On Wed, Nov 06, 2024 at 02:21:33PM +0100, Boris Brezillon wrote:
> On Wed, 6 Nov 2024 13:10:37 +0000
> Liviu Dudau <liviu.dudau@xxxxxxx> wrote:
>
> > panthor_heap_pool_release() does not take the panthor_heap_pool::lock, so the protection
> > is not really there. I could fix panthor_heap_pool_release() and then add a
> > lockdep_assert_held() before both calls to xa_load() if you think that's a better
> > solution.
>
> Hm, but panthor_heap_pool_release() doesn't release the heap contexts,
> it just calls xa_destroy(). If we have objects remaining in the xarray,
> they'll be leaked, but that's not a race. BTW, can we make this two
> separate patches. I feel like the thing on the vm is an actual fix,
> while the second one (adding a helper with a lockdep_assert()) is
> safety net that's worth having, but not necessarily something we need
> to backport.
I've decided to drop the panthor_heap.c changes as Boris is right, the pool->lock
should be enough. Adding a lockdep_assert() right after the down_write() call
also feels a bit silly, so I did not bother with it.
Best regards,
Liviu
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
