LinuxLists.cc - 2.6.15-rc1-mm1

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

Richard Knutsson <[email protected]> wrote:
>
> >ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
> >
> >
> Got a compiler-error:
>
> CC drivers/serial/jsm/jsm_tty.o
> drivers/serial/jsm/jsm_tty.c: In function `jsm_input':
> drivers/serial/jsm/jsm_tty.c:592: error: structure has no member named `flip'
> drivers/serial/jsm/jsm_tty.c:619: error: structure has no member named `flip'
> drivers/serial/jsm/jsm_tty.c:620: error: structure has no member named `flip'

Yes, sorry, JSM is known-to-be-bust-in-mm. The maintainers are working
(slowly) on fixing it up.

2005-11-18 04:02:35

by George Anzinger

[permalink] [raw]

Subject: Re: [-mm patch] kernel/signal.c: fix compile warning

Adrian Bunk wrote:
> On Thu, Nov 17, 2005 at 11:18:07AM -0800, Andrew Morton wrote:
>
>>...
>>Changes since 2.6.14-mm2:
>>...
>>+sigaction-should-clear-all-signals-on-sig_ign-not-just.patch
>>
>> Signal code fix
>>...
>
>
>
> Patches get bonus points when they don't introduce new compile
> warnings...

Even more when the fix them. My loss, your gain :). Still I wonder
why I did not see that.

George
--

--
George Anzinger [email protected]
HRT (High-res-timers): http://sourceforge.net/projects/high-res-timers/

2005-11-18 04:24:04

by Valdis Klētnieks

[permalink] [raw]

Subject: 2.6.15-rc1-mm1 - immediate system reset at boot

On Thu, 17 Nov 2005 17:23:30 EST, [email protected] said:

> Am chasing another issue - once I got past that, it wouldn't boot at all.
> Grub would act like it was loading, then 2 seconds or so later, grub would
> start up again. My first guess was CONFIG_DEBUG_RODATA=y, but I ruled that
> out. More detail once I've done some more binary searching and ruled out
> self-inflicted idiocy....

Rebuilt from a completely new and clean tree, 2.6.14-mm1 is fine, 2.6.15-rc1
will start booting OK, -rc1-mm1 chokes up almost instantly. It doesn't live
long enough for either/both earlyprintk=vga or initcall_debug to output
anything - grub says "loading", clears the screen, and 2 seconds later the
laptop gives the 'ka-chunk' noise it does on a system reset, and I'm looking at
the BIOS splash screen.

Any obvious places to look, or time to play bisection on the 600 patches
in -mm?

Attachments:

(No filename) (226.00 B)

2005-11-18 04:28:14

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1 - immediate system reset at boot

[email protected] wrote:
>
> On Thu, 17 Nov 2005 17:23:30 EST, [email protected] said:
>
> > Am chasing another issue - once I got past that, it wouldn't boot at all.
> > Grub would act like it was loading, then 2 seconds or so later, grub would
> > start up again. My first guess was CONFIG_DEBUG_RODATA=y, but I ruled that
> > out. More detail once I've done some more binary searching and ruled out
> > self-inflicted idiocy....
>
> Rebuilt from a completely new and clean tree, 2.6.14-mm1 is fine, 2.6.15-rc1
> will start booting OK, -rc1-mm1 chokes up almost instantly. It doesn't live
> long enough for either/both earlyprintk=vga or initcall_debug to output
> anything - grub says "loading", clears the screen, and 2 seconds later the
> laptop gives the 'ka-chunk' noise it does on a system reset, and I'm looking at
> the BIOS splash screen.
>
> Any obvious places to look, or time to play bisection on the 600 patches
> in -mm?

Can't think of anything, sorry. The obvious area to probe is around the
x86 and mm patch subserieses.

2005-11-18 07:20:40

by Reuben Farrelly

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

Hi,

On 18/11/2005 8:18 a.m., Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
>
> - reiser4 significantly updated
>
>
>
>
> Changes since 2.6.14-mm2:

This has been one of the best -mm releases in a while. No problems compiling
or running - and so far nearly 18 hours uptime without any surprises.

Following up on a posting from the last -mm release, I'm still seeing errors
loading multiple network drivers as modules (e100 and sky2) when
CONFIG_PREEMPT_BKL is enabled, with 2.6.14-mm1, 2.6.14-mm2 and now
2.6.15-rc1-mm1. Mainline git doesn't exhibit the problem, so it's -mm specific.

This is what is logged:

Nov 18 17:40:42 tornado kernel: e100: 0000:06:03.0: e100_eeprom_load: EEPROM
corrupted
Nov 18 17:40:42 tornado kernel: ACPI: PCI interrupt for device 0000:06:03.0
disabled
Nov 18 17:40:42 tornado kernel: e100: probe of 0000:06:03.0 failed with error -11
Nov 18 17:40:43 tornado kernel: ACPI: PCI Interrupt 0000:04:00.0[A] -> GSI 17
(level, low) -> IRQ 177
Nov 18 17:40:43 tornado kernel: sky2 0000:04:00.0: unsupported chip type 0xff
Nov 18 17:40:43 tornado kernel: ACPI: PCI interrupt for device 0000:04:00.0
disabled
Nov 18 17:40:43 tornado kernel: sky2: probe of 0000:04:00.0 failed with error -95

I'm certain that both of these NIC's are OK as they work fine with
CONFIG_PREEMPT_BKL not selected.

