Re: [RFC] kallsyms: embed source file:line info in kernel stack traces

On Tue, Mar 03, 2026 at 10:04:51AM +0100, Vlastimil Babka (SUSE) wrote:
>On 3/3/26 07:48, Tomasz Figa wrote:
>>> > Memory footprint measured with a simple KVM guest x86_64 config:
>>> >
>>> >  Table: 4,597,583 entries from 4,841 source files
>>> >    lineinfo_addrs[]     4,597,583 x u32  = 17.5 MiB
>>> >    lineinfo_file_ids[]  4,597,583 x u16  =  8.8 MiB
>>> >    lineinfo_lines[]     4,597,583 x u32  = 17.5 MiB
>>> >    file_offsets + filenames              ~  0.1 MiB
>>> >    Total .rodata increase:              ~ 44.0 MiB
>>> >
>>> >  vmlinux (stripped):  529 MiB -> 573 MiB  (+44 MiB / +8.3%)
>>>
>>> Hm, that's a significant increase.
>>
>> Random idea: Could this additional information (and I guess the code
>> that uses it too) be moved out to a loadable module?
>>
>> The obvious limitation would be that the user would need to have the
>> module loaded for the decoding to work, but that could be worked
>> around by marking it for autoload when a crash is noticed the first
>> time and then getting a better report the second time.
>
>IMHO that's too big of a disadvantage. Many crashes are rare and either you
>have this always enabled, or you missed your only chance.

If you already have it the crash trace from first time, and you obviously have
the decoding information, just run scripts/decode_stacktrace.sh in userspace?

-- 
Thanks,
Sasha