Re: [PATCH] mm/slub: use stackdepot to save stack trace in objects

[Andrew Morton]

On Wed, 12 May 2021 16:33:50 +0200 Vlastimil Babka <vbabka@xxxxxxx> wrote:

> On 5/10/21 6:46 AM, Andrew Morton wrote:
> > On Wed, 14 Apr 2021 18:34:34 +0200 glittao@xxxxxxxxx wrote:
> > 
> >> Many stack traces are similar so there are many similar arrays.
> >> Stackdepot saves each unique stack only once.
> >> 
> >> Replace field addrs in struct track with depot_stack_handle_t handle.
> >> Use stackdepot to save stack trace.
> >> 
> >> The benefits are smaller memory overhead and possibility to aggregate
> >> per-cache statistics in the future using the stackdepot handle
> >> instead of matching stacks manually.
> > 
> > Which tree was this prepared against?  5.12's kmem_obj_info() is
> > significantly different from the version you were working on.
> 
> It was based on -next at the time of submission, which contained patch in Paul's
> tree that expands kmem_obj_info to print also free call stack [1] so that also
> needs to be switched to stackdepot to work.

OK, sorry, I should have checked.

> > Please take a look, redo, retest and resend?  Thanks.
> 
> I expected [1] to be in 5.13-rc1, but as Paul explained to me, it's queued for
> 5.14. So if we (Oliver) rebase on current -next, can you queue it in the section
> of mmotm series that goes after -next?

I grabbed this version and queued it after the linux-next patches,
thanks.