2014-02-11 15:30:19

by Alexander Holler

[permalink] [raw]
Subject: work unwinding?

Hello,

Since the kernel has become more RT-aware and many stuff has moved into
works, dump_stack() has often become a bit useless, because it just
shows the call stack of a work, which quiet often is a bit short and
doesn't reveal the real source of a problem.

At least I would like it if there would be some mechanism to record the
call stack for works, so that dump_stack() could display what actually
queued a work (and it's call stack).

Not that I want to implement such, I'm just curious if someone already
has thought about implementing something like that, which I'm calling
here "work unwinding" ;)

Regards,

Alexander Holler