record/replay fixes: attempting to get avocado green

[PATCH v2 14/16] accel/tcg: define tlb_fill as a trace point

Posted by Alex Bennée 1 year, 11 months ago

While we do have a bunch of debugging we can turn on in cputlb there
is an argument for making tlb_fill a generally available trace point.
Any fault (via probe or access) will have to be preceded by a tlb_fill
for the address in question.

We don't bother logging the return address as that will be a host
address associated with translation and therefor can move around with
ASLR.

In my particular case I'm trying to track down a difference in memory
fault exception patterns between record and replay phases.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
---
 accel/tcg/cputlb.c     | 2 ++
 accel/tcg/trace-events | 1 +
 2 files changed, 3 insertions(+)

diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index db3f93fda9..c35df27caf 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -1308,6 +1308,8 @@ static void tlb_fill(CPUState *cpu, vaddr addr, int size,
 {
     bool ok;
 
+    trace_tlb_fill(addr, size, access_type, mmu_idx);
+
     /*
      * This is not a probe, so only valid return is success; failure
      * should result in exception + longjmp to the cpu loop.
diff --git a/accel/tcg/trace-events b/accel/tcg/trace-events
index 4e9b450520..31dda01c12 100644
--- a/accel/tcg/trace-events
+++ b/accel/tcg/trace-events
@@ -9,6 +9,7 @@ exec_tb_exit(void *last_tb, unsigned int flags) "tb:%p flags=0x%x"
 # cputlb.c
 memory_notdirty_write_access(uint64_t vaddr, uint64_t ram_addr, unsigned size) "0x%" PRIx64 " ram_addr 0x%" PRIx64 " size %u"
 memory_notdirty_set_dirty(uint64_t vaddr) "0x%" PRIx64
+tlb_fill(uint64_t vaddr, int size, int access_type, int mmu_idx) "0x%" PRIx64 "/%d %d %d"
 
 # translate-all.c
 translate_block(void *tb, uintptr_t pc, const void *tb_code) "tb:%p, pc:0x%"PRIxPTR", tb_code:%p"
-- 
2.39.2

Re: [PATCH v2 14/16] accel/tcg: define tlb_fill as a trace point

Posted by Richard Henderson 1 year, 11 months ago

On 12/11/23 01:13, Alex Bennée wrote:
> While we do have a bunch of debugging we can turn on in cputlb there
> is an argument for making tlb_fill a generally available trace point.
> Any fault (via probe or access) will have to be preceded by a tlb_fill
> for the address in question.
> 
> We don't bother logging the return address as that will be a host
> address associated with translation and therefor can move around with
> ASLR.
> 
> In my particular case I'm trying to track down a difference in memory
> fault exception patterns between record and replay phases.
> 
> Signed-off-by: Alex Bennée<alex.bennee@linaro.org>
> ---
>   accel/tcg/cputlb.c     | 2 ++
>   accel/tcg/trace-events | 1 +
>   2 files changed, 3 insertions(+)

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

r~

Re: [PATCH v2 14/16] accel/tcg: define tlb_fill as a trace point

Posted by Philippe Mathieu-Daudé 1 year, 11 months ago

On 11/12/23 10:13, Alex Bennée wrote:
> While we do have a bunch of debugging we can turn on in cputlb there
> is an argument for making tlb_fill a generally available trace point.
> Any fault (via probe or access) will have to be preceded by a tlb_fill
> for the address in question.
> 
> We don't bother logging the return address as that will be a host
> address associated with translation and therefor can move around with
> ASLR.
> 
> In my particular case I'm trying to track down a difference in memory
> fault exception patterns between record and replay phases.
> 
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> ---
>   accel/tcg/cputlb.c     | 2 ++
>   accel/tcg/trace-events | 1 +
>   2 files changed, 3 insertions(+)

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>