Willem de Bruijn
2021-Feb-23 19:13 UTC
[PATCH net v2 0/2] virtio-net: suppress bad irq warning for tx napi
On Tue, Feb 23, 2021 at 12:37 PM Wei Wang <weiwan at google.com> wrote:> > On Tue, Feb 23, 2021 at 6:26 AM Michael S. Tsirkin <mst at redhat.com> wrote: > > > > On Fri, Feb 19, 2021 at 05:44:34PM -0800, Wei Wang wrote: > > > With the implementation of napi-tx in virtio driver, we clean tx > > > descriptors from rx napi handler, for the purpose of reducing tx > > > complete interrupts. But this could introduce a race where tx complete > > > interrupt has been raised, but the handler found there is no work to do > > > because we have done the work in the previous rx interrupt handler. > > > This could lead to the following warning msg: > > > [ 3588.010778] irq 38: nobody cared (try booting with the > > > "irqpoll" option) > > > [ 3588.017938] CPU: 4 PID: 0 Comm: swapper/4 Not tainted > > > 5.3.0-19-generic #20~18.04.2-Ubuntu > > > [ 3588.017940] Call Trace: > > > [ 3588.017942] <IRQ> > > > [ 3588.017951] dump_stack+0x63/0x85 > > > [ 3588.017953] __report_bad_irq+0x35/0xc0 > > > [ 3588.017955] note_interrupt+0x24b/0x2a0 > > > [ 3588.017956] handle_irq_event_percpu+0x54/0x80 > > > [ 3588.017957] handle_irq_event+0x3b/0x60 > > > [ 3588.017958] handle_edge_irq+0x83/0x1a0 > > > [ 3588.017961] handle_irq+0x20/0x30 > > > [ 3588.017964] do_IRQ+0x50/0xe0 > > > [ 3588.017966] common_interrupt+0xf/0xf > > > [ 3588.017966] </IRQ> > > > [ 3588.017989] handlers: > > > [ 3588.020374] [<000000001b9f1da8>] vring_interrupt > > > [ 3588.025099] Disabling IRQ #38 > > > > > > This patch series contains 2 patches. The first one adds a new param to > > > struct vring_virtqueue to control if we want to suppress the bad irq > > > warning. And the second patch in virtio-net sets it for tx virtqueues if > > > napi-tx is enabled. > > > > I'm going to be busy until March, I think there should be a better > > way to fix this though. Will think about it and respond in about a week. > > > OK. Thanks.Yes, thanks for helping to think about a solution. The warning went unreported for years. I'm a bit hesitant to make actual datapath changes to suppress it, if those may actually have a higher risk of regressions for some workloads. Unless they actually might show a clear progression. Which may very well be possible given the high spurious interrupt rate. But the odd thing is that by virtue of the interrupt getting masked once the warning hits, it may actually be difficult to improve on the efficiency today. As you pointed out, just probabilistically throttling how often to steal work from the rx interrupt handler would be another low risk approach to reduce the incidence rate. Anyway, definitely no rush. This went unreported for a long time.