamoto
6 */ 6 */
7 7
8/* $Id$ */
9
10/* 8/*
11 * 'traps.c' handles hardware traps and faults after we have saved some 9 * 'traps.c' handles hardware traps and faults after we have saved some
12 * state in 'entry.S'. 10 * state in 'entry.S'.
@@ -35,6 +33,7 @@ asmlinkage void ei_handler(void);
35asmlinkage void rie_handler(void); 33asmlinkage void rie_handler(void);
36asmlinkage void debug_trap(void); 34asmlinkage void debug_trap(void);
37asmlinkage void cache_flushing_handler(void); 35asmlinkage void cache_flushing_handler(void);
36asmlinkage void ill_trap(void);
38 37
39#ifdef CONFIG_SMP 38#ifdef CONFIG_SMP
40extern void smp_reschedule_interrupt(void); 39extern void smp_reschedule_interrupt(void);
@@ -77,22 +76,22 @@ void set_eit_vector_entries(void)
77 eit_vector[5] = BRA_INSN(default_eit_handler, 5); 76 eit_vector[5] = BRA_INSN(default_eit_handler, 5);
78 eit_vector[8] = BRA_INSN(rie_handler, 8); 77 eit_vector[8] = BRA_INSN(rie_handler, 8);
79 eit_vector[12] = BRA_INSN(alignment_check, 12); 78 eit_vector[12] = BRA_INSN(alignment_check, 12);
80 eit_vector[16] = 0xff000000UL; 79 eit_vector[16] = BRA_INSN(ill_trap, 16);
81 eit_vector[17] = BRA_INSN(debug_trap, 17); 80 eit_vector[17] = BRA_INSN(debug_trap, 17);
82 eit_vector[18] = BRA_INSN(system_call, 18); 81 eit_vector[18] = BRA_INSN(system_call, 18);
83 eit_vector[19] = 0xff000000UL; 82 eit_vector[19] = BRA_INSN(ill_trap, 19);
84 eit_vector[20] = 0xff000000UL; 83 eit_vector[20] = BRA_INSN(ill_trap, 20);
85 eit_vector[21] = 0xff000000UL; 84 eit_vector[21] = BRA_INSN(ill_trap, 21);
86 eit_vector[22] = 0xff000000UL; 85 eit_vector[22] = BRA_INSN(ill_trap, 22);
87 eit_vector[23] = 0xff000000UL; 86 eit_vector[23] = BRA_INSN(ill_trap, 23);
88 eit_vector[24] = 0xff000000UL; 87 eit_vector[24] = BRA_INSN(ill_trap, 24);
89 eit_vector[25] = 0xff000000UL; 88 eit_vector[25] = BRA_INSN(ill_trap, 25);
90 eit_vector[26] = 0xff000000UL; 89 eit_vector[26] = BRA_INSN(ill_trap, 26);
91 eit_vector[27] = 0xff000000UL; 90 eit_vector[27] = BRA_INSN(ill_trap, 27);
92 eit_vector[28] = BRA_INSN(cache_flushing_handler, 28); 91 eit_vector[28] = BRA_INSN(cache_flushing_handler, 28);
93 eit_vector[29] = 0xff000000UL; 92 eit_vector[29] = BRA_INSN(ill_trap, 29);
94 eit_vector[30] = 0xff000000UL; 93 eit_vector[30] = BRA_INSN(ill_trap, 30);
95 eit_vector[31] = 0xff000000UL; 94 eit_vector[31] = BRA_INSN(ill_trap, 31);
96 eit_vector[32] = BRA_INSN(ei_handler, 32); 95 eit_vector[32] = BRA_INSN(ei_handler, 32);
97 eit_vector[64] = BRA_INSN(pie_handler, 64); 96 eit_vector[64] = BRA_INSN(pie_handler, 64);
98#ifdef CONFIG_MMU 97#ifdef CONFIG_MMU
@@ -286,7 +285,8 @@ asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
286 285
287DO_ERROR( 1, SIGTRAP, "debug trap", debug_trap) 286DO_ERROR( 1, SIGTRAP, "debug trap", debug_trap)
288DO_ERROR_INFO(0x20, SIGILL, "reserved instruction ", rie_handler, ILL_ILLOPC, regs->bpc) 287DO_ERROR_INFO(0x20, SIGILL, "reserved instruction ", rie_handler, ILL_ILLOPC, regs->bpc)
289DO_ERROR_INFO(0x100, SIGILL, "privilege instruction", pie_handler, ILL_PRVOPC, regs->bpc) 288DO_ERROR_INFO(0x100, SIGILL, "privileged instruction", pie_handler, ILL_PRVOPC, regs->bpc)
289DO_ERROR_INFO(-1, SIGILL, "illegal trap", ill_trap, ILL_ILLTRP, regs->bpc)
290 290
291extern int handle_unaligned_access(unsigned long, struct pt_regs *); 291extern int handle_unaligned_access(unsigned long, struct pt_regs *);
292 292
@@ -329,4 +329,3 @@ asmlinkage void do_alignment_check(struct pt_regs *regs, long error_code)
329 set_fs(oldfs); 329 set_fs(oldfs);
330 } 330 }
331} 331}
332
diff --git a/arch/mips/pci/fixup-tb0226.c b/arch/mips/pci/fixup-tb0226.c
index 61513d5d97da..b5d42b12de10 100644
--- a/arch/mips/pci/fixup-tb0226.c
+++ b/arch/mips/pci/fixup-tb0226.c
@@ -1,7 +1,7 @@
1/* 1/*
2 * fixup-tb0226.c, The TANBAC TB0226 specific PCI fixups. 2 * fixup-tb0226.c, The TANBAC TB0226 specific PCI fixups.
3 * 3 *
4 * Copyright (C) 2002-2004 Yoichi Yuasa <yuasa@hh.iij4u.or.jp> 4 * Copyright (C) 2002-2005 Yoichi Yuasa <yuasa@hh.iij4u.or.jp>
5 * 5 *
6 * This program is free software; you can redistribute it and/or modify 6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by 7 * it under the terms of the GNU General Public License as published by
@@ -20,6 +20,7 @@
20#include <linux/init.h> 20#include <linux/init.h>
21#include <linux/pci.h> 21#include <linux/pci.h>
22 22
23#include <asm/vr41xx/giu.h>
23#include <asm/vr41xx/tb0226.h> 24#include <asm/vr41xx/tb0226.h>
24 25
25int __init pcibios_map_irq(struct pci_dev *dev, u8 slot, u8 pin) 26int __init pcibios_map_irq(struct pci_dev *dev, u8 slot, u8 pin)
@@ -29,42 +30,42 @@ int __init pcibios_map_irq(struct pci_dev *dev, u8 slot, u8 pin)
29 switch (slot) { 30 switch (slot) {
30 case 12: 31 case 12:
31 vr41xx_set_irq_trigger(GD82559_1_PIN, 32 vr41xx_set_irq_trigger(GD82559_1_PIN,
32 TRIGGER_LEVEL, 33 IRQ_TRIGGER_LEVEL,
33 SIGNAL_THROUGH); 34 IRQ_SIGNAL_THROUGH);
34 vr41xx_set_irq_level(GD82559_1_PIN, LEVEL_LOW); 35 vr41xx_set_irq_level(GD82559_1_PIN, IRQ_LEVEL_LOW);
35 irq = GD82559_1_IRQ; 36 irq = GD82559_1_IRQ;
36 break; 37 break;
37 case 13: 38 case 13:
38 vr41xx_set_irq_trigger(GD82559_2_PIN, 39 vr41xx_set_irq_trigger(GD82559_2_PIN,
39 TRIGGER_LEVEL, 40 IRQ_TRIGGER_LEVEL,
40 SIGNAL_THROUGH); 41 IRQ_SIGNAL_THROUGH);
41 vr41xx_set_irq_level(GD82559_2_PIN, LEVEL_LOW); 42 vr41xx_set_irq_level(GD82559_2_PIN, IRQ_LEVEL_LOW);
42 irq = GD82559_2_IRQ; 43 irq = GD82559_2_IRQ;
43 break; 44 break;
44 case 14: 45 case 14:
45 switch (pin) { 46 switch (pin) {
46 case 1: 47 case 1:
47 vr41xx_set_irq_trigger(UPD720100_INTA_PIN, 48 vr41xx_set_irq_trigger(UPD720100_INTA_PIN,
48 TRIGGER_LEVEL, 49 IRQ_TRIGGER_LEVEL,
49 SIGNAL_THROUGH); 50 IRQ_SIGNAL_THROUGH);
50 vr41xx_set_irq_level(UPD720100_INTA_PIN, 51 vr41xx_set_irq_level(UPD720100_INTA_PIN,
51 LEVEL_LOW); 52 IRQ_LEVEL_LOW);
52 irq = UPD720100_INTA_IRQ; 53 irq = UPD720100_INTA_IRQ;
53 break; 54 break;
54 case 2: 55 case 2:
55 vr41xx_set_irq_trigger(UPD720100_INTB_PIN, 56 vr41xx_set_irq_trigger(UPD720100_INTB_PIN,
56 TRIGGER_LEVEL, 57 IRQ_TRIGGER_LEVEL,
57 SIGNAL_THROUGH); 58 IRQ_SIGNAL_THROUGH);
58 vr41xx_set_irq_level(UPD720100_INTB_PIN, 59 vr41xx_set_irq_level(UPD720100_INTB_PIN,
59 LEVEL_LOW); 60 IRQ_LEVEL_LOW);
60 irq = UPD720100_INTB_IRQ; 61 irq = UPD720100_INTB_IRQ;
61 break; 62 break;
62 case 3: 63 case 3:
63 vr41xx_set_irq_trigger(UPD720100_INTC_PIN, 64 vr41xx_set_irq_trigger(UPD720100_INTC_PIN,
64 TRIGGER_LEVEL, 65 IRQ_TRIGGER_LEVEL,
65 SIGNAL_THROUGH); 66 IRQ_SIGNAL_THROUGH);
66 vr41xx_set_irq_level(UPD720100_INTC_PIN, 67 vr41xx_set_irq_level(UPD720100_INTC_PIN,
67 LEVEL_LOW); 68 IRQ_LEVEL_LOW);
68 irq = UPD720100_INTC_IRQ; 69 irq = UPD720100_INTC_IRQ;
69 break; 70 break;
70 default: 71 default:
diff --git a/arch/ppc/kernel/cputable.c b/arch/ppc/kernel/cputable.c
index 546e1ea4cafa..6b76cf58d9e0 100644
--- a/arch/ppc/kernel/cputable.c
+++ b/arch/ppc/kernel/cputable.c
@@ -91,7 +91,7 @@ struct cpu_spec cpu_specs[] = {
91 .cpu_features = CPU_FTR_COMMON | CPU_FTR_601 | 91 .cpu_features = CPU_FTR_COMMON | CPU_FTR_601 |
92 CPU_FTR_HPTE_TABLE, 92 CPU_FTR_HPTE_TABLE,
93 .cpu_user_features = COMMON_PPC | PPC_FEATURE_601_INSTR | 93 .cpu_user_features = COMMON_PPC | PPC_FEATURE_601_INSTR |
94 PPC_FEATURE_UNIFIED_CACHE, 94 PPC_FEATURE_UNIFIED_CACHE | PPC_FEATURE_NO_TB,
95 .icache_bsize = 32, 95 .icache_bsize = 32,
96 .dcache_bsize = 32, 96 .dcache_bsize = 32,
97 .cpu_setup = __setup_cpu_601 97 .cpu_setup = __setup_cpu_601
@@ -745,7 +745,8 @@ struct cpu_spec cpu_specs[] = {
745 .cpu_name = "403GCX", 745 .cpu_name = "403GCX",
746 .cpu_features = CPU_FTR_SPLIT_ID_CACHE | 746 .cpu_features = CPU_FTR_SPLIT_ID_CACHE |
747 CPU_FTR_USE_TB, 747 CPU_FTR_USE_TB,
748 .cpu_user_features = PPC_FEATURE_32 | PPC_FEATURE_HAS_MMU, 748 .cpu_user_features = PPC_FEATURE_32 |
749 PPC_FEATURE_HAS_MMU | PPC_FEATURE_NO_TB,
749 .icache_bsize = 16, 750 .icache_bsize = 16,
750 .dcache_bsize = 16, 751 .dcache_bsize = 16,
751 }, 752 },
diff --git a/arch/ppc/kernel/dma-mapping.c b/arch/ppc/kernel/dma-mapping.c
index b566d982806c..8edee806dae7 100644
--- a/arch/ppc/kernel/dma-mapping.c
+++ b/arch/ppc/kernel/dma-mapping.c
@@ -401,10 +401,10 @@ EXPORT_SYMBOL(__dma_sync);
401static inline void __dma_sync_page_highmem(struct page *page, 401static inline void __dma_sync_page_highmem(struct page *page,
402 unsigned long offset, size_t size, int direction) 402 unsigned long offset, size_t size, int direction)
403{ 403{
404 size_t seg_size = min((size_t)PAGE_SIZE, size) - offset; 404 size_t seg_size = min((size_t)(PAGE_SIZE - offset), size);
405 size_t cur_size = seg_size; 405 size_t cur_size = seg_size;
406 unsigned long flags, start, seg_offset = offset; 406 unsigned long flags, start, seg_offset = offset;
407 int nr_segs = PAGE_ALIGN(size + (PAGE_SIZE - offset))/PAGE_SIZE; 407 int nr_segs = 1 + ((size - seg_size) + PAGE_SIZE - 1)/PAGE_SIZE;
408 int seg_nr = 0; 408 int seg_nr = 0;
409 409
410 local_irq_save(flags); 410 local_irq_save(flags);
diff --git a/arch/ppc64/kernel/module.c b/arch/ppc64/kernel/module.c
index c683bf88e690..928b8581fcb0 100644
--- a/arch/ppc64/kernel/module.c
+++ b/arch/ppc64/kernel/module.c
@@ -341,6 +341,19 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
341 *(unsigned long *)location = my_r2(sechdrs, me); 341 *(unsigned long *)location = my_r2(sechdrs, me);
342 break; 342 break;
343 343
344 case R_PPC64_TOC16:
345 /* Subtact TOC pointer */
346 value -= my_r2(sechdrs, me);
347 if (value + 0x8000 > 0xffff) {
348 printk("%s: bad TOC16 relocation (%lu)\n",
349 me->name, value);
350 return -ENOEXEC;
351 }
352 *((uint16_t *) location)
353 = (*((uint16_t *) location) & ~0xffff)
354 | (value & 0xffff);
355 break;
356
344 case R_PPC64_TOC16_DS: 357 case R_PPC64_TOC16_DS:
345 /* Subtact TOC pointer */ 358 /* Subtact TOC pointer */
346 value -= my_r2(sechdrs, me); 359 value -= my_r2(sechdrs, me);
diff --git a/arch/ppc64/kernel/pSeries_pci.c b/arch/ppc64/kernel/pSeries_pci.c
index 1f5f141fb7a1..928f8febdb3b 100644
--- a/arch/ppc64/kernel/pSeries_pci.c
+++ b/arch/ppc64/kernel/pSeries_pci.c
@@ -32,7 +32,7 @@
32 32
33#include "pci.h" 33#include "pci.h"
34 34
35static int __initdata s7a_workaround = -1; 35static int __devinitdata s7a_workaround = -1;
36 36
37#if 0 37#if 0
38void pcibios_name_device(struct pci_dev *dev) 38void pcibios_name_device(struct pci_dev *dev)
@@ -60,7 +60,7 @@ void pcibios_name_device(struct pci_dev *dev)
60DECLARE_PCI_FIXUP_HEADER(PCI_ANY_ID, PCI_ANY_ID, pcibios_name_device); 60DECLARE_PCI_FIXUP_HEADER(PCI_ANY_ID, PCI_ANY_ID, pcibios_name_device);
61#endif 61#endif
62 62
63static void __init check_s7a(void) 63static void __devinit check_s7a(void)
64{ 64{
65 struct device_node *root; 65 struct device_node *root;
66 char *model; 66 char *model;
diff --git a/arch/ppc64/kernel/pmac_setup.c b/arch/ppc64/kernel/pmac_setup.c
index 25755252067a..fa8121d53b89 100644
--- a/arch/ppc64/kernel/pmac_setup.c
+++ b/arch/ppc64/kernel/pmac_setup.c
@@ -115,7 +115,7 @@ static void __pmac pmac_show_cpuinfo(struct seq_file *m)
115 115
116 /* find motherboard type */ 116 /* find motherboard type */
117 seq_printf(m, "machine\t\t: "); 117 seq_printf(m, "machine\t\t: ");
118 np = find_devices("device-tree"); 118 np = of_find_node_by_path("/");
119 if (np != NULL) { 119 if (np != NULL) {
120 pp = (char *) get_property(np, "model", NULL); 120 pp = (char *) get_property(np, "model", NULL);
121 if (pp != NULL) 121 if (pp != NULL)
@@ -133,6 +133,7 @@ static void __pmac pmac_show_cpuinfo(struct seq_file *m)
133 } 133 }
134 seq_printf(m, "\n"); 134 seq_printf(m, "\n");
135 } 135 }
136 of_node_put(np);
136 } else 137 } else
137 seq_printf(m, "PowerMac\n"); 138 seq_printf(m, "PowerMac\n");
138 139
diff --git a/arch/sh/kernel/smp.c b/arch/sh/kernel/smp.c
index 56a39d69e080..5ecefc02896a 100644
--- a/arch/sh/kernel/smp.c
+++ b/arch/sh/kernel/smp.c
@@ -22,6 +22,7 @@
22#include <linux/time.h> 22#include <linux/time.h>
23#include <linux/timex.h> 23#include <linux/timex.h>
24#include <linux/sched.h> 24#include <linux/sched.h>
25#include <linux/module.h>
25 26
26#include <asm/atomic.h> 27#include <asm/atomic.h>
27#include <asm/processor.h> 28#include <asm/processor.h>
@@ -39,6 +40,8 @@ struct sh_cpuinfo cpu_data[NR_CPUS];
39extern void per_cpu_trap_init(void); 40extern void per_cpu_trap_init(void);
40 41
41cpumask_t cpu_possible_map; 42cpumask_t cpu_possible_map;
43EXPORT_SYMBOL(cpu_possible_map);
44
42cpumask_t cpu_online_map; 45cpumask_t cpu_online_map;
43static atomic_t cpus_booted = ATOMIC_INIT(0); 46static atomic_t cpus_booted = ATOMIC_INIT(0);
44 47
diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig
index f7c51b869049..6537445dac0e 100644
--- a/arch/sparc/Kconfig
+++ b/arch/sparc/Kconfig
@@ -21,10 +21,6 @@ config GENERIC_ISA_DMA
21 bool 21 bool
22 default y 22 default y
23 23
24config GENERIC_IOMAP
25 bool
26 default y
27
28source "init/Kconfig" 24source "init/Kconfig"
29 25
30menu "General machine setup" 26menu "General machine setup"
diff --git a/arch/sparc/defconfig b/arch/sparc/defconfig
index 8a3aef1e22f5..a69856263009 100644
--- a/arch/sparc/defconfig
+++ b/arch/sparc/defconfig
@@ -5,7 +5,6 @@ CONFIG_MMU=y
5CONFIG_UID16=y 5CONFIG_UID16=y
6CONFIG_HIGHMEM=y 6CONFIG_HIGHMEM=y
7CONFIG_GENERIC_ISA_DMA=y 7CONFIG_GENERIC_ISA_DMA=y
8CONFIG_GENERIC_IOMAP=y
9 8
10# 9#
11# Code maturity level options 10# Code maturity level options
diff --git a/arch/sparc64/kernel/dtlb_base.S b/arch/sparc64/kernel/dtlb_base.S
index 702d349c1e88..6528786840c0 100644
--- a/arch/sparc64/kernel/dtlb_base.S
+++ b/arch/sparc64/kernel/dtlb_base.S
@@ -53,19 +53,18 @@
53 * be guaranteed to be 0 ... mmu_context.h does guarantee this 53 * be guaranteed to be 0 ... mmu_context.h does guarantee this
54 * by only using 10 bits in the hwcontext value. 54 * by only using 10 bits in the hwcontext value.
