Please test my alpha tree

Please test my alpha tree

[Matt Turner]

After a few weeks of no work on the kernel, I've vacuumed up the
outstanding patches.

They include the removal of __do_IRQ, so please give it a test on your hardware.

On boot-up of my UP1500, I see

HWRPB cycle frequency bogus.  Estimated 796423805 Hz
Trying to install interrupt handler for IRQ8
------------[ cut here ]------------
WARNING: at kernel/softirq.c:138 __local_bh_enable+0xa8/0xc0()
Modules linked in:
fffffc0000973cf0 ffffffffffffffff fffffc000032ff78 0000000000000200
       0000000000000008 0000000000000008 0000000000000001 0000000000000001
       fffffc00003306e8 fffffc000099be18 fffffc0000315604 fffffc000035fab0
       fffffc00003157a4 fffffc000097bd38 0000000000000000 fffffc0002c08240
       fffffc0000310e10 fffffc0000974f48 fffffc000096e5c0 0000000000000038
       0000000000000001 0000000000000007 0000000000000000 0000000000000005
Trace:
[<fffffc000032ff78>] __local_bh_enable+0xa8/0xc0
[<fffffc00003306e8>] irq_enter+0x68/0x90
[<fffffc0000315604>] handle_irq+0x64/0xf0
[<fffffc000035fab0>] handle_bad_irq+0x0/0x390
[<fffffc00003157a4>] do_entInt+0xc4/0x1e0
[<fffffc0000310e10>] ret_from_sys_call+0x0/0x10
[<fffffc00004e694c>] vgacon_startup+0x1cc/0x450
[<fffffc00003309c0>] request_resource+0x0/0x30
[<fffffc00004e6980>] vgacon_startup+0x200/0x450
[<fffffc000031001c>] _stext+0x1c/0x20

---[ end trace 4eaa2a86a8e2da22 ]---
Console: colour VGA+ 80x25
console [ttyS0] enabled

So something else needs to be done.

Ivan, the last patch in my tree is a breakage fix that you proposed.
It seems fine to me; can you please give your sign-off?

Thanks,
Matt

Re: Please test my alpha tree

[Michael Cree]

On 12/01/11 18:29, Matt Turner wrote:
> After a few weeks of no work on the kernel, I've vacuumed up the
> outstanding patches.

Nice to see.

> They include the removal of __do_IRQ, so please give it a test on your hardware.
>
> On boot-up of my UP1500, I see
>
> HWRPB cycle frequency bogus.  Estimated 796423805 Hz
> Trying to install interrupt handler for IRQ8
> ------------[ cut here ]------------
> WARNING: at kernel/softirq.c:138 __local_bh_enable+0xa8/0xc0()

I don't think that's related to the __do_IRQ removal.  It occurs in the 
2.6.37 kernel just released.  Was discussed on linux-alpha with no 
resolution; see
http://marc.info/?l=linux-alpha&m=128859282018188&w=2 and following 
messages in thread.

> Ivan, the last patch in my tree is a breakage fix that you proposed.
> It seems fine to me; can you please give your sign-off?

The commit message should give more information.  How about the following?

Commit df9ee29270 made arch_local_irq_save and arch_local_irq_restore 
static inline which with -Werror trips up on __set_hae() and _set_hae() 
which are extern inline.  The naive solution is to make __set_hae() and 
set_hae() static inline but for reasons described in commit 
d559d4a24a3fe this breaks the generic kernel build.  Instead, since this 
is architecture specific code, this patch hard wires in the architecture 
specific method of disabling and enabling interrupts.

You can also add my Tested-by (I tested on generic, miata and dp264 
kernel builds).

The author information is mucked up on commits 0c9e7396d8d and 
3cf2323f0b74 and should be repaired.

The message on commit 3cf2323f0b74 is not very informative.  What "flag 
variables"?  Might be fixed by saying "Makefile flag variables".

