[Qemu-devel] [PULL 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB

Richard Henderson posted 18 patches 7 years, 8 months ago
There is a newer version of this series
[Qemu-devel] [PULL 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Richard Henderson 7 years, 8 months ago
From: "Emilio G. Cota" <cota@braap.org>

Use the recently-gained QHT feature of returning the matching TB if it
already exists. This allows us to get rid of the lookup we perform
right after acquiring tb_lock.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/cpu-exec.c            | 14 ++-------
 accel/tcg/translate-all.c       | 50 +++++++++++++++++++++++++++------
 docs/devel/multi-thread-tcg.txt |  3 ++
 3 files changed, 46 insertions(+), 21 deletions(-)

diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c
index d75c35380a..45f6ebc65e 100644
--- a/accel/tcg/cpu-exec.c
+++ b/accel/tcg/cpu-exec.c
@@ -245,10 +245,7 @@ void cpu_exec_step_atomic(CPUState *cpu)
         if (tb == NULL) {
             mmap_lock();
             tb_lock();
-            tb = tb_htable_lookup(cpu, pc, cs_base, flags, cf_mask);
-            if (likely(tb == NULL)) {
-                tb = tb_gen_code(cpu, pc, cs_base, flags, cflags);
-            }
+            tb = tb_gen_code(cpu, pc, cs_base, flags, cflags);
             tb_unlock();
             mmap_unlock();
         }
@@ -398,14 +395,7 @@ static inline TranslationBlock *tb_find(CPUState *cpu,
         tb_lock();
         acquired_tb_lock = true;
 
-        /* There's a chance that our desired tb has been translated while
-         * taking the locks so we check again inside the lock.
-         */
-        tb = tb_htable_lookup(cpu, pc, cs_base, flags, cf_mask);
-        if (likely(tb == NULL)) {
-            /* if no translated code available, then translate it now */
-            tb = tb_gen_code(cpu, pc, cs_base, flags, cf_mask);
-        }
+        tb = tb_gen_code(cpu, pc, cs_base, flags, cf_mask);
 
         mmap_unlock();
         /* We add the TB in the virtual pc hash table for the fast lookup */
diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
index c75298d08a..2585e6fd3e 100644
--- a/accel/tcg/translate-all.c
+++ b/accel/tcg/translate-all.c
@@ -1581,18 +1581,30 @@ static inline void tb_page_add(PageDesc *p, TranslationBlock *tb,
  * (-1) to indicate that only one page contains the TB.
  *
  * Called with mmap_lock held for user-mode emulation.
+ *
+ * Returns a pointer @tb, or a pointer to an existing TB that matches @tb.
+ * Note that in !user-mode, another thread might have already added a TB
+ * for the same block of guest code that @tb corresponds to. In that case,
+ * the caller should discard the original @tb, and use instead the returned TB.
  */
-static void tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
-                         tb_page_addr_t phys_page2)
+static TranslationBlock *
+tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
+             tb_page_addr_t phys_page2)
 {
     PageDesc *p;
     PageDesc *p2 = NULL;
+    void *existing_tb = NULL;
     uint32_t h;
 
     assert_memory_lock();
 
     /*
      * Add the TB to the page list, acquiring first the pages's locks.
+     * We keep the locks held until after inserting the TB in the hash table,
+     * so that if the insertion fails we know for sure that the TBs are still
+     * in the page descriptors.
+     * Note that inserting into the hash table first isn't an option, since
+     * we can only insert TBs that are fully initialized.
      */
     page_lock_pair(&p, phys_pc, &p2, phys_page2, 1);
     tb_page_add(p, tb, 0, phys_pc & TARGET_PAGE_MASK);
@@ -1602,21 +1614,33 @@ static void tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
         tb->page_addr[1] = -1;
     }
 
+    /* add in the hash table */
+    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
+                     tb->trace_vcpu_dstate);
+    qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
+
+    /* remove TB from the page(s) if we couldn't insert it */
+    if (unlikely(existing_tb)) {
+        tb_page_remove(p, tb);
+        invalidate_page_bitmap(p);
+        if (p2) {
+            tb_page_remove(p2, tb);
+            invalidate_page_bitmap(p2);
+        }
+        tb = existing_tb;
+    }
+
     if (p2) {
         page_unlock(p2);
     }
     page_unlock(p);
 
