Re: [PATCH bpf v4 3/3] bpf: Cache build IDs in sleepable stackmap path

[Ihor Solodrai <ihor.solodrai@linux.dev>]

On 5/14/26 5:24 PM, Alexei Starovoitov wrote:
> On Thu, May 14, 2026 at 11:48 AM Ihor Solodrai <ihor.solodrai@linux.dev> wrote:
>>
>> Stack traces often contain adjacent IPs from the same VMA or from
>> different VMAs backed by the same ELF file. Cache the last successfully
>> parsed build ID together with the resolved VMA range and backing file
>> so the sleepable build-ID path can avoid repeated VMA locking and file
>> parsing in common cases.
>>
>> Suggested-by: Mykyta Yatsenko <yatsenko@meta.com>
>> Acked-by: Mykyta Yatsenko <yatsenko@meta.com>
>> Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
>> ---
>>  kernel/bpf/stackmap.c | 56 ++++++++++++++++++++++++++++++++++++++++---
>>  1 file changed, 53 insertions(+), 3 deletions(-)
>>
>> diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
>> index c1e96df360c3..8a18c9645933 100644
>> --- a/kernel/bpf/stackmap.c
>> +++ b/kernel/bpf/stackmap.c
>> @@ -226,13 +226,33 @@ static void stack_map_get_build_id_offset_sleepable(struct bpf_stack_build_id *i
>>                 .vma = NULL,
>>                 .mm = mm,
>>         };
>> -       unsigned long vm_pgoff, vm_start;
>> +       struct {
>> +               struct file *file;
>> +               const unsigned char *build_id;
>> +               unsigned long vm_start;
>> +               unsigned long vm_end;
>> +               unsigned long vm_pgoff;
>> +       } cache = {};
>> +       unsigned long vm_pgoff, vm_start, vm_end;
> 
> cache logic cannot go into bpf. it's not a fix.
> So either drop it or the whole thing into bpf-next.
> Which is probably better.

I'll send a v5 targeting bpf-next shortly. Thanks.