systemtap on 4.16 kernel - stack unwinding problems

systemtap on 4.16 kernel - stack unwinding problems

[Phil Kirk]

hi,

i've been trying to use systemtap against a freshly built 4.16 kernel,
systemtap was also freshly built, but having problems with stack unwinding,
as an example, when running:

stap -e 'probe kernel.function("invalidate_mapping_pages")
{printf("stack:%s\n", print_backtrace());}'

i was only getting the top stack frame. when tracing inside the ip stack i
was getting the same results. after asking on irc i rebuilt my kernel
without CONFIG_RANDOMIZE_BASE set to y and now i get a more complete stack
trace, i need to try and verify if the stack i'm getting when tracing
invalidate_mapping_pages is accurate but for tracing in the ip stack it
looks much better.

the kernel was built using gcc 5.4.0, i'll try using a more up to date gcc
tomorrow.

phil

Re: systemtap on 4.16 kernel - stack unwinding problems

[Phil Kirk]

i've tried re-building with gcc-7 (7.2.0) but stack unwinding still fails
with CONFIG_RANDOMIZE_BASE set. for now i'll just build without it so i can
get the stack traces i want.

thanks

phil

On Fri, Mar 23, 2018 at 11:04 PM, Phil Kirk <pk34663@gmail.com> wrote:

> hi,
>
> i've been trying to use systemtap against a freshly built 4.16 kernel,
> systemtap was also freshly built, but having problems with stack unwinding,
> as an example, when running:
>
> stap -e 'probe kernel.function("invalidate_mapping_pages")
> {printf("stack:%s\n", print_backtrace());}'
>
> i was only getting the top stack frame. when tracing inside the ip stack i
> was getting the same results. after asking on irc i rebuilt my kernel
> without CONFIG_RANDOMIZE_BASE set to y and now i get a more complete stack
> trace, i need to try and verify if the stack i'm getting when tracing
> invalidate_mapping_pages is accurate but for tracing in the ip stack it
> looks much better.
>
> the kernel was built using gcc 5.4.0, i'll try using a more up to date gcc
> tomorrow.
>
> phil
>