[ARM] 3168/1: Update ARM signal delivery and masking
authorDaniel Jacobowitz <drow@false.org>
Sat, 19 Nov 2005 10:01:07 +0000 (10:01 +0000)
committerRussell King <rmk+kernel@arm.linux.org.uk>
Sat, 19 Nov 2005 10:01:07 +0000 (10:01 +0000)
Patch from Daniel Jacobowitz

After delivering a signal (creating its stack frame) we must check for
additional pending unblocked signals before returning to userspace.
Otherwise signals may be delayed past the next syscall or reschedule.

Once that was fixed it became obvious that the ARM signal mask manipulation
was broken.  It was a little bit broken before the recent SA_NODEFER
changes, and then very broken after them.  We must block the requested
signals before starting the handler or the same signal can be delivered
again before the handler even gets a chance to run.

Signed-off-by: Daniel Jacobowitz <dan@codesourcery.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
arch/arm/kernel/entry-common.S
arch/arm/kernel/signal.c
arch/arm/mm/Makefile
arch/arm/mm/blockops.c [deleted file]

index 066597f..f7f1830 100644 (file)
@@ -48,8 +48,7 @@ work_pending:
        mov     r0, sp                          @ 'regs'
        mov     r2, why                         @ 'syscall'
        bl      do_notify_resume
-       disable_irq                             @ disable interrupts
-       b       no_work_pending
+       b       ret_slow_syscall                @ Check work again
 
 work_resched:
        bl      schedule
index a917e3d..765922b 100644 (file)
@@ -595,23 +595,22 @@ handle_signal(unsigned long sig, struct k_sigaction *ka,
         */
        ret |= !valid_user_regs(regs);
 
-       /*
-        * Block the signal if we were unsuccessful.
-        */
        if (ret != 0) {
-               spin_lock_irq(&tsk->sighand->siglock);
-               sigorsets(&tsk->blocked, &tsk->blocked,
-                         &ka->sa.sa_mask);
-               if (!(ka->sa.sa_flags & SA_NODEFER))
-                       sigaddset(&tsk->blocked, sig);
-               recalc_sigpending();
-               spin_unlock_irq(&tsk->sighand->siglock);
+               force_sigsegv(sig, tsk);
+               return;
        }
 
-       if (ret == 0)
-               return;
+       /*
+        * Block the signal if we were successful.
+        */
+       spin_lock_irq(&tsk->sighand->siglock);
+       sigorsets(&tsk->blocked, &tsk->blocked,
+                 &ka->sa.sa_mask);
+       if (!(ka->sa.sa_flags & SA_NODEFER))
+               sigaddset(&tsk->blocked, sig);
+       recalc_sigpending();
+       spin_unlock_irq(&tsk->sighand->siglock);
 