With CONFIG_PREEMPT_BKL disabled and an otherwise identical config, the
driver modules load up just fine.

A known good kernel with this config was 2.6.14-rc5-mm1.
I have backed out git-netdev-all but it made no difference, as well as backed
out the e100 changes in -mm on 2.6.15-mm2, again no difference. So I suspect
it's not a netdev driver problem.

What else can I do to help narrow down the problem? What other trees or patches
would be worth backing out to try and narrow it down?

.config and system details are up on http://www.reub.net/kernel/

Reuben

2005-11-18 07:43:21

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

Reuben Farrelly <[email protected]> wrote:
>
> Hi,
>
> On 18/11/2005 8:18 a.m., Andrew Morton wrote:
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
> >
> > - reiser4 significantly updated
> >
> >
> >
> >
> > Changes since 2.6.14-mm2:
>
> This has been one of the best -mm releases in a while. No problems compiling
> or running - and so far nearly 18 hours uptime without any surprises.

We'll have to try harder. -mm2 is up there now, to break everything again.

> Following up on a posting from the last -mm release, I'm still seeing errors
> loading multiple network drivers as modules (e100 and sky2) when
> CONFIG_PREEMPT_BKL is enabled, with 2.6.14-mm1, 2.6.14-mm2 and now
> 2.6.15-rc1-mm1. Mainline git doesn't exhibit the problem, so it's -mm specific.
>
> This is what is logged:
>
> Nov 18 17:40:42 tornado kernel: e100: 0000:06:03.0: e100_eeprom_load: EEPROM
> corrupted
> Nov 18 17:40:42 tornado kernel: ACPI: PCI interrupt for device 0000:06:03.0
> disabled
> Nov 18 17:40:42 tornado kernel: e100: probe of 0000:06:03.0 failed with error -11
> Nov 18 17:40:43 tornado kernel: ACPI: PCI Interrupt 0000:04:00.0[A] -> GSI 17
> (level, low) -> IRQ 177
> Nov 18 17:40:43 tornado kernel: sky2 0000:04:00.0: unsupported chip type 0xff
> Nov 18 17:40:43 tornado kernel: ACPI: PCI interrupt for device 0000:04:00.0
> disabled
> Nov 18 17:40:43 tornado kernel: sky2: probe of 0000:04:00.0 failed with error -95
>
> I'm certain that both of these NIC's are OK as they work fine with
> CONFIG_PREEMPT_BKL not selected.
>
> With CONFIG_PREEMPT_BKL disabled and an otherwise identical config, the
> driver modules load up just fine.
>
> A known good kernel with this config was 2.6.14-rc5-mm1.
> I have backed out git-netdev-all but it made no difference, as well as backed
> out the e100 changes in -mm on 2.6.15-mm2, again no difference. So I suspect
> it's not a netdev driver problem.
>
> What else can I do to help narrow down the problem? What other trees or patches
> would be worth backing out to try and narrow it down?

I'd be suspecting the PCI changes firstly. That's gregkh-pci-*.

Conceivably git-acpi, but that hasn't changed in quite some time. In fact,
no ACPI changes since 2.6.14-rc5-mm1.

After that I don't know, sorry. Binary search time?

2005-11-18 07:57:07

by Sander

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

Andrew Morton wrote (ao):
> Reuben Farrelly <[email protected]> wrote:
> > On 18/11/2005 8:18 a.m., Andrew Morton wrote:
> > > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
> > >
> > > - reiser4 significantly updated
> > >
> > >
> > >
> > >
> > > Changes since 2.6.14-mm2:
> >
> > This has been one of the best -mm releases in a while. No problems compiling
> > or running - and so far nearly 18 hours uptime without any surprises.

I second this. It is stable under (my) load where the 2.6.14-mm releases
crashed. Most likely due to Reiser4, but couldn't capture anything yet
(netconsole troubles, have to investigate more) so not sure.

Reiser4 is very slow though, as is discussed on the reiserfs list atm.

> We'll have to try harder. -mm2 is up there now, to break everything again.

Nice :-)

--
Humilis IT Services and Solutions
http://www.humilis.net

2005-11-18 12:43:53

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Thursday 17 November 2005 14:18, Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1

Hi,

Any ideas why the mousedev module does not load automaticly with 15-rc1-mm1?

With 14-rc4-mm1 all is ok. With 15-rc1-mm1 I have no /dev/input which causes Xorg fun. The
mouse is connected to a usb hub and is detected when I unplug and replug it. I used make oldconfig
from a 14-rc4-mm1 config. Dist is debian amd64 unstable and is up to date.

When I manually modprobe mousedev things work as expected

DRM works fine and dri is enabled.

