Hi Hans,
oh - I read your patch the wrong way round - your fix is indeed contained
within the Ubuntu 9.04 kernel, I erroneously assumed that the handling of
IVTV_IOC_DMA_FRAME has to be within 'ivtv_serialized_ioctl()'.
But then the fact remains that the decoder crash occurs also with your
patch. The crash only happened once for me - here is the excerpt from the
syslog:
BUG: unable to handle kernel paging request at acd95ca8
IP: [<f7f24d87>] ivtv_decoder_ioctls+0x6b7/0x870 [ivtv]
*pde = 30e7f067 *pte = 2321b065
Oops: 0003 [#1] SMP
last sysfs file: /sys/devices/pci0000:00/0000:00:13.4/pools
Dumping ftrace buffer:
(ftrace buffer empty)
Modules linked in: binfmt_misc radeon drm vmnet vmblock vmci vmmon video output
cx8800 cx88xx videobuf_dvb bttv ir_common
lirc_dev dm_crypt powernow_k8 ivtvfb it87 hwmon_vid tuner_simple tuner_types
msp3400 saa7127 saa7115 tuner qt1010 snd_usb
dummy snd_hda_intel compat_ioctl32 snd_seq_oss i2c_algo_bit snd_seq_midi
dvb_usb_gl861 cx2341x snd_pcm_oss snd_mixer_oss
at snd_seq_midi_event snd_pcm dvb_core tveeprom usbhid snd_seq snd_seq_device
k8temp ppdev ati_agp agpgart i2c_piix4 pcsp
e shpchp parport_pc parport usblp usb_storage firewire_ohci firewire_core
crc_itu_t ohci1394 ieee1394 ehci_hcd ohci_hcd r
or
Pid: 12834, comm: mythfrontend.re Not tainted (2.6.28-4-generic #11-Ubuntu)
System Product Name
EIP: 0060:[<f7f24d87>] EFLAGS: 00210246 CPU: 1
EIP is at ivtv_decoder_ioctls+0x6b7/0x870 [ivtv]
EAX: f5f00079 EBX: f4ba39e0 ECX: acd95ca8 EDX: 00000000
ESI: 000051d4 EDI: f5f0f288 EBP: f5ea9f10 ESP: f5ea9d88
DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
Process mythfrontend.re (pid: 12834, ti=f5ea8000 task=f5c7f110 task.ti=f5ea8000)
Stack:
f5ea9ec4 c5f1d63c 00000001 00000800 acd95ca8 f5ea9de0 f5f00000 f5ea9dc4
c01500cb f5ea9ddc 0c3a629d 00000000 0027ac6c f5ea9dc4 c0113913 f5ea9df0
c01537fa f5ea9e0c c012e08a f5ea9ddc f5ea9dec 00200086 f5f003cc 00000000
Call Trace:
[<c01500cb>] ? getnstimeofday+0x4b/0x100
[<c0113913>] ? lapic_next_event+0x13/0x20
[<c01537fa>] ? clockevents_program_event+0x9a/0x150
[<c012e08a>] ? rebalance_domains+0xda/0x150
[<f7f2851f>] ? ivtv_queue_move+0x23f/0x280 [ivtv]
[<c012e134>] ? run_rebalance_domains+0x34/0xc0
[<f7f25555>] ? dma_post+0x215/0x360 [ivtv]
[<c01500cb>] ? getnstimeofday+0x4b/0x100
[<f7f28639>] ? ivtv_dequeue+0x59/0x70 [ivtv]
[<f7f24f92>] ? ivtv_serialized_ioctl+0x52/0x370 [ivtv]
[<f7f252ea>] ? ivtv_v4l2_ioctl+0x3a/0x60 [ivtv]
[<f7f252b0>] ? ivtv_v4l2_ioctl+0x0/0x60 [ivtv]
[<c01c3728>] ? vfs_ioctl+0x28/0x90
[<c01c3c0e>] ? do_vfs_ioctl+0x5e/0x200
[<c01c3e13>] ? sys_ioctl+0x63/0x70
[<c0103f6b>] ? sysenter_do_call+0x12/0x2f
Code: fe ff ff 81 c7 7c f2 00 00 a5 a5 a5 8b 85 90 fe ff ff 8b 95 90 fe ff ff
83 c0 79 f0 80 4a 79 40 8b 75 b4 31 d2 8b 89 ff ff 8b bd 90 fe ff ff 8b
EIP: [<f7f24d87>] ivtv_decoder_ioctls+0x6b7/0x870 [ivtv] SS:ESP 0068:f5ea9d88
---[ end trace 9976b7b9f13cea3d ]---
ivtv0: All encoder VBI stream buffers are full. Dropping data.
ivtv0: Cause: the application is not reading fast enough.
ivtv0: All encoder MPG stream buffers are full. Dropping data.
...
Regards,
Johannes
Hans Verkuil wrote:
> On Tuesday 28 April 2009 21:58:19 Johannes Reisinger wrote:
>> Hi Hans,
>>
>> sorry to bother you again - but maybe you can help me:
>>
>> I recently installed the new Ubuntu 9.04 release containing kernel
>> 2.6.28-11. The ChangeLog contains the info, that upstream kernel changes
>> from 2.6.28.8 were incorporated, amongst others the line:
>> V4L: ivtv: fix decoder crash regression
>>
>> I thought this was the patch for the problem you solved in February.
>> But after installing the new release, I had once a crash with the same
>> message as previously ('unable to handle kernel paging request'). It is
>> not as reproducible as before, but it seems to be the same error.
>>
>> I had a look at 'ivtv-ioctl.c' and recognized that the code sequence
>> 'case IVTV_IOC_DMA_FRAME: ... return ivtv_decoder_ioctls(...' is still in
>> function 'ivtv_default()' and not in 'ivtv_serialized_ioctl()'.
>>
>> Am I right in assuming that the Ubuntu guys have failed to incorporate
>> the patch or may I have missed something?
>
> I downloaded the vanilla 2.6.28.8 kernel and IVTV_IOC_DMA_FRAME and friends
> are definitely handled in ivtv_default, so it does indeed seem that Ubuntu
> didn't merge that patch.
>
> Regards,
>
> Hans
>
>> Best Regards,
>> Johannes
>>
>> Hans Verkuil wrote:
>>> On Saturday 14 February 2009 23:46:36 [email protected] wrote:
>>>> Hi Hans,
>>>>
>>>> you seem to have solved the problem!!
>>>> I am not able to reproduce the crashes with your latest version.
>>>> Because I usually avoid to use development drivers on my system:
>>>> Is it possible to apply your latest patch also to the ivtv-ioctl.c
>>>> of a 2.6.27 or 2.6.28 kernel?
>>> Sure, the attached patch applies nicely to 2.6.27 and 2.6.28.
>>>
>>> Thanks for your help!
>>>
>>> Regards,
>>>
>>> Hans
>
>
>
_______________________________________________
ivtv-users mailing list
[email protected]
http://ivtvdriver.org/mailman/listinfo/ivtv-users