-    /* add in the hash table */
-    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
-                     tb->trace_vcpu_dstate);
-    qht_insert(&tb_ctx.htable, tb, h, NULL);
-
 #ifdef CONFIG_USER_ONLY
     if (DEBUG_TB_CHECK_GATE) {
         tb_page_check();
     }
 #endif
+    return tb;
 }
 
 /* Called with mmap_lock held for user mode emulation.  */
@@ -1625,7 +1649,7 @@ TranslationBlock *tb_gen_code(CPUState *cpu,
                               uint32_t flags, int cflags)
 {
     CPUArchState *env = cpu->env_ptr;
-    TranslationBlock *tb;
+    TranslationBlock *tb, *existing_tb;
     tb_page_addr_t phys_pc, phys_page2;
     target_ulong virt_page2;
     tcg_insn_unit *gen_code_buf;
@@ -1773,7 +1797,15 @@ TranslationBlock *tb_gen_code(CPUState *cpu,
      * memory barrier is required before tb_link_page() makes the TB visible
      * through the physical hash table and physical page list.
      */
-    tb_link_page(tb, phys_pc, phys_page2);
+    existing_tb = tb_link_page(tb, phys_pc, phys_page2);
+    /* if the TB already exists, discard what we just translated */
+    if (unlikely(existing_tb != tb)) {
+        uintptr_t orig_aligned = (uintptr_t)gen_code_buf;
+
+        orig_aligned -= ROUND_UP(sizeof(*tb), qemu_icache_linesize);
+        atomic_set(&tcg_ctx->code_gen_ptr, orig_aligned);
+        return existing_tb;
+    }
     tcg_tb_insert(tb);
     return tb;
 }
diff --git a/docs/devel/multi-thread-tcg.txt b/docs/devel/multi-thread-tcg.txt
index faf8918b23..faf09c6069 100644
--- a/docs/devel/multi-thread-tcg.txt
+++ b/docs/devel/multi-thread-tcg.txt
@@ -140,6 +140,9 @@ to atomically insert new elements.
 The lookup caches are updated atomically and the lookup hash uses QHT
 which is designed for concurrent safe lookup.
 
+Parallel code generation is supported. QHT is used at insertion time
+as the synchronization point across threads, thereby ensuring that we only
+keep track of a single TranslationBlock for each guest code block.
 
 Memory maps and TLBs
 --------------------
-- 
2.17.1


Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Pavel Dovgalyuk 7 years, 7 months ago
This patch breaks record/replay.

I run execution recording of the WindowsXP machine with the following script:

./bin/qemu-system-i386 -d in_asm,exec -D xp_save.log -global apic-common.vapic=off \
  -icount shift=7,rr=record,rrfile=xp0.replay \
  -drive file=./images/xp_sp2.qcow2,if=none,id=img-direct,snapshot \
  -drive driver=blkreplay,if=none,image=img-direct,id=img-replay \
  -device ide-hd,drive=img-replay -net none -m 512M

QEMU fails at some moment. Here are the contents of the log:

----------------
IN: 
0x806ee2d0:  33 c0                    xorl     %eax, %eax
0x806ee2d2:  8a c1                    movb     %cl, %al
0x806ee2d4:  33 c9                    xorl     %ecx, %ecx
0x806ee2d6:  8a 88 58 e2 6e 80        movb     -0x7f911da8(%eax), %cl
0x806ee2dc:  89 0d 80 00 fe ff        movl     %ecx, 0xfffe0080
0x806ee2e2:  a1 80 00 fe ff           movl     0xfffe0080, %eax
0x806ee2e7:  c3                       retl     

Trace 0: 0x7fdc103b16a0 [00000000/806ee2d0/0x4000b0]
qemu: fatal: cpu_io_recompile: could not find TB for pc=0x7fec24fde2de
EAX=00000001 EBX=00006901 ECX=0000003d EDX=00000ffc
ESI=040d78c0 EDI=0000031f EBP=f878fc60 ESP=f878fc54
EIP=806ee2d0 EFL=00000202 [-------] CPL=0 II=0 A20=1 SMM=0 HLT=0
....