Ideas,
Ed Tomlinson

2005-11-18 20:53:01

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Fri, Nov 18, 2005 at 07:43:44AM -0500, Ed Tomlinson wrote:
> On Thursday 17 November 2005 14:18, Andrew Morton wrote:
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
>
> Hi,
>
> Any ideas why the mousedev module does not load automaticly with 15-rc1-mm1?

Are you using debian?
If so, what version of udev are you using? There are some known
reported problems with this, so I would suggest referring to the udev
bug list.

If you aren't using Debian, please let us know. Also, what version of
udev are you using?

thanks,

greg k-h

2005-11-18 20:58:47

by Reuben Farrelly

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On 18/11/2005 8:42 p.m., Andrew Morton wrote:
> Reuben Farrelly <[email protected]> wrote:
>> Hi,
>>
>> On 18/11/2005 8:18 a.m., Andrew Morton wrote:
>>> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.15-rc1/2.6.15-rc1-mm1
>>>
>>> - reiser4 significantly updated
>>>
>>>
>>>
>>>
>>> Changes since 2.6.14-mm2:
>> This has been one of the best -mm releases in a while. No problems compiling
>> or running - and so far nearly 18 hours uptime without any surprises.
>
> We'll have to try harder. -mm2 is up there now, to break everything again.
>
>> Following up on a posting from the last -mm release, I'm still seeing errors
>> loading multiple network drivers as modules (e100 and sky2) when
>> CONFIG_PREEMPT_BKL is enabled, with 2.6.14-mm1, 2.6.14-mm2 and now
>> 2.6.15-rc1-mm1. Mainline git doesn't exhibit the problem, so it's -mm specific.
>>
>> This is what is logged:
>>
>> Nov 18 17:40:42 tornado kernel: e100: 0000:06:03.0: e100_eeprom_load: EEPROM
>> corrupted
>> Nov 18 17:40:42 tornado kernel: ACPI: PCI interrupt for device 0000:06:03.0
>> disabled
>> Nov 18 17:40:42 tornado kernel: e100: probe of 0000:06:03.0 failed with error -11
>> Nov 18 17:40:43 tornado kernel: ACPI: PCI Interrupt 0000:04:00.0[A] -> GSI 17
>> (level, low) -> IRQ 177
>> Nov 18 17:40:43 tornado kernel: sky2 0000:04:00.0: unsupported chip type 0xff
>> Nov 18 17:40:43 tornado kernel: ACPI: PCI interrupt for device 0000:04:00.0
>> disabled
>> Nov 18 17:40:43 tornado kernel: sky2: probe of 0000:04:00.0 failed with error -95
>>
>> I'm certain that both of these NIC's are OK as they work fine with
>> CONFIG_PREEMPT_BKL not selected.
>>
>> With CONFIG_PREEMPT_BKL disabled and an otherwise identical config, the
>> driver modules load up just fine.
>>
>> A known good kernel with this config was 2.6.14-rc5-mm1.
>> I have backed out git-netdev-all but it made no difference, as well as backed
>> out the e100 changes in -mm on 2.6.15-mm2, again no difference. So I suspect
>> it's not a netdev driver problem.
>>
>> What else can I do to help narrow down the problem? What other trees or patches
>> would be worth backing out to try and narrow it down?
>
> I'd be suspecting the PCI changes firstly. That's gregkh-pci-*.
>
> Conceivably git-acpi, but that hasn't changed in quite some time. In fact,
> no ACPI changes since 2.6.14-rc5-mm1.
>
> After that I don't know, sorry. Binary search time?

Still looking into this one. It appears that backing out the gregkh-pci patches
in -rc1-mm2 made no difference.

I don't think I should be able to make this happen so easily though:

[root@tornado ~]# rmmod e100
[root@tornado ~]# rmmod sky2
[root@tornado ~]# strace modprobe e100
Unable to handle kernel NULL pointer dereference at virtual address 00000010
printing eip:
c0124fc7
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP
last sysfs file: /class/net/eth0/flags
Modules linked in: nfsd exportfs lockd sunrpc autofs4 lm85 hwmon_vid eeprom ipv6
binfmt_misc hw_random crc32 piix i2c_i801
CPU: 0
EIP: 0060:[<c0124fc7>] Not tainted VLI
EFLAGS: 00010202 (2.6.15-rc1-mm2-preempt)
EIP is at ptrace_check_attach+0x14/0xaf
eax: 00000001 ebx: 00000000 ecx: 00000001 edx: c0417c00
esi: 00000000 edi: 00000000 ebp: e93b0f9c esp: e93b0f90
ds: 007b es: 007b ss: 0068
Process strace (pid: 3386, threadinfo=e93b0000 task=ec4c0130)
Stack: 00000000 00000000 00000018 e93b0fb4 c0125840 00000000 00000018 00000000
4521bff4 e93b0000 c0102cb7 00000018 00000d3b 00000001 00000000 4521bff4
bf819eb8 0000001a 0000007b c010007b 0000001a ffffe410 00000073 00000202
Call Trace:
[<c0103c09>] show_stack+0x94/0xca
[<c0103dd2>] show_registers+0x17a/0x210
[<c0104008>] die+0x116/0x19d
[<c03394dd>] do_page_fault+0x1ed/0x63d
[<c01038af>] error_code+0x4f/0x54
[<c0125840>] sys_ptrace+0x50/0xb9
[<c0102cb7>] sysenter_past_esp+0x54/0x75
---------------------------
| preempt count: 00000002 ]
| 2 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)
.. [<c0338938>] .... _spin_lock_irqsave+0x11/0x71
.....[<c0103f36>] .. ( <= die+0x44/0x19d)

