2010-08-14 06:20:31

by Sedat Dilek

[permalink] [raw]
Subject: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

Hi Greg,

looks like 2.6.35.2 misses some fixes from 2.6.35.2-rc1?

sd@tbox:~/src/linux-2.6/patches/upstream$ cat patch-2.6.35.1
patch-2.6.35.2-rc1 > new.patch

sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git" new.patch | wc -l
132

sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git"
patch-2.6.35.2 | wc -l
128

Unfortunately, a kernel patched w/ 2.6.35.2 causes Call traces here.

Attached are dmeg outputs for both kernels.


Kind Regards,
- Sedat -


Attachments:
dmesg_2.6.35-2.sd.1-686.txt (89.61 kB)
dmesg_2.6.35-2.sd.rc1-686.txt (46.94 kB)
Download all attachments

2010-08-14 06:36:32

by Sedat Dilek

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

Hope this helps to give an overview what has changed:

$ grep "diff --git" patch-2.6.35.2 | sort > list.patch-2.6.35.2
$ grep "diff --git" new.patch | sort > list.patch-2.6.35.1+patch-2.6.35.2-rc1


$ diff -Naur list.patch-2.6.35.2
list.patch-2.6.35.1+patch-2.6.35.2-rc1 | grep ^+diff
+diff --git a/arch/arm/plat-mxc/include/mach/gpio.h
b/arch/arm/plat-mxc/include/mach/gpio.h
+diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
+diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
+diff --git a/fs/block_dev.c b/fs/block_dev.c
+diff --git a/include/linux/notifier.h b/include/linux/notifier.h
+diff --git a/Makefile b/Makefile

$ diff -Naur list.patch-2.6.35.2
list.patch-2.6.35.1+patch-2.6.35.2-rc1 | grep ^-diff
-diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
-diff --git a/mm/memory.c b/mm/memory.c

$ ls -l list.patch-2.6.35.*
-rw-r--r-- 1 sd sd 9004 14. Aug 08:32 list.patch-2.6.35.1+patch-2.6.35.2-rc1
-rw-r--r-- 1 sd sd 8724 14. Aug 08:32 list.patch-2.6.35.2

- Sedat -

On Sat, Aug 14, 2010 at 8:20 AM, Sedat Dilek <[email protected]> wrote:
> Hi Greg,
>
> looks like 2.6.35.2 misses some fixes from 2.6.35.2-rc1?
>
> sd@tbox:~/src/linux-2.6/patches/upstream$ cat patch-2.6.35.1
> patch-2.6.35.2-rc1 > new.patch
>
> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git" new.patch | wc -l
> 132
>
> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git"
> patch-2.6.35.2 | wc -l
> 128
>
> Unfortunately, a kernel patched w/ 2.6.35.2 causes Call traces here.
>
> Attached are dmeg outputs for both kernels.
>
>
> Kind Regards,
> - Sedat -
>

2010-08-14 13:43:46

by Greg KH

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

On Sat, Aug 14, 2010 at 08:20:26AM +0200, Sedat Dilek wrote:
> Hi Greg,
>
> looks like 2.6.35.2 misses some fixes from 2.6.35.2-rc1?

Huh? I don't understand what you are trying to show here at all.

> sd@tbox:~/src/linux-2.6/patches/upstream$ cat patch-2.6.35.1
> patch-2.6.35.2-rc1 > new.patch
>
> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git" new.patch | wc -l
> 132
>
> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git"
> patch-2.6.35.2 | wc -l
> 128
>
> Unfortunately, a kernel patched w/ 2.6.35.2 causes Call traces here.

And these are also showing up on Linus's tree right now as well? Have
you reported them?

thanks,

greg k-h

2010-08-14 13:43:48

by Greg KH

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

On Sat, Aug 14, 2010 at 08:36:30AM +0200, Sedat Dilek wrote:
> Hope this helps to give an overview what has changed:
>
> $ grep "diff --git" patch-2.6.35.2 | sort > list.patch-2.6.35.2
> $ grep "diff --git" new.patch | sort > list.patch-2.6.35.1+patch-2.6.35.2-rc1
>
>
> $ diff -Naur list.patch-2.6.35.2
> list.patch-2.6.35.1+patch-2.6.35.2-rc1 | grep ^+diff
> +diff --git a/arch/arm/plat-mxc/include/mach/gpio.h
> b/arch/arm/plat-mxc/include/mach/gpio.h
> +diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
> +diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
> +diff --git a/fs/block_dev.c b/fs/block_dev.c
> +diff --git a/include/linux/notifier.h b/include/linux/notifier.h
> +diff --git a/Makefile b/Makefile
>
> $ diff -Naur list.patch-2.6.35.2
> list.patch-2.6.35.1+patch-2.6.35.2-rc1 | grep ^-diff
> -diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
> -diff --git a/mm/memory.c b/mm/memory.c
>
> $ ls -l list.patch-2.6.35.*
> -rw-r--r-- 1 sd sd 9004 14. Aug 08:32 list.patch-2.6.35.1+patch-2.6.35.2-rc1
> -rw-r--r-- 1 sd sd 8724 14. Aug 08:32 list.patch-2.6.35.2