Pavel Dovgalyuk

> -----Original Message-----
> From: Richard Henderson [mailto:richard.henderson@linaro.org]
> Sent: Thursday, June 14, 2018 10:32 PM
> To: qemu-devel@nongnu.org
> Cc: peter.maydell@linaro.org; Emilio G. Cota
> Subject: [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing
> matching TB
> 
> From: "Emilio G. Cota" <cota@braap.org>
> 
> Use the recently-gained QHT feature of returning the matching TB if it
> already exists. This allows us to get rid of the lookup we perform
> right after acquiring tb_lock.
> 
> Suggested-by: Richard Henderson <richard.henderson@linaro.org>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: Emilio G. Cota <cota@braap.org>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  accel/tcg/cpu-exec.c            | 14 ++-------
>  accel/tcg/translate-all.c       | 50 +++++++++++++++++++++++++++------
>  docs/devel/multi-thread-tcg.txt |  3 ++
>  3 files changed, 46 insertions(+), 21 deletions(-)
> 
> diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c
> index d75c35380a..45f6ebc65e 100644
> --- a/accel/tcg/cpu-exec.c
> +++ b/accel/tcg/cpu-exec.c
> @@ -245,10 +245,7 @@ void cpu_exec_step_atomic(CPUState *cpu)
>          if (tb == NULL) {
>              mmap_lock();
>              tb_lock();
> -            tb = tb_htable_lookup(cpu, pc, cs_base, flags, cf_mask);
> -            if (likely(tb == NULL)) {
> -                tb = tb_gen_code(cpu, pc, cs_base, flags, cflags);
> -            }
> +            tb = tb_gen_code(cpu, pc, cs_base, flags, cflags);
>              tb_unlock();
>              mmap_unlock();
>          }
> @@ -398,14 +395,7 @@ static inline TranslationBlock *tb_find(CPUState *cpu,
>          tb_lock();
>          acquired_tb_lock = true;
> 
> -        /* There's a chance that our desired tb has been translated while
> -         * taking the locks so we check again inside the lock.
> -         */
> -        tb = tb_htable_lookup(cpu, pc, cs_base, flags, cf_mask);
> -        if (likely(tb == NULL)) {
> -            /* if no translated code available, then translate it now */
> -            tb = tb_gen_code(cpu, pc, cs_base, flags, cf_mask);
> -        }
> +        tb = tb_gen_code(cpu, pc, cs_base, flags, cf_mask);
> 
>          mmap_unlock();
>          /* We add the TB in the virtual pc hash table for the fast lookup */
> diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
> index c75298d08a..2585e6fd3e 100644
> --- a/accel/tcg/translate-all.c
> +++ b/accel/tcg/translate-all.c
> @@ -1581,18 +1581,30 @@ static inline void tb_page_add(PageDesc *p, TranslationBlock *tb,
>   * (-1) to indicate that only one page contains the TB.
>   *
>   * Called with mmap_lock held for user-mode emulation.
> + *
> + * Returns a pointer @tb, or a pointer to an existing TB that matches @tb.
> + * Note that in !user-mode, another thread might have already added a TB
> + * for the same block of guest code that @tb corresponds to. In that case,
> + * the caller should discard the original @tb, and use instead the returned TB.
>   */
> -static void tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
> -                         tb_page_addr_t phys_page2)
> +static TranslationBlock *
> +tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
> +             tb_page_addr_t phys_page2)
>  {
>      PageDesc *p;
>      PageDesc *p2 = NULL;
> +    void *existing_tb = NULL;
>      uint32_t h;
> 
>      assert_memory_lock();
> 
>      /*
>       * Add the TB to the page list, acquiring first the pages's locks.
> +     * We keep the locks held until after inserting the TB in the hash table,
> +     * so that if the insertion fails we know for sure that the TBs are still
> +     * in the page descriptors.
> +     * Note that inserting into the hash table first isn't an option, since
> +     * we can only insert TBs that are fully initialized.
>       */
>      page_lock_pair(&p, phys_pc, &p2, phys_page2, 1);
>      tb_page_add(p, tb, 0, phys_pc & TARGET_PAGE_MASK);
> @@ -1602,21 +1614,33 @@ static void tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
>          tb->page_addr[1] = -1;
>      }
> 
> +    /* add in the hash table */
> +    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
> +                     tb->trace_vcpu_dstate);
> +    qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
> +
> +    /* remove TB from the page(s) if we couldn't insert it */
> +    if (unlikely(existing_tb)) {
> +        tb_page_remove(p, tb);
> +        invalidate_page_bitmap(p);
> +        if (p2) {
> +            tb_page_remove(p2, tb);
> +            invalidate_page_bitmap(p2);
> +        }
> +        tb = existing_tb;
> +    }
> +
>      if (p2) {
>          page_unlock(p2);
>      }
>      page_unlock(p);
> 
> -    /* add in the hash table */
> -    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
> -                     tb->trace_vcpu_dstate);
> -    qht_insert(&tb_ctx.htable, tb, h, NULL);
> -
>  #ifdef CONFIG_USER_ONLY
>      if (DEBUG_TB_CHECK_GATE) {
>          tb_page_check();
>      }
>  #endif
> +    return tb;
>  }
> 
>  /* Called with mmap_lock held for user mode emulation.  */
> @@ -1625,7 +1649,7 @@ TranslationBlock *tb_gen_code(CPUState *cpu,
>                                uint32_t flags, int cflags)
>  {
>      CPUArchState *env = cpu->env_ptr;
> -    TranslationBlock *tb;
> +    TranslationBlock *tb, *existing_tb;
>      tb_page_addr_t phys_pc, phys_page2;
>      target_ulong virt_page2;
>      tcg_insn_unit *gen_code_buf;
> @@ -1773,7 +1797,15 @@ TranslationBlock *tb_gen_code(CPUState *cpu,
>       * memory barrier is required before tb_link_page() makes the TB visible
>       * through the physical hash table and physical page list.
>       */
> -    tb_link_page(tb, phys_pc, phys_page2);
> +    existing_tb = tb_link_page(tb, phys_pc, phys_page2);
> +    /* if the TB already exists, discard what we just translated */
> +    if (unlikely(existing_tb != tb)) {
> +        uintptr_t orig_aligned = (uintptr_t)gen_code_buf;
> +
> +        orig_aligned -= ROUND_UP(sizeof(*tb), qemu_icache_linesize);
> +        atomic_set(&tcg_ctx->code_gen_ptr, orig_aligned);
> +        return existing_tb;
> +    }
>      tcg_tb_insert(tb);
>      return tb;
>  }
> diff --git a/docs/devel/multi-thread-tcg.txt b/docs/devel/multi-thread-tcg.txt
> index faf8918b23..faf09c6069 100644
> --- a/docs/devel/multi-thread-tcg.txt
> +++ b/docs/devel/multi-thread-tcg.txt
> @@ -140,6 +140,9 @@ to atomically insert new elements.
>  The lookup caches are updated atomically and the lookup hash uses QHT
>  which is designed for concurrent safe lookup.
> 
> +Parallel code generation is supported. QHT is used at insertion time
> +as the synchronization point across threads, thereby ensuring that we only
> +keep track of a single TranslationBlock for each guest code block.
> 
>  Memory maps and TLBs
>  --------------------


Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Emilio G. Cota 7 years, 7 months ago
On Fri, Jun 29, 2018 at 10:25:03 +0300, Pavel Dovgalyuk wrote:
> This patch breaks record/replay.
> 
> I run execution recording of the WindowsXP machine with the following script:
> 
> ./bin/qemu-system-i386 -d in_asm,exec -D xp_save.log -global apic-common.vapic=off \
>   -icount shift=7,rr=record,rrfile=xp0.replay \
>   -drive file=./images/xp_sp2.qcow2,if=none,id=img-direct,snapshot \
>   -drive driver=blkreplay,if=none,image=img-direct,id=img-replay \
>   -device ide-hd,drive=img-replay -net none -m 512M
> 
> QEMU fails at some moment. Here are the contents of the log:
> 
> ----------------
> IN: 
> 0x806ee2d0:  33 c0                    xorl     %eax, %eax
> 0x806ee2d2:  8a c1                    movb     %cl, %al
> 0x806ee2d4:  33 c9                    xorl     %ecx, %ecx
> 0x806ee2d6:  8a 88 58 e2 6e 80        movb     -0x7f911da8(%eax), %cl
> 0x806ee2dc:  89 0d 80 00 fe ff        movl     %ecx, 0xfffe0080
> 0x806ee2e2:  a1 80 00 fe ff           movl     0xfffe0080, %eax
> 0x806ee2e7:  c3                       retl     
> 
> Trace 0: 0x7fdc103b16a0 [00000000/806ee2d0/0x4000b0]
> qemu: fatal: cpu_io_recompile: could not find TB for pc=0x7fec24fde2de