Code: 4c 5b 39 c0 a3 4c 5b 39 c0 c7 43 68 48 5b 39 c0 89 50 04 89 02 eb a8 55 89
e5 57 56 53 89 c3 89 d7 b8 00 7c 41 c0 e8 f5 39 21 00 <8b> 53 10 f6 c2 01 75 21
be fd ff ff ff b8 00 7c 41 c0 e8 43 3c
<3>Debug: sleeping function called from invalid context at
include/linux/rwsem.h:43
in_atomic():1, irqs_disabled():0
[<c0103c56>] dump_stack+0x17/0x19
[<c011a173>] __might_sleep+0x9c/0xae
[<c011da06>] profile_task_exit+0x16/0x49
[<c011f6f7>] do_exit+0x1c/0x47a
[<c010408f>] do_divide_error+0x0/0x9e
[<c03394dd>] do_page_fault+0x1ed/0x63d
[<c01038af>] error_code+0x4f/0x54
[<c0125840>] sys_ptrace+0x50/0xb9
[<c0102cb7>] sysenter_past_esp+0x54/0x75
---------------------------
| preempt count: 00000001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: strace[3386] exited with nonzero preempt_count 1!
---------------------------
| preempt count: 00000001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: soft lockup detected on CPU#0!

Pid: 3386, comm: strace
EIP: 0060:[<c0338b8c>] CPU: 0
EIP is at _write_lock_irqsave+0x65/0x7b
EFLAGS: 00000202 Not tainted (2.6.15-rc1-mm2-preempt)
EAX: 00ffffff EBX: 00000286 ECX: 00000000 EDX: 00000001
ESI: c0417c00 EDI: ec4c05ec EBP: e93b0e6c DS: 007b ES: 007b
CR0: 8005003b CR2: 00000010 CR3: 00456000 CR4: 000006d0
---------------------------
| preempt count: 00010001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: soft lockup detected on CPU#0!

Pid: 3386, comm: strace
EIP: 0060:[<c0338b8c>] CPU: 0
EIP is at _write_lock_irqsave+0x65/0x7b
EFLAGS: 00000202 Not tainted (2.6.15-rc1-mm2-preempt)
EAX: 00ffffff EBX: 00000286 ECX: 00000000 EDX: 00000001
ESI: c0417c00 EDI: ec4c05ec EBP: e93b0e6c DS: 007b ES: 007b
CR0: 8005003b CR2: 00000010 CR3: 00456000 CR4: 000006d0
---------------------------
| preempt count: 00010001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: soft lockup detected on CPU#0!

Pid: 3386, comm: strace
EIP: 0060:[<c0338b8c>] CPU: 0
EIP is at _write_lock_irqsave+0x65/0x7b
EFLAGS: 00000202 Not tainted (2.6.15-rc1-mm2-preempt)
EAX: 00ffffff EBX: 00000286 ECX: 00000000 EDX: 00000001
ESI: c0417c00 EDI: ec4c05ec EBP: e93b0e6c DS: 007b ES: 007b
CR0: 8005003b CR2: 00000010 CR3: 00456000 CR4: 000006d0
---------------------------
| preempt count: 00010001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: soft lockup detected on CPU#0!

Pid: 3386, comm: strace
EIP: 0060:[<c0338b8a>] CPU: 0
EIP is at _write_lock_irqsave+0x63/0x7b
EFLAGS: 00000202 Not tainted (2.6.15-rc1-mm2-preempt)
EAX: 00ffffff EBX: 00000286 ECX: 00000000 EDX: 00000001
ESI: c0417c00 EDI: ec4c05ec EBP: e93b0e6c DS: 007b ES: 007b
CR0: 8005003b CR2: 00000010 CR3: 00456000 CR4: 000006d0
---------------------------
| preempt count: 00010001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

BUG: soft lockup detected on CPU#0!

Pid: 3386, comm: strace
EIP: 0060:[<c0338b8a>] CPU: 0
EIP is at _write_lock_irqsave+0x63/0x7b
EFLAGS: 00000202 Not tainted (2.6.15-rc1-mm2-preempt)
EAX: 00ffffff EBX: 00000286 ECX: 00000000 EDX: 00000001
ESI: c0417c00 EDI: ec4c05ec EBP: e93b0e6c DS: 007b ES: 007b
CR0: 8005003b CR2: 00000010 CR3: 00456000 CR4: 000006d0
---------------------------
| preempt count: 00010001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<c03389cc>] .... _read_lock+0x10/0x6b
.....[<c0124fc7>] .. ( <= ptrace_check_attach+0x14/0xaf)

reuben

2005-11-18 21:14:05

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On 11/19/05, Greg KH <[email protected]> wrote:
> Are you using debian?
> If so, what version of udev are you using? There are some known
> reported problems with this, so I would suggest referring to the udev
> bug list.
>
In particular check the version requirements for udev - you need to be
on a version greater than or equal to 71. Sarge/stable has a really
old version. In particular I am running unstable as I had too many
funny errors (including this one) - but etch should be fine.

