Merge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  arch/tile: don't validate CROSS_COMPILE needlessly
  arch/tile: export only COMMAND_LINE_SIZE to userspace.
  arch/tile: rename ARCH_KMALLOC_MINALIGN to ARCH_DMA_MINALIGN
  arch/tile: Rename the hweight() implementations to __arch_hweight()
  arch/tile: extend syscall ABI to set r1 on return as well.
  arch/tile: Various cleanups.
  arch/tile: support backtracing on TILE-Gx
  arch/tile: Fix a couple of issues with the COMPAT code for TILE-Gx.
  arch/tile: Use separate, better minsec values for clocksource and sched_clock.
  arch/tile: correct a bug in freeing bootmem by VA for the optional second initrd.
  arch: tile: mm: pgtable.c: Removed duplicated #include
  arch: tile: kernel/proc.c Removed duplicated #include
  Add fanotify syscalls to <asm-generic/unistd.h>.
  arch/tile: support new kunmap_atomic() naming convention.
  tile: remove unused ISA_DMA_THRESHOLD define

Conflicts in arch/tile/configs/tile_defconfig (pick the mainline version
with the reduced defconfig).

8 files changed, 138 insertions, 67 deletions

diff --git a/arch/tile/kernel/backtrace.c b/arch/tile/kernel/backtrace.c
index 77265f3b58d6..d3c41c1ff6bd 100644
--- a/arch/tile/kernel/backtrace.c
+++ b/arch/tile/kernel/backtrace.c
@@ -19,9 +19,6 @@
 
 #include <arch/chip.h>
 
-#if TILE_CHIP < 10
-
-
 #include <asm/opcode-tile.h>
 
 
@@ -29,6 +26,27 @@
 #define TREG_LR 55
 
 
+#if TILE_CHIP >= 10
+#define tile_bundle_bits tilegx_bundle_bits
+#define TILE_MAX_INSTRUCTIONS_PER_BUNDLE TILEGX_MAX_INSTRUCTIONS_PER_BUNDLE
+#define TILE_BUNDLE_ALIGNMENT_IN_BYTES TILEGX_BUNDLE_ALIGNMENT_IN_BYTES
+#define tile_decoded_instruction tilegx_decoded_instruction
+#define tile_mnemonic tilegx_mnemonic
+#define parse_insn_tile parse_insn_tilegx
+#define TILE_OPC_IRET TILEGX_OPC_IRET
+#define TILE_OPC_ADDI TILEGX_OPC_ADDI
+#define TILE_OPC_ADDLI TILEGX_OPC_ADDLI
+#define TILE_OPC_INFO TILEGX_OPC_INFO
+#define TILE_OPC_INFOL TILEGX_OPC_INFOL
+#define TILE_OPC_JRP TILEGX_OPC_JRP
+#define TILE_OPC_MOVE TILEGX_OPC_MOVE
+#define OPCODE_STORE TILEGX_OPC_ST
+typedef long long bt_int_reg_t;
+#else
+#define OPCODE_STORE TILE_OPC_SW
+typedef int bt_int_reg_t;
+#endif
+
 /** A decoded bundle used for backtracer analysis. */
 struct BacktraceBundle {
         tile_bundle_bits bits;
@@ -41,7 +59,7 @@ struct BacktraceBundle {
 /* This implementation only makes sense for native tools. */
 /** Default function to read memory. */
 static bool bt_read_memory(void *result, VirtualAddress addr,
-                           size_t size, void *extra)
+                           unsigned int size, void *extra)
 {
         /* FIXME: this should do some horrible signal stuff to catch
          * SEGV cleanly and fail.
@@ -106,6 +124,12 @@ static bool bt_has_addi_sp(const struct BacktraceBundle *bundle, int *adjust)
                 find_matching_insn(bundle, TILE_OPC_ADDI, vals, 2);
         if (insn == NULL)
                 insn = find_matching_insn(bundle, TILE_OPC_ADDLI, vals, 2);
+#if TILE_CHIP >= 10
+        if (insn == NULL)
+                insn = find_matching_insn(bundle, TILEGX_OPC_ADDXLI, vals, 2);
+        if (insn == NULL)
+                insn = find_matching_insn(bundle, TILEGX_OPC_ADDXI, vals, 2);
+#endif
         if (insn == NULL)
                 return false;
 