Thanks for reporting.

From code inspection I can see how this could happen: we're calling
tcg_tb_remove for a TB that we did not just generate--we got an
existing one instead. Note that CF_NOCACHE is not part of
the CF_HASH mask, so this might explain why the problem only
occurs for r/r.

Can you reproduce this with any other guest? If not, I'd be
happy to use your windows qcow2 file if you could share it
with me off-list.

Thanks,

		Emilio

Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Pavel Dovgalyuk 7 years, 7 months ago
> From: Emilio G. Cota [mailto:cota@braap.org]
> On Fri, Jun 29, 2018 at 10:25:03 +0300, Pavel Dovgalyuk wrote:
> > This patch breaks record/replay.
> >
> > I run execution recording of the WindowsXP machine with the following script:
> >
> > ./bin/qemu-system-i386 -d in_asm,exec -D xp_save.log -global apic-common.vapic=off \
> >   -icount shift=7,rr=record,rrfile=xp0.replay \
> >   -drive file=./images/xp_sp2.qcow2,if=none,id=img-direct,snapshot \
> >   -drive driver=blkreplay,if=none,image=img-direct,id=img-replay \
> >   -device ide-hd,drive=img-replay -net none -m 512M
> >
> > QEMU fails at some moment. Here are the contents of the log:
> >
> > ----------------
> > IN:
> > 0x806ee2d0:  33 c0                    xorl     %eax, %eax
> > 0x806ee2d2:  8a c1                    movb     %cl, %al
> > 0x806ee2d4:  33 c9                    xorl     %ecx, %ecx
> > 0x806ee2d6:  8a 88 58 e2 6e 80        movb     -0x7f911da8(%eax), %cl
> > 0x806ee2dc:  89 0d 80 00 fe ff        movl     %ecx, 0xfffe0080
> > 0x806ee2e2:  a1 80 00 fe ff           movl     0xfffe0080, %eax
> > 0x806ee2e7:  c3                       retl
> >
> > Trace 0: 0x7fdc103b16a0 [00000000/806ee2d0/0x4000b0]
> > qemu: fatal: cpu_io_recompile: could not find TB for pc=0x7fec24fde2de
> 
> Thanks for reporting.
> 
> From code inspection I can see how this could happen: we're calling
> tcg_tb_remove for a TB that we did not just generate--we got an
> existing one instead. Note that CF_NOCACHE is not part of
> the CF_HASH mask, so this might explain why the problem only
> occurs for r/r.

Thanks.

> Can you reproduce this with any other guest? If not, I'd be
> happy to use your windows qcow2 file if you could share it
> with me off-list.

The same failure can be reproduced with linux-0.2.img, which was
downloaded from QEMU site.
I can't find it now, but I can upload this file if needed.

Pavel Dovgalyuk


Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Emilio G. Cota 7 years, 7 months ago
On Mon, Jul 02, 2018 at 08:52:14 +0300, Pavel Dovgalyuk wrote:
> The same failure can be reproduced with linux-0.2.img, which was
> downloaded from QEMU site.
> I can't find it now, but I can upload this file if needed.