55 */ 55 */
56#define CREATE_VPTE_OFFSET1(r1, r2) 56#define CREATE_VPTE_OFFSET1(r1, r2) nop
57#define CREATE_VPTE_OFFSET2(r1, r2) \ 57#define CREATE_VPTE_OFFSET2(r1, r2) \
58 srax r1, 10, r2 58 srax r1, 10, r2
59#define CREATE_VPTE_NOP nop
60#else 59#else
61#define CREATE_VPTE_OFFSET1(r1, r2) \ 60#define CREATE_VPTE_OFFSET1(r1, r2) \
62 srax r1, PAGE_SHIFT, r2 61 srax r1, PAGE_SHIFT, r2
63#define CREATE_VPTE_OFFSET2(r1, r2) \ 62#define CREATE_VPTE_OFFSET2(r1, r2) \
64 sllx r2, 3, r2 63 sllx r2, 3, r2
65#define CREATE_VPTE_NOP
66#endif 64#endif
67 65
68/* DTLB ** ICACHE line 1: Quick user TLB misses */ 66/* DTLB ** ICACHE line 1: Quick user TLB misses */
67 mov TLB_SFSR, %g1
69 ldxa [%g1 + %g1] ASI_DMMU, %g4 ! Get TAG_ACCESS 68 ldxa [%g1 + %g1] ASI_DMMU, %g4 ! Get TAG_ACCESS
70 andcc %g4, TAG_CONTEXT_BITS, %g0 ! From Nucleus? 69 andcc %g4, TAG_CONTEXT_BITS, %g0 ! From Nucleus?
71from_tl1_trap: 70from_tl1_trap:
@@ -74,18 +73,16 @@ from_tl1_trap:
74 be,pn %xcc, kvmap ! Yep, special processing 73 be,pn %xcc, kvmap ! Yep, special processing
75 CREATE_VPTE_OFFSET2(%g4, %g6) ! Create VPTE offset 74 CREATE_VPTE_OFFSET2(%g4, %g6) ! Create VPTE offset
76 cmp %g5, 4 ! Last trap level? 75 cmp %g5, 4 ! Last trap level?
77 be,pn %xcc, longpath ! Yep, cannot risk VPTE miss
78 nop ! delay slot
79 76
80/* DTLB ** ICACHE line 2: User finish + quick kernel TLB misses */ 77/* DTLB ** ICACHE line 2: User finish + quick kernel TLB misses */
78 be,pn %xcc, longpath ! Yep, cannot risk VPTE miss
79 nop ! delay slot
81 ldxa [%g3 + %g6] ASI_S, %g5 ! Load VPTE 80 ldxa [%g3 + %g6] ASI_S, %g5 ! Load VPTE
821: brgez,pn %g5, longpath ! Invalid, branch out 811: brgez,pn %g5, longpath ! Invalid, branch out
83 nop ! Delay-slot 82 nop ! Delay-slot
849: stxa %g5, [%g0] ASI_DTLB_DATA_IN ! Reload TLB 839: stxa %g5, [%g0] ASI_DTLB_DATA_IN ! Reload TLB
85 retry ! Trap return 84 retry ! Trap return
86 nop 85 nop
87 nop
88 nop
89 86
90/* DTLB ** ICACHE line 3: winfixups+real_faults */ 87/* DTLB ** ICACHE line 3: winfixups+real_faults */
91longpath: 88longpath:
@@ -106,8 +103,7 @@ longpath:
106 nop 103 nop
107 nop 104 nop
108 nop 105 nop
109 CREATE_VPTE_NOP 106 nop
110 107
111#undef CREATE_VPTE_OFFSET1 108#undef CREATE_VPTE_OFFSET1
112#undef CREATE_VPTE_OFFSET2 109#undef CREATE_VPTE_OFFSET2
113#undef CREATE_VPTE_NOP
diff --git a/arch/sparc64/kernel/dtlb_prot.S b/arch/sparc64/kernel/dtlb_prot.S
index d848bb7374bb..e0a920162604 100644
--- a/arch/sparc64/kernel/dtlb_prot.S
+++ b/arch/sparc64/kernel/dtlb_prot.S
@@ -14,14 +14,14 @@
14 */ 14 */
15 15
16/* PROT ** ICACHE line 1: User DTLB protection trap */ 16/* PROT ** ICACHE line 1: User DTLB protection trap */
17 stxa %g0, [%g1] ASI_DMMU ! Clear SFSR FaultValid bit 17 mov TLB_SFSR, %g1
18 membar #Sync ! Synchronize ASI stores 18 stxa %g0, [%g1] ASI_DMMU ! Clear FaultValid bit
19 rdpr %pstate, %g5 ! Move into alternate globals 19 membar #Sync ! Synchronize stores
20 rdpr %pstate, %g5 ! Move into alt-globals
20 wrpr %g5, PSTATE_AG|PSTATE_MG, %pstate 21 wrpr %g5, PSTATE_AG|PSTATE_MG, %pstate
21 rdpr %tl, %g1 ! Need to do a winfixup? 22 rdpr %tl, %g1 ! Need a winfixup?
22 cmp %g1, 1 ! Trap level >1? 23 cmp %g1, 1 ! Trap level >1?
23 mov TLB_TAG_ACCESS, %g4 ! Prepare reload of vaddr 24 mov TLB_TAG_ACCESS, %g4 ! For reload of vaddr
24 nop
25 25
26/* PROT ** ICACHE line 2: More real fault processing */ 26/* PROT ** ICACHE line 2: More real fault processing */
27 bgu,pn %xcc, winfix_trampoline ! Yes, perform winfixup 27 bgu,pn %xcc, winfix_trampoline ! Yes, perform winfixup
diff --git a/arch/sparc64/kernel/head.S b/arch/sparc64/kernel/head.S
index f1dcdf8f7433..b49dcd4504b0 100644
--- a/arch/sparc64/kernel/head.S
+++ b/arch/sparc64/kernel/head.S
@@ -28,19 +28,14 @@
28#include <asm/mmu.h> 28#include <asm/mmu.h>
29 29
30/* This section from from _start to sparc64_boot_end should fit into 30/* This section from from _start to sparc64_boot_end should fit into
31 * 0x0000.0000.0040.4000 to 0x0000.0000.0040.8000 and will be sharing space 31 * 0x0000000000404000 to 0x0000000000408000.
32 * with bootup_user_stack, which is from 0x0000.0000.0040.4000 to
33 * 0x0000.0000.0040.6000 and empty_bad_page, which is from
34 * 0x0000.0000.0040.6000 to 0x0000.0000.0040.8000.
35 */ 32 */
36
37 .text 33 .text
38 .globl start, _start, stext, _stext 34 .globl start, _start, stext, _stext
39_start: 35_start:
40start: 36start:
41_stext: 37_stext:
42stext: 38stext:
43bootup_user_stack:
44! 0x0000000000404000 39! 0x0000000000404000
45 b sparc64_boot 40 b sparc64_boot
46 flushw /* Flush register file. */ 41 flushw /* Flush register file. */
@@ -191,8 +186,9 @@ prom_boot_mapping_phys_low:
191 stx %l3, [%sp + 2047 + 128 + 0x10] ! num_rets, 5 186 stx %l3, [%sp + 2047 + 128 + 0x10] ! num_rets, 5
192 stx %l2, [%sp + 2047 + 128 + 0x18] ! arg1: "translate" 187 stx %l2, [%sp + 2047 + 128 + 0x18] ! arg1: "translate"
193 stx %l5, [%sp + 2047 + 128 + 0x20] ! arg2: prom_mmu_ihandle_cache 188 stx %l5, [%sp + 2047 + 128 + 0x20] ! arg2: prom_mmu_ihandle_cache
194 srlx %l0, 22, %l3 189 /* PAGE align */
195 sllx %l3, 22, %l3 190 srlx %l0, 13, %l3
191 sllx %l3, 13, %l3
196 stx %l3, [%sp + 2047 + 128 + 0x28] ! arg3: vaddr, our PC 192 stx %l3, [%sp + 2047 + 128 + 0x28] ! arg3: vaddr, our PC
197 stx %g0, [%sp + 2047 + 128 + 0x30] ! res1 193 stx %g0, [%sp + 2047 + 128 + 0x30] ! res1
198 stx %g0, [%sp + 2047 + 128 + 0x38] ! res2 194 stx %g0, [%sp + 2047 + 128 + 0x38] ! res2
@@ -211,6 +207,9 @@ prom_boot_mapping_phys_low:
211 ldx [%sp + 2047 + 128 + 0x48], %l2 ! physaddr high 207 ldx [%sp + 2047 + 128 + 0x48], %l2 ! physaddr high
212 stx %l2, [%l4 + 0x0] 208 stx %l2, [%l4 + 0x0]
213 ldx [%sp + 2047 + 128 + 0x50], %l3 ! physaddr low 209 ldx [%sp + 2047 + 128 + 0x50], %l3 ! physaddr low
210 /* 4MB align */
211 srlx %l3, 22, %l3
212 sllx %l3, 22, %l3
214 stx %l3, [%l4 + 0x8] 213 stx %l3, [%l4 + 0x8]
215 214
216 /* Leave service as-is, "call-method" */ 215 /* Leave service as-is, "call-method" */
@@ -388,31 +387,30 @@ tlb_fixup_done:
388 * former does use this code, the latter does not yet due 387 * former does use this code, the latter does not yet due
389 * to some complexities. That should be fixed up at some 388 * to some complexities. That should be fixed up at some
390 * point. 389 * point.
390 *
391 * There used to be enormous complexity wrt. transferring
392 * over from the firwmare's trap table to the Linux kernel's.
393 * For example, there was a chicken & egg problem wrt. building
394 * the OBP page tables, yet needing to be on the Linux kernel
395 * trap table (to translate PAGE_OFFSET addresses) in order to
396 * do that.
397 *
398 * We now handle OBP tlb misses differently, via linear lookups
399 * into the prom_trans[] array. So that specific problem no
400 * longer exists. Yet, unfortunately there are still some issues
401 * preventing trampoline.S from using this code... ho hum.
391 */ 402 */
392 .globl setup_trap_table 403 .globl setup_trap_table
393setup_trap_table: 404setup_trap_table:
394 save %sp, -192, %sp 405 save %sp, -192, %sp
395 406
396 /* Force interrupts to be disabled. Transferring over to 407 /* Force interrupts to be disabled. */
397 * the Linux trap table is a very delicate operation.
398 * Until we are actually on the Linux trap table, we cannot
399 * get the PAGE_OFFSET linear mappings translated. We need
400 * that mapping to be setup in order to initialize the firmware
401 * page tables.
402 *
403 * So there is this window of time, from the return from
404 * prom_set_trap_table() until inherit_prom_mappings_post()
405 * (in arch/sparc64/mm/init.c) completes, during which no
406 * firmware address space accesses can be made.
407 */
408 rdpr %pstate, %o1 408 rdpr %pstate, %o1
409 andn %o1, PSTATE_IE, %o1 409 andn %o1, PSTATE_IE, %o1
410 wrpr %o1, 0x0, %pstate 410 wrpr %o1, 0x0, %pstate
411 wrpr %g0, 15, %pil 411 wrpr %g0, 15, %pil
412 412
413 /* Ok, now make the final valid firmware call to jump over 413 /* Make the firmware call to jump over to the Linux trap table. */
414 * to the Linux trap table.
415 */
416 call prom_set_trap_table 414 call prom_set_trap_table
417 sethi %hi(sparc64_ttable_tl0), %o0 415 sethi %hi(sparc64_ttable_tl0), %o0
418 416
@@ -536,15 +534,21 @@ setup_tba: /* i0 = is_starfire */
536 534
537 ret 535 ret
538 restore 536 restore
537sparc64_boot_end:
538
539#include "systbls.S"
540#include "ktlb.S"
541#include "etrap.S"
542#include "rtrap.S"
543#include "winfixup.S"
544#include "entry.S"
539 545
540/* 546/*
541 * The following skips make sure the trap table in ttable.S is aligned 547 * The following skip makes sure the trap table in ttable.S is aligned
542 * on a 32K boundary as required by the v9 specs for TBA register. 548 * on a 32K boundary as required by the v9 specs for TBA register.
543 */ 549 */
544sparc64_boot_end: 5501:
545 .skip 0x2000 + _start - sparc64_boot_end 551 .skip 0x4000 + _start - 1b
546bootup_user_stack_end:
547 .skip 0x2000
548 552
549#ifdef CONFIG_SBUS 553#ifdef CONFIG_SBUS
550/* This is just a hack to fool make depend config.h discovering 554/* This is just a hack to fool make depend config.h discovering
@@ -556,15 +560,6 @@ bootup_user_stack_end:
556! 0x0000000000408000 560! 0x0000000000408000
557 561
558#include "ttable.S" 562#include "ttable.S"
559#include "systbls.S"
560#include "ktlb.S"
561#include "etrap.S"
562#include "rtrap.S"
563#include "winfixup.S"
564#include "entry.S"
565
566 /* This is just anal retentiveness on my part... */
567 .align 16384
568 563
569 .data 564 .data
570 .align 8 565 .align 8
diff --git a/arch/sparc64/kernel/itlb_base.S b/arch/sparc64/kernel/itlb_base.S
index b5e32dfa4fbc..4951ff8f6877 100644
--- a/arch/sparc64/kernel/itlb_base.S
+++ b/arch/sparc64/kernel/itlb_base.S
@@ -15,14 +15,12 @@
15 */ 15 */
16#define CREATE_VPTE_OFFSET1(r1, r2) \ 16#define CREATE_VPTE_OFFSET1(r1, r2) \
17 srax r1, 10, r2 17 srax r1, 10, r2
18#define CREATE_VPTE_OFFSET2(r1, r2) 18#define CREATE_VPTE_OFFSET2(r1, r2) nop
19#define CREATE_VPTE_NOP nop
20#else /* PAGE_SHIFT */ 19#else /* PAGE_SHIFT */
21#define CREATE_VPTE_OFFSET1(r1, r2) \ 20#define CREATE_VPTE_OFFSET1(r1, r2) \
22 srax r1, PAGE_SHIFT, r2 21 srax r1, PAGE_SHIFT, r2
23#define CREATE_VPTE_OFFSET2(r1, r2) \ 22#define CREATE_VPTE_OFFSET2(r1, r2) \
24 sllx r2, 3, r2 23 sllx r2, 3, r2
25#define CREATE_VPTE_NOP
26#endif /* PAGE_SHIFT */ 24#endif /* PAGE_SHIFT */
27 25
28 26
@@ -36,6 +34,7 @@
36 */ 34 */
37 35
38/* ITLB ** ICACHE line 1: Quick user TLB misses */ 36/* ITLB ** ICACHE line 1: Quick user TLB misses */
37 mov TLB_SFSR, %g1
39 ldxa [%g1 + %g1] ASI_IMMU, %g4 ! Get TAG_ACCESS 38 ldxa [%g1 + %g1] ASI_IMMU, %g4 ! Get TAG_ACCESS
40 CREATE_VPTE_OFFSET1(%g4, %g6) ! Create VPTE offset 39 CREATE_VPTE_OFFSET1(%g4, %g6) ! Create VPTE offset
41 CREATE_VPTE_OFFSET2(%g4, %g6) ! Create VPTE offset 40 CREATE_VPTE_OFFSET2(%g4, %g6) ! Create VPTE offset
@@ -43,41 +42,38 @@
431: brgez,pn %g5, 3f ! Not valid, branch out 421: brgez,pn %g5, 3f ! Not valid, branch out
44 sethi %hi(_PAGE_EXEC), %g4 ! Delay-slot 43 sethi %hi(_PAGE_EXEC), %g4 ! Delay-slot
45 andcc %g5, %g4, %g0 ! Executable? 44 andcc %g5, %g4, %g0 ! Executable?
45
46/* ITLB ** ICACHE line 2: Real faults */
46 be,pn %xcc, 3f ! Nope, branch. 47 be,pn %xcc, 3f ! Nope, branch.
47 nop ! Delay-slot 48 nop ! Delay-slot
482: stxa %g5, [%g0] ASI_ITLB_DATA_IN ! Load PTE into TLB 492: stxa %g5, [%g0] ASI_ITLB_DATA_IN ! Load PTE into TLB
49 retry ! Trap return 50 retry ! Trap return
503: rdpr %pstate, %g4 ! Move into alternate globals 513: rdpr %pstate, %g4 ! Move into alt-globals
51
52/* ITLB ** ICACHE line 2: Real faults */
53 wrpr %g4, PSTATE_AG|PSTATE_MG, %pstate 52 wrpr %g4, PSTATE_AG|PSTATE_MG, %pstate
54 rdpr %tpc, %g5 ! And load faulting VA 53 rdpr %tpc, %g5 ! And load faulting VA
55 mov FAULT_CODE_ITLB, %g4 ! It was read from ITLB 54 mov FAULT_CODE_ITLB, %g4 ! It was read from ITLB
56sparc64_realfault_common: ! Called by TL0 dtlb_miss too 55
56/* ITLB ** ICACHE line 3: Finish faults */
57sparc64_realfault_common: ! Called by dtlb_miss
57 stb %g4, [%g6 + TI_FAULT_CODE] 58 stb %g4, [%g6 + TI_FAULT_CODE]
58 stx %g5, [%g6 + TI_FAULT_ADDR] 59 stx %g5, [%g6 + TI_FAULT_ADDR]
59 ba,pt %xcc, etrap ! Save state 60 ba,pt %xcc, etrap ! Save state
601: rd %pc, %g7 ! ... 611: rd %pc, %g7 ! ...