@@ -190,13 +214,52 @@ static inline bool bt_has_move_r52_sp(const struct BacktraceBundle *bundle)
         return find_matching_insn(bundle, TILE_OPC_MOVE, vals, 2) != NULL;
 }
 
-/** Does this bundle contain the instruction 'sw sp, lr'? */
+/** Does this bundle contain a store of lr to sp? */
 static inline bool bt_has_sw_sp_lr(const struct BacktraceBundle *bundle)
 {
         static const int vals[2] = { TREG_SP, TREG_LR };
-        return find_matching_insn(bundle, TILE_OPC_SW, vals, 2) != NULL;
+        return find_matching_insn(bundle, OPCODE_STORE, vals, 2) != NULL;
+}
+
+#if TILE_CHIP >= 10
+/** Track moveli values placed into registers. */
+static inline void bt_update_moveli(const struct BacktraceBundle *bundle,
+                                    int moveli_args[])
+{
+        int i;
+        for (i = 0; i < bundle->num_insns; i++) {
+                const struct tile_decoded_instruction *insn =
+                        &bundle->insns[i];
+
+                if (insn->opcode->mnemonic == TILEGX_OPC_MOVELI) {
+                        int reg = insn->operand_values[0];
+                        moveli_args[reg] = insn->operand_values[1];
+                }
+        }
 }
 
+/** Does this bundle contain an 'add sp, sp, reg' instruction
+ * from a register that we saw a moveli into, and if so, what
+ * is the value in the register?
+ */
+static bool bt_has_add_sp(const struct BacktraceBundle *bundle, int *adjust,
+                          int moveli_args[])
+{
+        static const int vals[2] = { TREG_SP, TREG_SP };
+
+        const struct tile_decoded_instruction *insn =
+                find_matching_insn(bundle, TILEGX_OPC_ADDX, vals, 2);
+        if (insn) {
+                int reg = insn->operand_values[2];
+                if (moveli_args[reg]) {
+                        *adjust = moveli_args[reg];
+                        return true;
+                }
+        }
+        return false;
+}
+#endif
+
 /** Locates the caller's PC and SP for a program starting at the
  * given address.
  */
@@ -227,6 +290,11 @@ static void find_caller_pc_and_caller_sp(CallerLocation *location,
         int next_bundle = 0;
         VirtualAddress pc;
 
+#if TILE_CHIP >= 10
+        /* Naively try to track moveli values to support addx for -m32. */
+        int moveli_args[TILEGX_NUM_REGISTERS] = { 0 };
+#endif
+
         /* Default to assuming that the caller's sp is the current sp.
          * This is necessary to handle the case where we start backtracing
          * right at the end of the epilog.
@@ -380,7 +448,11 @@ static void find_caller_pc_and_caller_sp(CallerLocation *location,
 
                 if (!sp_determined) {
                         int adjust;
-                        if (bt_has_addi_sp(&bundle, &adjust)) {
+                        if (bt_has_addi_sp(&bundle, &adjust)
+#if TILE_CHIP >= 10
+                            || bt_has_add_sp(&bundle, &adjust, moveli_args)
+#endif
+                                ) {
                                 location->sp_location = SP_LOC_OFFSET;
 
                                 if (adjust <= 0) {
@@ -427,6 +499,11 @@ static void find_caller_pc_and_caller_sp(CallerLocation *location,
                                         sp_determined = true;
                                 }
                         }
+
+#if TILE_CHIP >= 10
+                        /* Track moveli arguments for -m32 mode. */
+                        bt_update_moveli(&bundle, moveli_args);
+#endif
                 }
 
                 if (bt_has_iret(&bundle)) {
@@ -502,11 +579,10 @@ void backtrace_init(BacktraceIterator *state,
                 break;
         }
 
-        /* The frame pointer should theoretically be aligned mod 8. If
-         * it's not even aligned mod 4 then something terrible happened
-         * and we should mark it as invalid.
+        /* If the frame pointer is not aligned to the basic word size
+         * something terrible happened and we should mark it as invalid.
          */
-        if (fp % 4 != 0)
+        if (fp % sizeof(bt_int_reg_t) != 0)
                 fp = -1;
 
