From: Jakub Kicinski <[email protected]>
Date: Wed, 17 Apr 2019 10:51:19 -0700
> Unlike atomic_add(), refcount_add() does not deal well
> with a negative argument. TLS fallback code reallocates
> the skb and is very likely to shrink the truesize, leading to:
>
> [ 189.513254] WARNING: CPU: 5 PID: 0 at lib/refcount.c:81
> refcount_add_not_zero_checked+0x15c/0x180
> Call Trace:
> refcount_add_checked+0x6/0x40
> tls_enc_skb+0xb93/0x13e0 [tls]
>
> Once wmem_allocated count saturates the application can no longer
> send data on the socket. This is similar to Eric's fixes for GSO,
> TCP:
> commit 7ec318feeed1 ("tcp: gso: avoid refcount_t warning from
> tcp_gso_segment()")
> and UDP:
> commit 575b65bc5bff ("udp: avoid refcount_t saturation in
> __udp_gso_segment()").
>
> Unlike the GSO case, for TLS fallback it's likely that the skb has
> shrunk, so the "likely" annotation is the other way around (likely
> branch being "sub").
>
> Fixes: e8f69799810c ("net/tls: Add generic NIC offload infrastructure")
> Signed-off-by: Jakub Kicinski <[email protected]>
> Reviewed-by: John Hurley <[email protected]>
Applied and queued up for -stable, thanks Jakub.
> Should we add a helper for this in -next? Because with
> CONFIG_REFCOUNT_FULL=n we don't need the branch, we can just
> do an atomic_add() directly..
I think the value of that isn't so high. Are we counting branch miss
penalities in code pathes where we are copying SKB data or doing SW
crypto? :-)
