Messages in this thread | | | From | William Tu <> | Date | Wed, 3 Jul 2019 09:48:09 -0700 | Subject | Re: [PATCH bpf v2] xdp: fix race on generic receive path |
| |
On Wed, Jul 3, 2019 at 6:20 AM Magnus Karlsson <magnus.karlsson@gmail.com> wrote: > > On Wed, Jul 3, 2019 at 2:09 PM Ilya Maximets <i.maximets@samsung.com> wrote: > > > > Unlike driver mode, generic xdp receive could be triggered > > by different threads on different CPU cores at the same time > > leading to the fill and rx queue breakage. For example, this > > could happen while sending packets from two processes to the > > first interface of veth pair while the second part of it is > > open with AF_XDP socket. > > > > Need to take a lock for each generic receive to avoid race. > > I measured the performance degradation of rxdrop on my local machine > and it went from 2.19 to 2.08, so roughly a 5% drop. I think we can > live with this in XDP_SKB mode. If we at some later point in time need > to boost performance in this mode, let us look at it then from a > broader perspective and find the most low hanging fruit. > > Thanks Ilya for this fix. > > Acked-by: Magnus Karlsson <magnus.karlsson@intel.com> > > > Fixes: c497176cb2e4 ("xsk: add Rx receive functions and poll support") > > Signed-off-by: Ilya Maximets <i.maximets@samsung.com> > > ---
Tested on my machine and works ok. Tested-by: William Tu <u9012063@gmail.com>
| |