Re: Failure during Stack Depot allocating hash table of 1048576 entries with kvcalloc

[Andrew Morton]

On Sat, 28 Jan 2023 15:26:00 +0100 Borislav Petkov <bp@xxxxxxxxx> wrote:

> On Sat, Jan 28, 2023 at 02:55:58PM +0100, Linux kernel regression tracking (Thorsten Leemhuis) wrote:
> > On 28.01.23 12:03, Borislav Petkov wrote:
> > > On Sat, Jan 28, 2023 at 03:41:50AM +0100, Mirsad Goran Todorovac wrote:
> > >> This appears to be a duplicate of the report:
> > >> https://lore.kernel.org/linux-mm/2c677d85-820c-d41a-fc98-7d3974b49e42@xxxxxxxxxxxx/raw
> > > 
> > > Yah, looks like
> > > 
> > > 56a61617dd22 ("mm: use stack_depot for recording kmemleak's backtrace")
> > > 
> > > needs to be reverted.
> > 
> > Unless I'm missing something (which might easily be the case) there is a
> > patch for that issue in -mm already:
> > 
> > https://lore.kernel.org/all/20230119224022.80752C433F0@xxxxxxxxxxxxxxx/
> > 
> > Or where two different issues discussed in the thread Mirsad mentioned
> > above?
> 
> Probably the same issue. This one fixes the issue on my machine - thanks!
> 
> Tested-by: Borislav Petkov (AMD) <bp@xxxxxxxxx>
> 

OK, thanks, I didn't realize this issue was so serious.

I reordered Zhaoyang Huang's series so that "mm: use
stack_depot_early_init for kmemleak" comes ahead of "mm: move
KMEMLEAK's Kconfig items from lib to mm" and I've staged "mm: use
stack_depot_early_init for kmemleak" in the mm-hotfixes branch for
upstream merging in this -rc cycle.