61 nop
62
63/* ITLB ** ICACHE line 3: Finish faults + window fixups */
64 call do_sparc64_fault ! Call fault handler 62 call do_sparc64_fault ! Call fault handler
65 add %sp, PTREGS_OFF, %o0! Compute pt_regs arg 63 add %sp, PTREGS_OFF, %o0! Compute pt_regs arg
66 ba,pt %xcc, rtrap_clr_l6 ! Restore cpu state 64 ba,pt %xcc, rtrap_clr_l6 ! Restore cpu state
67 nop 65 nop
66
67/* ITLB ** ICACHE line 4: Window fixups */
68winfix_trampoline: 68winfix_trampoline:
69 rdpr %tpc, %g3 ! Prepare winfixup TNPC 69 rdpr %tpc, %g3 ! Prepare winfixup TNPC
70 or %g3, 0x7c, %g3 ! Compute offset to branch 70 or %g3, 0x7c, %g3 ! Compute branch offset
71 wrpr %g3, %tnpc ! Write it into TNPC 71 wrpr %g3, %tnpc ! Write it into TNPC
72 done ! Do it to it 72 done ! Do it to it
73
74/* ITLB ** ICACHE line 4: Unused... */
75 nop 73 nop
76 nop 74 nop
77 nop 75 nop
78 nop 76 nop
79 CREATE_VPTE_NOP
80 77
81#undef CREATE_VPTE_OFFSET1 78#undef CREATE_VPTE_OFFSET1
82#undef CREATE_VPTE_OFFSET2 79#undef CREATE_VPTE_OFFSET2
83#undef CREATE_VPTE_NOP
diff --git a/arch/sparc64/kernel/ktlb.S b/arch/sparc64/kernel/ktlb.S
index 7796b37f478c..d9244d3c9f73 100644
--- a/arch/sparc64/kernel/ktlb.S
+++ b/arch/sparc64/kernel/ktlb.S
@@ -58,9 +58,6 @@ vpte_noent:
58 done 58 done
59 59
60vpte_insn_obp: 60vpte_insn_obp:
61 sethi %hi(prom_pmd_phys), %g5
62 ldx [%g5 + %lo(prom_pmd_phys)], %g5
63
64 /* Behave as if we are at TL0. */ 61 /* Behave as if we are at TL0. */
65 wrpr %g0, 1, %tl 62 wrpr %g0, 1, %tl
66 rdpr %tpc, %g4 /* Find original faulting iaddr */ 63 rdpr %tpc, %g4 /* Find original faulting iaddr */
@@ -71,58 +68,57 @@ vpte_insn_obp:
71 mov TLB_SFSR, %g1 68 mov TLB_SFSR, %g1
72 stxa %g4, [%g1 + %g1] ASI_IMMU 69 stxa %g4, [%g1 + %g1] ASI_IMMU
73 70
74 /* Get PMD offset. */ 71 sethi %hi(prom_trans), %g5
75 srlx %g4, 23, %g6 72 or %g5, %lo(prom_trans), %g5
76 and %g6, 0x7ff, %g6 73
77 sllx %g6, 2, %g6 741: ldx [%g5 + 0x00], %g6 ! base
78 75 brz,a,pn %g6, longpath ! no more entries, fail
79 /* Load PMD, is it valid? */ 76 mov TLB_SFSR, %g1 ! and restore %g1
80 lduwa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 77 ldx [%g5 + 0x08], %g1 ! len
81 brz,pn %g5, longpath 78 add %g6, %g1, %g1 ! end
82 sllx %g5, 11, %g5 79 cmp %g6, %g4
83 80 bgu,pt %xcc, 2f
84 /* Get PTE offset. */ 81 cmp %g4, %g1
85 srlx %g4, 13, %g6 82 bgeu,pt %xcc, 2f
86 and %g6, 0x3ff, %g6 83 ldx [%g5 + 0x10], %g1 ! PTE
87 sllx %g6, 3, %g6 84
88 85 /* TLB load, restore %g1, and return from trap. */
89 /* Load PTE. */ 86 sub %g4, %g6, %g6
90 ldxa [%g5 + %g6] ASI_PHYS_USE_EC, %g5 87 add %g1, %g6, %g5
91 brgez,pn %g5, longpath 88 mov TLB_SFSR, %g1
92 nop
93
94 /* TLB load and return from trap. */
95 stxa %g5, [%g0] ASI_ITLB_DATA_IN 89 stxa %g5, [%g0] ASI_ITLB_DATA_IN
96 retry 90 retry
97 91
98kvmap_do_obp: 922: ba,pt %xcc, 1b
99 sethi %hi(prom_pmd_phys), %g5 93 add %g5, (3 * 8), %g5 ! next entry
100 ldx [%g5 + %lo(prom_pmd_phys)], %g5
101
102 /* Get PMD offset. */
103 srlx %g4, 23, %g6
104 and %g6, 0x7ff, %g6
105 sllx %g6, 2, %g6
106
107 /* Load PMD, is it valid? */
108 lduwa [%g5 + %g6] ASI_PHYS_USE_EC, %g5
109 brz,pn %g5, longpath
110 sllx %g5, 11, %g5
111
112 /* Get PTE offset. */
113 srlx %g4, 13, %g6
114 and %g6, 0x3ff, %g6
115 sllx %g6, 3, %g6
116
117 /* Load PTE. */
118 ldxa [%g5 + %g6] ASI_PHYS_USE_EC, %g5
119 brgez,pn %g5, longpath
120 nop
121 94
122 /* TLB load and return from trap. */ 95kvmap_do_obp:
96 sethi %hi(prom_trans), %g5
97 or %g5, %lo(prom_trans), %g5
98 srlx %g4, 13, %g4
99 sllx %g4, 13, %g4
100
1011: ldx [%g5 + 0x00], %g6 ! base
102 brz,a,pn %g6, longpath ! no more entries, fail
103 mov TLB_SFSR, %g1 ! and restore %g1
104 ldx [%g5 + 0x08], %g1 ! len
105 add %g6, %g1, %g1 ! end
106 cmp %g6, %g4
107 bgu,pt %xcc, 2f
108 cmp %g4, %g1
109 bgeu,pt %xcc, 2f
110 ldx [%g5 + 0x10], %g1 ! PTE
111
112 /* TLB load, restore %g1, and return from trap. */
113 sub %g4, %g6, %g6
114 add %g1, %g6, %g5
115 mov TLB_SFSR, %g1
123 stxa %g5, [%g0] ASI_DTLB_DATA_IN 116 stxa %g5, [%g0] ASI_DTLB_DATA_IN
124 retry 117 retry
125 118
1192: ba,pt %xcc, 1b
120 add %g5, (3 * 8), %g5 ! next entry
121
126/* 122/*
127 * On a first level data miss, check whether this is to the OBP range (note 123 * On a first level data miss, check whether this is to the OBP range (note
128 * that such accesses can be made by prom, as well as by kernel using 124 * that such accesses can be made by prom, as well as by kernel using
diff --git a/arch/sparc64/kernel/pci_iommu.c b/arch/sparc64/kernel/pci_iommu.c
index 425c60cfea19..a11910be1013 100644
--- a/arch/sparc64/kernel/pci_iommu.c
+++ b/arch/sparc64/kernel/pci_iommu.c
@@ -49,12 +49,6 @@ static void __iommu_flushall(struct pci_iommu *iommu)
49 49
50 /* Ensure completion of previous PIO writes. */ 50 /* Ensure completion of previous PIO writes. */
51 (void) pci_iommu_read(iommu->write_complete_reg); 51 (void) pci_iommu_read(iommu->write_complete_reg);
52
53 /* Now update everyone's flush point. */
54 for (entry = 0; entry < PBM_NCLUSTERS; entry++) {
55 iommu->alloc_info[entry].flush =
56 iommu->alloc_info[entry].next;
57 }
58} 52}
59 53
60#define IOPTE_CONSISTENT(CTX) \ 54#define IOPTE_CONSISTENT(CTX) \
@@ -80,120 +74,117 @@ static void inline iopte_make_dummy(struct pci_iommu *iommu, iopte_t *iopte)
80 iopte_val(*iopte) = val; 74 iopte_val(*iopte) = val;
81} 75}
82 76
83void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize) 77/* Based largely upon the ppc64 iommu allocator. */
78static long pci_arena_alloc(struct pci_iommu *iommu, unsigned long npages)
84{ 79{
85 int i; 80 struct pci_iommu_arena *arena = &iommu->arena;
86 81 unsigned long n, i, start, end, limit;
87 tsbsize /= sizeof(iopte_t); 82 int pass;
88 83
89 for (i = 0; i < tsbsize; i++) 84 limit = arena->limit;
90 iopte_make_dummy(iommu, &iommu->page_table[i]); 85 start = arena->hint;
91} 86 pass = 0;
92 87
93static iopte_t *alloc_streaming_cluster(struct pci_iommu *iommu, unsigned long npages) 88again:
94{ 89 n = find_next_zero_bit(arena->map, limit, start);
95 iopte_t *iopte, *limit, *first; 90 end = n + npages;
96 unsigned long cnum, ent, flush_point; 91 if (unlikely(end >= limit)) {
97 92 if (likely(pass < 1)) {
98 cnum = 0; 93 limit = start;
99 while ((1UL << cnum) < npages) 94 start = 0;
100 cnum++; 95 __iommu_flushall(iommu);
101 iopte = (iommu->page_table + 96 pass++;
102 (cnum << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); 97 goto again;
103 98 } else {
104 if (cnum == 0) 99 /* Scanned the whole thing, give up. */
105 limit = (iommu->page_table + 100 return -1;
106 iommu->lowest_consistent_map);
107 else
108 limit = (iopte +
109 (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS)));
110
111 iopte += ((ent = iommu->alloc_info[cnum].next) << cnum);
112 flush_point = iommu->alloc_info[cnum].flush;
113
114 first = iopte;
115 for (;;) {
116 if (IOPTE_IS_DUMMY(iommu, iopte)) {
117 if ((iopte + (1 << cnum)) >= limit)
118 ent = 0;
119 else
120 ent = ent + 1;
121 iommu->alloc_info[cnum].next = ent;
122 if (ent == flush_point)
123 __iommu_flushall(iommu);
124 break;
125 } 101 }
126 iopte += (1 << cnum); 102 }
127 ent++; 103
128 if (iopte >= limit) { 104 for (i = n; i < end; i++) {
129 iopte = (iommu->page_table + 105 if (test_bit(i, arena->map)) {
130 (cnum << 106 start = i + 1;
131 (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); 107 goto again;
132 ent = 0;
133 } 108 }
134 if (ent == flush_point)
135 __iommu_flushall(iommu);
136 if (iopte == first)
137 goto bad;
138 } 109 }
139 110
140 /* I've got your streaming cluster right here buddy boy... */ 111 for (i = n; i < end; i++)
141 return iopte; 112 __set_bit(i, arena->map);
142 113
143bad: 114 arena->hint = end;
144 printk(KERN_EMERG "pci_iommu: alloc_streaming_cluster of npages(%ld) failed!\n", 115
145 npages); 116 return n;
146 return NULL;
147} 117}
148 118
149static void free_streaming_cluster(struct pci_iommu *iommu, dma_addr_t base, 119static void pci_arena_free(struct pci_iommu_arena *arena, unsigned long base, unsigned long npages)
150 unsigned long npages, unsigned long ctx)
151{ 120{
152 unsigned long cnum, ent; 121 unsigned long i;
153 122
154 cnum = 0; 123 for (i = base; i < (base + npages); i++)
155 while ((1UL << cnum) < npages) 124 __clear_bit(i, arena->map);
156 cnum++; 125}
157 126
158 ent = (base << (32 - IO_PAGE_SHIFT + PBM_LOGCLUSTERS - iommu->page_table_sz_bits)) 127void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize, u32 dma_offset, u32 dma_addr_mask)
159 >> (32 + PBM_LOGCLUSTERS + cnum - iommu->page_table_sz_bits); 128{
129 unsigned long i, tsbbase, order, sz, num_tsb_entries;
130
131 num_tsb_entries = tsbsize / sizeof(iopte_t);
132
133 /* Setup initial software IOMMU state. */
134 spin_lock_init(&iommu->lock);
135 iommu->ctx_lowest_free = 1;
136 iommu->page_table_map_base = dma_offset;
137 iommu->dma_addr_mask = dma_addr_mask;
138
139 /* Allocate and initialize the free area map. */
140 sz = num_tsb_entries / 8;
141 sz = (sz + 7UL) & ~7UL;
142 iommu->arena.map = kmalloc(sz, GFP_KERNEL);
143 if (!iommu->arena.map) {
144 prom_printf("PCI_IOMMU: Error, kmalloc(arena.map) failed.\n");
145 prom_halt();
146 }
147 memset(iommu->arena.map, 0, sz);
148 iommu->arena.limit = num_tsb_entries;
160 149
161 /* If the global flush might not have caught this entry, 150 /* Allocate and initialize the dummy page which we
162 * adjust the flush point such that we will flush before 151 * set inactive IO PTEs to point to.
163 * ever trying to reuse it.
164 */ 152 */
165#define between(X,Y,Z) (((Z) - (Y)) >= ((X) - (Y))) 153 iommu->dummy_page = __get_free_pages(GFP_KERNEL, 0);
166 if (between(ent, iommu->alloc_info[cnum].next, iommu->alloc_info[cnum].flush)) 154 if (!iommu->dummy_page) {
167 iommu->alloc_info[cnum].flush = ent; 155 prom_printf("PCI_IOMMU: Error, gfp(dummy_page) failed.\n");
168#undef between 156 prom_halt();
157 }
158 memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
159 iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);
160
161 /* Now allocate and setup the IOMMU page table itself. */
162 order = get_order(tsbsize);
163 tsbbase = __get_free_pages(GFP_KERNEL, order);
164 if (!tsbbase) {
165 prom_printf("PCI_IOMMU: Error, gfp(tsb) failed.\n");
166 prom_halt();
167 }
168 iommu->page_table = (iopte_t *)tsbbase;
169
170 for (i = 0; i < num_tsb_entries; i++)
171 iopte_make_dummy(iommu, &iommu->page_table[i]);
169} 172}
170 173
171/* We allocate consistent mappings from the end of cluster zero. */ 174static inline iopte_t *alloc_npages(struct pci_iommu *iommu, unsigned long npages)
172static iopte_t *alloc_consistent_cluster(struct pci_iommu *iommu, unsigned long npages)
173{ 175{
174 iopte_t *iopte; 176 long entry;
175 177
176 iopte = iommu->page_table + (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS)); 178 entry = pci_arena_alloc(iommu, npages);
177 while (iopte > iommu->page_table) { 179 if (unlikely(entry < 0))
178 iopte--; 180 return NULL;
179 if (IOPTE_IS_DUMMY(iommu, iopte)) {
180 unsigned long tmp = npages;
181 181
182 while (--tmp) { 182 return iommu->page_table + entry;
183 iopte--; 183}
184 if (!IOPTE_IS_DUMMY(iommu, iopte))
185 break;
186 }
187 if (tmp == 0) {
188 u32 entry = (iopte - iommu->page_table);
189 184
190 if (entry < iommu->lowest_consistent_map) 185static inline void free_npages(struct pci_iommu *iommu, dma_addr_t base, unsigned long npages)
191 iommu->lowest_consistent_map = entry; 186{
192 return iopte; 187 pci_arena_free(&iommu->arena, base >> IO_PAGE_SHIFT, npages);
193 }
194 }
195 }
196 return NULL;
197} 188}
198 189
199static int iommu_alloc_ctx(struct pci_iommu *iommu) 190static int iommu_alloc_ctx(struct pci_iommu *iommu)
@@ -233,7 +224,7 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad
233 struct pcidev_cookie *pcp; 224 struct pcidev_cookie *pcp;
234 struct pci_iommu *iommu; 225 struct pci_iommu *iommu;
235 iopte_t *iopte; 226 iopte_t *iopte;
236 unsigned long flags, order, first_page, ctx; 227 unsigned long flags, order, first_page;
237 void *ret; 228 void *ret;
238 int npages; 229 int npages;
239 230
@@ -251,9 +242,10 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad
251 iommu = pcp->pbm->iommu; 242 iommu = pcp->pbm->iommu;
252 243
253 spin_lock_irqsave(&iommu->lock, flags); 244 spin_lock_irqsave(&iommu->lock, flags);
254 iopte = alloc_consistent_cluster(iommu, size >> IO_PAGE_SHIFT); 245 iopte = alloc_npages(iommu, size >> IO_PAGE_SHIFT);
255 if (iopte == NULL) { 246 spin_unlock_irqrestore(&iommu->lock, flags);
256 spin_unlock_irqrestore(&iommu->lock, flags); 247
248 if (unlikely(iopte == NULL)) {
257 free_pages(first_page, order); 249 free_pages(first_page, order);
258 return NULL; 250 return NULL;
259 } 251 }
@@ -262,31 +254,15 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad
262 ((iopte - iommu->page_table) << IO_PAGE_SHIFT)); 254 ((iopte - iommu->page_table) << IO_PAGE_SHIFT));
263 ret = (void *) first_page; 255 ret = (void *) first_page;
264 npages = size >> IO_PAGE_SHIFT; 256 npages = size >> IO_PAGE_SHIFT;
265 ctx = 0;
266 if (iommu->iommu_ctxflush)
267 ctx = iommu_alloc_ctx(iommu);
268 first_page = __pa(first_page); 257 first_page = __pa(first_page);
269 while (npages--) { 258 while (npages--) {
270 iopte_val(*iopte) = (IOPTE_CONSISTENT(ctx) | 259 iopte_val(*iopte) = (IOPTE_CONSISTENT(0UL) |
271 IOPTE_WRITE | 260 IOPTE_WRITE |
272 (first_page & IOPTE_PAGE)); 261 (first_page & IOPTE_PAGE));
273 iopte++; 262 iopte++;
274 first_page += IO_PAGE_SIZE; 263 first_page += IO_PAGE_SIZE;
275 } 264 }
276 265
277 {
278 int i;
279 u32 daddr = *dma_addrp;
280
281 npages = size >> IO_PAGE_SHIFT;
282 for (i = 0; i < npages; i++) {
283 pci_iommu_write(iommu->iommu_flush, daddr);
284 daddr += IO_PAGE_SIZE;
285 }
286 }
287
288 spin_unlock_irqrestore(&iommu->lock, flags);
289
290 return ret; 266 return ret;
291} 267}
292 268
@@ -296,7 +272,7 @@ void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_
296 struct pcidev_cookie *pcp; 272 struct pcidev_cookie *pcp;
297 struct pci_iommu *iommu; 273 struct pci_iommu *iommu;
298 iopte_t *iopte; 274 iopte_t *iopte;
299 unsigned long flags, order, npages, i, ctx; 275 unsigned long flags, order, npages;
300 276
301 npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT; 277 npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT;
302 pcp = pdev->sysdata; 278 pcp = pdev->sysdata;
@@ -306,46 +282,7 @@ void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_
306 282
307 spin_lock_irqsave(&iommu->lock, flags); 283 spin_lock_irqsave(&iommu->lock, flags);
308 284
309 if ((iopte - iommu->page_table) == 285 free_npages(iommu, dvma, npages);
310 iommu->lowest_consistent_map) {
311 iopte_t *walk = iopte + npages;
312 iopte_t *limit;
313
314 limit = (iommu->page_table +
315 (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS)));
316 while (walk < limit) {
317 if (!IOPTE_IS_DUMMY(iommu, walk))
318 break;
319 walk++;
320 }
321 iommu->lowest_consistent_map =
322 (walk - iommu->page_table);
323 }
324
325 /* Data for consistent mappings cannot enter the streaming
326 * buffers, so we only need to update the TSB. We flush
327 * the IOMMU here as well to prevent conflicts with the
328 * streaming mapping deferred tlb flush scheme.