If running another distribution check this also as it is a real requirement.

To find the latest version of udev required check Documentation/Changes

Ian
--
Ian McDonald
http://wand.net.nz/~iam4
WAND Network Research Group
University of Waikato
New Zealand

2005-11-18 21:20:21

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

Reuben Farrelly <[email protected]> wrote:
>
> I don't think I should be able to make this happen so easily though:
>
>
> [root@tornado ~]# rmmod e100
> [root@tornado ~]# rmmod sky2
> [root@tornado ~]# strace modprobe e100
> Unable to handle kernel NULL pointer dereference at virtual address 00000010
> printing eip:
> c0124fc7
> *pde = 00000000
> Oops: 0000 [#1]
> PREEMPT SMP
> last sysfs file: /class/net/eth0/flags
> Modules linked in: nfsd exportfs lockd sunrpc autofs4 lm85 hwmon_vid eeprom ipv6
> binfmt_misc hw_random crc32 piix i2c_i801
> CPU: 0
> EIP: 0060:[<c0124fc7>] Not tainted VLI
> EFLAGS: 00010202 (2.6.15-rc1-mm2-preempt)
> EIP is at ptrace_check_attach+0x14/0xaf

This might help..

Begin forwarded message:

Date: Fri, 18 Nov 2005 18:07:43 +0000
From: Christoph Hellwig <[email protected]>
To: Badari Pulavarty <[email protected]>
Cc: [email protected], lkml <[email protected]>
Subject: Re: 2.6.15-rc1-mm1 panic in ptrace_check_attach()

On Fri, Nov 18, 2005 at 09:56:40AM -0800, Badari Pulavarty wrote:
> Hi Andrew,
>
> I am not sure if its already reported. I get panic in
> ptrace_check_attach() while trying to run UML on 2.6.15-rc1-mm1.
>
> Going to try 2.6.15-rc1-mm2 now.

Looks like 2.6.15-rc1-mm1 has total crap in ptrace_get_task_struct
(and it looks like my fault because I sent out a wrong patch).

The patch below should fix it:

Index: linux-2.6/kernel/ptrace.c
===================================================================
--- linux-2.6.orig/kernel/ptrace.c 2005-11-18 10:25:35.000000000 +0100
+++ linux-2.6/kernel/ptrace.c 2005-11-18 10:25:54.000000000 +0100
@@ -459,7 +459,7 @@
read_unlock(&tasklist_lock);
if (!child)
return ERR_PTR(-ESRCH);
- return 0;
+ return child;
}

#ifndef __ARCH_SYS_PTRACE

2005-11-18 23:16:25

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Friday 18 November 2005 16:14, Ian McDonald wrote:
> On 11/19/05, Greg KH <[email protected]> wrote:
> > Are you using debian?
> > If so, what version of udev are you using? There are some known
> > reported problems with this, so I would suggest referring to the udev
> > bug list.
> >
> In particular check the version requirements for udev - you need to be
> on a version greater than or equal to 71. Sarge/stable has a really
> old version. In particular I am running unstable as I had too many
> funny errors (including this one) - but etch should be fine.
>
> If running another distribution check this also as it is a real requirement.
>
> To find the latest version of udev required check Documentation/Changes

devinfo -v
udevinfo, version 074

dpkg -s
Package: udev
Status: install ok installed
Priority: extra
Section: admin
Installed-Size: 1072
Maintainer: Marco d'Itri <[email protected]>
Architecture: amd64
Version: 0.074-3

Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
bugs.

Thanks
Ed Tomlinson

2005-11-18 23:19:09

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Fri, Nov 18, 2005 at 06:16:01PM -0500, Ed Tomlinson wrote:
> On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > On 11/19/05, Greg KH <[email protected]> wrote:
> > > Are you using debian?
> > > If so, what version of udev are you using? There are some known
> > > reported problems with this, so I would suggest referring to the udev
> > > bug list.
> > >
> > In particular check the version requirements for udev - you need to be
> > on a version greater than or equal to 71. Sarge/stable has a really
> > old version. In particular I am running unstable as I had too many
> > funny errors (including this one) - but etch should be fine.
> >
> > If running another distribution check this also as it is a real requirement.
> >
> > To find the latest version of udev required check Documentation/Changes
>
> devinfo -v
> udevinfo, version 074
>
> dpkg -s
> Package: udev
> Status: install ok installed
> Priority: extra
> Section: admin
> Installed-Size: 1072
> Maintainer: Marco d'Itri <[email protected]>
> Architecture: amd64
> Version: 0.074-3
>
> Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> bugs.

