Re: linux-next: manual merge of the slab tree with the rcu tree

[Vlastimil Babka (SUSE)]

On 8/4/26 10:26, Alice Ryhl wrote:
> On Fri, Jul 31, 2026 at 03:01:37PM -0700, Nathan Chancellor wrote:
>> On Fri, Jul 31, 2026 at 11:14:42PM +0200, Alice Ryhl wrote:
>> > I like Nathan's fix as well.
>> > 
>> > Not sure if it needs to be done in the merge commit, or if we can
>> > apply it as a patch to char-misc despite the unknown CFG_*?
>> 
>> It would need to be applied at the merge of the two trees because
>> CONFIG_KVFREE_RCU_BATCHED was introduced back in 6.15 in commit
>> c9f8f1242a4c ("slab: don't batch kvfree_rcu() with SLUB_TINY").
> 
> Is that required? Yes, the cfg(CONFIG_KVFREE_RCU_BATCHED) will fail in
> Greg's tree because the CFG doesn't exist and is hence not enabled ...
> but the cfg not applying is what we want.

But it does exist, CONFIG_KVFREE_RCU_BATCHED itself is old. What's new in
slab tree is kvfree_call_rcu() taking a "struct kvfree_rcu_head *" parameter
instead of "struct rcu_head *head" (which is actually "callback_head").

With Gary's suggestion [1] CONFIG_KVFREE_RCU_BATCHED will go out of the
picture completely anyway. What remains is an unconditional change of the
parameter. Can it be "resolved in advance" or only in the merge commit?

https://lore.kernel.org/all/11a8bf75-8fc1-4d32-a350-6bf8156dd8ab@xxxxxxxxxx/

> The main risk I can think of is a warning about an unknown CFG being
> used. Perhaps we could silence the warning?
> 
> Or I mean, we can also just fix it in the merge commit if that works for
> everyone.
> 
> Alice