329 */
330
331 ctx = 0;
332 if (iommu->iommu_ctxflush)
333 ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL;
334
335 for (i = 0; i < npages; i++, iopte++)
336 iopte_make_dummy(iommu, iopte);
337
338 if (iommu->iommu_ctxflush) {
339 pci_iommu_write(iommu->iommu_ctxflush, ctx);
340 } else {
341 for (i = 0; i < npages; i++) {
342 u32 daddr = dvma + (i << IO_PAGE_SHIFT);
343
344 pci_iommu_write(iommu->iommu_flush, daddr);
345 }
346 }
347
348 iommu_free_ctx(iommu, ctx);
349 286
350 spin_unlock_irqrestore(&iommu->lock, flags); 287 spin_unlock_irqrestore(&iommu->lock, flags);
351 288
@@ -372,25 +309,27 @@ dma_addr_t pci_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direct
372 iommu = pcp->pbm->iommu; 309 iommu = pcp->pbm->iommu;
373 strbuf = &pcp->pbm->stc; 310 strbuf = &pcp->pbm->stc;
374 311
375 if (direction == PCI_DMA_NONE) 312 if (unlikely(direction == PCI_DMA_NONE))
376 BUG(); 313 goto bad_no_ctx;
377 314
378 oaddr = (unsigned long)ptr; 315 oaddr = (unsigned long)ptr;
379 npages = IO_PAGE_ALIGN(oaddr + sz) - (oaddr & IO_PAGE_MASK); 316 npages = IO_PAGE_ALIGN(oaddr + sz) - (oaddr & IO_PAGE_MASK);
380 npages >>= IO_PAGE_SHIFT; 317 npages >>= IO_PAGE_SHIFT;
381 318
382 spin_lock_irqsave(&iommu->lock, flags); 319 spin_lock_irqsave(&iommu->lock, flags);
320 base = alloc_npages(iommu, npages);
321 ctx = 0;
322 if (iommu->iommu_ctxflush)
323 ctx = iommu_alloc_ctx(iommu);
324 spin_unlock_irqrestore(&iommu->lock, flags);
383 325
384 base = alloc_streaming_cluster(iommu, npages); 326 if (unlikely(!base))
385 if (base == NULL)
386 goto bad; 327 goto bad;
328
387 bus_addr = (iommu->page_table_map_base + 329 bus_addr = (iommu->page_table_map_base +
388 ((base - iommu->page_table) << IO_PAGE_SHIFT)); 330 ((base - iommu->page_table) << IO_PAGE_SHIFT));
389 ret = bus_addr | (oaddr & ~IO_PAGE_MASK); 331 ret = bus_addr | (oaddr & ~IO_PAGE_MASK);
390 base_paddr = __pa(oaddr & IO_PAGE_MASK); 332 base_paddr = __pa(oaddr & IO_PAGE_MASK);
391 ctx = 0;
392 if (iommu->iommu_ctxflush)
393 ctx = iommu_alloc_ctx(iommu);
394 if (strbuf->strbuf_enabled) 333 if (strbuf->strbuf_enabled)
395 iopte_protection = IOPTE_STREAMING(ctx); 334 iopte_protection = IOPTE_STREAMING(ctx);
396 else 335 else
@@ -401,12 +340,13 @@ dma_addr_t pci_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direct
401 for (i = 0; i < npages; i++, base++, base_paddr += IO_PAGE_SIZE) 340 for (i = 0; i < npages; i++, base++, base_paddr += IO_PAGE_SIZE)
402 iopte_val(*base) = iopte_protection | base_paddr; 341 iopte_val(*base) = iopte_protection | base_paddr;
403 342
404 spin_unlock_irqrestore(&iommu->lock, flags);
405
406 return ret; 343 return ret;
407 344
408bad: 345bad:
409 spin_unlock_irqrestore(&iommu->lock, flags); 346 iommu_free_ctx(iommu, ctx);
347bad_no_ctx:
348 if (printk_ratelimit())
349 WARN_ON(1);
410 return PCI_DMA_ERROR_CODE; 350 return PCI_DMA_ERROR_CODE;
411} 351}
412 352
@@ -481,10 +421,13 @@ void pci_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int
481 struct pci_iommu *iommu; 421 struct pci_iommu *iommu;
482 struct pci_strbuf *strbuf; 422 struct pci_strbuf *strbuf;
483 iopte_t *base; 423 iopte_t *base;
484 unsigned long flags, npages, ctx; 424 unsigned long flags, npages, ctx, i;
485 425
486 if (direction == PCI_DMA_NONE) 426 if (unlikely(direction == PCI_DMA_NONE)) {
487 BUG(); 427 if (printk_ratelimit())
428 WARN_ON(1);
429 return;
430 }
488 431
489 pcp = pdev->sysdata; 432 pcp = pdev->sysdata;
490 iommu = pcp->pbm->iommu; 433 iommu = pcp->pbm->iommu;
@@ -510,13 +453,14 @@ void pci_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int
510 453
511 /* Step 1: Kick data out of streaming buffers if necessary. */ 454 /* Step 1: Kick data out of streaming buffers if necessary. */
512 if (strbuf->strbuf_enabled) 455 if (strbuf->strbuf_enabled)
513 pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction); 456 pci_strbuf_flush(strbuf, iommu, bus_addr, ctx,
457 npages, direction);
514 458
515 /* Step 2: Clear out first TSB entry. */ 459 /* Step 2: Clear out TSB entries. */
516 iopte_make_dummy(iommu, base); 460 for (i = 0; i < npages; i++)
461 iopte_make_dummy(iommu, base + i);
517 462
518 free_streaming_cluster(iommu, bus_addr - iommu->page_table_map_base, 463 free_npages(iommu, bus_addr - iommu->page_table_map_base, npages);
519 npages, ctx);
520 464
521 iommu_free_ctx(iommu, ctx); 465 iommu_free_ctx(iommu, ctx);
522 466
@@ -621,6 +565,8 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int
621 pci_map_single(pdev, 565 pci_map_single(pdev,
622 (page_address(sglist->page) + sglist->offset), 566 (page_address(sglist->page) + sglist->offset),
623 sglist->length, direction); 567 sglist->length, direction);
568 if (unlikely(sglist->dma_address == PCI_DMA_ERROR_CODE))
569 return 0;
624 sglist->dma_length = sglist->length; 570 sglist->dma_length = sglist->length;
625 return 1; 571 return 1;
626 } 572 }
@@ -629,21 +575,29 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int
629 iommu = pcp->pbm->iommu; 575 iommu = pcp->pbm->iommu;
630 strbuf = &pcp->pbm->stc; 576 strbuf = &pcp->pbm->stc;
631 577
632 if (direction == PCI_DMA_NONE) 578 if (unlikely(direction == PCI_DMA_NONE))
633 BUG(); 579 goto bad_no_ctx;
634 580
635 /* Step 1: Prepare scatter list. */ 581 /* Step 1: Prepare scatter list. */
636 582
637 npages = prepare_sg(sglist, nelems); 583 npages = prepare_sg(sglist, nelems);
638 584
639 /* Step 2: Allocate a cluster. */ 585 /* Step 2: Allocate a cluster and context, if necessary. */
640 586
641 spin_lock_irqsave(&iommu->lock, flags); 587 spin_lock_irqsave(&iommu->lock, flags);
642 588
643 base = alloc_streaming_cluster(iommu, npages); 589 base = alloc_npages(iommu, npages);
590 ctx = 0;
591 if (iommu->iommu_ctxflush)
592 ctx = iommu_alloc_ctx(iommu);
593
594 spin_unlock_irqrestore(&iommu->lock, flags);
595
644 if (base == NULL) 596 if (base == NULL)
645 goto bad; 597 goto bad;
646 dma_base = iommu->page_table_map_base + ((base - iommu->page_table) << IO_PAGE_SHIFT); 598
599 dma_base = iommu->page_table_map_base +
600 ((base - iommu->page_table) << IO_PAGE_SHIFT);
647 601
648 /* Step 3: Normalize DMA addresses. */ 602 /* Step 3: Normalize DMA addresses. */
649 used = nelems; 603 used = nelems;
@@ -656,30 +610,28 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int
656 } 610 }
657 used = nelems - used; 611 used = nelems - used;
658 612
659 /* Step 4: Choose a context if necessary. */ 613 /* Step 4: Create the mappings. */
660 ctx = 0;
661 if (iommu->iommu_ctxflush)
662 ctx = iommu_alloc_ctx(iommu);
663
664 /* Step 5: Create the mappings. */
665 if (strbuf->strbuf_enabled) 614 if (strbuf->strbuf_enabled)
666 iopte_protection = IOPTE_STREAMING(ctx); 615 iopte_protection = IOPTE_STREAMING(ctx);
667 else 616 else
668 iopte_protection = IOPTE_CONSISTENT(ctx); 617 iopte_protection = IOPTE_CONSISTENT(ctx);
669 if (direction != PCI_DMA_TODEVICE) 618 if (direction != PCI_DMA_TODEVICE)
670 iopte_protection |= IOPTE_WRITE; 619 iopte_protection |= IOPTE_WRITE;
671 fill_sg (base, sglist, used, nelems, iopte_protection); 620
621 fill_sg(base, sglist, used, nelems, iopte_protection);
622
672#ifdef VERIFY_SG 623#ifdef VERIFY_SG
673 verify_sglist(sglist, nelems, base, npages); 624 verify_sglist(sglist, nelems, base, npages);
674#endif 625#endif
675 626
676 spin_unlock_irqrestore(&iommu->lock, flags);
677
678 return used; 627 return used;
679 628
680bad: 629bad:
681 spin_unlock_irqrestore(&iommu->lock, flags); 630 iommu_free_ctx(iommu, ctx);
682 return PCI_DMA_ERROR_CODE; 631bad_no_ctx:
632 if (printk_ratelimit())
633 WARN_ON(1);
634 return 0;
683} 635}
684 636
685/* Unmap a set of streaming mode DMA translations. */ 637/* Unmap a set of streaming mode DMA translations. */
@@ -692,8 +644,10 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems,
692 unsigned long flags, ctx, i, npages; 644 unsigned long flags, ctx, i, npages;
693 u32 bus_addr; 645 u32 bus_addr;
694 646
695 if (direction == PCI_DMA_NONE) 647 if (unlikely(direction == PCI_DMA_NONE)) {
696 BUG(); 648 if (printk_ratelimit())
649 WARN_ON(1);
650 }
697 651
698 pcp = pdev->sysdata; 652 pcp = pdev->sysdata;
699 iommu = pcp->pbm->iommu; 653 iommu = pcp->pbm->iommu;
@@ -705,7 +659,8 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems,
705 if (sglist[i].dma_length == 0) 659 if (sglist[i].dma_length == 0)
706 break; 660 break;
707 i--; 661 i--;
708 npages = (IO_PAGE_ALIGN(sglist[i].dma_address + sglist[i].dma_length) - bus_addr) >> IO_PAGE_SHIFT; 662 npages = (IO_PAGE_ALIGN(sglist[i].dma_address + sglist[i].dma_length) -
663 bus_addr) >> IO_PAGE_SHIFT;
709 664
710 base = iommu->page_table + 665 base = iommu->page_table +
711 ((bus_addr - iommu->page_table_map_base) >> IO_PAGE_SHIFT); 666 ((bus_addr - iommu->page_table_map_base) >> IO_PAGE_SHIFT);
@@ -726,11 +681,11 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems,
726 if (strbuf->strbuf_enabled) 681 if (strbuf->strbuf_enabled)
727 pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction); 682 pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction);
728 683
729 /* Step 2: Clear out first TSB entry. */ 684 /* Step 2: Clear out the TSB entries. */
730 iopte_make_dummy(iommu, base); 685 for (i = 0; i < npages; i++)
686 iopte_make_dummy(iommu, base + i);
731 687
732 free_streaming_cluster(iommu, bus_addr - iommu->page_table_map_base, 688 free_npages(iommu, bus_addr - iommu->page_table_map_base, npages);
733 npages, ctx);
734 689
735 iommu_free_ctx(iommu, ctx); 690 iommu_free_ctx(iommu, ctx);
736 691
diff --git a/arch/sparc64/kernel/pci_psycho.c b/arch/sparc64/kernel/pci_psycho.c
index 6ed1ef25e0ac..c03ed5f49d31 100644
--- a/arch/sparc64/kernel/pci_psycho.c
+++ b/arch/sparc64/kernel/pci_psycho.c
@@ -1207,13 +1207,9 @@ static void psycho_scan_bus(struct pci_controller_info *p)
1207static void psycho_iommu_init(struct pci_controller_info *p) 1207static void psycho_iommu_init(struct pci_controller_info *p)
1208{ 1208{
1209 struct pci_iommu *iommu = p->pbm_A.iommu; 1209 struct pci_iommu *iommu = p->pbm_A.iommu;
1210 unsigned long tsbbase, i; 1210 unsigned long i;
1211 u64 control; 1211 u64 control;
1212 1212
1213 /* Setup initial software IOMMU state. */
1214 spin_lock_init(&iommu->lock);
1215 iommu->ctx_lowest_free = 1;
1216
1217 /* Register addresses. */ 1213 /* Register addresses. */
1218 iommu->iommu_control = p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL; 1214 iommu->iommu_control = p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL;
1219 iommu->iommu_tsbbase = p->pbm_A.controller_regs + PSYCHO_IOMMU_TSBBASE; 1215 iommu->iommu_tsbbase = p->pbm_A.controller_regs + PSYCHO_IOMMU_TSBBASE;
@@ -1240,40 +1236,10 @@ static void psycho_iommu_init(struct pci_controller_info *p)
1240 /* Leave diag mode enabled for full-flushing done 1236 /* Leave diag mode enabled for full-flushing done
1241 * in pci_iommu.c 1237 * in pci_iommu.c
1242 */ 1238 */
1239 pci_iommu_table_init(iommu, IO_TSB_SIZE, 0xc0000000, 0xffffffff);
1243 1240
1244 iommu->dummy_page = __get_free_pages(GFP_KERNEL, 0); 1241 psycho_write(p->pbm_A.controller_regs + PSYCHO_IOMMU_TSBBASE,
1245 if (!iommu->dummy_page) { 1242 __pa(iommu->page_table));
1246 prom_printf("PSYCHO_IOMMU: Error, gfp(dummy_page) failed.\n");
1247 prom_halt();
1248 }
1249 memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
1250 iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);
1251
1252 /* Using assumed page size 8K with 128K entries we need 1MB iommu page
1253 * table (128K ioptes * 8 bytes per iopte). This is
1254 * page order 7 on UltraSparc.
