PB1176 broken in 2.6.35-rc1, bisected

PB1176 broken in 2.6.35-rc1, bisected

[Linus Walleij]

Hi,

my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
and hangs:

(...)
Switching to clocksource timer3
NET: Registered protocol family 2
IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
TCP established hash table entries: 4096 (order: 3, 32768 bytes)
TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
TCP: Hash tables configured (established 4096 bind 4096)
TCP reno registered
UDP hash table entries: 256 (order: 0, 4096 bytes)
UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
NET: Registered protocol family 1
RPC: Registered udp transport module.
RPC: Registered tcp transport module.
RPC: Registered tcp NFSv4.1 backchannel transport module.


I bisected until I had this as the first commit causing the trouble:

commit ac1d426e825ab5778995f2f6f053ca2e6b45c622
Merge: fda0e18 a3685f0
Author: Russell King <rmk+kernel@arm.linux.org.uk>
Date:   Mon May 17 17:24:04 2010 +0100

    Merge branch 'devel-stable' into devel

    Conflicts:
        arch/arm/Kconfig
        arch/arm/include/asm/system.h
        arch/arm/mm/Kconfig


This doesn't say much so I suspect it's this latter one causing it:

commit fda0e18c8a7a3e02747c2b045b4fcd2c920410b9
Merge: 98830bc d1e86d6
Author: Russell King <rmk+kernel@arm.linux.org.uk>
Date:   Mon May 17 17:21:23 2010 +0100

    Merge branch 'devel-pmu' into devel


Any hints on how to fix this? Some PMU issue on this board
perhaps?

Yours,
Linus Walleij

PB1176 broken in 2.6.35-rc1, bisected

[Catalin Marinas]

On Wed, 2010-06-02 at 23:33 +0100, Linus Walleij wrote:
> my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
> and hangs:
> 
> (...)
> Switching to clocksource timer3
> NET: Registered protocol family 2
> IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
> TCP established hash table entries: 4096 (order: 3, 32768 bytes)
> TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
> TCP: Hash tables configured (established 4096 bind 4096)
> TCP reno registered
> UDP hash table entries: 256 (order: 0, 4096 bytes)
> UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
> NET: Registered protocol family 1
> RPC: Registered udp transport module.
> RPC: Registered tcp transport module.
> RPC: Registered tcp NFSv4.1 backchannel transport module.

Have you tried disabling CONFIG_CACHE_L2X0?

-- 
Catalin

PB1176 broken in 2.6.35-rc1, bisected

[Russell King - ARM Linux]

On Thu, Jun 03, 2010 at 12:33:57AM +0200, Linus Walleij wrote:
> I bisected until I had this as the first commit causing the trouble:
> 
> commit ac1d426e825ab5778995f2f6f053ca2e6b45c622
> Merge: fda0e18 a3685f0
> Author: Russell King <rmk+kernel@arm.linux.org.uk>
> Date:   Mon May 17 17:24:04 2010 +0100
> 
>     Merge branch 'devel-stable' into devel
> 
>     Conflicts:
>         arch/arm/Kconfig
>         arch/arm/include/asm/system.h
>         arch/arm/mm/Kconfig

You should be able to bisect further - did bisect just give up at this
point?

Do you know which half of that merge was also tested?

PB1176 broken in 2.6.35-rc1, bisected

[Linus Walleij]

2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
> On Wed, 2010-06-02 at 23:33 +0100, Linus Walleij wrote:
>> my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
>> and hangs:
>>
>> (...)
>> Switching to clocksource timer3
>> NET: Registered protocol family 2
>> IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
>> TCP established hash table entries: 4096 (order: 3, 32768 bytes)
>> TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
>> TCP: Hash tables configured (established 4096 bind 4096)
>> TCP reno registered
>> UDP hash table entries: 256 (order: 0, 4096 bytes)
>> UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
>> NET: Registered protocol family 1
>> RPC: Registered udp transport module.
>> RPC: Registered tcp transport module.
>> RPC: Registered tcp NFSv4.1 backchannel transport module.
>
> Have you tried disabling CONFIG_CACHE_L2X0?

Hey it works! :-)

I suppose your latest patches will hopefully  get the defconfig with
l2x0 working again in -rc2 if they're merged?

