[Qemu-devel] [PATCH v2 05/11] tcg/i386: Make direct jump patching thread-safe

[Sergey Fedorov <sergey.fedorov@linaro.org>]

From: Sergey Fedorov <serge.fdrv@gmail.com>

Ensure direct jump patching in i386 is atomic by:
 * naturally aligning a location of direct jump address;
 * using atomic_read()/atomic_set() for code patching.

tcg_out_nopn() implementation:
Suggested-by: Richard Henderson <rth@twiddle.net>.

Signed-off-by: Sergey Fedorov <serge.fdrv@gmail.com>
Signed-off-by: Sergey Fedorov <sergey.fedorov@linaro.org>
---

Changes in v2:
 * Use tcg_out_nopn() implementation suggested by Richard Henderson
 * Rework alignment checking and gap calculation

 include/exec/exec-all.h   |  2 +-
 tcg/i386/tcg-target.inc.c | 23 +++++++++++++++++++++++
 2 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h
index 59709c9dd5c9..82399175fe80 100644
--- a/include/exec/exec-all.h
+++ b/include/exec/exec-all.h
@@ -312,7 +312,7 @@ void ppc_tb_set_jmp_target(uintptr_t jmp_addr, uintptr_t addr);
 static inline void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr)
 {
     /* patch the branch destination */
-    stl_le_p((void*)jmp_addr, addr - (jmp_addr + 4));
+    atomic_set((int32_t *)jmp_addr, addr - (jmp_addr + 4));
     /* no need to flush icache explicitly */
 }
 #elif defined(__s390x__)
diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c
index 9187d34caf6d..3ffc81a1d168 100644
--- a/tcg/i386/tcg-target.inc.c
+++ b/tcg/i386/tcg-target.inc.c
@@ -1123,6 +1123,21 @@ static void tcg_out_jmp(TCGContext *s, tcg_insn_unit *dest)
     tcg_out_branch(s, 0, dest);
 }
 
+static void tcg_out_nopn(TCGContext *s, int n)
+{
+    int i;
+    /* Emit 1 or 2 operand size prefixes for the standard one byte nop,
+     * "xchg %eax,%eax", forming "xchg %ax,%ax". All cores accept the
+     * duplicate prefix, and all of the interesting recent cores can
+     * decode and discard the duplicates in a single cycle.
+     */
+    tcg_debug_assert(n >= 1);
+    for (i = 1; i < n; ++i) {
+        tcg_out8(s, 0x66);
+    }
+    tcg_out8(s, 0x90);
+}
+
 #if defined(CONFIG_SOFTMMU)
 /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr,
  *                                     int mmu_idx, uintptr_t ra)
@@ -1777,6 +1792,14 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
     case INDEX_op_goto_tb:
         if (s->tb_jmp_offset) {
             /* direct jump method */
+            int gap;
+            /* jump displacement must be aligned for atomic patching;
+             * see if we need to add extra nops before jump
+             */
+            gap = tcg_pcrel_diff(s, QEMU_ALIGN_PTR_UP(s->code_ptr + 1, 4));
+            if (gap != 1) {
+                tcg_out_nopn(s, gap - 1);
+            }
             tcg_out8(s, OPC_JMP_long); /* jmp im */
             s->tb_jmp_offset[args[0]] = tcg_current_code_size(s);
             tcg_out32(s, 0);
-- 
2.8.1