Lots of things have changed in the input core since that kernel version.
Others have reported this same issue, see the thread from Ted Tso on
lkml a while ago when 2.6.15-rc1 came out.

thanks,

greg k-h

2005-11-18 23:35:20

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > On 11/19/05, Greg KH <[email protected]> wrote:
> > > Are you using debian?
> > > If so, what version of udev are you using? There are some known
> > > reported problems with this, so I would suggest referring to the udev
> > > bug list.
> > >
> > In particular check the version requirements for udev - you need to be
> > on a version greater than or equal to 71. Sarge/stable has a really
> > old version. In particular I am running unstable as I had too many
> > funny errors (including this one) - but etch should be fine.
> >
> > If running another distribution check this also as it is a real requirement.
> >
> > To find the latest version of udev required check Documentation/Changes
>
> devinfo -v
> udevinfo, version 074
>
> dpkg -s
> Package: udev
> Status: install ok installed
> Priority: extra
> Section: admin
> Installed-Size: 1072
> Maintainer: Marco d'Itri <[email protected]>
> Architecture: amd64
> Version: 0.074-3
>
> Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> bugs.

There does not seem to be anything that fits this reported as a debian bug. Where
is the udev bugs list?

TIA
Ed Tomlinson

2005-11-19 00:11:42

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Fri, Nov 18, 2005 at 06:35:11PM -0500, Ed Tomlinson wrote:
> On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> > On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > > On 11/19/05, Greg KH <[email protected]> wrote:
> > > > Are you using debian?
> > > > If so, what version of udev are you using? There are some known
> > > > reported problems with this, so I would suggest referring to the udev
> > > > bug list.
> > > >
> > > In particular check the version requirements for udev - you need to be
> > > on a version greater than or equal to 71. Sarge/stable has a really
> > > old version. In particular I am running unstable as I had too many
> > > funny errors (including this one) - but etch should be fine.
> > >
> > > If running another distribution check this also as it is a real requirement.
> > >
> > > To find the latest version of udev required check Documentation/Changes
> >
> > devinfo -v
> > udevinfo, version 074
> >
> > dpkg -s
> > Package: udev
> > Status: install ok installed
> > Priority: extra
> > Section: admin
> > Installed-Size: 1072
> > Maintainer: Marco d'Itri <[email protected]>
> > Architecture: amd64
> > Version: 0.074-3
> >
> > Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> > bugs.
>
> There does not seem to be anything that fits this reported as a debian bug. Where
> is the udev bugs list?

For Debian? I have no idea as I do not use it :)

For general udev issues/queries try the linux-hotplug-devel mailing
list.

Oh, and are you sure you actually have the proper module loaded?

thanks,

greg k-h

2005-11-19 01:24:50

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Friday 18 November 2005 18:51, Greg KH wrote:
> On Fri, Nov 18, 2005 at 06:35:11PM -0500, Ed Tomlinson wrote:
> > On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> > > On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > > > On 11/19/05, Greg KH <[email protected]> wrote:
> > > > > Are you using debian?
> > > > > If so, what version of udev are you using? There are some known
> > > > > reported problems with this, so I would suggest referring to the udev
> > > > > bug list.
> > > > >
> > > > In particular check the version requirements for udev - you need to be
> > > > on a version greater than or equal to 71. Sarge/stable has a really
> > > > old version. In particular I am running unstable as I had too many
> > > > funny errors (including this one) - but etch should be fine.
> > > >
> > > > If running another distribution check this also as it is a real requirement.
> > > >
> > > > To find the latest version of udev required check Documentation/Changes
> > >
> > > devinfo -v
> > > udevinfo, version 074
> > >
> > > dpkg -s
> > > Package: udev
> > > Status: install ok installed
> > > Priority: extra
> > > Section: admin
> > > Installed-Size: 1072
> > > Maintainer: Marco d'Itri <[email protected]>
> > > Architecture: amd64
> > > Version: 0.074-3
> > >
> > > Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> > > bugs.
> >
> > There does not seem to be anything that fits this reported as a debian bug. Where
> > is the udev bugs list?
>
> For Debian? I have no idea as I do not use it :)
>
> For general udev issues/queries try the linux-hotplug-devel mailing
> list.
>
> Oh, and are you sure you actually have the proper module loaded?

Think only the mousedev module is not loaded. Once I modprobe it the mouse works
and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
acting normal.

Ed

2005-11-19 01:40:11

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

> Think only the mousedev module is not loaded. Once I modprobe it the mouse works
> and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
> acting normal.
>
I was having this on one of my machines. I'll try and track this down
next week probably if it still occurs and use git-bisect on it if
nobody else does first though. I'll check out the thread mentioned as
well.

