From: David Lebrun <david.leb...@uclouvain.be> Date: Tue, 10 Jan 2017 14:39:22 +0100
> When CONFIG_PREEMPT=y, CONFIG_IPV6=m and CONFIG_SEG6_HMAC=y, > seg6_hmac_init() is called during the initialization of the ipv6 module. > This causes a subsequent call to smp_processor_id() with preemption > enabled, resulting in the following trace. ... > This patch disable BHs in seg6_hmac_init() and seg6_hmac_exit(). BHs are > re-enabled for the call to crypto_alloc_shash. Indeed, this function might > call try_module_get(), which cannot be called in atomic context. > > Signed-off-by: David Lebrun <david.leb...@uclouvain.be> ... > - char *ring = kzalloc(SEG6_HMAC_RING_SIZE, GFP_KERNEL); > + char *ring = kzalloc(SEG6_HMAC_RING_SIZE, GFP_ATOMIC); I agree with Eric that moving to GFP_ATOMIC in this spots is a major step backwards, and that fixing the per-cpu access during this initializer is a much better direction to go. Thanks.