Please upload it somewhere and share the full QEMU invocation
needed to replicate.

Thanks,

		Emilio

Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Pavel Dovgalyuk 7 years, 7 months ago
> From: Emilio G. Cota [mailto:cota@braap.org]
> On Mon, Jul 02, 2018 at 08:52:14 +0300, Pavel Dovgalyuk wrote:
> > The same failure can be reproduced with linux-0.2.img, which was
> > downloaded from QEMU site.
> > I can't find it now, but I can upload this file if needed.
> 
> Please upload it somewhere and share the full QEMU invocation
> needed to replicate.

https://github.com/Dovgalyuk/qemu-images/blob/master/linux-0.2.img

qemu-system-i386 -drive file=images/linux-0.2.img,if=none,snapshot,id=img -drive
driver=blkreplay,if=none,id=rr,image=img -device ide-hd,drive=rr -net none -icount
shift=5,rr=record,rrfile=linux02.rr

Pavel Dovgalyuk


Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Emilio G. Cota 7 years, 7 months ago
On Tue, Jul 03, 2018 at 08:38:52 +0300, Pavel Dovgalyuk wrote:
> > From: Emilio G. Cota [mailto:cota@braap.org]
> > On Mon, Jul 02, 2018 at 08:52:14 +0300, Pavel Dovgalyuk wrote:
> > > The same failure can be reproduced with linux-0.2.img, which was
> > > downloaded from QEMU site.
> > > I can't find it now, but I can upload this file if needed.
> > 
> > Please upload it somewhere and share the full QEMU invocation
> > needed to replicate.
> 
> https://github.com/Dovgalyuk/qemu-images/blob/master/linux-0.2.img
> 
> qemu-system-i386 -drive file=images/linux-0.2.img,if=none,snapshot,id=img -drive
> driver=blkreplay,if=none,id=rr,image=img -device ide-hd,drive=rr -net none -icount
> shift=5,rr=record,rrfile=linux02.rr

The appended patch fixes it for me. Can you please test on your
windows image?

The rationale is to honour CF_NOCACHE, so that we always return
a new TB from tb_gen_code.

Thanks,

		Emilio

---
diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
index 170b957..49d77fa 100644
--- a/accel/tcg/translate-all.c
+++ b/accel/tcg/translate-all.c
@@ -1446,7 +1446,8 @@ static void do_tb_phys_invalidate(TranslationBlock *tb, bool rm_from_page_list)
     phys_pc = tb->page_addr[0] + (tb->pc & ~TARGET_PAGE_MASK);
     h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb_cflags(tb) & CF_HASH_MASK,
                      tb->trace_vcpu_dstate);
-    if (!qht_remove(&tb_ctx.htable, tb, h)) {
+    if (!(tb->cflags & CF_NOCACHE) &&
+        !qht_remove(&tb_ctx.htable, tb, h)) {
         return;
     }
 
@@ -1604,8 +1605,6 @@ tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
 {
     PageDesc *p;
     PageDesc *p2 = NULL;
-    void *existing_tb = NULL;
-    uint32_t h;
 
     assert_memory_lock();
 
@@ -1625,20 +1624,25 @@ tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
         tb->page_addr[1] = -1;
     }
 
-    /* add in the hash table */
-    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
-                     tb->trace_vcpu_dstate);
-    qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
+    if (!(tb->cflags & CF_NOCACHE)) {
+        void *existing_tb = NULL;
+        uint32_t h;
 
-    /* remove TB from the page(s) if we couldn't insert it */
-    if (unlikely(existing_tb)) {
-        tb_page_remove(p, tb);
-        invalidate_page_bitmap(p);
-        if (p2) {
-            tb_page_remove(p2, tb);
-            invalidate_page_bitmap(p2);
+        /* add in the hash table */
+        h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
+                         tb->trace_vcpu_dstate);
+        qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
+
+        /* remove TB from the page(s) if we couldn't insert it */
+        if (unlikely(existing_tb)) {
+            tb_page_remove(p, tb);
+            invalidate_page_bitmap(p);
+            if (p2) {
+                tb_page_remove(p2, tb);
+                invalidate_page_bitmap(p2);
+            }
+            tb = existing_tb;
         }
-        tb = existing_tb;
     }
 
