search for: virtnet_poll

On Sun, Aug 02, 2020 at 01:56:33PM +0800, Mao Wenan wrote: > The loop may exist if vq->broken is true, > virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split > will return NULL, so virtnet_poll will reschedule napi to > receive packet, it will lead cpu usage(si) to 100%. > > call trace as below: > virtnet_poll > virtnet_receive > virtqueue_get_buf_ctx > virtqueue_get_buf_ctx_packed > virtqueue_get_buf_ctx_split > virtqueue_napi_complete > virtque...

Just noticed the subject is wrong: this is no longer a virtio_net patch. On Sun, Aug 02, 2020 at 01:56:33PM +0800, Mao Wenan wrote: > The loop may exist if vq->broken is true, > virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split > will return NULL, so virtnet_poll will reschedule napi to > receive packet, it will lead cpu usage(si) to 100%. > > call trace as below: > virtnet_poll > virtnet_receive > virtqueue_get_buf_ctx > virtqueue_get_buf_ctx_packed > virtqueue_get_buf_ctx_split > virtqueue_napi_complete > virtque...

virtnet_poll is called from soft IRQ and it tries to allocate GFP_ATOMIC memory (through try_fill_recv). This allocation can fail and we are falling back to schedule_delayed_work in that case. Let's add __GFP_NOWARN to the allocation flags to get rid of the allocator complains for failed allocations: [227...

virtnet_poll is called from soft IRQ and it tries to allocate GFP_ATOMIC memory (through try_fill_recv). This allocation can fail and we are falling back to schedule_delayed_work in that case. Let's add __GFP_NOWARN to the allocation flags to get rid of the allocator complains for failed allocations: [227...

...s. > Hi, Our verification team encountered the following degradation, introduced by this series. Please check KASAN bug report below. We tested before and after the series, but didn't bisect within it. We can do if necessary. Thanks, Tariq [ 46.166544] BUG: KASAN: use-after-free in virtnet_poll+0xd9c/0xfd1 [virtio_net] [ 46.166593] Read of size 8 at addr ffff8803400da608 by task ip/1013 [ 46.166603] CPU: 3 PID: 1013 Comm: ip Not tainted 4.18.0-rc6-for-upstream-dbg-2018-07-26_19-45-52-64 #1 [ 46.166606] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 0...

...s. > Hi, Our verification team encountered the following degradation, introduced by this series. Please check KASAN bug report below. We tested before and after the series, but didn't bisect within it. We can do if necessary. Thanks, Tariq [ 46.166544] BUG: KASAN: use-after-free in virtnet_poll+0xd9c/0xfd1 [virtio_net] [ 46.166593] Read of size 8 at addr ffff8803400da608 by task ip/1013 [ 46.166603] CPU: 3 PID: 1013 Comm: ip Not tainted 4.18.0-rc6-for-upstream-dbg-2018-07-26_19-45-52-64 #1 [ 46.166606] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 0...

...e and after the series, but didn't bisect within it. We > can do if necessary. > > Thanks, > Tariq > I see this might already be fixed, here: https://marc.info/?l=linux-netdev&m=153335713407532&w=2 Verifying... > > [?? 46.166544] BUG: KASAN: use-after-free in virtnet_poll+0xd9c/0xfd1 > [virtio_net] > [?? 46.166593] Read of size 8 at addr ffff8803400da608 by task ip/1013 > > [?? 46.166603] CPU: 3 PID: 1013 Comm: ip Not tainted > 4.18.0-rc6-for-upstream-dbg-2018-07-26_19-45-52-64 #1 > [?? 46.166606] Hardware name: QEMU Standard PC (i440FX + PIIX,...

...type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/> > </interface> > > Panic log: > > <1>BUG: unable to handle kernel NULL pointer dereference at 0000000000000010 > <1>IP: [<ffffffffa0079469>] virtnet_poll+0x4f9/0x910 [virtio_net] > <4>PGD 1aa2f4067 PUD 1aa2f5067 PMD 0 > <4>Oops: 0000 [#1] SMP > <4>last sysfs file: /sys/devices/pci0000:00/0000:00:03.0/virtio0/net/eth9/ifindex > <4>CPU 0 > <4>Modules linked in: 8021q garp stp llc ipt_REJECT nf_conntrack_...

...type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/> > </interface> > > Panic log: > > <1>BUG: unable to handle kernel NULL pointer dereference at 0000000000000010 > <1>IP: [<ffffffffa0079469>] virtnet_poll+0x4f9/0x910 [virtio_net] > <4>PGD 1aa2f4067 PUD 1aa2f5067 PMD 0 > <4>Oops: 0000 [#1] SMP > <4>last sysfs file: /sys/devices/pci0000:00/0000:00:03.0/virtio0/net/eth9/ifindex > <4>CPU 0 > <4>Modules linked in: 8021q garp stp llc ipt_REJECT nf_conntrack_...

...b mlx4 cards. With both busy_poll and busy_read are set to 50 in guest, 1 byte netperf tcp_rr shows 127% improvement: transaction rate was increased from 8353.33 to 18966.87. Changes from V2: - Avoid introducing new states and spinlocks by reusuing the NAPI state - Fix the budget calculation in virtnet_poll() - Drop patch 1/3 from V2 since it was useless Changes from V1: - split the patch info smaller ones - add more details about test setup/configuration Jason Wang (2): virtio-net: introduce virtnet_receive() virtio-net: rx busy polling support drivers/net/virtio_net.c | 67 ++++++++++++++++++...

