On Mon, Jun 02, 2025 at 04:15:56PM +0200, Michal Koutný [off-list ref] wrote:

I'd say this might be relevant (although I don't see the possibly
incorrect error handlnig path) but it doesn't mean this commit fixes it,
it'd rather require the reproducer to adjust the N on this path.

Hm, possibly syzbot caught up here [1]:

-mkdir(&(0x7f0000000000)='./cgroup/file0\x00', 0xd0939199c36b4d28) (fail_nth: 8)
+mkdirat$cgroup_root(0xffffffffffffff9c, &(0x7f00000005c0)='./cgroup.net/syz0\x00', 0x1ff) (fail_nth: 23)

So there's something fishy in the error handling.

HTH,
Michal

[1] https://lore.kernel.org/lkml/68403875.a00a0220.d4325.000a.GAE@google.com/ (local)