Re: Interrupt storms with wm(4)
On Fri, Oct 15, 2021 at 08:04:22PM +0200, Joerg Sonnenberger wrote: > Is it the only device using that pin? If so, can you add an event > counter or so to wm_intr to check what percentage of calls to wm_intr > return handled == 0? I did not have the opportunity to investigate the software. I tried moving the network interface to a PCI board that works fine in another machine, the interrupt storms remained. I swaped the machine itself and the interrupt storms disapeared with the same software, hence the culprit was the machine's motherboard. -- Emmanuel Dreyfus m...@netbsd.org
Re: Interrupt storms with wm(4)
On Fri, Oct 15, 2021 at 02:10:32AM +, Emmanuel Dreyfus wrote: > The wm0 interface gets a lot of interrupts. On low usage, CPU > is spent around 10% in interrupts. It can rise to more than 80%. > See below what systat vm says when the machine is quiet. > ioapic0 pin 16 is the wm0 interface. Is it the only device using that pin? If so, can you add an event counter or so to wm_intr to check what percentage of calls to wm_intr return handled == 0? Joerg
Re: Interrupt storms with wm(4)
> The wm0 interface gets a lot of interrupts. On low usage, CPU is > spent around 10% in interrupts. It can rise to more than 80%. !! > Any idea of what could I try to improve the situation? As a possible data point, I have a machine with six wm interfaces in it (ifconfig -l reports wm0 wm1 wm2 wm3 wm4 wm5 bge0 lo0 bridge0 tap1 bridge1 tap2 bridge2 tap3 bridge3 tun0 tun1 tun2 srt0 tap0 tap4 tap5) and it doesn't exhibit anything like that. It is, however, running much older NetBSD (my 5.2 derivative), and it's amd64, not i386. They also are not 82801s: wm0 at pci1 dev 0 function 0: 82580 1000BaseT Ethernet, rev. 1 wm1 at pci1 dev 0 function 1: 82580 1000BaseT Ethernet, rev. 1 wm2 at pci1 dev 0 function 2: 82580 1000BaseT Ethernet, rev. 1 wm3 at pci1 dev 0 function 3: 82580 1000BaseT Ethernet, rev. 1 wm4 at pci3 dev 1 function 0: Intel i82546GB 1000BASE-T Ethernet, rev. 3 wm5 at pci3 dev 1 function 1: Intel i82546GB 1000BASE-T Ethernet, rev. 3 I mention it in case there _is_ something of use in knowing that older NetBSD doesn't necessarily do that. Maybe it's worth trying something older, maybe even as old as 5.2, just to see if it helps with the interrupt storms on your hardware? The 5.2 wm driver does appear to be prepared to recognize 82801I interfaces. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTMLmo...@rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
Interrupt storms with wm(4)
Hello After batteling with a Xen virtulalized IPfilter machine that froze (see http://mail-index.netbsd.org/port-xen/2021/10/07/msg010185.html), I tried to unvirtualize it and moved it to real hardware. It now exhibits interrupts storms that sometimes lead to a freeze. The wm0 interface gets a lot of interrupts. On low usage, CPU is spent around 10% in interrupts. It can rise to more than 80%. See below what systat vm says when the machine is quiet. ioapic0 pin 16 is the wm0 interface. I tried NetBSD 9.2/i386 with options NO_PCI_MSI_MSIX, WM_DISABLE_MSI=1 and WM_DISABLE_MSIX=1 without improvement. I also tried NetBSD-current/i386 GENERIC, same behavior. Any idea of what could I try to improve the situation? Here is what dmesg said about wm0: wm0 at pci0 dev 25 function 0: 82801I (AMT) LAN Controller (rev. 0x02) wm0: interrupting at ioapic0 pin 16 wm0: PCI-Express bus wm0: 2048 words FLASH, version 1.1.0 wm0: Ethernet address 00:30:48:de:cc:4c wm0: 0x2a4480 And here is an idle vmstat: 3 usersLoad 0.21 0.18 0.09 Fri Oct 15 04:02:42 Proc:r d sCsw Traps SysCal Intr Soft Fault PAGING SWAPPING 9 20 211 1315 1399in out in out ops 0.0% Sy 0.0% Us 0.0% Ni 12.0% In 88.0% Idpages ||||||||||| %%forks fkppw Anon 109976 3% zero 2744312 1415 Interrupts fksvm Exec30116% wired 11104 TLB shootdownpwait File75312 2% inact 100 cpu0 timer relck Meta 9293% bufs 646 ioapic0 pin 9rlkok (kB)real swaponly free 1147 ioapic0 pin 16 noram Active 204300 2797228 msi1 vec 0 ndcpy Namei Sys-cache Proc-cache 168 ioapic0 pin 4fltcp Calls hits% hits %zfod 66 100 cow 512 fmin Disks: wd0 wd1 cd0 raid0 682 ftarg seeksitarg xfersflnan bytespdfre -- Emmanuel Dreyfus m...@netbsd.org