...b mlx4 cards. With both busy_poll and busy_read are set to 50 in guest, 1 byte netperf tcp_rr shows 127% improvement: transaction rate was increased from 8353.33 to 18966.87. Changes from V2: - Avoid introducing new states and spinlocks by reusuing the NAPI state - Fix the budget calculation in virtnet_poll() - Drop patch 1/3 from V2 since it was useless Changes from V1: - split the patch info smaller ones - add more details about test setup/configuration Jason Wang (2): virtio-net: introduce virtnet_receive() virtio-net: rx busy polling support drivers/net/virtio_net.c | 67 ++++++++++++++++++...

I had to add an explicit tag to suppress compiler warning: gcc isn't smart enough to notice that len is always initialized since function is called with size > 0. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> Reviewed-by: Cornelia Huck <cornelia.huck at de.ibm.com> --- drivers/vhost/net.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff

I had to add an explicit tag to suppress compiler warning: gcc isn't smart enough to notice that len is always initialized since function is called with size > 0. Signed-off-by: Michael S. Tsirkin <mst at redhat.com> Reviewed-by: Cornelia Huck <cornelia.huck at de.ibm.com> --- drivers/vhost/net.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff

On Mon, Jul 23, 2018 at 11:36:03PM +0900, Toshiaki Makita wrote: > From: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> > > Add some ethtool stat items useful for performance analysis. > > Signed-off-by: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> Series: Acked-by: Michael S. Tsirkin <mst at redhat.com> Patch 1 seems appropriate for stable, even

On Mon, Jul 23, 2018 at 11:36:03PM +0900, Toshiaki Makita wrote: > From: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> > > Add some ethtool stat items useful for performance analysis. > > Signed-off-by: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> Series: Acked-by: Michael S. Tsirkin <mst at redhat.com> Patch 1 seems appropriate for stable, even

...tic void skb_recv_done(struct virtque } } +static void virtnet_napi_enable(struct virtnet_info *vi) +{ + napi_enable(&vi->napi); + + /* If all buffers were filled by other side before we napi_enabled, we + * won't get another interrupt, so process any outstanding packets + * now. virtnet_poll wants re-enable the queue, so we disable here. + * We synchronize against interrupts via NAPI_STATE_SCHED */ + if (napi_schedule_prep(&vi->napi)) { + virtqueue_disable_cb(vi->rvq); + __napi_schedule(&vi->napi); + } +} + static void refill_work(struct work_struct *work) { str...

...tic void skb_recv_done(struct virtque } } +static void virtnet_napi_enable(struct virtnet_info *vi) +{ + napi_enable(&vi->napi); + + /* If all buffers were filled by other side before we napi_enabled, we + * won't get another interrupt, so process any outstanding packets + * now. virtnet_poll wants re-enable the queue, so we disable here. + * We synchronize against interrupts via NAPI_STATE_SCHED */ + if (napi_schedule_prep(&vi->napi)) { + virtqueue_disable_cb(vi->rvq); + __napi_schedule(&vi->napi); + } +} + static void refill_work(struct work_struct *work) { str...

On 2020/8/2 ??3:44, Mao Wenan wrote: > The loop may exist if vq->broken is true, > virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split > will return NULL, so virtnet_poll will reschedule napi to > receive packet, it will lead cpu usage(si) to 100%. > > call trace as below: > virtnet_poll > virtnet_receive > virtqueue_get_buf_ctx > virtqueue_get_buf_ctx_packed > virtqueue_get_buf_ctx_split > virtqueue_napi_complete > virtqueu...

...000000000045df2e: e31020300004 lg %r1,48(%r2) 000000000045df34: a7480000 lhi %r4,0 000000000045df38: 96011001 oi 1(%r1),1 000000000045df3c: a7f4ffef brc 15,45df1a Call Trace: ([<000000000045c016>] virtnet_poll+0x96/0x42c) [<000000000048cda2>] net_rx_action+0xca/0x150 [<0000000000137f7a>] __do_softirq+0x9e/0x130 [<00000000001105d6>] do_softirq+0xae/0xb4 [<0000000000138182>] irq_exit+0x96/0x9c [<000000000010d710>] do_extint+0xcc/0xf8 [<00000000001135d0>] ext_no_vti...

...000000000045df2e: e31020300004 lg %r1,48(%r2) 000000000045df34: a7480000 lhi %r4,0 000000000045df38: 96011001 oi 1(%r1),1 000000000045df3c: a7f4ffef brc 15,45df1a Call Trace: ([<000000000045c016>] virtnet_poll+0x96/0x42c) [<000000000048cda2>] net_rx_action+0xca/0x150 [<0000000000137f7a>] __do_softirq+0x9e/0x130 [<00000000001105d6>] do_softirq+0xae/0xb4 [<0000000000138182>] irq_exit+0x96/0x9c [<000000000010d710>] do_extint+0xcc/0xf8 [<00000000001135d0>] ext_no_vti...