Cheers
Michael.

Re: Please test my alpha tree

[Ivan Kokshaysky]

On Thu, Jan 13, 2011 at 10:10:12AM +1300, Michael Cree wrote:
> On 12/01/11 18:29, Matt Turner wrote:
>> On boot-up of my UP1500, I see
>>
>> HWRPB cycle frequency bogus.  Estimated 796423805 Hz
>> Trying to install interrupt handler for IRQ8
>> ------------[ cut here ]------------
>> WARNING: at kernel/softirq.c:138 __local_bh_enable+0xa8/0xc0()
>
> I don't think that's related to the __do_IRQ removal.  It occurs in the 
> 2.6.37 kernel just released.  Was discussed on linux-alpha with no 
> resolution; see
> http://marc.info/?l=linux-alpha&m=128859282018188&w=2 and following 
> messages in thread.

Fixed, the patch will follow shortly.

>> Ivan, the last patch in my tree is a breakage fix that you proposed.
>> It seems fine to me; can you please give your sign-off?
>
> The commit message should give more information.  How about the following?
>
> Commit df9ee29270 made arch_local_irq_save and arch_local_irq_restore 
> static inline which with -Werror trips up on __set_hae() and _set_hae() 
> which are extern inline.  The naive solution is to make __set_hae() and 
> set_hae() static inline but for reasons described in commit d559d4a24a3fe 
> this breaks the generic kernel build.  Instead, since this is architecture 
> specific code, this patch hard wires in the architecture specific method of 
> disabling and enabling interrupts.

Agreed, that would be nice description.

> You can also add my Tested-by (I tested on generic, miata and dp264 kernel 
> builds).

Matt, feel free to add
 Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>

Ivan.

Re: Please test my alpha tree

[Matt Turner]

On Wed, Jan 12, 2011 at 9:54 PM, Ivan Kokshaysky
<ink@jurassic.park.msu.ru> wrote:
> On Thu, Jan 13, 2011 at 10:10:12AM +1300, Michael Cree wrote:
>> On 12/01/11 18:29, Matt Turner wrote:
>>> On boot-up of my UP1500, I see
>>>
>>> HWRPB cycle frequency bogus.  Estimated 796423805 Hz
>>> Trying to install interrupt handler for IRQ8
>>> ------------[ cut here ]------------
>>> WARNING: at kernel/softirq.c:138 __local_bh_enable+0xa8/0xc0()
>>
>> I don't think that's related to the __do_IRQ removal.  It occurs in the
>> 2.6.37 kernel just released.  Was discussed on linux-alpha with no
>> resolution; see
>> http://marc.info/?l=linux-alpha&m=128859282018188&w=2 and following
>> messages in thread.
>
> Fixed, the patch will follow shortly.
>
>>> Ivan, the last patch in my tree is a breakage fix that you proposed.
>>> It seems fine to me; can you please give your sign-off?
>>
>> The commit message should give more information.  How about the following?
>>
>> Commit df9ee29270 made arch_local_irq_save and arch_local_irq_restore
>> static inline which with -Werror trips up on __set_hae() and _set_hae()
>> which are extern inline.  The naive solution is to make __set_hae() and
>> set_hae() static inline but for reasons described in commit d559d4a24a3fe
>> this breaks the generic kernel build.  Instead, since this is architecture
>> specific code, this patch hard wires in the architecture specific method of
>> disabling and enabling interrupts.
>
> Agreed, that would be nice description.
>
>> You can also add my Tested-by (I tested on generic, miata and dp264 kernel
>> builds).
>
> Matt, feel free to add
>  Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
>
> Ivan.

Thanks Ivan! I've updated this patch with your tag.