         /* -1 means "don't know initial_frame_caller_pc". */
@@ -547,9 +623,16 @@ void backtrace_init(BacktraceIterator *state,
         state->read_memory_func_extra = read_memory_func_extra;
 }
 
+/* Handle the case where the register holds more bits than the VA. */
+static bool valid_addr_reg(bt_int_reg_t reg)
+{
+        return ((VirtualAddress)reg == reg);
+}
+
 bool backtrace_next(BacktraceIterator *state)
 {
-        VirtualAddress next_fp, next_pc, next_frame[2];
+        VirtualAddress next_fp, next_pc;
+        bt_int_reg_t next_frame[2];
 
         if (state->fp == -1) {
                 /* No parent frame. */
@@ -563,11 +646,9 @@ bool backtrace_next(BacktraceIterator *state)
         }
 
         next_fp = next_frame[1];
-        if (next_fp % 4 != 0) {
-                /* Caller's frame pointer is suspect, so give up.
-                 * Technically it should be aligned mod 8, but we will
-                 * be forgiving here.
-                 */
+        if (!valid_addr_reg(next_frame[1]) ||
+            next_fp % sizeof(bt_int_reg_t) != 0) {
+                /* Caller's frame pointer is suspect, so give up. */
                 return false;
         }
 
@@ -585,7 +666,7 @@ bool backtrace_next(BacktraceIterator *state)
         } else {
                 /* Get the caller PC from the frame linkage area. */
                 next_pc = next_frame[0];
-                if (next_pc == 0 ||
+                if (!valid_addr_reg(next_frame[0]) || next_pc == 0 ||
                     next_pc % TILE_BUNDLE_ALIGNMENT_IN_BYTES != 0) {
                         /* The PC is suspect, so give up. */
                         return false;
@@ -599,23 +680,3 @@ bool backtrace_next(BacktraceIterator *state)
 
         return true;
 }
-
-#else /* TILE_CHIP < 10 */
-
-void backtrace_init(BacktraceIterator *state,
-                    BacktraceMemoryReader read_memory_func,
-                    void *read_memory_func_extra,
-                    VirtualAddress pc, VirtualAddress lr,
-                    VirtualAddress sp, VirtualAddress r52)
-{
-        state->pc = pc;
-        state->sp = sp;
-        state->fp = -1;
-        state->initial_frame_caller_pc = -1;
-        state->read_memory_func = read_memory_func;
-        state->read_memory_func_extra = read_memory_func_extra;
-}
-
-bool backtrace_next(BacktraceIterator *state) { return false; }
-
-#endif /* TILE_CHIP < 10 */
diff --git a/arch/tile/kernel/compat_signal.c b/arch/tile/kernel/compat_signal.c
index d5efb215dd5f..9c710db43f13 100644
--- a/arch/tile/kernel/compat_signal.c
+++ b/arch/tile/kernel/compat_signal.c
@@ -56,13 +56,15 @@ struct compat_ucontext {
         sigset_t          uc_sigmask;   /* mask last for extensibility */
 };
 
+#define COMPAT_SI_PAD_SIZE      ((SI_MAX_SIZE - 3 * sizeof(int)) / sizeof(int))
+
 struct compat_siginfo {
         int si_signo;
         int si_errno;
         int si_code;
 
         union {
-                int _pad[SI_PAD_SIZE];
+                int _pad[COMPAT_SI_PAD_SIZE];
 