1255 */
1256 tsbbase = __get_free_pages(GFP_KERNEL, get_order(IO_TSB_SIZE));
1257 if (!tsbbase) {
1258 prom_printf("PSYCHO_IOMMU: Error, gfp(tsb) failed.\n");
1259 prom_halt();
1260 }
1261 iommu->page_table = (iopte_t *)tsbbase;
1262 iommu->page_table_sz_bits = 17;
1263 iommu->page_table_map_base = 0xc0000000;
1264 iommu->dma_addr_mask = 0xffffffff;
1265 pci_iommu_table_init(iommu, IO_TSB_SIZE);
1266
1267 /* We start with no consistent mappings. */
1268 iommu->lowest_consistent_map =
1269 1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS);
1270
1271 for (i = 0; i < PBM_NCLUSTERS; i++) {
1272 iommu->alloc_info[i].flush = 0;
1273 iommu->alloc_info[i].next = 0;
1274 }
1275
1276 psycho_write(p->pbm_A.controller_regs + PSYCHO_IOMMU_TSBBASE, __pa(tsbbase));
1277 1243
1278 control = psycho_read(p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL); 1244 control = psycho_read(p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL);
1279 control &= ~(PSYCHO_IOMMU_CTRL_TSBSZ | PSYCHO_IOMMU_CTRL_TBWSZ); 1245 control &= ~(PSYCHO_IOMMU_CTRL_TSBSZ | PSYCHO_IOMMU_CTRL_TBWSZ);
@@ -1281,7 +1247,7 @@ static void psycho_iommu_init(struct pci_controller_info *p)
1281 psycho_write(p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL, control); 1247 psycho_write(p->pbm_A.controller_regs + PSYCHO_IOMMU_CONTROL, control);
1282 1248
1283 /* If necessary, hook us up for starfire IRQ translations. */ 1249 /* If necessary, hook us up for starfire IRQ translations. */
1284 if(this_is_starfire) 1250 if (this_is_starfire)
1285 p->starfire_cookie = starfire_hookup(p->pbm_A.portid); 1251 p->starfire_cookie = starfire_hookup(p->pbm_A.portid);
1286 else 1252 else
1287 p->starfire_cookie = NULL; 1253 p->starfire_cookie = NULL;
diff --git a/arch/sparc64/kernel/pci_sabre.c b/arch/sparc64/kernel/pci_sabre.c
index 0ee6bd5b9ac6..da8e1364194f 100644
--- a/arch/sparc64/kernel/pci_sabre.c
+++ b/arch/sparc64/kernel/pci_sabre.c
@@ -1267,13 +1267,9 @@ static void sabre_iommu_init(struct pci_controller_info *p,
1267 u32 dma_mask) 1267 u32 dma_mask)
1268{ 1268{
1269 struct pci_iommu *iommu = p->pbm_A.iommu; 1269 struct pci_iommu *iommu = p->pbm_A.iommu;
1270 unsigned long tsbbase, i, order; 1270 unsigned long i;
1271 u64 control; 1271 u64 control;
1272 1272
1273 /* Setup initial software IOMMU state. */
1274 spin_lock_init(&iommu->lock);
1275 iommu->ctx_lowest_free = 1;
1276
1277 /* Register addresses. */ 1273 /* Register addresses. */
1278 iommu->iommu_control = p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL; 1274 iommu->iommu_control = p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL;
1279 iommu->iommu_tsbbase = p->pbm_A.controller_regs + SABRE_IOMMU_TSBBASE; 1275 iommu->iommu_tsbbase = p->pbm_A.controller_regs + SABRE_IOMMU_TSBBASE;
@@ -1295,26 +1291,10 @@ static void sabre_iommu_init(struct pci_controller_info *p,
1295 /* Leave diag mode enabled for full-flushing done 1291 /* Leave diag mode enabled for full-flushing done
1296 * in pci_iommu.c 1292 * in pci_iommu.c
1297 */ 1293 */
1294 pci_iommu_table_init(iommu, tsbsize * 1024 * 8, dvma_offset, dma_mask);
1298 1295
1299 iommu->dummy_page = __get_free_pages(GFP_KERNEL, 0); 1296 sabre_write(p->pbm_A.controller_regs + SABRE_IOMMU_TSBBASE,
1300 if (!iommu->dummy_page) { 1297 __pa(iommu->page_table));
1301 prom_printf("PSYCHO_IOMMU: Error, gfp(dummy_page) failed.\n");
1302 prom_halt();
1303 }
1304 memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
1305 iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);
1306
1307 tsbbase = __get_free_pages(GFP_KERNEL, order = get_order(tsbsize * 1024 * 8));
1308 if (!tsbbase) {
1309 prom_printf("SABRE_IOMMU: Error, gfp(tsb) failed.\n");
1310 prom_halt();
1311 }
1312 iommu->page_table = (iopte_t *)tsbbase;
1313 iommu->page_table_map_base = dvma_offset;
1314 iommu->dma_addr_mask = dma_mask;
1315 pci_iommu_table_init(iommu, PAGE_SIZE << order);
1316
1317 sabre_write(p->pbm_A.controller_regs + SABRE_IOMMU_TSBBASE, __pa(tsbbase));
1318 1298
1319 control = sabre_read(p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL); 1299 control = sabre_read(p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL);
1320 control &= ~(SABRE_IOMMUCTRL_TSBSZ | SABRE_IOMMUCTRL_TBWSZ); 1300 control &= ~(SABRE_IOMMUCTRL_TSBSZ | SABRE_IOMMUCTRL_TBWSZ);
@@ -1322,11 +1302,9 @@ static void sabre_iommu_init(struct pci_controller_info *p,
1322 switch(tsbsize) { 1302 switch(tsbsize) {
1323 case 64: 1303 case 64:
1324 control |= SABRE_IOMMU_TSBSZ_64K; 1304 control |= SABRE_IOMMU_TSBSZ_64K;
1325 iommu->page_table_sz_bits = 16;
1326 break; 1305 break;
1327 case 128: 1306 case 128:
1328 control |= SABRE_IOMMU_TSBSZ_128K; 1307 control |= SABRE_IOMMU_TSBSZ_128K;
1329 iommu->page_table_sz_bits = 17;
1330 break; 1308 break;
1331 default: 1309 default:
1332 prom_printf("iommu_init: Illegal TSB size %d\n", tsbsize); 1310 prom_printf("iommu_init: Illegal TSB size %d\n", tsbsize);
@@ -1334,15 +1312,6 @@ static void sabre_iommu_init(struct pci_controller_info *p,
1334 break; 1312 break;
1335 } 1313 }
1336 sabre_write(p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL, control); 1314 sabre_write(p->pbm_A.controller_regs + SABRE_IOMMU_CONTROL, control);
1337
1338 /* We start with no consistent mappings. */
1339 iommu->lowest_consistent_map =
1340 1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS);
1341
1342 for (i = 0; i < PBM_NCLUSTERS; i++) {
1343 iommu->alloc_info[i].flush = 0;
1344 iommu->alloc_info[i].next = 0;
1345 }
1346} 1315}
1347 1316
1348static void pbm_register_toplevel_resources(struct pci_controller_info *p, 1317static void pbm_register_toplevel_resources(struct pci_controller_info *p,
diff --git a/arch/sparc64/kernel/pci_schizo.c b/arch/sparc64/kernel/pci_schizo.c
index cae5b61fe2f0..d8c4e0919b4e 100644
--- a/arch/sparc64/kernel/pci_schizo.c
+++ b/arch/sparc64/kernel/pci_schizo.c
@@ -1765,7 +1765,7 @@ static void schizo_pbm_strbuf_init(struct pci_pbm_info *pbm)
1765static void schizo_pbm_iommu_init(struct pci_pbm_info *pbm) 1765static void schizo_pbm_iommu_init(struct pci_pbm_info *pbm)
1766{ 1766{
1767 struct pci_iommu *iommu = pbm->iommu; 1767 struct pci_iommu *iommu = pbm->iommu;
1768 unsigned long tsbbase, i, tagbase, database, order; 1768 unsigned long i, tagbase, database;
1769 u32 vdma[2], dma_mask; 1769 u32 vdma[2], dma_mask;
1770 u64 control; 1770 u64 control;
1771 int err, tsbsize; 1771 int err, tsbsize;
@@ -1800,10 +1800,6 @@ static void schizo_pbm_iommu_init(struct pci_pbm_info *pbm)
1800 prom_halt(); 1800 prom_halt();
1801 }; 1801 };
1802 1802
1803 /* Setup initial software IOMMU state. */
1804 spin_lock_init(&iommu->lock);
1805 iommu->ctx_lowest_free = 1;
1806
1807 /* Register addresses, SCHIZO has iommu ctx flushing. */ 1803 /* Register addresses, SCHIZO has iommu ctx flushing. */
1808 iommu->iommu_control = pbm->pbm_regs + SCHIZO_IOMMU_CONTROL; 1804 iommu->iommu_control = pbm->pbm_regs + SCHIZO_IOMMU_CONTROL;
1809 iommu->iommu_tsbbase = pbm->pbm_regs + SCHIZO_IOMMU_TSBBASE; 1805 iommu->iommu_tsbbase = pbm->pbm_regs + SCHIZO_IOMMU_TSBBASE;
@@ -1832,56 +1828,9 @@ static void schizo_pbm_iommu_init(struct pci_pbm_info *pbm)
1832 /* Leave diag mode enabled for full-flushing done 1828 /* Leave diag mode enabled for full-flushing done
1833 * in pci_iommu.c 1829 * in pci_iommu.c
1834 */ 1830 */
1831 pci_iommu_table_init(iommu, tsbsize * 8 * 1024, vdma[0], dma_mask);
1835 1832
1836 iommu->dummy_page = __get_free_pages(GFP_KERNEL, 0); 1833 schizo_write(iommu->iommu_tsbbase, __pa(iommu->page_table));
1837 if (!iommu->dummy_page) {
1838 prom_printf("PSYCHO_IOMMU: Error, gfp(dummy_page) failed.\n");
1839 prom_halt();
1840 }
1841 memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
1842 iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);
1843
1844 /* Using assumed page size 8K with 128K entries we need 1MB iommu page
1845 * table (128K ioptes * 8 bytes per iopte). This is
1846 * page order 7 on UltraSparc.
1847 */
1848 order = get_order(tsbsize * 8 * 1024);
1849 tsbbase = __get_free_pages(GFP_KERNEL, order);
1850 if (!tsbbase) {
1851 prom_printf("%s: Error, gfp(tsb) failed.\n", pbm->name);
1852 prom_halt();
1853 }
1854
1855 iommu->page_table = (iopte_t *)tsbbase;
1856 iommu->page_table_map_base = vdma[0];
1857 iommu->dma_addr_mask = dma_mask;
1858 pci_iommu_table_init(iommu, PAGE_SIZE << order);
1859
1860 switch (tsbsize) {
1861 case 64:
1862 iommu->page_table_sz_bits = 16;
1863 break;
1864
1865 case 128:
1866 iommu->page_table_sz_bits = 17;
1867 break;
1868
1869 default:
1870 prom_printf("iommu_init: Illegal TSB size %d\n", tsbsize);
1871 prom_halt();
1872 break;
1873 };
1874
1875 /* We start with no consistent mappings. */
1876 iommu->lowest_consistent_map =
1877 1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS);
1878
1879 for (i = 0; i < PBM_NCLUSTERS; i++) {
1880 iommu->alloc_info[i].flush = 0;
1881 iommu->alloc_info[i].next = 0;
1882 }
1883
1884 schizo_write(iommu->iommu_tsbbase, __pa(tsbbase));
1885 1834
1886 control = schizo_read(iommu->iommu_control); 1835 control = schizo_read(iommu->iommu_control);
1887 control &= ~(SCHIZO_IOMMU_CTRL_TSBSZ | SCHIZO_IOMMU_CTRL_TBWSZ); 1836 control &= ~(SCHIZO_IOMMU_CTRL_TSBSZ | SCHIZO_IOMMU_CTRL_TBWSZ);
diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index 590df5a16f5a..b137fd63f5e1 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -1001,13 +1001,6 @@ void smp_penguin_jailcell(int irq, struct pt_regs *regs)
1001 preempt_enable(); 1001 preempt_enable();
1002} 1002}
1003 1003
1004extern unsigned long xcall_promstop;
1005
1006void smp_promstop_others(void)
1007{
1008 smp_cross_call(&xcall_promstop, 0, 0, 0);
1009}
1010
1011#define prof_multiplier(__cpu) cpu_data(__cpu).multiplier 1004#define prof_multiplier(__cpu) cpu_data(__cpu).multiplier
1012#define prof_counter(__cpu) cpu_data(__cpu).counter 1005#define prof_counter(__cpu) cpu_data(__cpu).counter
1013 1006
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 0d2e967c7200..1e44ee26cee8 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -105,7 +105,7 @@ static void __init read_obp_memory(const char *property,
105 regs[i].phys_addr = base; 105 regs[i].phys_addr = base;
106 regs[i].reg_size = size; 106 regs[i].reg_size = size;
107 } 107 }
108 sort(regs, ents, sizeof(struct linux_prom64_registers), 108 sort(regs, ents, sizeof(struct linux_prom64_registers),
109 cmp_p64, NULL); 109 cmp_p64, NULL);
110} 110}
111 111
@@ -367,8 +367,11 @@ struct linux_prom_translation {
367 unsigned long size; 367 unsigned long size;
368 unsigned long data; 368 unsigned long data;
369}; 369};
370static struct linux_prom_translation prom_trans[512] __initdata; 370
371static unsigned int prom_trans_ents __initdata; 371/* Exported for kernel TLB miss handling in ktlb.S */
372struct linux_prom_translation prom_trans[512] __read_mostly;
373unsigned int prom_trans_ents __read_mostly;
374unsigned int swapper_pgd_zero __read_mostly;
372 375
373extern unsigned long prom_boot_page; 376extern unsigned long prom_boot_page;
374extern void prom_remap(unsigned long physpage, unsigned long virtpage, int mmu_ihandle); 377extern void prom_remap(unsigned long physpage, unsigned long virtpage, int mmu_ihandle);
@@ -378,122 +381,57 @@ extern void register_prom_callbacks(void);
378/* Exported for SMP bootup purposes. */ 381/* Exported for SMP bootup purposes. */
379unsigned long kern_locked_tte_data; 382unsigned long kern_locked_tte_data;
380 383
381/* Exported for kernel TLB miss handling in ktlb.S */
382unsigned long prom_pmd_phys __read_mostly;
383unsigned int swapper_pgd_zero __read_mostly;
384
385static pmd_t *prompmd __read_mostly;
386
387#define BASE_PAGE_SIZE 8192
388
389/* 384/*
390 * Translate PROM's mapping we capture at boot time into physical address. 385 * Translate PROM's mapping we capture at boot time into physical address.
391 * The second parameter is only set from prom_callback() invocations. 386 * The second parameter is only set from prom_callback() invocations.
392 */ 387 */
393unsigned long prom_virt_to_phys(unsigned long promva, int *error) 388unsigned long prom_virt_to_phys(unsigned long promva, int *error)
394{ 389{
395 pmd_t *pmdp = prompmd + ((promva >> 23) & 0x7ff); 390 int i;
396 pte_t *ptep;
397 unsigned long base;
398
399 if (pmd_none(*pmdp)) {
400 if (error)
401 *error = 1;
402 return 0;
403 }
404 ptep = (pte_t *)__pmd_page(*pmdp) + ((promva >> 13) & 0x3ff);
405 if (!pte_present(*ptep)) {
406 if (error)
407 *error = 1;
408 return 0;
409 }
410 if (error) {
411 *error = 0;
412 return pte_val(*ptep);
413 }
414 base = pte_val(*ptep) & _PAGE_PADDR;
415
416 return base + (promva & (BASE_PAGE_SIZE - 1));
417}
418 391
419/* The obp translations are saved based on 8k pagesize, since obp can 392 for (i = 0; i < prom_trans_ents; i++) {
420 * use a mixture of pagesizes. Misses to the LOW_OBP_ADDRESS -> 393 struct linux_prom_translation *p = &prom_trans[i];
421 * HI_OBP_ADDRESS range are handled in entry.S and do not use the vpte
422 * scheme (also, see rant in inherit_locked_prom_mappings()).
423 */
424static void __init build_obp_range(unsigned long start, unsigned long end, unsigned long data)
425{
426 unsigned long vaddr;
427 394
428 for (vaddr = start; vaddr < end; vaddr += BASE_PAGE_SIZE) { 395 if (promva >= p->virt &&
429 unsigned long val; 396 promva < (p->virt + p->size)) {
430 pmd_t *pmd; 397 unsigned long base = p->data & _PAGE_PADDR;
431 pte_t *pte;
432 398
433 pmd = prompmd + ((vaddr >> 23) & 0x7ff); 399 if (error)
434 if (pmd_none(*pmd)) { 400 *error = 0;
435 pte = __alloc_bootmem(BASE_PAGE_SIZE, BASE_PAGE_SIZE, 401 return base + (promva & (8192 - 1));
436 PAGE_SIZE);
437 if (!pte)
438 prom_halt();
439 memset(pte, 0, BASE_PAGE_SIZE);
440 pmd_set(pmd, pte);
441 } 402 }
442 pte = (pte_t *) __pmd_page(*pmd) + ((vaddr >> 13) & 0x3ff);
443
444 val = data;
445
446 /* Clear diag TTE bits. */
447 if (tlb_type == spitfire)
448 val &= ~0x0003fe0000000000UL;
449
450 set_pte_at(&init_mm, vaddr, pte,
451 __pte(val | _PAGE_MODIFIED));
452
453 data += BASE_PAGE_SIZE;
454 } 403 }
404 if (error)
405 *error = 1;
406 return 0UL;
455} 407}
456 408
409/* The obp translations are saved based on 8k pagesize, since obp can
410 * use a mixture of pagesizes. Misses to the LOW_OBP_ADDRESS ->
411 * HI_OBP_ADDRESS range are handled in ktlb.S and do not use the vpte
412 * scheme (also, see rant in inherit_locked_prom_mappings()).
413 */
457static inline int in_obp_range(unsigned long vaddr) 414static inline int in_obp_range(unsigned long vaddr)
458{ 415{
459 return (vaddr >= LOW_OBP_ADDRESS && 416 return (vaddr >= LOW_OBP_ADDRESS &&
460 vaddr < HI_OBP_ADDRESS); 417 vaddr < HI_OBP_ADDRESS);
461} 418}
462 419
463#define OBP_PMD_SIZE 2048 420static int cmp_ptrans(const void *a, const void *b)
464static void __init build_obp_pgtable(void)
465{ 421{
466 unsigned long i; 422 const struct linux_prom_translation *x = a, *y = b;
467
468 prompmd = __alloc_bootmem(OBP_PMD_SIZE, OBP_PMD_SIZE, PAGE_SIZE);
469 if (!prompmd)
470 prom_halt();
471
472 memset(prompmd, 0, OBP_PMD_SIZE);
473
474 prom_pmd_phys = __pa(prompmd);
475
476 for (i = 0; i < prom_trans_ents; i++) {
477 unsigned long start, end;
478
479 if (!in_obp_range(prom_trans[i].virt))
480 continue;
481 423
482 start = prom_trans[i].virt; 424 if (x->virt > y->virt)
483 end = start + prom_trans[i].size; 425 return 1;
484 if (end > HI_OBP_ADDRESS) 426 if (x->virt < y->virt)
485 end = HI_OBP_ADDRESS; 427 return -1;
486 428 return 0;
487 build_obp_range(start, end, prom_trans[i].data);
488 }
489} 429}
490 430
491/* Read OBP translations property into 'prom_trans[]'. 431/* Read OBP translations property into 'prom_trans[]'. */
492 * Return the number of entries.
493 */
494static void __init read_obp_translations(void) 432static void __init read_obp_translations(void)
495{ 433{
496 int n, node; 434 int n, node, ents, first, last, i;
497 435
498 node = prom_finddevice("/virtual-memory"); 436 node = prom_finddevice("/virtual-memory");
499 n = prom_getproplen(node, "translations"); 437 n = prom_getproplen(node, "translations");
@@ -515,7 +453,41 @@ static void __init read_obp_translations(void)
515 453
516 n = n / sizeof(struct linux_prom_translation); 454 n = n / sizeof(struct linux_prom_translation);
517 455
518 prom_trans_ents = n; 456 ents = n;
457
458 sort(prom_trans, ents, sizeof(struct linux_prom_translation),
459 cmp_ptrans, NULL);
460
461 /* Now kick out all the non-OBP entries. */
462 for (i = 0; i < ents; i++) {
463 if (in_obp_range(prom_trans[i].virt))
464 break;
465 }
466 first = i;
467 for (; i < ents; i++) {
468 if (!in_obp_range(prom_trans[i].virt))
469 break;
470 }
471 last = i;
472
473 for (i = 0; i < (last - first); i++) {
474 struct linux_prom_translation *src = &prom_trans[i + first];
475 struct linux_prom_translation *dest = &prom_trans[i];
476
477 *dest = *src;
478 }
479 for (; i < ents; i++) {
480 struct linux_prom_translation *dest = &prom_trans[i];
481 dest->virt = dest->size = dest->data = 0x0UL;
482 }
483
484 prom_trans_ents = last - first;
485
486 if (tlb_type == spitfire) {
487 /* Clear diag TTE bits. */
488 for (i = 0; i < prom_trans_ents; i++)
489 prom_trans[i].data &= ~0x0003fe0000000000UL;
490 }
519} 491}
520 492
521static void __init remap_kernel(void) 493static void __init remap_kernel(void)
@@ -553,21 +525,18 @@ static void __init remap_kernel(void)
553} 525}
554 526
555 527
556static void __init inherit_prom_mappings_pre(void) 528static void __init inherit_prom_mappings(void)
557{ 529{
558 read_obp_translations(); 530 read_obp_translations();
559 531
560 /* Now fixup OBP's idea about where we really are mapped. */ 532 /* Now fixup OBP's idea about where we really are mapped. */
561 prom_printf("Remapping the kernel... "); 533 prom_printf("Remapping the kernel... ");
562 remap_kernel(); 534 remap_kernel();
563
564 prom_printf("done.\n"); 535 prom_printf("done.\n");
565}
566 536
567static void __init inherit_prom_mappings_post(void) 537 prom_printf("Registering callbacks... ");
568{
569 build_obp_pgtable();
570 register_prom_callbacks(); 538 register_prom_callbacks();
539 prom_printf("done.\n");
571} 540}
572 541
573/* The OBP specifications for sun4u mark 0xfffffffc00000000 and 542/* The OBP specifications for sun4u mark 0xfffffffc00000000 and
@@ -1519,7 +1488,7 @@ void __init paging_init(void)
1519 1488
1520 swapper_pgd_zero = pgd_val(swapper_pg_dir[0]); 1489 swapper_pgd_zero = pgd_val(swapper_pg_dir[0]);
1521 1490
1522 inherit_prom_mappings_pre(); 1491 inherit_prom_mappings();
1523 1492
1524 /* Ok, we can use our TLB miss and window trap handlers safely. 1493 /* Ok, we can use our TLB miss and window trap handlers safely.
