Messages in this thread | | | Date | Thu, 15 Sep 2016 11:27:29 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH 08/12] x86/dumpstack: Pin the target stack in save_stack_trace_tsk() |
| |
* Linus Torvalds <torvalds@linux-foundation.org> wrote:
> It absolutely is - compared to the vmalloc costs. Andy has quite noticeable > slowdowns without the stack reuse - and the stack reuse requires the delayed > freeing. With the stack reuse, fork/exit actually sped up.
Ok, great, that's convincing!
I'm still applying it two-phase and I'll wait for v2 of the second half of the series, to have the fix Josh alluded to.
Thanks,
Ingo
| |