                 /* kill() */
                 struct {
diff --git a/arch/tile/kernel/intvec_32.S b/arch/tile/kernel/intvec_32.S
index 3404c75f8e64..84f296ca9e63 100644
--- a/arch/tile/kernel/intvec_32.S
+++ b/arch/tile/kernel/intvec_32.S
@@ -952,7 +952,7 @@ STD_ENTRY(interrupt_return)
          * able to safely read all the remaining words on those cache
          * lines without waiting for the memory subsystem.
          */
-        pop_reg_zero r0, r1, sp, PTREGS_OFFSET_REG(30) - PTREGS_OFFSET_REG(0)
+        pop_reg_zero r0, r28, sp, PTREGS_OFFSET_REG(30) - PTREGS_OFFSET_REG(0)
         pop_reg_zero r30, r2, sp, PTREGS_OFFSET_PC - PTREGS_OFFSET_REG(30)
         pop_reg_zero r21, r3, sp, PTREGS_OFFSET_EX1 - PTREGS_OFFSET_PC
         pop_reg_zero lr, r4, sp, PTREGS_OFFSET_REG(52) - PTREGS_OFFSET_EX1
@@ -1017,7 +1017,17 @@ STD_ENTRY(interrupt_return)
         { move r22, zero; move r23, zero }
         { move r24, zero; move r25, zero }
         { move r26, zero; move r27, zero }
-        { move r28, zero; move r29, zero }
+
+        /* Set r1 to errno if we are returning an error, otherwise zero. */
+        {
+         moveli r29, 1024
+         sub    r1, zero, r0
+        }
+        slt_u   r29, r1, r29
+        {
+         mnz    r1, r29, r1
+         move   r29, zero
+        }
         iret
 
         /*
diff --git a/arch/tile/kernel/proc.c b/arch/tile/kernel/proc.c
index 92ef925d2f8d..2e02c41ddf3b 100644
--- a/arch/tile/kernel/proc.c
+++ b/arch/tile/kernel/proc.c
@@ -23,7 +23,6 @@
 #include <linux/sysctl.h>
 #include <linux/hardirq.h>
 #include <linux/mman.h>
-#include <linux/smp.h>
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/sections.h>
diff --git a/arch/tile/kernel/setup.c b/arch/tile/kernel/setup.c
index 4dd21c1e6d5e..e7d54c73d5c1 100644
--- a/arch/tile/kernel/setup.c
+++ b/arch/tile/kernel/setup.c
@@ -953,7 +953,7 @@ static void __init load_hv_initrd(void)
         if (rc != stat.size) {
                 pr_err("Error reading %d bytes from hvfs file '%s': %d\n",
                        stat.size, initramfs_file, rc);
-                free_bootmem((unsigned long) initrd, stat.size);
+                free_initrd_mem((unsigned long) initrd, stat.size);
                 return;
         }
         initrd_start = (unsigned long) initrd;
@@ -962,7 +962,7 @@ static void __init load_hv_initrd(void)
 
 void __init free_initrd_mem(unsigned long begin, unsigned long end)
 {
-        free_bootmem(begin, end - begin);
+        free_bootmem(__pa(begin), end - begin);
 }
 
 static void __init validate_hv(void)
diff --git a/arch/tile/kernel/stack.c b/arch/tile/kernel/stack.c
index b6268d3ae869..38a68b0b4581 100644
--- a/arch/tile/kernel/stack.c
+++ b/arch/tile/kernel/stack.c
@@ -108,7 +108,6 @@ static bool read_memory_func(void *result, VirtualAddress address,
 /* Return a pt_regs pointer for a valid fault handler frame */
 static struct pt_regs *valid_fault_handler(struct KBacktraceIterator* kbt)
 {
-#ifndef __tilegx__
         const char *fault = NULL;  /* happy compiler */
         char fault_buf[64];
         VirtualAddress sp = kbt->it.sp;
@@ -146,7 +145,6 @@ static struct pt_regs *valid_fault_handler(struct KBacktraceIterator* kbt)
         }
         if (!kbt->profile || (INT_MASK(p->faultnum) & QUEUED_INTERRUPTS) == 0)
                 return p;
-#endif
         return NULL;
 }
 
@@ -351,12 +349,6 @@ void tile_show_stack(struct KBacktraceIterator *kbt, int headers)
                        kbt->task->pid, kbt->task->tgid, kbt->task->comm,
                        smp_processor_id(), get_cycles());
         }
-#ifdef __tilegx__
-        if (kbt->is_current) {
-                __insn_mtspr(SPR_SIM_CONTROL,
-                             SIM_DUMP_SPR_ARG(SIM_DUMP_BACKTRACE));
-        }
-#endif
         kbt->verbose = 1;
         i = 0;
         for (; !KBacktraceIterator_end(kbt); KBacktraceIterator_next(kbt)) {
diff --git a/arch/tile/kernel/time.c b/arch/tile/kernel/time.c
index b9ab25a889b5..6bed820e1421 100644
--- a/arch/tile/kernel/time.c
+++ b/arch/tile/kernel/time.c
@@ -36,16 +36,6 @@
 /* How many cycles per second we are running at. */
 static cycles_t cycles_per_sec __write_once;
 