1525 * We need to do a quick peek here to see if we are on StarFire 1494 * We need to do a quick peek here to see if we are on StarFire
@@ -1530,23 +1499,15 @@ void __init paging_init(void)
1530 extern void setup_tba(int); 1499 extern void setup_tba(int);
1531 setup_tba(this_is_starfire); 1500 setup_tba(this_is_starfire);
1532 } 1501 }
1533 __flush_tlb_all();
1534 1502
1535 /* Everything from this point forward, until we are done with 1503 inherit_locked_prom_mappings(1);
1536 * inherit_prom_mappings_post(), must complete successfully 1504
1537 * without calling into the firmware. The firwmare page tables 1505 __flush_tlb_all();
1538 * have not been built, but we are running on the Linux kernel's
1539 * trap table.
1540 */
1541 1506
1542 /* Setup bootmem... */ 1507 /* Setup bootmem... */
1543 pages_avail = 0; 1508 pages_avail = 0;
1544 last_valid_pfn = end_pfn = bootmem_init(&pages_avail); 1509 last_valid_pfn = end_pfn = bootmem_init(&pages_avail);
1545 1510
1546 inherit_prom_mappings_post();
1547
1548 inherit_locked_prom_mappings(1);
1549
1550#ifdef CONFIG_DEBUG_PAGEALLOC 1511#ifdef CONFIG_DEBUG_PAGEALLOC
1551 kernel_physical_mapping_init(); 1512 kernel_physical_mapping_init();
1552#endif 1513#endif
diff --git a/arch/sparc64/mm/ultra.S b/arch/sparc64/mm/ultra.S
index 058b8126c1a7..e4c9151fa116 100644
--- a/arch/sparc64/mm/ultra.S
+++ b/arch/sparc64/mm/ultra.S
@@ -453,22 +453,6 @@ xcall_flush_dcache_page_spitfire: /* %g1 == physical page address
453 nop 453 nop
454 nop 454 nop
455 455
456 .globl xcall_promstop
457xcall_promstop:
458 rdpr %pstate, %g2
459 wrpr %g2, PSTATE_IG | PSTATE_AG, %pstate
460 rdpr %pil, %g2
461 wrpr %g0, 15, %pil
462 sethi %hi(109f), %g7
463 b,pt %xcc, etrap_irq
464109: or %g7, %lo(109b), %g7
465 flushw
466 call prom_stopself
467 nop
468 /* We should not return, just spin if we do... */
4691: b,a,pt %xcc, 1b
470 nop
471
472 .data 456 .data
473 457
474errata32_hwbug: 458errata32_hwbug:
diff --git a/arch/sparc64/prom/misc.c b/arch/sparc64/prom/misc.c
index 9b895faf077b..87f5cfce23bb 100644
--- a/arch/sparc64/prom/misc.c
+++ b/arch/sparc64/prom/misc.c
@@ -68,19 +68,11 @@ void prom_cmdline(void)
68 local_irq_restore(flags); 68 local_irq_restore(flags);
69} 69}
70 70
71#ifdef CONFIG_SMP
72extern void smp_promstop_others(void);
73#endif
74
75/* Drop into the prom, but completely terminate the program. 71/* Drop into the prom, but completely terminate the program.
76 * No chance of continuing. 72 * No chance of continuing.
77 */ 73 */
78void prom_halt(void) 74void prom_halt(void)
79{ 75{
80#ifdef CONFIG_SMP
81 smp_promstop_others();
82 udelay(8000);
83#endif
84again: 76again:
85 p1275_cmd("exit", P1275_INOUT(0, 0)); 77 p1275_cmd("exit", P1275_INOUT(0, 0));
86 goto again; /* PROM is out to get me -DaveM */ 78 goto again; /* PROM is out to get me -DaveM */
@@ -88,10 +80,6 @@ again:
88 80
89void prom_halt_power_off(void) 81void prom_halt_power_off(void)
90{ 82{
91#ifdef CONFIG_SMP
92 smp_promstop_others();
93 udelay(8000);
94#endif
95 p1275_cmd("SUNW,power-off", P1275_INOUT(0, 0)); 83 p1275_cmd("SUNW,power-off", P1275_INOUT(0, 0));
96 84
97 /* if nothing else helps, we just halt */ 85 /* if nothing else helps, we just halt */
diff --git a/arch/um/drivers/Makefile b/arch/um/drivers/Makefile
index 783e18cae090..de17d4c6e02d 100644
--- a/arch/um/drivers/Makefile
+++ b/arch/um/drivers/Makefile
@@ -13,7 +13,7 @@ mcast-objs := mcast_kern.o mcast_user.o
13net-objs := net_kern.o net_user.o 13net-objs := net_kern.o net_user.o
14mconsole-objs := mconsole_kern.o mconsole_user.o 14mconsole-objs := mconsole_kern.o mconsole_user.o
15hostaudio-objs := hostaudio_kern.o 15hostaudio-objs := hostaudio_kern.o
16ubd-objs := ubd_kern.o 16ubd-objs := ubd_kern.o ubd_user.o
17port-objs := port_kern.o port_user.o 17port-objs := port_kern.o port_user.o
18harddog-objs := harddog_kern.o harddog_user.o 18harddog-objs := harddog_kern.o harddog_user.o
19 19
diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c
index e77a38da4350..f73134333f64 100644
--- a/arch/um/drivers/ubd_kern.c
+++ b/arch/um/drivers/ubd_kern.c
@@ -35,7 +35,6 @@
35#include "linux/blkpg.h" 35#include "linux/blkpg.h"
36#include "linux/genhd.h" 36#include "linux/genhd.h"
37#include "linux/spinlock.h" 37#include "linux/spinlock.h"
38#include "asm/atomic.h"
39#include "asm/segment.h" 38#include "asm/segment.h"
40#include "asm/uaccess.h" 39#include "asm/uaccess.h"
41#include "asm/irq.h" 40#include "asm/irq.h"
@@ -54,21 +53,20 @@
54#include "mem.h" 53#include "mem.h"
55#include "mem_kern.h" 54#include "mem_kern.h"
56#include "cow.h" 55#include "cow.h"
57#include "aio.h"
58 56
59enum ubd_req { UBD_READ, UBD_WRITE }; 57enum ubd_req { UBD_READ, UBD_WRITE };
60 58
61struct io_thread_req { 59struct io_thread_req {
62 enum aio_type op; 60 enum ubd_req op;
63 int fds[2]; 61 int fds[2];
64 unsigned long offsets[2]; 62 unsigned long offsets[2];
65 unsigned long long offset; 63 unsigned long long offset;
66 unsigned long length; 64 unsigned long length;
67 char *buffer; 65 char *buffer;
68 int sectorsize; 66 int sectorsize;
69 int bitmap_offset; 67 unsigned long sector_mask;
70 long bitmap_start; 68 unsigned long long cow_offset;
71 long bitmap_end; 69 unsigned long bitmap_words[2];
72 int error; 70 int error;
73}; 71};
74 72
@@ -82,31 +80,28 @@ extern int create_cow_file(char *cow_file, char *backing_file,
82 unsigned long *bitmap_len_out, 80 unsigned long *bitmap_len_out,
83 int *data_offset_out); 81 int *data_offset_out);
84extern int read_cow_bitmap(int fd, void *buf, int offset, int len); 82extern int read_cow_bitmap(int fd, void *buf, int offset, int len);
85extern void do_io(struct io_thread_req *req, struct request *r, 83extern void do_io(struct io_thread_req *req);
86 unsigned long *bitmap);
87 84
88static inline int ubd_test_bit(__u64 bit, void *data) 85static inline int ubd_test_bit(__u64 bit, unsigned char *data)
89{ 86{
90 unsigned char *buffer = data;
91 __u64 n; 87 __u64 n;
92 int bits, off; 88 int bits, off;
93 89
94 bits = sizeof(buffer[0]) * 8; 90 bits = sizeof(data[0]) * 8;
95 n = bit / bits; 91 n = bit / bits;
96 off = bit % bits; 92 off = bit % bits;
97 return((buffer[n] & (1 << off)) != 0); 93 return((data[n] & (1 << off)) != 0);
98} 94}
99 95
100static inline void ubd_set_bit(__u64 bit, void *data) 96static inline void ubd_set_bit(__u64 bit, unsigned char *data)
101{ 97{
102 unsigned char *buffer = data;
103 __u64 n; 98 __u64 n;
104 int bits, off; 99 int bits, off;
105 100
106 bits = sizeof(buffer[0]) * 8; 101 bits = sizeof(data[0]) * 8;
107 n = bit / bits; 102 n = bit / bits;
108 off = bit % bits; 103 off = bit % bits;
109 buffer[n] |= (1 << off); 104 data[n] |= (1 << off);
110} 105}
111/*End stuff from ubd_user.h*/ 106/*End stuff from ubd_user.h*/
112 107
@@ -115,6 +110,8 @@ static inline void ubd_set_bit(__u64 bit, void *data)
115static DEFINE_SPINLOCK(ubd_io_lock); 110static DEFINE_SPINLOCK(ubd_io_lock);
116static DEFINE_SPINLOCK(ubd_lock); 111static DEFINE_SPINLOCK(ubd_lock);
117 112
113static void (*do_ubd)(void);
114
118static int ubd_open(struct inode * inode, struct file * filp); 115static int ubd_open(struct inode * inode, struct file * filp);
119static int ubd_release(struct inode * inode, struct file * file); 116static int ubd_release(struct inode * inode, struct file * file);
120static int ubd_ioctl(struct inode * inode, struct file * file, 117static int ubd_ioctl(struct inode * inode, struct file * file,
@@ -161,8 +158,6 @@ struct cow {
161 int data_offset; 158 int data_offset;
162}; 159};
163 160
164#define MAX_SG 64
165
166struct ubd { 161struct ubd {
167 char *file; 162 char *file;
168 int count; 163 int count;
@@ -173,7 +168,6 @@ struct ubd {
173 int no_cow; 168 int no_cow;
174 struct cow cow; 169 struct cow cow;
175 struct platform_device pdev; 170 struct platform_device pdev;
176 struct scatterlist sg[MAX_SG];
177}; 171};
178 172
179#define DEFAULT_COW { \ 173#define DEFAULT_COW { \
@@ -466,114 +460,81 @@ __uml_help(fakehd,
466); 460);
467 461
468static void do_ubd_request(request_queue_t * q); 462static void do_ubd_request(request_queue_t * q);
469static int in_ubd; 463
464/* Only changed by ubd_init, which is an initcall. */
465int thread_fd = -1;
470 466
471/* Changed by ubd_handler, which is serialized because interrupts only 467/* Changed by ubd_handler, which is serialized because interrupts only
472 * happen on CPU 0. 468 * happen on CPU 0.
473 */ 469 */
474int intr_count = 0; 470int intr_count = 0;
475 471
476static void ubd_end_request(struct request *req, int bytes, int uptodate) 472/* call ubd_finish if you need to serialize */
473static void __ubd_finish(struct request *req, int error)
477{ 474{
478 if (!end_that_request_first(req, uptodate, bytes >> 9)) { 475 int nsect;
479 add_disk_randomness(req->rq_disk); 476
480 end_that_request_last(req); 477 if(error){
478 end_request(req, 0);
479 return;
481 } 480 }
481 nsect = req->current_nr_sectors;
482 req->sector += nsect;
483 req->buffer += nsect << 9;
484 req->errors = 0;
485 req->nr_sectors -= nsect;
486 req->current_nr_sectors = 0;
487 end_request(req, 1);
482} 488}
483 489
484/* call ubd_finish if you need to serialize */ 490static inline void ubd_finish(struct request *req, int error)
485static void __ubd_finish(struct request *req, int bytes)
486{ 491{
487 if(bytes < 0){ 492 spin_lock(&ubd_io_lock);
488 ubd_end_request(req, 0, 0); 493 __ubd_finish(req, error);
489 return; 494 spin_unlock(&ubd_io_lock);
490 }
491
492 ubd_end_request(req, bytes, 1);
493} 495}
494 496
495static inline void ubd_finish(struct request *req, int bytes) 497/* Called without ubd_io_lock held */
498static void ubd_handler(void)
496{ 499{
497 spin_lock(&ubd_io_lock); 500 struct io_thread_req req;
498 __ubd_finish(req, bytes); 501 struct request *rq = elv_next_request(ubd_queue);
499 spin_unlock(&ubd_io_lock); 502 int n;
503
504 do_ubd = NULL;
505 intr_count++;
506 n = os_read_file(thread_fd, &req, sizeof(req));
507 if(n != sizeof(req)){
508 printk(KERN_ERR "Pid %d - spurious interrupt in ubd_handler, "
509 "err = %d\n", os_getpid(), -n);
510 spin_lock(&ubd_io_lock);
511 end_request(rq, 0);
512 spin_unlock(&ubd_io_lock);
513 return;
514 }
515
516 ubd_finish(rq, req.error);
517 reactivate_fd(thread_fd, UBD_IRQ);
518 do_ubd_request(ubd_queue);
500} 519}
501 520
502struct bitmap_io {
503 atomic_t count;
504 struct aio_context aio;
505};
506
507struct ubd_aio {
508 struct aio_context aio;
509 struct request *req;
510 int len;
511 struct bitmap_io *bitmap;
512 void *bitmap_buf;
513};
514
515static int ubd_reply_fd = -1;
516
517static irqreturn_t ubd_intr(int irq, void *dev, struct pt_regs *unused) 521static irqreturn_t ubd_intr(int irq, void *dev, struct pt_regs *unused)
518{ 522{
519 struct aio_thread_reply reply; 523 ubd_handler();
520 struct ubd_aio *aio; 524 return(IRQ_HANDLED);
521 struct request *req; 525}
522 int err, n, fd = (int) (long) dev;
523
524 while(1){
525 err = os_read_file(fd, &reply, sizeof(reply));
526 if(err == -EAGAIN)
527 break;
528 if(err < 0){
529 printk("ubd_aio_handler - read returned err %d\n",
530 -err);
531 break;
532 }
533
534 aio = container_of(reply.data, struct ubd_aio, aio);
535 n = reply.err;
536
537 if(n == 0){
538 req = aio->req;
539 req->nr_sectors -= aio->len >> 9;
540
541 if((aio->bitmap != NULL) &&
542 (atomic_dec_and_test(&aio->bitmap->count))){
543 aio->aio = aio->bitmap->aio;
544 aio->len = 0;
545 kfree(aio->bitmap);
546 aio->bitmap = NULL;
547 submit_aio(&aio->aio);
548 }
549 else {
550 if((req->nr_sectors == 0) &&
551 (aio->bitmap == NULL)){
552 int len = req->hard_nr_sectors << 9;
553 ubd_finish(req, len);
554 }
555
556 if(aio->bitmap_buf != NULL)
557 kfree(aio->bitmap_buf);
558 kfree(aio);
559 }
560 }
561 else if(n < 0){
562 ubd_finish(aio->req, n);
563 if(aio->bitmap != NULL)
564 kfree(aio->bitmap);
565 if(aio->bitmap_buf != NULL)
566 kfree(aio->bitmap_buf);
567 kfree(aio);
568 }
569 }
570 reactivate_fd(fd, UBD_IRQ);
571 526
572 do_ubd_request(ubd_queue); 527/* Only changed by ubd_init, which is an initcall. */
528static int io_pid = -1;
573 529
574 return(IRQ_HANDLED); 530void kill_io_thread(void)
531{
532 if(io_pid != -1)
533 os_kill_process(io_pid, 1);
575} 534}
576 535
536__uml_exitcall(kill_io_thread);
537
577static int ubd_file_size(struct ubd *dev, __u64 *size_out) 538static int ubd_file_size(struct ubd *dev, __u64 *size_out)
578{ 539{
579 char *file; 540 char *file;
@@ -608,7 +569,7 @@ static int ubd_open_dev(struct ubd *dev)
608 &dev->cow.data_offset, create_ptr); 569 &dev->cow.data_offset, create_ptr);
609 570
610 if((dev->fd == -ENOENT) && create_cow){ 571 if((dev->fd == -ENOENT) && create_cow){
611 dev->fd = create_cow_file(dev->file, dev->cow.file, 572 dev->fd = create_cow_file(dev->file, dev->cow.file,
612 dev->openflags, 1 << 9, PAGE_SIZE, 573 dev->openflags, 1 << 9, PAGE_SIZE,
613 &dev->cow.bitmap_offset, 574 &dev->cow.bitmap_offset,
614 &dev->cow.bitmap_len, 575 &dev->cow.bitmap_len,
@@ -870,10 +831,6 @@ int ubd_init(void)
870{ 831{
871 int i; 832 int i;
872 833
873 ubd_reply_fd = init_aio_irq(UBD_IRQ, "ubd", ubd_intr);
874 if(ubd_reply_fd < 0)
875 printk("Setting up ubd AIO failed, err = %d\n", ubd_reply_fd);
876
877 devfs_mk_dir("ubd"); 834 devfs_mk_dir("ubd");
878 if (register_blkdev(MAJOR_NR, "ubd")) 835 if (register_blkdev(MAJOR_NR, "ubd"))
879 return -1; 836 return -1;
@@ -884,7 +841,6 @@ int ubd_init(void)
884 return -1; 841 return -1;
885 } 842 }
886 843
887 blk_queue_max_hw_segments(ubd_queue, MAX_SG);
888 if (fake_major != MAJOR_NR) { 844 if (fake_major != MAJOR_NR) {
889 char name[sizeof("ubd_nnn\0")]; 845 char name[sizeof("ubd_nnn\0")];
890 846
@@ -896,12 +852,40 @@ int ubd_init(void)
896 driver_register(&ubd_driver); 852 driver_register(&ubd_driver);
897 for (i = 0; i < MAX_DEV; i++) 853 for (i = 0; i < MAX_DEV; i++)
898 ubd_add(i); 854 ubd_add(i);
899
900 return 0; 855 return 0;
901} 856}
902 857
903late_initcall(ubd_init); 858late_initcall(ubd_init);
904 859
860int ubd_driver_init(void){
861 unsigned long stack;
862 int err;
863
864 /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/
865 if(global_openflags.s){
866 printk(KERN_INFO "ubd: Synchronous mode\n");
867 /* Letting ubd=sync be like using ubd#s= instead of ubd#= is
868 * enough. So use anyway the io thread. */
869 }
870 stack = alloc_stack(0, 0);
871 io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *),
872 &thread_fd);
873 if(io_pid < 0){
874 printk(KERN_ERR
875 "ubd : Failed to start I/O thread (errno = %d) - "
876 "falling back to synchronous I/O\n", -io_pid);
877 io_pid = -1;
878 return(0);
879 }
880 err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr,
881 SA_INTERRUPT, "ubd", ubd_dev);
882 if(err != 0)
883 printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err);
884 return(err);
885}
886
887device_initcall(ubd_driver_init);
888
905static int ubd_open(struct inode *inode, struct file *filp) 889static int ubd_open(struct inode *inode, struct file *filp)
906{ 890{
907 struct gendisk *disk = inode->i_bdev->bd_disk; 891 struct gendisk *disk = inode->i_bdev->bd_disk;
@@ -939,55 +923,105 @@ static int ubd_release(struct inode * inode, struct file * file)
939 return(0); 923 return(0);
940} 924}
941 925
942static void cowify_bitmap(struct io_thread_req *req, unsigned long *bitmap) 926static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask,
927 __u64 *cow_offset, unsigned long *bitmap,
928 __u64 bitmap_offset, unsigned long *bitmap_words,
929 __u64 bitmap_len)
943{ 930{
944 __u64 sector = req->offset / req->sectorsize; 931 __u64 sector = io_offset >> 9;
945 int i; 932 int i, update_bitmap = 0;
933
934 for(i = 0; i < length >> 9; i++){
935 if(cow_mask != NULL)
936 ubd_set_bit(i, (unsigned char *) cow_mask);
937 if(ubd_test_bit(sector + i, (unsigned char *) bitmap))
938 continue;
946 939
947 for(i = 0; i < req->length / req->sectorsize; i++){ 940 update_bitmap = 1;
948 if(ubd_test_bit(sector + i, bitmap)) 941 ubd_set_bit(sector + i, (unsigned char *) bitmap);
949 continue; 942 }
943
944 if(!update_bitmap)
945 return;
950 946
951 if(req->bitmap_start == -1) 947 *cow_offset = sector / (sizeof(unsigned long) * 8);
952 req->bitmap_start = sector + i;
953 req->bitmap_end = sector + i + 1;
954 948
955 ubd_set_bit(sector + i, bitmap); 949 /* This takes care of the case where we're exactly at the end of the
956 } 950 * device, and *cow_offset + 1 is off the end. So, just back it up
951 * by one word. Thanks to Lynn Kerby for the fix and James McMechan
952 * for the original diagnosis.
