On 2026-01-13 15:25:26 [-0800], Alexei Starovoitov wrote:

On Tue, Jan 13, 2026 at 3:55 AM Sebastian Andrzej Siewior
[off-list ref] wrote:

quoted

The last one does local_lock_irqsave() on PREEMPT_RT which does a
spin_lock(). That means atomic context is not possible. Where did I make
a wrong turn? Or did this change recently? I do remember that Alexei
reworked parts of the allocator to make the local_lock based trylock
allocation work.

Are you forgetting about local_lock_is_locked() in __slab_alloc() ?

Yeah but this just checks it. Further down the road there is
local_lock_cpu_slab() for the allocation and there is no try-lock on RT.

With sheaves the whole thing will be very different.

Yes.

Sebastian