Matts tree, IRQ related warning

Matts tree, IRQ related warning

[Morten Holst Larsen]

Hi,

When I build and run a kernel for my DS15 from Matts tree, I get a warning.
Regardless of the warning everything seems to work ok.

/Morten

Linux version 2.6.36+ (root@ds15) (gcc version 4.2.4 (Debian 4.2.4-6)) #4 
Sun Oct 31 12:12:07 CET 2010
Booting on Titan variation DEFAULT using machine vector TITAN from SRM
Major Options: EV67 MAGIC_SYSRQ
** SNIP **
Trying to install interrupt handler for IRQ8
------------[ cut here ]------------
WARNING: at kernel/softirq.c:138 __local_bh_enable+0xac/0xc0()
Modules linked in:
fffffc0001677d60 fffffc000167ceb8 fffffc000103506c 0000000000000200
0000000000000008 fffffc0001677e00 0000000001234567 0000000000000000
fffffc00010354e8 fffffc000169aed0 fffffc00010154a8 fffffc000169aed0
fffffc00010158a4 fffffc0001686a20 0000000000000000 fffffc0001677e18
fffffc0001010d70 fffffc0001668e0c fffffc000166ad78 fffffc0001695260
0000000000000007 0000000000000000 0000000000000440 0000000000000200
Trace:
[<fffffc000103506c>] __local_bh_enable+0xac/0xc0
[<fffffc00010354e8>] irq_enter+0x68/0x80
[<fffffc00010154a8>] handle_irq+0x58/0x110
[<fffffc00010158a4>] do_entInt+0xc4/0x1d0
[<fffffc0001010d70>] ret_from_sys_call+0x0/0x10
[<fffffc000106d860>] irq_to_desc+0x0/0x30
[<fffffc000101001c>] _stext+0x1c/0x20
---[ end trace 4eaa2a86a8e2da22 ]--- 

Re: Matts tree, IRQ related warning

[Michael Cree]

On 01/11/10 18:58, Morten Holst Larsen wrote:
> When I build and run a kernel for my DS15 from Matts tree, I get a warning.
> Regardless of the warning everything seems to work ok.
>
> /Morten
>
> Linux version 2.6.36+ (root@ds15) (gcc version 4.2.4 (Debian 4.2.4-6))
> #4 Sun Oct 31 12:12:07 CET 2010
> Booting on Titan variation DEFAULT using machine vector TITAN from SRM
> Major Options: EV67 MAGIC_SYSRQ
> ** SNIP **
> Trying to install interrupt handler for IRQ8
> ------------[ cut here ]------------
> WARNING: at kernel/softirq.c:138 __local_bh_enable+0xac/0xc0()
> Modules linked in:
> fffffc0001677d60 fffffc000167ceb8 fffffc000103506c 0000000000000200
> 0000000000000008 fffffc0001677e00 0000000001234567 0000000000000000
> fffffc00010354e8 fffffc000169aed0 fffffc00010154a8 fffffc000169aed0
> fffffc00010158a4 fffffc0001686a20 0000000000000000 fffffc0001677e18
> fffffc0001010d70 fffffc0001668e0c fffffc000166ad78 fffffc0001695260
> 0000000000000007 0000000000000000 0000000000000440 0000000000000200
> Trace:
> [<fffffc000103506c>] __local_bh_enable+0xac/0xc0
> [<fffffc00010354e8>] irq_enter+0x68/0x80
> [<fffffc00010154a8>] handle_irq+0x58/0x110
> [<fffffc00010158a4>] do_entInt+0xc4/0x1d0
> [<fffffc0001010d70>] ret_from_sys_call+0x0/0x10
> [<fffffc000106d860>] irq_to_desc+0x0/0x30
> [<fffffc000101001c>] _stext+0x1c/0x20
> ---[ end trace 4eaa2a86a8e2da22 ]---

Yes, I have seen that warning too.  It's now present in Linus' tree.  I 
see that commit 75e1056f5 did some work in this area and I am 
speculating that it (or follow on commits) might be the cause.  CCed to 
the authors of the commit to bring it to their attention.

Cheers
Michael.

Re: Matts tree, IRQ related warning

[Peter Zijlstra]

On Mon, 2010-11-01 at 20:28 +1300, Michael Cree wrote:
> > WARNING: at kernel/softirq.c:138 __local_bh_enable+0xac/0xc0()

That's WARN_ON(!irqs_disabled()), right?

How can that be, you're in the irq_enter path, the hardware should have
disabled IRQs already, right?

Re: Matts tree, IRQ related warning

[Venkatesh Pallipadi]

On Mon, Nov 1, 2010 at 4:48 AM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> On Mon, 2010-11-01 at 20:28 +1300, Michael Cree wrote:
>> > WARNING: at kernel/softirq.c:138 __local_bh_enable+0xac/0xc0()
>
> That's WARN_ON(!irqs_disabled()), right?
>
> How can that be, you're in the irq_enter path, the hardware should have
> disabled IRQs already, right?
>
>

Code snippet from arch/alpha/kernel/irq.c:handle_irq

        irq_enter();
        /*
         * __do_IRQ() must be called with IPL_MAX. Note that we do not
         * explicitly enable interrupts afterwards - some MILO PALcode
         * (namely LX164 one) seems to have severe problems with RTI
         * at IPL 0.
         */
        local_irq_disable();
        __do_IRQ(irq);
        irq_exit();


So, interrupt seems to be getting disabled after irq_enter and WARN_ON
with this change is coming from irq_enter.

Thanks,
Venki