953 */
954 if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) /
955 sizeof(unsigned long) - 1))
956 (*cow_offset)--;
957
958 bitmap_words[0] = bitmap[*cow_offset];
959 bitmap_words[1] = bitmap[*cow_offset + 1];
960
961 *cow_offset *= sizeof(unsigned long);
962 *cow_offset += bitmap_offset;
963}
964
965static void cowify_req(struct io_thread_req *req, unsigned long *bitmap,
966 __u64 bitmap_offset, __u64 bitmap_len)
967{
968 __u64 sector = req->offset >> 9;
969 int i;
970
971 if(req->length > (sizeof(req->sector_mask) * 8) << 9)
972 panic("Operation too long");
973
974 if(req->op == UBD_READ) {
975 for(i = 0; i < req->length >> 9; i++){
976 if(ubd_test_bit(sector + i, (unsigned char *) bitmap))
977 ubd_set_bit(i, (unsigned char *)
978 &req->sector_mask);
979 }
980 }
981 else cowify_bitmap(req->offset, req->length, &req->sector_mask,
982 &req->cow_offset, bitmap, bitmap_offset,
983 req->bitmap_words, bitmap_len);
957} 984}
958 985
959/* Called with ubd_io_lock held */ 986/* Called with ubd_io_lock held */
960static int prepare_request(struct request *req, struct io_thread_req *io_req, 987static int prepare_request(struct request *req, struct io_thread_req *io_req)
961 unsigned long long offset, int page_offset,
962 int len, struct page *page)
963{ 988{
964 struct gendisk *disk = req->rq_disk; 989 struct gendisk *disk = req->rq_disk;
965 struct ubd *dev = disk->private_data; 990 struct ubd *dev = disk->private_data;
991 __u64 offset;
992 int len;
993
994 if(req->rq_status == RQ_INACTIVE) return(1);
966 995
967 /* This should be impossible now */ 996 /* This should be impossible now */
968 if((rq_data_dir(req) == WRITE) && !dev->openflags.w){ 997 if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
969 printk("Write attempted on readonly ubd device %s\n", 998 printk("Write attempted on readonly ubd device %s\n",
970 disk->disk_name); 999 disk->disk_name);
971 ubd_end_request(req, 0, 0); 1000 end_request(req, 0);
972 return(1); 1001 return(1);
973 } 1002 }
974 1003
1004 offset = ((__u64) req->sector) << 9;
1005 len = req->current_nr_sectors << 9;
1006
975 io_req->fds[0] = (dev->cow.file != NULL) ? dev->cow.fd : dev->fd; 1007 io_req->fds[0] = (dev->cow.file != NULL) ? dev->cow.fd : dev->fd;
976 io_req->fds[1] = dev->fd; 1008 io_req->fds[1] = dev->fd;
1009 io_req->cow_offset = -1;
977 io_req->offset = offset; 1010 io_req->offset = offset;
978 io_req->length = len; 1011 io_req->length = len;
979 io_req->error = 0; 1012 io_req->error = 0;
980 io_req->op = (rq_data_dir(req) == READ) ? AIO_READ : AIO_WRITE; 1013 io_req->sector_mask = 0;
1014
1015 io_req->op = (rq_data_dir(req) == READ) ? UBD_READ : UBD_WRITE;
981 io_req->offsets[0] = 0; 1016 io_req->offsets[0] = 0;
982 io_req->offsets[1] = dev->cow.data_offset; 1017 io_req->offsets[1] = dev->cow.data_offset;
983 io_req->buffer = page_address(page) + page_offset; 1018 io_req->buffer = req->buffer;
984 io_req->sectorsize = 1 << 9; 1019 io_req->sectorsize = 1 << 9;
985 io_req->bitmap_offset = dev->cow.bitmap_offset;
986 io_req->bitmap_start = -1;
987 io_req->bitmap_end = -1;
988 1020
989 if((dev->cow.file != NULL) && (io_req->op == UBD_WRITE)) 1021 if(dev->cow.file != NULL)
990 cowify_bitmap(io_req, dev->cow.bitmap); 1022 cowify_req(io_req, dev->cow.bitmap, dev->cow.bitmap_offset,
1023 dev->cow.bitmap_len);
1024
991 return(0); 1025 return(0);
992} 1026}
993 1027
@@ -996,36 +1030,30 @@ static void do_ubd_request(request_queue_t *q)
996{ 1030{
997 struct io_thread_req io_req; 1031 struct io_thread_req io_req;
998 struct request *req; 1032 struct request *req;
999 __u64 sector; 1033 int err, n;
1000 int err; 1034
1001 1035 if(thread_fd == -1){
1002 if(in_ubd) 1036 while((req = elv_next_request(q)) != NULL){
1003 return; 1037 err = prepare_request(req, &io_req);
1004 in_ubd = 1; 1038 if(!err){
1005 while((req = elv_next_request(q)) != NULL){ 1039 do_io(&io_req);
1006 struct gendisk *disk = req->rq_disk; 1040 __ubd_finish(req, io_req.error);
1007 struct ubd *dev = disk->private_data; 1041 }
1008 int n, i; 1042 }
1009 1043 }
1010 blkdev_dequeue_request(req); 1044 else {
1011 1045 if(do_ubd || (req = elv_next_request(q)) == NULL)
1012 sector = req->sector; 1046 return;
1013 n = blk_rq_map_sg(q, req, dev->sg); 1047 err = prepare_request(req, &io_req);
1014 1048 if(!err){
1015 for(i = 0; i < n; i++){ 1049 do_ubd = ubd_handler;
1016 struct scatterlist *sg = &dev->sg[i]; 1050 n = os_write_file(thread_fd, (char *) &io_req,
1017 1051 sizeof(io_req));
1018 err = prepare_request(req, &io_req, sector << 9, 1052 if(n != sizeof(io_req))
1019 sg->offset, sg->length, 1053 printk("write to io thread failed, "
1020 sg->page); 1054 "errno = %d\n", -n);
1021 if(err)
1022 continue;
1023
1024 sector += sg->length >> 9;
1025 do_io(&io_req, req, dev->cow.bitmap);
1026 } 1055 }
1027 } 1056 }
1028 in_ubd = 0;
1029} 1057}
1030 1058
1031static int ubd_ioctl(struct inode * inode, struct file * file, 1059static int ubd_ioctl(struct inode * inode, struct file * file,
@@ -1241,95 +1269,131 @@ int create_cow_file(char *cow_file, char *backing_file, struct openflags flags,
1241 return(err); 1269 return(err);
1242} 1270}
1243 1271
1244void do_io(struct io_thread_req *req, struct request *r, unsigned long *bitmap) 1272static int update_bitmap(struct io_thread_req *req)
1245{ 1273{
1246 struct ubd_aio *aio; 1274 int n;
1247 struct bitmap_io *bitmap_io = NULL;
1248 char *buf;
1249 void *bitmap_buf = NULL;
1250 unsigned long len, sector;
1251 int nsectors, start, end, bit, err;
1252 __u64 off;
1253
1254 if(req->bitmap_start != -1){
1255 /* Round up to the nearest word */
1256 int round = sizeof(unsigned long);
1257 len = (req->bitmap_end - req->bitmap_start +
1258 round * 8 - 1) / (round * 8);
1259 len *= round;
1260
1261 off = req->bitmap_start / (8 * round);
1262 off *= round;
1263
1264 bitmap_io = kmalloc(sizeof(*bitmap_io), GFP_KERNEL);
1265 if(bitmap_io == NULL){
1266 printk("Failed to kmalloc bitmap IO\n");
1267 req->error = 1;
1268 return;
1269 }
1270 1275
1271 bitmap_buf = kmalloc(len, GFP_KERNEL); 1276 if(req->cow_offset == -1)
1272 if(bitmap_buf == NULL){ 1277 return(0);
1273 printk("do_io : kmalloc of bitmap chunk "
1274 "failed\n");
1275 kfree(bitmap_io);
1276 req->error = 1;
1277 return;
1278 }
1279 memcpy(bitmap_buf, &bitmap[off / sizeof(bitmap[0])], len);
1280
1281 *bitmap_io = ((struct bitmap_io)
1282 { .count = ATOMIC_INIT(0),
1283 .aio = INIT_AIO(AIO_WRITE, req->fds[1],
1284 bitmap_buf, len,
1285 req->bitmap_offset + off,
1286 ubd_reply_fd) } );
1287 }
1288 1278
1289 nsectors = req->length / req->sectorsize; 1279 n = os_seek_file(req->fds[1], req->cow_offset);
1290 start = 0; 1280 if(n < 0){
1291 end = nsectors; 1281 printk("do_io - bitmap lseek failed : err = %d\n", -n);
1292 bit = 0; 1282 return(1);
1293 do { 1283 }
1294 if(bitmap != NULL){
1295 sector = req->offset / req->sectorsize;
1296 bit = ubd_test_bit(sector + start, bitmap);
1297 end = start;
1298 while((end < nsectors) &&
1299 (ubd_test_bit(sector + end, bitmap) == bit))
1300 end++;
1301 }
1302 1284
1303 off = req->offsets[bit] + req->offset + 1285 n = os_write_file(req->fds[1], &req->bitmap_words,
1304 start * req->sectorsize; 1286 sizeof(req->bitmap_words));
1305 len = (end - start) * req->sectorsize; 1287 if(n != sizeof(req->bitmap_words)){
1306 buf = &req->buffer[start * req->sectorsize]; 1288 printk("do_io - bitmap update failed, err = %d fd = %d\n", -n,
1289 req->fds[1]);
1290 return(1);
1291 }
1307 1292
1308 aio = kmalloc(sizeof(*aio), GFP_KERNEL); 1293 return(0);
1309 if(aio == NULL){ 1294}
1310 req->error = 1;
1311 return;
1312 }
1313 1295
1314 *aio = ((struct ubd_aio) 1296void do_io(struct io_thread_req *req)
1315 { .aio = INIT_AIO(req->op, req->fds[bit], buf, 1297{
1316 len, off, ubd_reply_fd), 1298 char *buf;
1317 .len = len, 1299 unsigned long len;
1318 .req = r, 1300 int n, nsectors, start, end, bit;
1319 .bitmap = bitmap_io, 1301 int err;
1320 .bitmap_buf = bitmap_buf }); 1302 __u64 off;
1321 1303
1322 if(aio->bitmap != NULL) 1304 nsectors = req->length / req->sectorsize;
1323 atomic_inc(&aio->bitmap->count); 1305 start = 0;
1324 1306 do {
1325 err = submit_aio(&aio->aio); 1307 bit = ubd_test_bit(start, (unsigned char *) &req->sector_mask);
1326 if(err){ 1308 end = start;
1327 printk("do_io - submit_aio failed, " 1309 while((end < nsectors) &&
1328 "err = %d\n", err); 1310 (ubd_test_bit(end, (unsigned char *)
1329 req->error = 1; 1311 &req->sector_mask) == bit))
1330 return; 1312 end++;
1331 } 1313
1314 off = req->offset + req->offsets[bit] +
1315 start * req->sectorsize;
1316 len = (end - start) * req->sectorsize;
1317 buf = &req->buffer[start * req->sectorsize];
1318
1319 err = os_seek_file(req->fds[bit], off);
1320 if(err < 0){
1321 printk("do_io - lseek failed : err = %d\n", -err);
1322 req->error = 1;
1323 return;
1324 }
1325 if(req->op == UBD_READ){
1326 n = 0;
1327 do {
1328 buf = &buf[n];
1329 len -= n;
1330 n = os_read_file(req->fds[bit], buf, len);
1331 if (n < 0) {
1332 printk("do_io - read failed, err = %d "
1333 "fd = %d\n", -n, req->fds[bit]);
1334 req->error = 1;
1335 return;
1336 }
1337 } while((n < len) && (n != 0));
1338 if (n < len) memset(&buf[n], 0, len - n);
1339 } else {
1340 n = os_write_file(req->fds[bit], buf, len);
1341 if(n != len){
1342 printk("do_io - write failed err = %d "
1343 "fd = %d\n", -n, req->fds[bit]);
1344 req->error = 1;
1345 return;
1346 }
1347 }
1348
1349 start = end;
1350 } while(start < nsectors);
1332 1351
1333 start = end; 1352 req->error = update_bitmap(req);
1334 } while(start < nsectors);
1335} 1353}
1354
1355/* Changed in start_io_thread, which is serialized by being called only
1356 * from ubd_init, which is an initcall.
1357 */
1358int kernel_fd = -1;
1359
1360/* Only changed by the io thread */
1361int io_count = 0;
1362
1363int io_thread(void *arg)
1364{
1365 struct io_thread_req req;
1366 int n;
1367
1368 ignore_sigwinch_sig();
1369 while(1){
1370 n = os_read_file(kernel_fd, &req, sizeof(req));
1371 if(n != sizeof(req)){
1372 if(n < 0)
1373 printk("io_thread - read failed, fd = %d, "
1374 "err = %d\n", kernel_fd, -n);
1375 else {
1376 printk("io_thread - short read, fd = %d, "
1377 "length = %d\n", kernel_fd, n);
1378 }
1379 continue;
1380 }
1381 io_count++;
1382 do_io(&req);
1383 n = os_write_file(kernel_fd, &req, sizeof(req));
1384 if(n != sizeof(req))
1385 printk("io_thread - write failed, fd = %d, err = %d\n",
1386 kernel_fd, -n);
1387 }
1388}
1389
1390/*
1391 * Overrides for Emacs so that we follow Linus's tabbing style.
1392 * Emacs will notice this stuff at the end of the file and automatically
1393 * adjust the settings for this buffer only. This must remain at the end
1394 * of the file.
1395 * ---------------------------------------------------------------------------
1396 * Local variables:
1397 * c-file-style: "linux"
1398 * End:
1399 */
diff --git a/arch/um/drivers/ubd_user.c b/arch/um/drivers/ubd_user.c
new file mode 100644
index 000000000000..b94d2bc4fe06
--- /dev/null
+++ b/arch/um/drivers/ubd_user.c
@@ -0,0 +1,75 @@
1/*
2 * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com)
3 * Copyright (C) 2001 Ridgerun,Inc (glonnon@ridgerun.com)
4 * Licensed under the GPL
5 */
6
7#include <stddef.h>
8#include <unistd.h>
9#include <errno.h>
10#include <sched.h>
11#include <signal.h>
12#include <string.h>
13#include <netinet/in.h>
14#include <sys/time.h>
15#include <sys/socket.h>
16#include <sys/mman.h>
17#include <sys/param.h>
18#include "asm/types.h"
19#include "user_util.h"
20#include "kern_util.h"
21#include "user.h"
22#include "ubd_user.h"
23#include "os.h"
24#include "cow.h"
25
26#include <endian.h>
27#include <byteswap.h>
28
29void ignore_sigwinch_sig(void)
30{
31 signal(SIGWINCH, SIG_IGN);
32}
33
34int start_io_thread(unsigned long sp, int *fd_out)
35{
36 int pid, fds[2], err;
37
38 err = os_pipe(fds, 1, 1);
39 if(err < 0){
40 printk("start_io_thread - os_pipe failed, err = %d\n", -err);
41 goto out;
42 }
43
44 kernel_fd = fds[0];
45 *fd_out = fds[1];
46
47 pid = clone(io_thread, (void *) sp, CLONE_FILES | CLONE_VM | SIGCHLD,
48 NULL);
49 if(pid < 0){
50 printk("start_io_thread - clone failed : errno = %d\n", errno);
51 err = -errno;
52 goto out_close;
53 }
54
55 return(pid);
56
57 out_close:
58 os_close_file(fds[0]);
59 os_close_file(fds[1]);
60 kernel_fd = -1;
61 *fd_out = -1;
62 out:
63 return(err);
64}
65
66/*
67 * Overrides for Emacs so that we follow Linus's tabbing style.
