Re: [PATCH v2] memcg: add charging of already allocated slab objects

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Aug 28, 2024 at 5:20 PM Shakeel Butt <shakeel.butt@xxxxxxxxx> wrote:
>
> On Wed, Aug 28, 2024 at 04:25:30PM GMT, Yosry Ahmed wrote:
> > On Tue, Aug 27, 2024 at 4:52 PM Shakeel Butt <shakeel.butt@xxxxxxxxx> wrote:
> > >
> [...]
> > > +
> > > +       /* Ignore KMALLOC_NORMAL cache to avoid circular dependency. */
> > > +       if ((s->flags & KMALLOC_TYPE) == SLAB_KMALLOC)
> > > +               return true;
> >
> > Taking a step back here, why do we need this? Which circular
> > dependency are we avoiding here?
>
> commit 494c1dfe855ec1f70f89552fce5eadf4a1717552
> Author: Waiman Long <longman@xxxxxxxxxx>
> Date:   Mon Jun 28 19:37:38 2021 -0700
>
>     mm: memcg/slab: create a new set of kmalloc-cg-<n> caches
>
>     There are currently two problems in the way the objcg pointer array
>     (memcg_data) in the page structure is being allocated and freed.
>
>     On its allocation, it is possible that the allocated objcg pointer
>     array comes from the same slab that requires memory accounting. If this
>     happens, the slab will never become empty again as there is at least
>     one object left (the obj_cgroup array) in the slab.
>
>     When it is freed, the objcg pointer array object may be the last one
>     in its slab and hence causes kfree() to be called again. With the
>     right workload, the slab cache may be set up in a way that allows the
>     recursive kfree() calling loop to nest deep enough to cause a kernel
>     stack overflow and panic the system.
>     ...

Thanks for the reference, this makes sense.

Wouldn't it be easier to special case the specific slab cache used for
the objcg vector or use a dedicated cache for it instead of using
kmalloc caches to begin with?

Anyway, I am fine with any approach you and/or the slab maintainers
prefer, as long as we make things clear. If you keep the following
approach as-is, please expand the comment or refer to the commit you
just referenced.

Personally, I prefer either explicitly special casing the slab cache
used for the objcgs vector, explicitly tagging KMALLOC_NORMAL
allocations, or having a dedicated documented helper that finds the
slab cache kmalloc type (if any) or checks if it is a KMALLOC_NORMAL
cache.





[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux