k2.6.12-8 irqpoll and old IDE ALI15X3 chipset (IBM Desktop machine)

Thu Sep 8 03:13:08 UTC 2005

Hi,

skoal a écrit :
> 
>>What does irqpoll actually do?, and if it's an integral part is
> 
> 
>>preventing irq conflicts, why is it not active by default? Does it
> 
> take
> 
> 
>>up heaps of resources, or cause other problems that I should be aware
> 
> of?
> 
> 
> 
> First, I don't know what kernel version you are using, but that kernel
> boot time option was just a very recent patch to 2.6.12 (and not part
> of a release either, I believe).  I saw no mention of that option in
> the latest 2.6.12 release I'm currently using.  It may be part of the
> 2.6.13 release, but I sure couldn't find any documentation of 'irqpoll'
> in any version prior to 2.6.12.  So, assuming you are using the stock
> 5.04 kernel (2.6.10), I really don't know how this 'irqpoll' option is
> helping you.  It must be a well hidden secret in the code as well...
> 
> 
> 
> With that said, the irqpoll option basically toggles a handler in the
> kernel irq.c (I think) source which basically says, "hey! I just
> received an IRQ request out here on the bus, and I don't know who it's
> from and what I should do with it."  So, the kernel will poll all it's
> IRQ handling routines to find someone to service it.
> 
> 
> 
> Well, a little digging through gdm and gnome-session source, I think
> what's happening is that when you logoff (and since you are doing so
> from a remote SSH, right? or as a local admin with remote access to
> GDM), I think one of your network cards is throwing up an IRQ.  I don't
> know whether it's your ADSL NIC or your intranet NIC doing it though.
> 
> 
> 
> Ok, so what does that mean to you? Well, since you did _not_ experience
> these same problems with warty and only hoary, several things (quite
> possibly) could be causing this.  My first inclination is kernel IRQ
> handler changes.  Next, would be changes in HAL.  HAL is a beast, and I
> really don't understand all the places it sticks it's dirty little
> fingers into.  Anyways, it could even be a more recent NIC driver
> change.  Pure speculation...my specialty!
> 
> 
> 
> Apparently, that 'irqpoll' kernel option fixed it for you, but if it
> persists (or returns at random), the easiest sol'n might just be a $2
> 3Com NIC off eBay.  One of your NIC's could be causing it, and maybe it
> was just ignored somehow in prior (2.6.8?) wary kernel handlers.  I
> dunno.  The only other thing I could think of is if you recently added
> some other piece of hardware on your box when you upgraded to Hoary. 
> Wish I could provide you better information than my pure speculation
> here.  Either way, I found your original post informative and
> intriguing.

FYI, when moving to breezy kernel 2.6.12 on old AMD K6-3 machine, I was
forced to add "irqpoll" option in grub to boot up properly.

else IRQ 14 just get dropped !

Here not because of an ethernet card but because of the IDE chipset...
far less funny... ;)

here's dmesg

[   39.167406] ALI15X3: IDE controller at PCI slot 0000:00:0f.0
[   39.167435] ACPI: PCI Interrupt 0000:00:0f.0[A]: no GSI - using IRQ 14
[   39.167451] PCI: setting IRQ 14 as level-triggered
[   39.167477] ALI15X3: chipset revision 32
[   39.167486] ALI15X3: not 100% native mode: will probe irqs later
[   39.167528]     ide0: BM-DMA at 0x7030-0x7037, BIOS settings:
hda:DMA, hdb:pio
[   39.167569]     ide1: BM-DMA at 0x7038-0x703f, BIOS settings:
hdc:DMA, hdd:pio
[   39.167607] Probing IDE interface ide0...
[   39.431490] hda: IBM-DARA-212000, ATA DISK drive
[   43.012213] irq 14: nobody cared (try booting with the "irqpoll" option.
[   43.012263]  [<c012de33>] __report_bad_irq+0x31/0x74
[   43.012299]  [<c012df0b>] note_interrupt+0x7d/0xa2
[   43.012321]  [<c012da30>] __do_IRQ+0x85/0xb1
[   43.012342]  [<c0104ca5>] do_IRQ+0x19/0x24
[   43.012365]  [<c01038b6>] common_interrupt+0x1a/0x20
[   43.012388]  [<c0119058>] __do_softirq+0x2c/0x7d
[   43.012415]  [<c01190cb>] do_softirq+0x22/0x26
[   43.012432]  [<c0104caa>] do_IRQ+0x1e/0x24
[   43.012447]  [<c01038b6>] common_interrupt+0x1a/0x20
[   43.012469]  [<c012dc0e>] setup_irq+0x9d/0xc1
[   43.012487]  [<cc8aa4a0>] ide_intr+0x0/0xed [ide_core]
[   43.012590]  [<c012dd27>] request_irq+0x72/0x8b
[   43.012611]  [<cc8ad646>] init_irq+0x1d4/0x302 [ide_core]
[   43.012673]  [<cc8aa4a0>] ide_intr+0x0/0xed [ide_core]
[   43.012736]  [<cc8adb54>] hwif_init+0xfe/0x1db [ide_core]
[   43.012797]  [<cc8ad2e9>] probe_hwif_init_with_fixup+0x1f/0x6c [ide_core]
[   43.012860]  [<cc8b009e>] ide_setup_pci_device+0x47/0x7d [ide_core]
[   43.012929]  [<cc8b0205>] ide_scan_pcidev+0x34/0x4d [ide_core]
[   43.012992]  [<cc8b0245>] ide_scan_pcibus+0x27/0x8e [ide_core]
[   43.013054]  [<cc858005>] ide_generic_init+0x5/0x12 [ide_generic]
[   43.013073]  [<c012967a>] sys_init_module+0xb5/0x172
[   43.013103]  [<c0102ef9>] syscall_call+0x7/0xb
[   43.013129] handlers:
[   43.013142] [<cc8aa4a0>] (ide_intr+0x0/0xed [ide_core])
[   43.013213] Disabling IRQ #14
[   43.013420] ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
[   43.013723] Probing IDE interface ide1...

-- 
Best regards,

Guillaume

"The Desert, it's a tremendous place to cross!
you only see people you like and who truly like you."
Nicolas Sarkozy.

http://www.grhq.net/homer.txt
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3389 bytes
Desc: S/MIME Cryptographic Signature
URL: <https://lists.ubuntu.com/archives/ubuntu-users/attachments/20050908/7070678a/attachment.bin>