68 * Emacs will notice this stuff at the end of the file and automatically
69 * adjust the settings for this buffer only. This must remain at the end
70 * of the file.
71 * ---------------------------------------------------------------------------
72 * Local variables:
73 * c-file-style: "linux"
74 * End:
75 */
diff --git a/arch/um/include/aio.h b/arch/um/include/aio.h
index 83f16877ab08..423bae9153f8 100644
--- a/arch/um/include/aio.h
+++ b/arch/um/include/aio.h
@@ -14,27 +14,15 @@ struct aio_thread_reply {
14}; 14};
15 15
16struct aio_context { 16struct aio_context {
17 enum aio_type type;
18 int fd;
19 void *data;
20 int len;
21 unsigned long long offset;
22 int reply_fd; 17 int reply_fd;
23 struct aio_context *next; 18 struct aio_context *next;
24}; 19};
25 20
26#define INIT_AIO(aio_type, aio_fd, aio_data, aio_len, aio_offset, \
27 aio_reply_fd) \
28 { .type = aio_type, \
29 .fd = aio_fd, \
30 .data = aio_data, \
31 .len = aio_len, \
32 .offset = aio_offset, \
33 .reply_fd = aio_reply_fd }
34
35#define INIT_AIO_CONTEXT { .reply_fd = -1, \ 21#define INIT_AIO_CONTEXT { .reply_fd = -1, \
36 .next = NULL } 22 .next = NULL }
37 23
38extern int submit_aio(struct aio_context *aio); 24extern int submit_aio(enum aio_type type, int fd, char *buf, int len,
25 unsigned long long offset, int reply_fd,
26 struct aio_context *aio);
39 27
40#endif 28#endif
diff --git a/arch/um/include/os.h b/arch/um/include/os.h
index 6f766e1faecc..2e58e304b8be 100644
--- a/arch/um/include/os.h
+++ b/arch/um/include/os.h
@@ -6,6 +6,7 @@
6#ifndef __OS_H__ 6#ifndef __OS_H__
7#define __OS_H__ 7#define __OS_H__
8 8
9#include "uml-config.h"
9#include "asm/types.h" 10#include "asm/types.h"
10#include "../os/include/file.h" 11#include "../os/include/file.h"
11 12
@@ -159,7 +160,11 @@ extern int can_do_skas(void);
159 160
160/* Make sure they are clear when running in TT mode. Required by 161/* Make sure they are clear when running in TT mode. Required by
161 * SEGV_MAYBE_FIXABLE */ 162 * SEGV_MAYBE_FIXABLE */
163#ifdef UML_CONFIG_MODE_SKAS
162#define clear_can_do_skas() do { ptrace_faultinfo = proc_mm = 0; } while (0) 164#define clear_can_do_skas() do { ptrace_faultinfo = proc_mm = 0; } while (0)
165#else
166#define clear_can_do_skas() do {} while (0)
167#endif
163 168
164/* mem.c */ 169/* mem.c */
165extern int create_mem_file(unsigned long len); 170extern int create_mem_file(unsigned long len);
diff --git a/arch/um/os-Linux/aio.c b/arch/um/os-Linux/aio.c
index f6e64026f995..41cfb0944201 100644
--- a/arch/um/os-Linux/aio.c
+++ b/arch/um/os-Linux/aio.c
@@ -6,7 +6,6 @@
6#include <stdlib.h> 6#include <stdlib.h>
7#include <unistd.h> 7#include <unistd.h>
8#include <signal.h> 8#include <signal.h>
9#include <string.h>
10#include <errno.h> 9#include <errno.h>
11#include <sched.h> 10#include <sched.h>
12#include <sys/syscall.h> 11#include <sys/syscall.h>
@@ -17,31 +16,18 @@
17#include "user.h" 16#include "user.h"
18#include "mode.h" 17#include "mode.h"
19 18
19struct aio_thread_req {
20 enum aio_type type;
21 int io_fd;
22 unsigned long long offset;
23 char *buf;
24 int len;
25 struct aio_context *aio;
26};
27
20static int aio_req_fd_r = -1; 28static int aio_req_fd_r = -1;
21static int aio_req_fd_w = -1; 29static int aio_req_fd_w = -1;
22 30
23static int update_aio(struct aio_context *aio, int res)
24{
25 if(res < 0)
26 aio->len = res;
27 else if((res == 0) && (aio->type == AIO_READ)){
28 /* This is the EOF case - we have hit the end of the file
29 * and it ends in a partial block, so we fill the end of
30 * the block with zeros and claim success.
31 */
32 memset(aio->data, 0, aio->len);
33 aio->len = 0;
34 }
35 else if(res > 0){
36 aio->len -= res;
37 aio->data += res;
38 aio->offset += res;
39 return aio->len;
40 }
41
42 return 0;
43}
44
45#if defined(HAVE_AIO_ABI) 31#if defined(HAVE_AIO_ABI)
46#include <linux/aio_abi.h> 32#include <linux/aio_abi.h>
47 33
@@ -80,7 +66,8 @@ static long io_getevents(aio_context_t ctx_id, long min_nr, long nr,
80 * that it now backs the mmapped area. 66 * that it now backs the mmapped area.
81 */ 67 */
82 68
83static int do_aio(aio_context_t ctx, struct aio_context *aio) 69static int do_aio(aio_context_t ctx, enum aio_type type, int fd, char *buf,
70 int len, unsigned long long offset, struct aio_context *aio)
84{ 71{
85 struct iocb iocb, *iocbp = &iocb; 72 struct iocb iocb, *iocbp = &iocb;
86 char c; 73 char c;
@@ -88,39 +75,40 @@ static int do_aio(aio_context_t ctx, struct aio_context *aio)
88 75
89 iocb = ((struct iocb) { .aio_data = (unsigned long) aio, 76 iocb = ((struct iocb) { .aio_data = (unsigned long) aio,
90 .aio_reqprio = 0, 77 .aio_reqprio = 0,
91 .aio_fildes = aio->fd, 78 .aio_fildes = fd,
92 .aio_buf = (unsigned long) aio->data, 79 .aio_buf = (unsigned long) buf,
93 .aio_nbytes = aio->len, 80 .aio_nbytes = len,
94 .aio_offset = aio->offset, 81 .aio_offset = offset,
95 .aio_reserved1 = 0, 82 .aio_reserved1 = 0,
96 .aio_reserved2 = 0, 83 .aio_reserved2 = 0,
97 .aio_reserved3 = 0 }); 84 .aio_reserved3 = 0 });
98 85
99 switch(aio->type){ 86 switch(type){
100 case AIO_READ: 87 case AIO_READ:
101 iocb.aio_lio_opcode = IOCB_CMD_PREAD; 88 iocb.aio_lio_opcode = IOCB_CMD_PREAD;
89 err = io_submit(ctx, 1, &iocbp);
102 break; 90 break;
103 case AIO_WRITE: 91 case AIO_WRITE:
104 iocb.aio_lio_opcode = IOCB_CMD_PWRITE; 92 iocb.aio_lio_opcode = IOCB_CMD_PWRITE;
93 err = io_submit(ctx, 1, &iocbp);
105 break; 94 break;
106 case AIO_MMAP: 95 case AIO_MMAP:
107 iocb.aio_lio_opcode = IOCB_CMD_PREAD; 96 iocb.aio_lio_opcode = IOCB_CMD_PREAD;
108 iocb.aio_buf = (unsigned long) &c; 97 iocb.aio_buf = (unsigned long) &c;
109 iocb.aio_nbytes = sizeof(c); 98 iocb.aio_nbytes = sizeof(c);
99 err = io_submit(ctx, 1, &iocbp);
110 break; 100 break;
111 default: 101 default:
112 printk("Bogus op in do_aio - %d\n", aio->type); 102 printk("Bogus op in do_aio - %d\n", type);
113 err = -EINVAL; 103 err = -EINVAL;
114 goto out; 104 break;
115 } 105 }
116 106
117 err = io_submit(ctx, 1, &iocbp);
118 if(err > 0) 107 if(err > 0)
119 err = 0; 108 err = 0;
120 else 109 else
121 err = -errno; 110 err = -errno;
122 111
123 out:
124 return err; 112 return err;
125} 113}
126 114
@@ -129,9 +117,8 @@ static aio_context_t ctx = 0;
129static int aio_thread(void *arg) 117static int aio_thread(void *arg)
130{ 118{
131 struct aio_thread_reply reply; 119 struct aio_thread_reply reply;
132 struct aio_context *aio;
133 struct io_event event; 120 struct io_event event;
134 int err, n; 121 int err, n, reply_fd;
135 122
136 signal(SIGWINCH, SIG_IGN); 123 signal(SIGWINCH, SIG_IGN);
137 124
@@ -144,22 +131,14 @@ static int aio_thread(void *arg)
144 "errno = %d\n", errno); 131 "errno = %d\n", errno);
145 } 132 }
146 else { 133 else {
147 /* This is safe as we've just a pointer here. */
148 aio = (struct aio_context *) (long) event.data;
149 if(update_aio(aio, event.res)){
150 do_aio(ctx, aio);
151 continue;
152 }
153
154 reply = ((struct aio_thread_reply) 134 reply = ((struct aio_thread_reply)
155 { .data = aio, 135 { .data = (void *) (long) event.data,
156 .err = aio->len }); 136 .err = event.res });
157 err = os_write_file(aio->reply_fd, &reply, 137 reply_fd = ((struct aio_context *) reply.data)->reply_fd;
158 sizeof(reply)); 138 err = os_write_file(reply_fd, &reply, sizeof(reply));
159 if(err != sizeof(reply)) 139 if(err != sizeof(reply))
160 printk("aio_thread - write failed, " 140 printk("aio_thread - write failed, fd = %d, "
161 "fd = %d, err = %d\n", aio->reply_fd, 141 "err = %d\n", aio_req_fd_r, -err);
162 -err);
163 } 142 }
164 } 143 }
165 return 0; 144 return 0;
@@ -167,35 +146,35 @@ static int aio_thread(void *arg)
167 146
168#endif 147#endif
169 148
170static int do_not_aio(struct aio_context *aio) 149static int do_not_aio(struct aio_thread_req *req)
171{ 150{
172 char c; 151 char c;
173 int err; 152 int err;
174 153
175 switch(aio->type){ 154 switch(req->type){
176 case AIO_READ: 155 case AIO_READ:
177 err = os_seek_file(aio->fd, aio->offset); 156 err = os_seek_file(req->io_fd, req->offset);
178 if(err) 157 if(err)
179 goto out; 158 goto out;
180 159
181 err = os_read_file(aio->fd, aio->data, aio->len); 160 err = os_read_file(req->io_fd, req->buf, req->len);
182 break; 161 break;
183 case AIO_WRITE: 162 case AIO_WRITE:
184 err = os_seek_file(aio->fd, aio->offset); 163 err = os_seek_file(req->io_fd, req->offset);
185 if(err) 164 if(err)
186 goto out; 165 goto out;
187 166
188 err = os_write_file(aio->fd, aio->data, aio->len); 167 err = os_write_file(req->io_fd, req->buf, req->len);
189 break; 168 break;
190 case AIO_MMAP: 169 case AIO_MMAP:
191 err = os_seek_file(aio->fd, aio->offset); 170 err = os_seek_file(req->io_fd, req->offset);
192 if(err) 171 if(err)
193 goto out; 172 goto out;
194 173
195 err = os_read_file(aio->fd, &c, sizeof(c)); 174 err = os_read_file(req->io_fd, &c, sizeof(c));
196 break; 175 break;
197 default: 176 default:
198 printk("do_not_aio - bad request type : %d\n", aio->type); 177 printk("do_not_aio - bad request type : %d\n", req->type);
199 err = -EINVAL; 178 err = -EINVAL;
200 break; 179 break;
201 } 180 }
@@ -206,14 +185,14 @@ static int do_not_aio(struct aio_context *aio)
206 185
207static int not_aio_thread(void *arg) 186static int not_aio_thread(void *arg)
208{ 187{
209 struct aio_context *aio; 188 struct aio_thread_req req;
210 struct aio_thread_reply reply; 189 struct aio_thread_reply reply;
211 int err; 190 int err;
212 191
213 signal(SIGWINCH, SIG_IGN); 192 signal(SIGWINCH, SIG_IGN);
214 while(1){ 193 while(1){
215 err = os_read_file(aio_req_fd_r, &aio, sizeof(aio)); 194 err = os_read_file(aio_req_fd_r, &req, sizeof(req));
216 if(err != sizeof(aio)){ 195 if(err != sizeof(req)){
217 if(err < 0) 196 if(err < 0)
218 printk("not_aio_thread - read failed, " 197 printk("not_aio_thread - read failed, "
219 "fd = %d, err = %d\n", aio_req_fd_r, 198 "fd = %d, err = %d\n", aio_req_fd_r,
@@ -224,34 +203,17 @@ static int not_aio_thread(void *arg)
224 } 203 }
225 continue; 204 continue;
226 } 205 }
227 again: 206 err = do_not_aio(&req);
228 err = do_not_aio(aio); 207 reply = ((struct aio_thread_reply) { .data = req.aio,
229 208 .err = err });
230 if(update_aio(aio, err)) 209 err = os_write_file(req.aio->reply_fd, &reply, sizeof(reply));
231 goto again;
232
233 reply = ((struct aio_thread_reply) { .data = aio,
234 .err = aio->len });
235 err = os_write_file(aio->reply_fd, &reply, sizeof(reply));
236 if(err != sizeof(reply)) 210 if(err != sizeof(reply))
237 printk("not_aio_thread - write failed, fd = %d, " 211 printk("not_aio_thread - write failed, fd = %d, "
238 "err = %d\n", aio_req_fd_r, -err); 212 "err = %d\n", aio_req_fd_r, -err);
239 } 213 }
240} 214}
241 215
242static int submit_aio_24(struct aio_context *aio)
243{
244 int err;
245
246 err = os_write_file(aio_req_fd_w, &aio, sizeof(aio));
247 if(err == sizeof(aio))
248 err = 0;
249
250 return err;
251}
252
253static int aio_pid = -1; 216static int aio_pid = -1;
254static int (*submit_proc)(struct aio_context *aio);
255 217
256static int init_aio_24(void) 218static int init_aio_24(void)
257{ 219{
@@ -283,33 +245,11 @@ static int init_aio_24(void)
283#endif 245#endif
284 printk("2.6 host AIO support not used - falling back to I/O " 246 printk("2.6 host AIO support not used - falling back to I/O "
285 "thread\n"); 247 "thread\n");
286
287 submit_proc = submit_aio_24;
288
289 return 0; 248 return 0;
290} 249}
291 250
292#ifdef HAVE_AIO_ABI 251#ifdef HAVE_AIO_ABI
293#define DEFAULT_24_AIO 0 252#define DEFAULT_24_AIO 0
294static int submit_aio_26(struct aio_context *aio)
295{
296 struct aio_thread_reply reply;
297 int err;
298
299 err = do_aio(ctx, aio);
300 if(err){
301 reply = ((struct aio_thread_reply) { .data = aio,
302 .err = err });
303 err = os_write_file(aio->reply_fd, &reply, sizeof(reply));
304 if(err != sizeof(reply))
305 printk("submit_aio_26 - write failed, "
306 "fd = %d, err = %d\n", aio->reply_fd, -err);
307 else err = 0;
308 }
309
310 return err;
311}
312
313static int init_aio_26(void) 253static int init_aio_26(void)
314{ 254{
315 unsigned long stack; 255 unsigned long stack;
@@ -330,22 +270,39 @@ static int init_aio_26(void)
330 aio_pid = err; 270 aio_pid = err;
331 271
332 printk("Using 2.6 host AIO\n"); 272 printk("Using 2.6 host AIO\n");
273 return 0;
274}
275
276static int submit_aio_26(enum aio_type type, int io_fd, char *buf, int len,
277 unsigned long long offset, struct aio_context *aio)
278{
279 struct aio_thread_reply reply;
280 int err;
333 281
334 submit_proc = submit_aio_26; 282 err = do_aio(ctx, type, io_fd, buf, len, offset, aio);
283 if(err){
284 reply = ((struct aio_thread_reply) { .data = aio,
285 .err = err });
286 err = os_write_file(aio->reply_fd, &reply, sizeof(reply));
287 if(err != sizeof(reply))
288 printk("submit_aio_26 - write failed, "
289 "fd = %d, err = %d\n", aio->reply_fd, -err);
290 else err = 0;
291 }
335 292
336 return 0; 293 return err;
337} 294}
338 295
339#else 296#else
340#define DEFAULT_24_AIO 1 297#define DEFAULT_24_AIO 1
341static int submit_aio_26(struct aio_context *aio) 298static int init_aio_26(void)
342{ 299{
343 return -ENOSYS; 300 return -ENOSYS;
344} 301}
345 302
346static int init_aio_26(void) 303static int submit_aio_26(enum aio_type type, int io_fd, char *buf, int len,
304 unsigned long long offset, struct aio_context *aio)
347{ 305{
348 submit_proc = submit_aio_26;
349 return -ENOSYS; 306 return -ENOSYS;
350} 307}
351#endif 308#endif
@@ -412,7 +369,33 @@ static void exit_aio(void)
412 369
413__uml_exitcall(exit_aio); 370__uml_exitcall(exit_aio);
414 371
415int submit_aio(struct aio_context *aio) 372static int submit_aio_24(enum aio_type type, int io_fd, char *buf, int len,
373 unsigned long long offset, struct aio_context *aio)
416{ 374{
417 return (*submit_proc)(aio); 375 struct aio_thread_req req = { .type = type,
376 .io_fd = io_fd,
377 .offset = offset,
378 .buf = buf,
379 .len = len,
380 .aio = aio,
381 };
382 int err;
383
384 err = os_write_file(aio_req_fd_w, &req, sizeof(req));
385 if(err == sizeof(req))
386 err = 0;
387
388 return err;
389}
390
391int submit_aio(enum aio_type type, int io_fd, char *buf, int len,
392 unsigned long long offset, int reply_fd,
393 struct aio_context *aio)
394{
395 aio->reply_fd = reply_fd;
396 if(aio_24)
397 return submit_aio_24(type, io_fd, buf, len, offset, aio);
398 else {
399 return submit_aio_26(type, io_fd, buf, len, offset, aio);
400 }
418} 401}