I still fail to understand what you are trying to show here at all.

confused,

greg k-h

2010-08-14 15:28:41

by Sedat Dilek

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

On Sat, Aug 14, 2010 at 3:43 PM, Greg KH <[email protected]> wrote:
> On Sat, Aug 14, 2010 at 08:36:30AM +0200, Sedat Dilek wrote:
>> Hope this helps to give an overview what has changed:
[...]
>
> I still fail to understand what you are trying to show here at all.
>
> confused,
>
> greg k-h
>

I was trying to show you that the patches in .35.2-rc1 and .35.2 are
not the same.
Normally, the rcX patches become the final ones.

Unfortunately, I was too lazy to check the names of the patches and
simply compared the files that have been modified (lines beginning
with "diff --git a/file b/file").

In the meantime, Mr./Mrs.(?) "wylda" (sorry, don't know if first or
last name) bisected the problem [1].

The confusing part for me is why "fresh" upstream patches entered a
2.6.35.y fix-series.
I can't say if only i386 is concerned, but as

mm: fix missing page table unmap for stack guard page failure case
mm: keep a guard page below a grow-down stack segment

are also upstream patches, I think the both cause problems for
.35-gitX also [2].

- Sedat -

[1] https://bugzilla.kernel.org/show_bug.cgi?id=16588
[2] https://bugzilla.kernel.org/show_bug.cgi?id=16588#c5
[3] [1] http://lkml.org/lkml/2010/8/14/54

2010-08-14 15:40:57

by Sedat Dilek

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

On Sat, Aug 14, 2010 at 3:42 PM, Greg KH <[email protected]> wrote:
> On Sat, Aug 14, 2010 at 08:20:26AM +0200, Sedat Dilek wrote:
>> Hi Greg,
>>
>> looks like 2.6.35.2 misses some fixes from 2.6.35.2-rc1?
>
> Huh?  I don't understand what you are trying to show here at all.
>
>> sd@tbox:~/src/linux-2.6/patches/upstream$ cat patch-2.6.35.1
>> patch-2.6.35.2-rc1 > new.patch
>>
>> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git" new.patch | wc -l
>> 132
>>
>> sd@tbox:~/src/linux-2.6/patches/upstream$ grep "diff --git"
>> patch-2.6.35.2 | wc -l
>> 128
>>
>> Unfortunately, a kernel patched w/ 2.6.35.2 causes Call traces here.
>
> And these are also showing up on Linus's tree right now as well?  Have
> you reported them?
>

Shall I open a separate BR for 2.6.35-gitX?
I can confirm the problem occured in the first time in git12 (saw it
in git14 and git15).

As I have currently a slow system, I won't do a bisect or X-times
kernel-builds, sorry for that.
Each build takes me around 2hrs.

- Sedat -

2010-08-14 16:11:38

by wylda

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

Hi,

> Shall I open a separate BR for 2.6.35-gitX?

No it's not necessary. Leaving a comment under a bisected bugreport
is enough. Notice comment #0 stating, that all the latest stable
come with this regression - no need to open a new bug for each
version.

W.

> - Sedat -

PS: Wylda ("He";) is a nickname, so none of First or Surname...

2010-08-14 17:33:54

by Greg KH

[permalink] [raw]
Subject: Re: Call traces w/ 2.6.35.2 (not w/ 2.6.35.1 plus 2.6.35.2-rc1)

On Sat, Aug 14, 2010 at 05:28:39PM +0200, Sedat Dilek wrote:
> On Sat, Aug 14, 2010 at 3:43 PM, Greg KH <[email protected]> wrote:
> > On Sat, Aug 14, 2010 at 08:36:30AM +0200, Sedat Dilek wrote:
> >> Hope this helps to give an overview what has changed:
> [...]
> >
> > I still fail to understand what you are trying to show here at all.
> >
> > confused,
> >
> > greg k-h
> >
>
> I was trying to show you that the patches in .35.2-rc1 and .35.2 are
> not the same.

Yes, I know that, I did the release :)

> Normally, the rcX patches become the final ones.
>
> Unfortunately, I was too lazy to check the names of the patches and
> simply compared the files that have been modified (lines beginning
> with "diff --git a/file b/file").
>
> In the meantime, Mr./Mrs.(?) "wylda" (sorry, don't know if first or
> last name) bisected the problem [1].
>
> The confusing part for me is why "fresh" upstream patches entered a
> 2.6.35.y fix-series.

They were deemed necessary for inclusion.

> I can't say if only i386 is concerned, but as
>
> mm: fix missing page table unmap for stack guard page failure case
> mm: keep a guard page below a grow-down stack segment
>
> are also upstream patches, I think the both cause problems for
> .35-gitX also [2].

Yes, Linus just posted a potential fix. Can you try that out?

thanks,

greg k-h