-/*
- * We set up shift and multiply values with a minsec of five seconds,
- * since our timer counter counts down 31 bits at a frequency of
- * no less than 500 MHz.  See @minsec for clocks_calc_mult_shift().
- * We could use a different value for the 64-bit free-running
- * cycle counter, but we use the same one for consistency, and since
- * we will be reasonably precise with this value anyway.
- */
-#define TILE_MINSEC 5
-
 cycles_t get_clock_rate(void)
 {
         return cycles_per_sec;
@@ -68,6 +58,14 @@ cycles_t get_cycles(void)
 }
 #endif
 
+/*
+ * We use a relatively small shift value so that sched_clock()
+ * won't wrap around very often.
+ */
+#define SCHED_CLOCK_SHIFT 10
+
+static unsigned long sched_clock_mult __write_once;
+
 static cycles_t clocksource_get_cycles(struct clocksource *cs)
 {
         return get_cycles();
@@ -78,6 +76,7 @@ static struct clocksource cycle_counter_cs = {
         .rating = 300,
         .read = clocksource_get_cycles,
         .mask = CLOCKSOURCE_MASK(64),
+        .shift = 22,   /* typical value, e.g. x86 tsc uses this */
         .flags = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
@@ -88,8 +87,10 @@ static struct clocksource cycle_counter_cs = {
 void __init setup_clock(void)
 {
         cycles_per_sec = hv_sysconf(HV_SYSCONF_CPU_SPEED);
-        clocksource_calc_mult_shift(&cycle_counter_cs, cycles_per_sec,
-                                    TILE_MINSEC);
+        sched_clock_mult =
+                clocksource_hz2mult(cycles_per_sec, SCHED_CLOCK_SHIFT);
+        cycle_counter_cs.mult =
+                clocksource_hz2mult(cycles_per_sec, cycle_counter_cs.shift);
 }
 
 void __init calibrate_delay(void)
@@ -117,9 +118,14 @@ void __init time_init(void)
  * counter, plus bit 31, which signifies that the counter has wrapped
  * from zero to (2**31) - 1.  The INT_TILE_TIMER interrupt will be
  * raised as long as bit 31 is set.
+ *
+ * The TILE_MINSEC value represents the largest range of real-time
+ * we can possibly cover with the timer, based on MAX_TICK combined
+ * with the slowest reasonable clock rate we might run at.
  */
 
 #define MAX_TICK 0x7fffffff   /* we have 31 bits of countdown timer */
+#define TILE_MINSEC 5         /* timer covers no more than 5 seconds */
 
 static int tile_timer_set_next_event(unsigned long ticks,
                                      struct clock_event_device *evt)
@@ -211,8 +217,7 @@ void do_timer_interrupt(struct pt_regs *regs, int fault_num)
 unsigned long long sched_clock(void)
 {
         return clocksource_cyc2ns(get_cycles(),
-                                  cycle_counter_cs.mult,
-                                  cycle_counter_cs.shift);
+                                  sched_clock_mult, SCHED_CLOCK_SHIFT);
 }
 
 int setup_profiling_timer(unsigned int multiplier)
diff --git a/arch/tile/kernel/traps.c b/arch/tile/kernel/traps.c
index 3870abbeeaa2..0f362dc2c57f 100644
--- a/arch/tile/kernel/traps.c
+++ b/arch/tile/kernel/traps.c
@@ -128,7 +128,9 @@ static int special_ill(bundle_bits bundle, int *sigp, int *codep)
 #ifdef __tilegx__
         if ((bundle & TILEGX_BUNDLE_MODE_MASK) != 0)
                 return 0;
-        if (get_Opcode_X1(bundle) != UNARY_OPCODE_X1)
+        if (get_Opcode_X1(bundle) != RRR_0_OPCODE_X1)
+                return 0;
+        if (get_RRROpcodeExtension_X1(bundle) != UNARY_RRR_0_OPCODE_X1)
                 return 0;
         if (get_UnaryOpcodeExtension_X1(bundle) != ILL_UNARY_OPCODE_X1)
                 return 0;