Re: [PATCH net] rhashtable: fix crash due to mm api change

[Michal Hocko <mhocko@suse.com>]

On Mon 26-09-22 10:31:39, Florian Westphal wrote:
> Martin Zaharinov reports BUG() in mm land for 5.19.10 kernel:
>  kernel BUG at mm/vmalloc.c:2437!
>  invalid opcode: 0000 [#1] SMP
>  CPU: 28 PID: 0 Comm: swapper/28 Tainted: G        W  O      5.19.9 #1
>  [..]
>  RIP: 0010:__get_vm_area_node+0x120/0x130
>   __vmalloc_node_range+0x96/0x1e0
>   kvmalloc_node+0x92/0xb0
>   bucket_table_alloc.isra.0+0x47/0x140
>   rhashtable_try_insert+0x3a4/0x440
>   rhashtable_insert_slow+0x1b/0x30
>  [..]
> 
> bucket_table_alloc uses kvzalloc(GPF_ATOMIC).  If kmalloc fails, this now
> falls through to vmalloc and hits code paths that assume GFP_KERNEL.
> 
> I sent a patch to restore GFP_ATOMIC support in kvmalloc but mm
> maintainers rejected it.
> 
> This patch is partial revert of
> commit 93f976b5190d ("lib/rhashtable: simplify bucket_table_alloc()"),
> to avoid kvmalloc for ATOMIC case.
> 
> As kvmalloc doesn't warn when used with ATOMIC, kernel will only crash
> once vmalloc fallback occurs, so we may see more crashes in other areas
> in the future.
> 
> Most other callers seem ok but kvm_mmu_topup_memory_cache looks like it
> might be affected by the same breakage, so Cc kvm@.
> 
> Reported-by: Martin Zaharinov <micron10@gmail.com>
> Fixes: a421ef303008 ("mm: allow !GFP_KERNEL allocations for kvmalloc")
> Link: https://lore.kernel.org/linux-mm/Yy3MS2uhSgjF47dy@pc636/T/#t
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: kvm@vger.kernel.org
> Signed-off-by: Florian Westphal <fw@strlen.de>

Please continue in the original email thread until we sort out the most
reasonable solution for this.
-- 
Michal Hocko
SUSE Labs