1. Patchew
  2. QEMU
  3. View series

[PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary

Alex Bennée posted 1 patch 5 years, 8 months ago
Test docker-mingw@fedora passed
Test checkpatch passed
Test asan passed
Test docker-quick@centos7 passed
Test FreeBSD passed
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20200603112442.22833-1-alex.bennee@linaro.org
exec.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
[PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary
Posted by Alex Bennée 5 years, 8 months ago 
There is no particular reason why you can't have a watchpoint in TCG
that covers a large chunk of the address space. We could be clever
about it but these cases are pretty rare and we can assume the user
will expect a little performance degradation.

NB: In my testing gdb will silently squash a watchpoint like:

  watch (char[0x7fffffffff]) *0x0

to a 4 byte watchpoint. Practically it will limit the maximum size
based on max-value-size. However given enough of a tweak the sky is
the limit.

Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

---
v2
  - use cleaner in_page = -(addr | TARGET_PAGE_MASK) logic per rth
---
 exec.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/exec.c b/exec.c
index 5162f0d12f9..65a4376df37 100644
--- a/exec.c
+++ b/exec.c
@@ -1036,6 +1036,7 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
                           int flags, CPUWatchpoint **watchpoint)
 {
     CPUWatchpoint *wp;
+    vaddr in_page;
 
     /* forbid ranges which are empty or run off the end of the address space */
     if (len == 0 || (addr + len - 1) < addr) {
@@ -1056,7 +1057,12 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
         QTAILQ_INSERT_TAIL(&cpu->watchpoints, wp, entry);
     }
 
-    tlb_flush_page(cpu, addr);
+    in_page = -(addr | TARGET_PAGE_MASK);
+    if (len <= in_page) {
+        tlb_flush_page(cpu, addr);
+    } else {
+        tlb_flush(cpu);
+    }
 
     if (watchpoint)
         *watchpoint = wp;
-- 
2.20.1
Re: [PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary
Posted by Philippe Mathieu-Daudé 5 years, 8 months ago 
On 6/3/20 1:24 PM, Alex Bennée wrote:
> There is no particular reason why you can't have a watchpoint in TCG
> that covers a large chunk of the address space. We could be clever
> about it but these cases are pretty rare and we can assume the user
> will expect a little performance degradation.
> 
> NB: In my testing gdb will silently squash a watchpoint like:
> 
>   watch (char[0x7fffffffff]) *0x0
> 
> to a 4 byte watchpoint. Practically it will limit the maximum size
> based on max-value-size. However given enough of a tweak the sky is
> the limit.
> 
> Reported-by: Alexander Bulekov <alxndr@bu.edu>
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> 
> ---
> v2
>   - use cleaner in_page = -(addr | TARGET_PAGE_MASK) logic per rth

Can we have a macro for this?
Maybe QEMU_IN_PAGE_OFFSET(addr, TARGET_PAGE_MASK)?
or QEMU_OFFSET_IN_PAGE()...

> ---
>  exec.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/exec.c b/exec.c
> index 5162f0d12f9..65a4376df37 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1036,6 +1036,7 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
>                            int flags, CPUWatchpoint **watchpoint)
>  {
>      CPUWatchpoint *wp;
> +    vaddr in_page;
>  
>      /* forbid ranges which are empty or run off the end of the address space */
>      if (len == 0 || (addr + len - 1) < addr) {
> @@ -1056,7 +1057,12 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
>          QTAILQ_INSERT_TAIL(&cpu->watchpoints, wp, entry);
>      }
>  
> -    tlb_flush_page(cpu, addr);
> +    in_page = -(addr | TARGET_PAGE_MASK);
> +    if (len <= in_page) {
> +        tlb_flush_page(cpu, addr);
> +    } else {
> +        tlb_flush(cpu);
> +    }
>  
>      if (watchpoint)
>          *watchpoint = wp;
>
Re: [PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary
Posted by Philippe Mathieu-Daudé 5 years, 8 months ago 
On 6/3/20 2:46 PM, Philippe Mathieu-Daudé wrote:
> On 6/3/20 1:24 PM, Alex Bennée wrote:
>> There is no particular reason why you can't have a watchpoint in TCG
>> that covers a large chunk of the address space. We could be clever
>> about it but these cases are pretty rare and we can assume the user
>> will expect a little performance degradation.
>>
>> NB: In my testing gdb will silently squash a watchpoint like:
>>
>>   watch (char[0x7fffffffff]) *0x0
>>
>> to a 4 byte watchpoint. Practically it will limit the maximum size
>> based on max-value-size. However given enough of a tweak the sky is
>> the limit.
>>
>> Reported-by: Alexander Bulekov <alxndr@bu.edu>
>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>>
>> ---
>> v2
>>   - use cleaner in_page = -(addr | TARGET_PAGE_MASK) logic per rth
> 
> Can we have a macro for this?
> Maybe QEMU_IN_PAGE_OFFSET(addr, TARGET_PAGE_MASK)?
> or QEMU_OFFSET_IN_PAGE()...

