kernel-packages team mailing list archive

[Launchpad Bug Tracker <1046326@xxxxxxxxxxxxxxxxxx>]

[Expired for linux (Ubuntu) because there has been no activity for 60
days.]

** Changed in: linux (Ubuntu)
       Status: Incomplete => Expired

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1046326

Title:
  Ubuntu Slow, crash, sluggish from version 10.04 upwards

Status in “linux” package in Ubuntu:
  Expired

Bug description:
  Ubuntu will at random points suddenly become very slow and sluggish.
  graphics run far behind, keyboard strokes take multiple seconds to
  show, graphics are updated in super slowmo mode, or if you're unlucky;
  the system dies completely.

  This is related to a problem with the interrupt handling. On my own
  machine (MSI i7, nvidia card, 16 Gb, 5 sata HDD's) I have had this
  problem from 10.04 (when I bough the hware) up to 12.04. I have since
  upgraded up to various kernels up to 3.5.3 with varying results. Some
  of the suggestions in posts make the periods between crashes shorter,
  some days it works fine, some days it happens every five minutes, but
  it NEVER goes away.

  If you consider this is interrupt related, that makes perfect sense.
  Every single 'crash' is marked by an unanswered interrupt. See below
  syslog:

  Sep  5 15:13:14 Server kernel: [112195.631364] irq 16: nobody cared (try booting with the "irqpoll" option)
  Sep  5 15:13:14 Server kernel: [112195.631368] Pid: 17901, comm: firefox Tainted: P           O 3.5.3-030503-generic #201208252335
  Sep  5 15:13:14 Server kernel: [112195.631369] Call Trace:
  Sep  5 15:13:14 Server kernel: [112195.631370]  <IRQ>  [<ffffffff810e5a5d>] __report_bad_irq+0x3d/0xe0
  Sep  5 15:13:14 Server kernel: [112195.631378]  [<ffffffff810e5ce5>] note_interrupt+0x135/0x190
  Sep  5 15:13:14 Server kernel: [112195.631380]  [<ffffffff810e3559>] handle_irq_event_percpu+0xa9/0x210
  Sep  5 15:13:14 Server kernel: [112195.631382]  [<ffffffff810e370e>] handle_irq_event+0x4e/0x80
  Sep  5 15:13:14 Server kernel: [112195.631384]  [<ffffffff810e6874>] handle_fasteoi_irq+0x64/0x120
  Sep  5 15:13:14 Server kernel: [112195.631388]  [<ffffffff81016632>] handle_irq+0x22/0x40
  Sep  5 15:13:14 Server kernel: [112195.631391]  [<ffffffff816a49ea>] do_IRQ+0x5a/0xe0
  Sep  5 15:13:14 Server kernel: [112195.631394]  [<ffffffff8169a86a>] common_interrupt+0x6a/0x6a
  Sep  5 15:13:14 Server kernel: [112195.631395]  <EOI>  [<ffffffff816a2c2d>] ? system_call_fastpath+0x1a/0x1f
  Sep  5 15:13:14 Server kernel: [112195.631398] handlers:
  Sep  5 15:13:14 Server kernel: [112195.631401] [<ffffffff814bc670>] usb_hcd_irq
  Sep  5 15:13:14 Server kernel: [112195.631462] [<ffffffffa0d65cb0>] nv_kern_isr [nvidia]
  Sep  5 15:13:14 Server kernel: [112195.631470] [<ffffffffa00640f0>] rhine_interrupt [via_rhine]
  Sep  5 15:13:14 Server kernel: [112195.631471] Disabling IRQ #16

  The problem is massively compounded by the fact that irqpoll and
  irqfixup options NO LONGER WORK AT ALL (
  https://bugs.launchpad.net/ubuntu/+source/linux/+bug/855199 ).

  No matter if you switch to Nouveau, diff kernel, or other drivers, the
  problem remains is varying severity. Again; logical if there is a
  problem with the kernel interrupt handling itself.

  So here's the kicker. It is easy to resolve! By forcing a restart of
  the interrupt handlers the system INSTANTLY comes back to 100%
  functional life:

  root@Server:~# service network-manager restart

  My theory (for what it's worth; I have programmed assembly for 30
  years) is that there is a multi-threading fault somewhere, which
  causes the kernel to miss an interrupt, or to miss allocating it.

  The one thing they have in common though: Shared interrupts (usually
  16) and LOAD on that interrupt (USB (KB+MS), VGA, ETH) on high speed
  machines.

  I have created a keyboard shortcut that does the service restart and
  this works perfectly 100% of the time. If any dev needs more info,
  contact me.

  Just to prove a point: Same machine on win7 works perfectly.

  Michael

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1046326/+subscriptions