Ian
--
Ian McDonald
http://wand.net.nz/~iam4
WAND Network Research Group
University of Waikato
New Zealand

2005-11-19 01:42:16

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Fri, Nov 18, 2005 at 08:24:33PM -0500, Ed Tomlinson wrote:
> On Friday 18 November 2005 18:51, Greg KH wrote:
> > On Fri, Nov 18, 2005 at 06:35:11PM -0500, Ed Tomlinson wrote:
> > > On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> > > > On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > > > > On 11/19/05, Greg KH <[email protected]> wrote:
> > > > > > Are you using debian?
> > > > > > If so, what version of udev are you using? There are some known
> > > > > > reported problems with this, so I would suggest referring to the udev
> > > > > > bug list.
> > > > > >
> > > > > In particular check the version requirements for udev - you need to be
> > > > > on a version greater than or equal to 71. Sarge/stable has a really
> > > > > old version. In particular I am running unstable as I had too many
> > > > > funny errors (including this one) - but etch should be fine.
> > > > >
> > > > > If running another distribution check this also as it is a real requirement.
> > > > >
> > > > > To find the latest version of udev required check Documentation/Changes
> > > >
> > > > devinfo -v
> > > > udevinfo, version 074
> > > >
> > > > dpkg -s
> > > > Package: udev
> > > > Status: install ok installed
> > > > Priority: extra
> > > > Section: admin
> > > > Installed-Size: 1072
> > > > Maintainer: Marco d'Itri <[email protected]>
> > > > Architecture: amd64
> > > > Version: 0.074-3
> > > >
> > > > Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> > > > bugs.
> > >
> > > There does not seem to be anything that fits this reported as a debian bug. Where
> > > is the udev bugs list?
> >
> > For Debian? I have no idea as I do not use it :)
> >
> > For general udev issues/queries try the linux-hotplug-devel mailing
> > list.
> >
> > Oh, and are you sure you actually have the proper module loaded?
>
> Think only the mousedev module is not loaded. Once I modprobe it the mouse works
> and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
> acting normal.

Then you just need to make sure that module is loaded properly, which
doesn't sound like a udev issue :)

thanks,

greg k-h

2005-11-19 01:49:54

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

> > Think only the mousedev module is not loaded. Once I modprobe it the mouse works
> > and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
> > acting normal.
>
> Then you just need to make sure that module is loaded properly, which
> doesn't sound like a udev issue :)
>
My personal experience was that all my modules loaded automatically,
including mousedev and then mousedev doesn't anymore (but haven't
checked very latest kernel tree) and then you have to manually load.

Ian
--
Ian McDonald
http://wand.net.nz/~iam4
WAND Network Research Group
University of Waikato
New Zealand

2005-11-19 03:24:16

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Friday 18 November 2005 20:26, Greg KH wrote:
> On Fri, Nov 18, 2005 at 08:24:33PM -0500, Ed Tomlinson wrote:
> > On Friday 18 November 2005 18:51, Greg KH wrote:
> > > On Fri, Nov 18, 2005 at 06:35:11PM -0500, Ed Tomlinson wrote:
> > > > On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> > > > > On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > > > > > On 11/19/05, Greg KH <[email protected]> wrote:
> > > > > > > Are you using debian?
> > > > > > > If so, what version of udev are you using? There are some known
> > > > > > > reported problems with this, so I would suggest referring to the udev
> > > > > > > bug list.
> > > > > > >
> > > > > > In particular check the version requirements for udev - you need to be
> > > > > > on a version greater than or equal to 71. Sarge/stable has a really
> > > > > > old version. In particular I am running unstable as I had too many
> > > > > > funny errors (including this one) - but etch should be fine.
> > > > > >
> > > > > > If running another distribution check this also as it is a real requirement.
> > > > > >
> > > > > > To find the latest version of udev required check Documentation/Changes
> > > > >
> > > > > devinfo -v
> > > > > udevinfo, version 074
> > > > >
> > > > > dpkg -s
> > > > > Package: udev
> > > > > Status: install ok installed
> > > > > Priority: extra
> > > > > Section: admin
> > > > > Installed-Size: 1072
> > > > > Maintainer: Marco d'Itri <[email protected]>
> > > > > Architecture: amd64
> > > > > Version: 0.074-3
> > > > >
> > > > > Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> > > > > bugs.
> > > >
> > > > There does not seem to be anything that fits this reported as a debian bug. Where
> > > > is the udev bugs list?
> > >
> > > For Debian? I have no idea as I do not use it :)
> > >
> > > For general udev issues/queries try the linux-hotplug-devel mailing
> > > list.
> > >
> > > Oh, and are you sure you actually have the proper module loaded?
> >
> > Think only the mousedev module is not loaded. Once I modprobe it the mouse works
> > and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
> > acting normal.
>
> Then you just need to make sure that module is loaded properly, which
> doesn't sound like a udev issue :)

Then its a kernel problem.

It works with the _same_ user space in 2.6.14-rc4-mm1 and fails with 15-rc1-mm2...
This implies either something is broken in the kernel or udev is not doing its job with
the new (input?) changes. Since I switched to udev this summer I have not had to load
many modules certainly not mousedev.

I'll do a bit of experimenting this weekend to see if I cannot figure out more about what is
happening.