     if (p2 && p2 != p) {

Re: [Qemu-devel] [PULL, 14/18] translate-all: discard TB when tb_link_page returns an existing matching TB
Posted by Pavel Dovgalyuk 7 years, 7 months ago
> From: Emilio G. Cota [mailto:cota@braap.org]
> On Tue, Jul 03, 2018 at 08:38:52 +0300, Pavel Dovgalyuk wrote:
> > > From: Emilio G. Cota [mailto:cota@braap.org]
> > > On Mon, Jul 02, 2018 at 08:52:14 +0300, Pavel Dovgalyuk wrote:
> > > > The same failure can be reproduced with linux-0.2.img, which was
> > > > downloaded from QEMU site.
> > > > I can't find it now, but I can upload this file if needed.
> > >
> > > Please upload it somewhere and share the full QEMU invocation
> > > needed to replicate.
> >
> > https://github.com/Dovgalyuk/qemu-images/blob/master/linux-0.2.img
> >
> > qemu-system-i386 -drive file=images/linux-0.2.img,if=none,snapshot,id=img -drive
> > driver=blkreplay,if=none,id=rr,image=img -device ide-hd,drive=rr -net none -icount
> > shift=5,rr=record,rrfile=linux02.rr
> 
> The appended patch fixes it for me. Can you please test on your
> windows image?
> 
> The rationale is to honour CF_NOCACHE, so that we always return
> a new TB from tb_gen_code.

Works for me, thank you.

Tested-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>

Pavel Dovgalyuk

> 
> ---
> diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
> index 170b957..49d77fa 100644
> --- a/accel/tcg/translate-all.c
> +++ b/accel/tcg/translate-all.c
> @@ -1446,7 +1446,8 @@ static void do_tb_phys_invalidate(TranslationBlock *tb, bool
> rm_from_page_list)
>      phys_pc = tb->page_addr[0] + (tb->pc & ~TARGET_PAGE_MASK);
>      h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb_cflags(tb) & CF_HASH_MASK,
>                       tb->trace_vcpu_dstate);
> -    if (!qht_remove(&tb_ctx.htable, tb, h)) {
> +    if (!(tb->cflags & CF_NOCACHE) &&
> +        !qht_remove(&tb_ctx.htable, tb, h)) {
>          return;
>      }
> 
> @@ -1604,8 +1605,6 @@ tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
>  {
>      PageDesc *p;
>      PageDesc *p2 = NULL;
> -    void *existing_tb = NULL;
> -    uint32_t h;
> 
>      assert_memory_lock();
> 
> @@ -1625,20 +1624,25 @@ tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
>          tb->page_addr[1] = -1;
>      }
> 
> -    /* add in the hash table */
> -    h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
> -                     tb->trace_vcpu_dstate);
> -    qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
> +    if (!(tb->cflags & CF_NOCACHE)) {
> +        void *existing_tb = NULL;
> +        uint32_t h;
> 
> -    /* remove TB from the page(s) if we couldn't insert it */
> -    if (unlikely(existing_tb)) {
> -        tb_page_remove(p, tb);
> -        invalidate_page_bitmap(p);
> -        if (p2) {
> -            tb_page_remove(p2, tb);
> -            invalidate_page_bitmap(p2);
> +        /* add in the hash table */
> +        h = tb_hash_func(phys_pc, tb->pc, tb->flags, tb->cflags & CF_HASH_MASK,
> +                         tb->trace_vcpu_dstate);
> +        qht_insert(&tb_ctx.htable, tb, h, &existing_tb);
> +
> +        /* remove TB from the page(s) if we couldn't insert it */
> +        if (unlikely(existing_tb)) {
> +            tb_page_remove(p, tb);
> +            invalidate_page_bitmap(p);
> +            if (p2) {
> +                tb_page_remove(p2, tb);
> +                invalidate_page_bitmap(p2);
> +            }
> +            tb = existing_tb;
>          }
> -        tb = existing_tb;
>      }
> 
>      if (p2 && p2 != p) {