svn commit: r347526 - head/sys/net
Mark Johnston
markj at freebsd.org
Mon May 13 19:38:02 UTC 2019
On Mon, May 13, 2019 at 01:45:29PM +0000, Andrey V. Elsukov wrote:
> Author: ae
> Date: Mon May 13 13:45:28 2019
> New Revision: 347526
> URL: https://svnweb.freebsd.org/changeset/base/347526
>
> Log:
> Rework locking in BPF code to remove rwlock from fast path.
>
> On high packets rate the contention on rwlock in bpf_*tap*() functions
> can lead to packets dropping. To avoid this, migrate this code to use
> epoch(9) KPI and ConcurrencyKit's lists.
>
> * all lists changed to use CK_LIST;
> * reference counting added to bpf_if and bpf_d;
> * now bpf_if references ifnet and releases this reference on destroy;
> * each bpf_d descriptor references bpf_if when it is attached;
> * new struct bpf_program_buffer introduced to keep BPF filter programs;
> * bpf_program_buffer, bpf_d and bpf_if structures are freed by
> epoch_call();
> * bpf_freelist and ifnet_departure event are no longer needed, thus
> both are removed;
>
> Reviewed by: melifaro
> Sponsored by: Yandex LLC
> Differential Revision: https://reviews.freebsd.org/D20224
This seems to introduce a possible recursion on the bpf cdev mutex. Now
bpfwrite() holds the mutex over the if_output call, which may enter
bpf_mtap() via bridge_transmit(), and bpf_mtap() may attempt to
reacquire the mutex.
More information about the svn-src-all
mailing list