Thanks for the help
Ed

2005-11-21 00:00:06

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

> It works with the _same_ user space in 2.6.14-rc4-mm1 and fails with 15-rc1-mm2...
> This implies either something is broken in the kernel or udev is not doing its job with
> the new (input?) changes. Since I switched to udev this summer I have not had to load
> many modules certainly not mousedev.
>
> I'll do a bit of experimenting this weekend to see if I cannot figure out more about what is
> happening.
>
> Thanks for the help
> Ed
>
I would suggest using git-bisect to find the problem - I do know that
the problem was/is in Linus Torvalds tree. It was fine around 2.6.13
timeframe and broke sometime after this. I find git-bisect very
powerful for testing these types of bugs.

I was going to test myself in this way but the machine which
experienced the problem caught fire!! I know this sounds like the dog
ate my homework but unfortunately it was true and it was a machine I
borrowed too....

Ian
--
Ian McDonald
http://wand.net.nz/~iam4
WAND Network Research Group
University of Waikato
New Zealand

2005-11-21 00:49:05

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Fri, Nov 18, 2005 at 10:24:09PM -0500, Ed Tomlinson wrote:
> On Friday 18 November 2005 20:26, Greg KH wrote:
> > On Fri, Nov 18, 2005 at 08:24:33PM -0500, Ed Tomlinson wrote:
> > > On Friday 18 November 2005 18:51, Greg KH wrote:
> > > > On Fri, Nov 18, 2005 at 06:35:11PM -0500, Ed Tomlinson wrote:
> > > > > On Friday 18 November 2005 18:16, Ed Tomlinson wrote:
> > > > > > On Friday 18 November 2005 16:14, Ian McDonald wrote:
> > > > > > > On 11/19/05, Greg KH <[email protected]> wrote:
> > > > > > > > Are you using debian?
> > > > > > > > If so, what version of udev are you using? There are some known
> > > > > > > > reported problems with this, so I would suggest referring to the udev
> > > > > > > > bug list.
> > > > > > > >
> > > > > > > In particular check the version requirements for udev - you need to be
> > > > > > > on a version greater than or equal to 71. Sarge/stable has a really
> > > > > > > old version. In particular I am running unstable as I had too many
> > > > > > > funny errors (including this one) - but etch should be fine.
> > > > > > >
> > > > > > > If running another distribution check this also as it is a real requirement.
> > > > > > >
> > > > > > > To find the latest version of udev required check Documentation/Changes
> > > > > >
> > > > > > devinfo -v
> > > > > > udevinfo, version 074
> > > > > >
> > > > > > dpkg -s
> > > > > > Package: udev
> > > > > > Status: install ok installed
> > > > > > Priority: extra
> > > > > > Section: admin
> > > > > > Installed-Size: 1072
> > > > > > Maintainer: Marco d'Itri <[email protected]>
> > > > > > Architecture: amd64
> > > > > > Version: 0.074-3
> > > > > >
> > > > > > Interestingly the same udev works fine with 14-rc4-mm1. I'll check the debian
> > > > > > bugs.
> > > > >
> > > > > There does not seem to be anything that fits this reported as a debian bug. Where
> > > > > is the udev bugs list?
> > > >
> > > > For Debian? I have no idea as I do not use it :)
> > > >
> > > > For general udev issues/queries try the linux-hotplug-devel mailing
> > > > list.
> > > >
> > > > Oh, and are you sure you actually have the proper module loaded?
> > >
> > > Think only the mousedev module is not loaded. Once I modprobe it the mouse works
> > > and the /dev/input/mice appears. The mouse works normally with all buttons and wheels
> > > acting normal.
> >
> > Then you just need to make sure that module is loaded properly, which
> > doesn't sound like a udev issue :)
>
> Then its a kernel problem.

Or a broken userspace configuration that just happened to work
previously :)

I really don't know which one this is, but as it seems it only is
showing up on Debian systems...

thanks,

greg k-h

2005-11-22 20:47:51

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

> Or a broken userspace configuration that just happened to work
> previously :)
>
> I really don't know which one this is, but as it seems it only is
> showing up on Debian systems...
>
> thanks,
>
> greg k-h
>
OK. Have got quite a bit further in tracking this down and will now
log with Debian somehow.

It is definitely userspace but I don't know enough about that to say
whether it is udev (Debian rules) or something else.

I used git-bisect with 2.6.14 and 2.6.15-rc1 as start and end points.
In the end the patch which caused it to break was the final one - just
altering the Makefile with the version number. I've reverified as well
by going back to 2.6.15-rc1 and just altering it to 2.6.14 and the
problem (mousedev not loaded automatically and no /dev/input /mice)
disappears. I almost thought I'd made a mistake in my testing until I
manually altered that makefile!!

It is definitely not a kernel issue in my opinion.

Ian
--
Ian McDonald
http://wand.net.nz/~iam4
WAND Network Research Group
University of Waikato
New Zealand

2005-11-22 21:00:36

[permalink] [raw]

Subject: Re: 2.6.15-rc1-mm1

On Wed, Nov 23, 2005 at 09:47:49AM +1300, Ian McDonald wrote:
> It is definitely not a kernel issue in my opinion.

Thank you for following up on this and letting us know.

greg k-h

2005-11-22 22:05:23