Linus Walleij

PB1176 broken in 2.6.35-rc1, bisected

[Catalin Marinas]

On Thu, 2010-06-03 at 09:48 +0100, Linus Walleij wrote:
> 2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
> > On Wed, 2010-06-02 at 23:33 +0100, Linus Walleij wrote:
> >> my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
> >> and hangs:
> >>
> >> (...)
> >> Switching to clocksource timer3
> >> NET: Registered protocol family 2
> >> IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
> >> TCP established hash table entries: 4096 (order: 3, 32768 bytes)
> >> TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
> >> TCP: Hash tables configured (established 4096 bind 4096)
> >> TCP reno registered
> >> UDP hash table entries: 256 (order: 0, 4096 bytes)
> >> UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
> >> NET: Registered protocol family 1
> >> RPC: Registered udp transport module.
> >> RPC: Registered tcp transport module.
> >> RPC: Registered tcp NFSv4.1 backchannel transport module.
> >
> > Have you tried disabling CONFIG_CACHE_L2X0?
> 
> Hey it works! :-)
> 
> I suppose your latest patches will hopefully  get the defconfig with
> l2x0 working again in -rc2 if they're merged?

I think there is an L2 erratum (and trying to get the right workaround).
It may be triggered by commit e7c5650f6.

I'll get back on the list with a patch.

-- 
Catalin

PB1176 broken in 2.6.35-rc1, bisected

[Linus Walleij]

2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
> On Thu, 2010-06-03 at 09:48 +0100, Linus Walleij wrote:
>> 2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
>> > On Wed, 2010-06-02 at 23:33 +0100, Linus Walleij wrote:
>> >> my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
>> >> and hangs:
>> >>
>> >> (...)
>> >> Switching to clocksource timer3
>> >> NET: Registered protocol family 2
>> >> IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
>> >> TCP established hash table entries: 4096 (order: 3, 32768 bytes)
>> >> TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
>> >> TCP: Hash tables configured (established 4096 bind 4096)
>> >> TCP reno registered
>> >> UDP hash table entries: 256 (order: 0, 4096 bytes)
>> >> UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
>> >> NET: Registered protocol family 1
>> >> RPC: Registered udp transport module.
>> >> RPC: Registered tcp transport module.
>> >> RPC: Registered tcp NFSv4.1 backchannel transport module.
>> >
>> > Have you tried disabling CONFIG_CACHE_L2X0?
>>
>> Hey it works! :-)
>>
>> I suppose your latest patches will hopefully ?get the defconfig with
>> l2x0 working again in -rc2 if they're merged?
>
> I think there is an L2 erratum (and trying to get the right workaround).
> It may be triggered by commit e7c5650f6.
>
> I'll get back on the list with a patch.

Have I missed this? My 2.6.35's and -next trees still require me to
disable the L2X0 cache on the PB1176 for it to boot.

It's a bit sad if we have a 2.6.35 which cannot even boot on this
card...

Can you send me a copy of the errata so I can look at it if
you're swamped with other stuff?

Yours,
Linus Walleij

PB1176 broken in 2.6.35-rc1, bisected

[Catalin Marinas]

On Tue, 2010-06-29 at 01:42 +0100, Linus Walleij wrote:
> 2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
> > On Thu, 2010-06-03 at 09:48 +0100, Linus Walleij wrote:
> >> 2010/6/3 Catalin Marinas <catalin.marinas@arm.com>:
> >> > On Wed, 2010-06-02 at 23:33 +0100, Linus Walleij wrote:
> >> >> my PB1176 doesn't boot properly in 2.6.35-rc1, it goes to this point
> >> >> and hangs:
> >> >>
> >> >> (...)
> >> >> Switching to clocksource timer3
> >> >> NET: Registered protocol family 2
> >> >> IP route cache hash table entries: 1024 (order: 0, 4096 bytes)
> >> >> TCP established hash table entries: 4096 (order: 3, 32768 bytes)
> >> >> TCP bind hash table entries: 4096 (order: 2, 16384 bytes)
> >> >> TCP: Hash tables configured (established 4096 bind 4096)
> >> >> TCP reno registered
> >> >> UDP hash table entries: 256 (order: 0, 4096 bytes)
> >> >> UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
> >> >> NET: Registered protocol family 1
> >> >> RPC: Registered udp transport module.
> >> >> RPC: Registered tcp transport module.
> >> >> RPC: Registered tcp NFSv4.1 backchannel transport module.
> >> >
> >> > Have you tried disabling CONFIG_CACHE_L2X0?
> >>
> >> Hey it works! :-)
> >>
> >> I suppose your latest patches will hopefully  get the defconfig with
> >> l2x0 working again in -rc2 if they're merged?
> >
> > I think there is an L2 erratum (and trying to get the right workaround).
> > It may be triggered by commit e7c5650f6.
> >
> > I'll get back on the list with a patch.
> 
> Have I missed this? My 2.6.35's and -next trees still require me to
> disable the L2X0 cache on the PB1176 for it to boot.
> 
> It's a bit sad if we have a 2.6.35 which cannot even boot on this
> card...

I posted some patches couple of times but I haven't received any
comments. I'm posting another version (now only affecting the RealView
code) and cc you. If it works for you as well, I'll send it to the patch
system.

-- 
Catalin