Date: Tue, 21 Jan 2025 16:02:48 +0000 From: bugzilla-noreply@freebsd.org To: net@FreeBSD.org Subject: [Bug 284057] vmxnet3/iflib: crash in vmxnet3_isc_txd_credits_update Message-ID: <bug-284057-7501-eyuDAEv8A9@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-284057-7501@https.bugs.freebsd.org/bugzilla/> References: <bug-284057-7501@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=284057 --- Comment #12 from Andriy Gapon <avg@FreeBSD.org> --- Another thing. There is a curious comment in vmxnet3_msix_intr_assign which I cannot really understand: /* * Don't provide the corresponding rxq irq for reference - * we want the transmit task to be attached to a task queue * that is different from the one used by the corresponding * rxq irq. That is because the TX doorbell writes are very * expensive as virtualized MMIO operations, so we want to * be able to defer them to another core when possible so * that they don't steal receive processing cycles during * stack turnarounds like TCP ACK generation. The other * piece to this approach is enabling the iflib abdicate * option (currently via an interface-specific * tunable/sysctl). */ Just out of curiosity I set dev.vmx.0.iflib.tx_abdicate=1. Since doing that I do not see any concurrent calls to vmxnet3_isc_txd_credits_update for the same queue. That is, each queue is always handled on its own CPU. Used this for checking: dtrace -n 'vmxnet3_isc_txd_credits_update:entry /args[1] != curthread->td_oncpu/ { @[stack()] = count(); }' -- You are receiving this mail because: You are the assignee for the bug.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-284057-7501-eyuDAEv8A9>