Matt
--
To unsubscribe from this list: send the line "unsubscribe linux-alpha" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Please test my alpha tree

[Matt Turner]

On Wed, Jan 12, 2011 at 9:10 PM, Michael Cree <mcree@orcon.net.nz> wrote:
> On 12/01/11 18:29, Matt Turner wrote:
>>
>> After a few weeks of no work on the kernel, I've vacuumed up the
>> outstanding patches.
>
> Nice to see.
>
>> They include the removal of __do_IRQ, so please give it a test on your
>> hardware.
>>
>> On boot-up of my UP1500, I see
>>
>> HWRPB cycle frequency bogus.  Estimated 796423805 Hz
>> Trying to install interrupt handler for IRQ8
>> ------------[ cut here ]------------
>> WARNING: at kernel/softirq.c:138 __local_bh_enable+0xa8/0xc0()
>
> I don't think that's related to the __do_IRQ removal.  It occurs in the
> 2.6.37 kernel just released.  Was discussed on linux-alpha with no
> resolution; see
> http://marc.info/?l=linux-alpha&m=128859282018188&w=2 and following messages
> in thread.
>
>> Ivan, the last patch in my tree is a breakage fix that you proposed.
>> It seems fine to me; can you please give your sign-off?
>
> The commit message should give more information.  How about the following?
>
> Commit df9ee29270 made arch_local_irq_save and arch_local_irq_restore static
> inline which with -Werror trips up on __set_hae() and _set_hae() which are
> extern inline.  The naive solution is to make __set_hae() and set_hae()
> static inline but for reasons described in commit d559d4a24a3fe this breaks
> the generic kernel build.  Instead, since this is architecture specific
> code, this patch hard wires in the architecture specific method of disabling
> and enabling interrupts.
>
> You can also add my Tested-by (I tested on generic, miata and dp264 kernel
> builds).
>
> The author information is mucked up on commits 0c9e7396d8d and 3cf2323f0b74
> and should be repaired.
>
> The message on commit 3cf2323f0b74 is not very informative.  What "flag
> variables"?  Might be fixed by saying "Makefile flag variables".
>
> Cheers
> Michael.
>

All updated. Thanks for testing and providing a better description. :)

Matt
--
To unsubscribe from this list: send the line "unsubscribe linux-alpha" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH] alpha: fix WARN_ON in __local_bh_enable()

[Ivan Kokshaysky]

Interrupts ought to be disabled _before_ irq_enter().

Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>

diff --git a/arch/alpha/kernel/irq.c b/arch/alpha/kernel/irq.c
index 5912900..9ab234f 100644
--- a/arch/alpha/kernel/irq.c
+++ b/arch/alpha/kernel/irq.c
@@ -160,15 +160,14 @@ handle_irq(int irq)
 		return;
 	}
 
-	irq_enter();
 	/*
-	 * handle_irq() must be called with IPL_MAX. Note that we do not
+	 * From here we must proceed 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();
+	irq_enter();
 	generic_handle_irq_desc(irq, desc);
 	irq_exit();
 }
-

Re: [PATCH] alpha: fix WARN_ON in __local_bh_enable()

[Matt Turner]

On Wed, Jan 12, 2011 at 10:02 PM, Ivan Kokshaysky
<ink@jurassic.park.msu.ru> wrote:
> Interrupts ought to be disabled _before_ irq_enter().
>
> Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
>
> diff --git a/arch/alpha/kernel/irq.c b/arch/alpha/kernel/irq.c
> index 5912900..9ab234f 100644
> --- a/arch/alpha/kernel/irq.c
> +++ b/arch/alpha/kernel/irq.c
> @@ -160,15 +160,14 @@ handle_irq(int irq)
>                return;
>        }
>
> -       irq_enter();
>        /*
> -        * handle_irq() must be called with IPL_MAX. Note that we do not
> +        * From here we must proceed 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();
> +       irq_enter();
>        generic_handle_irq_desc(irq, desc);
>        irq_exit();
>  }
> -

Thanks! Applied.

Matt
--
To unsubscribe from this list: send the line "unsubscribe linux-alpha" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html