As this is queued, I suppose the implicit answer is "no."

> 
>> ---
>>  exec.c | 8 +++++++-
>>  1 file changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/exec.c b/exec.c
>> index 5162f0d12f9..65a4376df37 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -1036,6 +1036,7 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
>>                            int flags, CPUWatchpoint **watchpoint)
>>  {
>>      CPUWatchpoint *wp;
>> +    vaddr in_page;
>>  
>>      /* forbid ranges which are empty or run off the end of the address space */
>>      if (len == 0 || (addr + len - 1) < addr) {
>> @@ -1056,7 +1057,12 @@ int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
>>          QTAILQ_INSERT_TAIL(&cpu->watchpoints, wp, entry);
>>      }
>>  
>> -    tlb_flush_page(cpu, addr);
>> +    in_page = -(addr | TARGET_PAGE_MASK);
>> +    if (len <= in_page) {
>> +        tlb_flush_page(cpu, addr);
>> +    } else {
>> +        tlb_flush(cpu);
>> +    }
>>  
>>      if (watchpoint)
>>          *watchpoint = wp;
>>
>
Re: [PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary
Posted by Alex Bennée 5 years, 8 months ago 
Philippe Mathieu-Daudé <philmd@redhat.com> writes:

> On 6/3/20 2:46 PM, Philippe Mathieu-Daudé wrote:
>> On 6/3/20 1:24 PM, Alex Bennée wrote:
>>> There is no particular reason why you can't have a watchpoint in TCG
>>> that covers a large chunk of the address space. We could be clever
>>> about it but these cases are pretty rare and we can assume the user
>>> will expect a little performance degradation.
>>>
>>> NB: In my testing gdb will silently squash a watchpoint like:
>>>
>>>   watch (char[0x7fffffffff]) *0x0
>>>
>>> to a 4 byte watchpoint. Practically it will limit the maximum size
>>> based on max-value-size. However given enough of a tweak the sky is
>>> the limit.
>>>
>>> Reported-by: Alexander Bulekov <alxndr@bu.edu>
>>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>>>
>>> ---
>>> v2
>>>   - use cleaner in_page = -(addr | TARGET_PAGE_MASK) logic per rth
>> 
>> Can we have a macro for this?
>> Maybe QEMU_IN_PAGE_OFFSET(addr, TARGET_PAGE_MASK)?
>> or QEMU_OFFSET_IN_PAGE()...
>
> As this is queued, I suppose the implicit answer is "no."

Richard took it into tcg/next as is. I think having a macro may well be
nice clean-up but I struggled to pick a good include location so left it
for a future clean-up series ;-)

-- 
Alex Bennée
Re: [PATCH v2] exec: flush the whole TLB if a watchpoint crosses a page boundary
Posted by Richard Henderson 5 years, 8 months ago 
On 6/3/20 4:24 AM, Alex Bennée wrote:
> There is no particular reason why you can't have a watchpoint in TCG
> that covers a large chunk of the address space. We could be clever
> about it but these cases are pretty rare and we can assume the user
> will expect a little performance degradation.
> 
> NB: In my testing gdb will silently squash a watchpoint like:
> 
>   watch (char[0x7fffffffff]) *0x0
> 
> to a 4 byte watchpoint. Practically it will limit the maximum size
> based on max-value-size. However given enough of a tweak the sky is
> the limit.
> 
> Reported-by: Alexander Bulekov <alxndr@bu.edu>
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> 
> ---
> v2
>   - use cleaner in_page = -(addr | TARGET_PAGE_MASK) logic per rth
> ---
>  exec.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

Queued to tcg-next.


r~

Patchew 2.1-devel-b67e4c2

© 2016 - 2026 Red Hat, Inc.