Bug 7518
| Summary: | Irq storm on Intel-based chipsets | ||
|---|---|---|---|
| Product: | Drivers | Reporter: | Vasily Averin (vasily.averin) |
| Component: | PCI | Assignee: | Tejun Heo (htejun) |
| Status: | CLOSED CODE_FIX | ||
| Severity: | normal | CC: | akpm, alan |
| Priority: | P2 | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.19-rc5 | Subsystem: | |
| Regression: | --- | Bisected commit-id: | |
| Attachments: |
failed bootlog on 2.6.19-rc5 kernel
lspci -vvvxxx output /proc/interrupts output failed bootlog with noapic acpi=off nolapic options in kernel commandline lspci -vvvxxx output for second boot /proc/interrupts output for second boot bootlog when IDE was switched back to legacy mode dmesg from kernel without IDE support 2.6.19-rc6 bootlog: USB support enabled /proc/interrupts output before loading uhci-usb module /proc/interrupts output after loading uhci-usb module lspci -vvvxxx output, USB support enabled |
||
Created attachment 9494 [details]
failed bootlog on 2.6.19-rc5 kernel
failed bootlog on 2.6.19-rc5 kernel:
booted with init=/bin/bash option in kernel commandline, then
# modprobe ide-cd
Created attachment 9495 [details]
lspci -vvvxxx output
Created attachment 9496 [details]
/proc/interrupts output
Created attachment 9497 [details]
failed bootlog with noapic acpi=off nolapic options in kernel commandline
Created attachment 9498 [details]
lspci -vvvxxx output for second boot
On Tue, 14 Nov 2006 00:49:30 -0800 bugme-daemon@bugzilla.kernel.org wrote: > Kernel Version: 2.6.19-rc5 > Most recent kernel where this bug did not occur: 2.6.19-rc5 huh? Created attachment 9499 [details]
/proc/interrupts output for second boot
Created attachment 9500 [details]
bootlog when IDE was switched back to legacy mode
I've reprduced this issue on Intel E7520/ICH5 node too, in this case IDE controller did shared the interrupt together with following devices: 00:04.0 0604: 8086:3597 (rev 0a) 00:05.0 0604: 8086:3598 (rev 0a) 00:04.0 PCI bridge: Intel Corporation E7525/E7520 PCI Express Port B (rev 0a) 00:05.0 PCI bridge: Intel Corporation E7520 PCI Express Port B1 (rev 0a) As far as I know this issue is reproduced on all old kernels too, at least on latest 2.6.9-based kernels from RHEL4. On Tue, 14 Nov 2006 00:49:30 -0800 bugme-daemon@bugzilla.kernel.org wrote: > Kernel Version: 2.6.19-rc5 > Most recent kernel where this bug did not occur: 2.6.19-rc5 huh? Sorry, it was mistake, this issue is present on the 2.6.19-rc5 kernels too and I do not know any kenrels where this issue is fixed. Could you please change this attribute? Please see some additional details in the following thread: http://thread.gmane.org/gmane.linux.kernel/460071/focus=13693 Problem tracked down. Please read the following message. http://thread.gmane.org/gmane.linux.kernel/460071/focus=466885 Vasily, can you verify that using libata solves the problem? Created attachment 9527 [details]
dmesg from kernel without IDE support
I've disabled IDE support in the config and recompiled the kernel.
It seems you are right, problem go away, new kernel was booted without any
problems. Please see dmesg in the attachment.
Workaround "hda=noprobe" helps: IDE do not detect CDROM and all other devices on the node works well. However when I enable USB support on the node, one of the USB controllers begins to use the same IRQ. And IRQ storm occurs again when I load uhci_hcd driver on the node. It looks very strange for me: we do not accessing any IDE devices in this case. I would note that I've reproduced this issue if I detach the CDROM manually. Created attachment 9543 [details]
2.6.19-rc6 bootlog: USB support enabled
2.6.19-rc6 kernel with IDE support and "hda=noprobe" option in kernel
commandline.
loaded with init=/bin/bash
# cat /proc/interrupts >interrupts.rc6.usb.before
# /sbin/modprobe uhci-hcd
# cat /proc/interrupts >interrupts.rc6.usb.after
# lspci -vvvxxx >lspci.rc6.usb
Created attachment 9544 [details]
/proc/interrupts output before loading uhci-usb module
Created attachment 9545 [details]
/proc/interrupts output after loading uhci-usb module
Created attachment 9546 [details]
lspci -vvvxxx output, USB support enabled
Fixed in upstream kernel for both libata and ide. Closing. Feel free to re-open if problem persists. |
Most recent kernel where this bug did not occur: 2.6.19-rc5 Hardware Environment: motherboards based on Intel 915G/ICH6 and Intel E7520/ICH5 chipsets Problem Description: IRQ storm occured when IDE controller mode is in native mode Steps to reproduce: 1) install the possible distributions on the node when IDE controller is in legacy mode and cd/dvd is attached into IDE. 2) install the kernel and check that it works well when IDE is in legacy mode 3) reboot, BIOS, switch ide settings to set IDE controler into native mode. For example on ASUSTeK P5GD1-VM motherboard, AMI BIOS Version 1010.005 released 03/29/2005 : BIOS -> Main -> IDE Configuration -> Onboard IDE Operate Mode: Enhanced mode Enhanced mode support on: S-ATA+P-ATA (or to P-ATA) and save the settings 4) booting the kernel, loads the ide-cd module ==> irq assigned for IDE controller (and for one of PCI Express ports) is disabled due irq storm. Comments: noapic, acpi=off, nolapic, irqpoll, pci=routeirq and other kernel commandline options does not help. I've checked Intel spec updates but don't found the similair known issues. I assume that this issue is PCI Express port-related.