Bug 9007

Summary: r8169: ifconfig eth3 up reboots instantly unless acpi=noirq
Product: ACPI Reporter: Xavier Bestel (xavier.bestel)
Component: Config-InterruptsAssignee: Zhang Rui (rui.zhang)
Status: REJECTED INSUFFICIENT_DATA    
Severity: high CC: acpi-bugzilla, romieu, StormByte
Priority: P1    
Hardware: All   
OS: Linux   
Kernel Version: 2.6.22 Subsystem:
Regression: --- Bisected commit-id:
Attachments: dmesg
lshw
acpidump with acpi=noirq
lspci -v with acpi=noirq
lspci -xxx with acpi=noirq

Description Xavier Bestel 2007-09-11 06:49:11 UTC
Distribution: Debian/stable, with stable and unstable kernel
Hardware Environment: DELL Poweredge + Realtek 8169 (see lshw.txt)

Problem Description: ifconfig eth3 up reboots instantly

Steps to reproduce:
Poweron the machine, look for which interface has r8169 (in my case it's eth3), and type the above command. Watch the machine reboot.

It should be boted that booting with "acpi=noirq" just stops the problem from appearing.
Comment 1 Xavier Bestel 2007-09-11 06:51:41 UTC
Created attachment 12792 [details]
dmesg
Comment 2 Xavier Bestel 2007-09-11 06:52:16 UTC
Created attachment 12793 [details]
lshw
Comment 3 Xavier Bestel 2007-09-11 06:54:26 UTC
Note: the dmesg is with "acpi=noirq". If requested I can try to provide a dmesg from a boot without option.
Comment 4 Arjan van de Ven 2007-09-11 07:43:50 UTC
does this also happen with an untainted kernel?
Comment 5 Xavier Bestel 2007-09-11 08:01:40 UTC
Yes, it does. After first seeing the problem with 2.6.18, I upgraded to 2.6.22 which doesn't contain madwifi, and the problem was still there. I only found about "acpi=noirq" after a few reboots. The dmesg.txt attached shows raid resyncing (because of a reboot) and later madwifi starting, because I had to download and compile it first.
Comment 6 Len Brown 2007-09-12 18:39:13 UTC
you mentioned this happened in 2.6.18 and 2.6.22
Did this not happen in any previous linux release?

re: reboot
is this an immediate hardware reset, or does the machine shut-down?

please paste the /proc/interrupts with and without "acpi=noirq".
It may be helpful to try this with "acpi=off" also.
Comment 7 ykzhao 2007-10-08 06:02:52 UTC
Hi, Xavier
Will you please upload the following files with boot option of acpi=noirq ?
a. acpidump
b. lspci -v
c. lspci -xxx
Thanks
Comment 8 Xavier Bestel 2007-10-08 06:17:45 UTC
Created attachment 13077 [details]
acpidump with acpi=noirq
Comment 9 Xavier Bestel 2007-10-08 06:18:19 UTC
Created attachment 13078 [details]
lspci -v with acpi=noirq
Comment 10 Xavier Bestel 2007-10-08 06:18:53 UTC
Created attachment 13079 [details]
lspci -xxx with acpi=noirq
Comment 11 Xavier Bestel 2007-10-08 06:20:46 UTC
in comment #9: s/lspci -c/lspci -v/
Comment 12 Xavier Bestel 2007-10-08 06:23:16 UTC
FWIW, here is /proc/interrupts with acpi=noirq:

           CPU0       CPU1
  0:         92          0   IO-APIC-edge      timer
  1:          4          4   IO-APIC-edge      i8042
  2:          0          0    XT-PIC-XT        cascade
  3:    5680487    5676821   IO-APIC-edge      serial
  6:          0          3   IO-APIC-edge      floppy
  7:          0          0   IO-APIC-edge      parport0
  8:          1          0   IO-APIC-edge      rtc
  9:          1          1   IO-APIC-edge      acpi
 12:          3          1   IO-APIC-edge      i8042
 14:    7916419    7949148   IO-APIC-edge      ide0
 15:    7897606    7924721   IO-APIC-edge      ide1
 17:   19254391   19256739   IO-APIC-fasteoi   uhci_hcd:usb3, eth0
 18:   30724494   30724673   IO-APIC-fasteoi   ehci_hcd:usb4, eth1
 19:    2642744    2641963   IO-APIC-fasteoi   aic7xxx, uhci_hcd:usb1
 20:  194199154  194118901   IO-APIC-fasteoi   wifi0
 21:   10441271   10477969   IO-APIC-fasteoi   ide2, ide3
 22:          0          0   IO-APIC-fasteoi   uhci_hcd:usb2
NMI:          0          0
LOC:  330979908  288123256
ERR:          0
MIS:          0


I'm sorry I still can't play with the machine right now to try with other acpi= settings. I'll try when I can.
Comment 13 Zhang Rui 2007-12-06 19:17:57 UTC
Hi,
We can't debug further if no info available.
Please attach the dmesg and /proc/interrupts without any kernel parameters..
Or esle we'll close this bug and mark it as INSUFFICIENT_DATA.
Comment 14 Len Brown 2008-01-01 11:44:59 UTC
please re-open when the system is available for debugging.
Comment 15 David 2016-10-17 06:12:11 UTC
I am affected by something like this, and over the internet there are plenty of reports of this r8169 freezing the machine when heavy network load.

I have a little server in my house affected by this issue and available for testing. For the moment, booting with acpi=noirq causes a very slow boot process, but seems to avoid this issue.

Not sure if the bug is the same, because I am not having a reboot, I am having just a freeze and computer unresponsive to even keyboard actions. If I should open a new bug please tell me so, otherwise, ask for required testings and I will provide.

Thanks.
Comment 16 David 2016-10-17 06:14:51 UTC
Btw, tested in 4.7.7+grsecurity patches (gentoo hardened-sources-4.7.7)