-       force_sigsegv(sig, tsk);
 }
 
 /*
index 59f47d4..ffe73ba 100644 (file)
@@ -51,4 +51,4 @@ obj-$(CONFIG_CPU_ARM1026)     += proc-arm1026.o
 obj-$(CONFIG_CPU_SA110)                += proc-sa110.o
 obj-$(CONFIG_CPU_SA1100)       += proc-sa1100.o
 obj-$(CONFIG_CPU_XSCALE)       += proc-xscale.o
-obj-$(CONFIG_CPU_V6)           += proc-v6.o blockops.o
+obj-$(CONFIG_CPU_V6)           += proc-v6.o
diff --git a/arch/arm/mm/blockops.c b/arch/arm/mm/blockops.c
deleted file mode 100644 (file)
index 4f5ee2d..0000000
+++ /dev/null
@@ -1,185 +0,0 @@
-#include <linux/kernel.h>
-#include <linux/init.h>
-#include <linux/errno.h>
-#include <linux/mm.h>
-
-#include <asm/memory.h>
-#include <asm/ptrace.h>
-#include <asm/cacheflush.h>
-#include <asm/traps.h>
-
-extern struct cpu_cache_fns blk_cache_fns;
-
-#define HARVARD_CACHE
-
-/*
- *     blk_flush_kern_dcache_page(kaddr)
- *
- *     Ensure that the data held in the page kaddr is written back
- *     to the page in question.
- *
- *     - kaddr   - kernel address (guaranteed to be page aligned)
- */
-static void __attribute__((naked))
-blk_flush_kern_dcache_page(void *kaddr)
-{
-       asm(
-       "add    r1, r0, %0                                                      \n\
-       sub     r1, r1, %1                                                      \n\
-1:     .word   0xec401f0e      @ mcrr  p15, 0, r0, r1, c14, 0  @ blocking      \n\
-       mov     r0, #0                                                          \n\
-       mcr     p15, 0, r0, c7, c5, 0                                           \n\
-       mcr     p15, 0, r0, c7, c10, 4                                          \n\
-       mov     pc, lr"
-       :
-       : "I" (PAGE_SIZE), "I" (L1_CACHE_BYTES));
-}
-
-/*
- *     blk_dma_inv_range(start,end)
- *
- *     Invalidate the data cache within the specified region; we will
- *     be performing a DMA operation in this region and we want to
- *     purge old data in the cache.
- *
- *     - start   - virtual start address of region
- *     - end     - virtual end address of region
- */
-static void __attribute__((naked))
-blk_dma_inv_range_unified(unsigned long start, unsigned long end)
-{
-       asm(
-       "tst    r0, %0                                                          \n\
-       mcrne   p15, 0, r0, c7, c11, 1          @ clean unified line            \n\
-       tst     r1, %0                                                          \n\
-       mcrne   p15, 0, r1, c7, c15, 1          @ clean & invalidate unified line\n\
-       .word   0xec401f06      @ mcrr  p15, 0, r1, r0, c6, 0   @ blocking      \n\
-       mov     r0, #0                                                          \n\
-       mcr     p15, 0, r0, c7, c10, 4          @ drain write buffer            \n\
-       mov     pc, lr"
-       :
-       : "I" (L1_CACHE_BYTES - 1));
-}
-
-static void __attribute__((naked))
-blk_dma_inv_range_harvard(unsigned long start, unsigned long end)
-{
-       asm(
-       "tst    r0, %0                                                          \n\
-       mcrne   p15, 0, r0, c7, c10, 1          @ clean D line                  \n\
-       tst     r1, %0                                                          \n\
-       mcrne   p15, 0, r1, c7, c14, 1          @ clean & invalidate D line     \n\
-       .word   0xec401f06      @ mcrr  p15, 0, r1, r0, c6, 0   @ blocking      \n\
-       mov     r0, #0                                                          \n\
-       mcr     p15, 0, r0, c7, c10, 4          @ drain write buffer            \n\
-       mov     pc, lr"
-       :
-       : "I" (L1_CACHE_BYTES - 1));
-}
-
-/*
- *     blk_dma_clean_range(start,end)
- *     - start   - virtual start address of region
- *     - end     - virtual end address of region
- */
-static void __attribute__((naked))
-blk_dma_clean_range(unsigned long start, unsigned long end)
-{
-       asm(
-       ".word  0xec401f0c      @ mcrr  p15, 0, r1, r0, c12, 0  @ blocking      \n\
-       mov     r0, #0                                                          \n\
-       mcr     p15, 0, r0, c7, c10, 4          @ drain write buffer            \n\
-       mov     pc, lr");
-}
-
-/*
- *     blk_dma_flush_range(start,end)
- *     - start   - virtual start address of region
- *     - end     - virtual end address of region
- */
-static void __attribute__((naked))
-blk_dma_flush_range(unsigned long start, unsigned long end)
-{
-       asm(
-       ".word  0xec401f0e      @ mcrr  p15, 0, r1, r0, c14, 0  @ blocking      \n\
-       mov     pc, lr");
-}
-
-static int blockops_trap(struct pt_regs *regs, unsigned int instr)
-{
-       regs->ARM_r4 |= regs->ARM_r2;
-       regs->ARM_pc += 4;
-       return 0;
-}
-
-static char *func[] = {
-       "Prefetch data range",
-       "Clean+Invalidate data range",
-       "Clean data range",
-       "Invalidate data range",
-       "Invalidate instr range"
-};
-
-static struct undef_hook blockops_hook __initdata = {
-       .instr_mask     = 0x0fffffd0,
-       .instr_val      = 0x0c401f00,
-       .cpsr_mask      = PSR_T_BIT,
-       .cpsr_val       = 0,
-       .fn             = blockops_trap,
-};
-
-static int __init blockops_check(void)
-{
-       register unsigned int err asm("r4") = 0;
-       unsigned int err_pos = 1;
-       unsigned int cache_type;
-       int i;
-
-       asm("mrc p15, 0, %0, c0, c0, 1" : "=r" (cache_type));
-
-       printk("Checking V6 block cache operations:\n");
-       register_undef_hook(&blockops_hook);
-
-       __asm__ ("mov   r0, %0\n\t"
-               "mov    r1, %1\n\t"
-               "mov    r2, #1\n\t"
-               ".word  0xec401f2c @ mcrr p15, 0, r1, r0, c12, 2\n\t"
-               "mov    r2, #2\n\t"
-               ".word  0xec401f0e @ mcrr p15, 0, r1, r0, c14, 0\n\t"
-               "mov    r2, #4\n\t"
-               ".word  0xec401f0c @ mcrr p15, 0, r1, r0, c12, 0\n\t"
-               "mov    r2, #8\n\t"
-               ".word  0xec401f06 @ mcrr p15, 0, r1, r0, c6, 0\n\t"
-               "mov    r2, #16\n\t"
-               ".word  0xec401f05 @ mcrr p15, 0, r1, r0, c5, 0\n\t"
-               :
-               : "r" (PAGE_OFFSET), "r" (PAGE_OFFSET + 128)
-               : "r0", "r1", "r2");
-
-       unregister_undef_hook(&blockops_hook);
-
-       for (i = 0; i < ARRAY_SIZE(func); i++, err_pos <<= 1)
-               printk("%30s: %ssupported\n", func[i], err & err_pos ? "not " : "");
-
-       if ((err & 8) == 0) {
-               printk(" --> Using %s block cache invalidate\n",
-                       cache_type & (1 << 24) ? "harvard" : "unified");
-               if (cache_type & (1 << 24))
-                       cpu_cache.dma_inv_range = blk_dma_inv_range_harvard;
-               else
-                       cpu_cache.dma_inv_range = blk_dma_inv_range_unified;
-       }
-       if ((err & 4) == 0) {
-               printk(" --> Using block cache clean\n");
-               cpu_cache.dma_clean_range        = blk_dma_clean_range;
-       }
-       if ((err & 2) == 0) {
-               printk(" --> Using block cache clean+invalidate\n");
-               cpu_cache.dma_flush_range        = blk_dma_flush_range;
-               cpu_cache.flush_kern_dcache_page = blk_flush_kern_dcache_page;
-       }
-
-       return 0;
-}
-
-__initcall(blockops_check);