udftools confuses hal

[Sergio Callegari]

Binary package hint: udftools

Bug in udftools or possibly in pktcdvd kernel module.

Using Ubuntu Intrepid 32bit on an older PC with an ATHLON CPU and Samsung SH-S202J dvdwriter.

with Linux version 2.6.27-11-generic (buildd@vernadsky) (gcc version 4.3.2 (Ubuntu 4.3.2-1ubuntu11) ) #1 SMP Wed Apr 1 20:57:48 UTC 2009

pktcdvd module
filename: /lib/modules/2.6.27-11-generic/kernel/drivers/block/pktcdvd.ko
license: GPL
author: Jens Axboe <email address hidden>
description: Packet writing layer for CD/DVD drives
srcversion: 4AF2F07A4806489F964C395
depends: cdrom
vermagic: 2.6.27-11-generic SMP mod_unload modversions 586

and udftools 1.0.0b3-14

Following problems happen:

1) When booting with a CDROM or DVDROM in drive, Xorg starts with no keyboard and mouse, claiming that it was impossible to get them from hal

2) When 1 happens dvdwriters makes seek-like noise. Data content on DVD can be lost.

3) Regardless of 1 and 2, whenever udftools service starts, kernel complains and kernel messages report sysfs: duplicate filename '252:0' can not be created and dump of kernel modules plus id: 8403, comm: pktsetup Tainted: P W 2.6.27-11-generic # and then a backtrace:

[ 1320.478690] [<c0131e65>] warn_slowpath+0x65/0x90
[ 1320.478703] [<c024e372>] ? idr_get_empty_slot+0xe2/0x270
[ 1320.478711] [<c024e584>] ? ida_get_new_above+0x84/0x1c0
[ 1320.478716] [<c01c772e>] ? find_inode+0xe/0x70
[ 1320.478721] [<c0200450>] ? sysfs_ilookup_test+0x0/0x20
[ 1320.478726] [<c0200450>] ? sysfs_ilookup_test+0x0/0x20
[ 1320.478730] [<c0254bd2>] ? strcmp+0x12/0x40
[ 1320.478737] [<c0200779>] ? sysfs_find_dirent+0x29/0x40
[ 1320.478742] [<c0200838>] ? __sysfs_add_one+0x18/0x90
[ 1320.478746] [<c0200a0e>] sysfs_add_one+0x4e/0x50
[ 1320.478750] [<c0201899>] sysfs_do_create_link+0xd9/0x130
[ 1320.478756] [<c0201927>] sysfs_create_link+0x17/0x20
[ 1320.478760] [<c02c2aa9>] device_add+0x1d9/0x400
[ 1320.478766] [<c02009d8>] ? sysfs_add_one+0x18/0x50
[ 1320.478771] [<c02c2ce7>] device_register+0x17/0x20
[ 1320.478775] [<c02c2d9a>] device_create_vargs+0xaa/0xc0
[ 1320.478779] [<c02c2ddd>] device_create+0x2d/0x30
[ 1320.478786] [<f902207f>] pkt_setup_dev+0x1ef/0x330 [pktcdvd]
[ 1320.478804] [<f902239f>] pkt_ctl_ioctl+0x15f/0x1f0 [pktcdvd]
[ 1320.478813] [<c01bf3c1>] vfs_ioctl+0x81/0x90
[ 1320.478820] [<c0185801>] ? generic_perform_write_2copy+0x121/0x320
[ 1320.478826] [<c01bf556>] do_vfs_ioctl+0x66/0x200
[ 1320.478830] [<c0215008>] ? cap_file_ioctl+0x8/0x10
[ 1320.478840] [<c01bf75b>] sys_ioctl+0x6b/0x70
[ 1320.478845] [<c0103f7b>] sysenter_do_call+0x12/0x2f
[ 1320.478850] [<c0185801>] ? generic_perform_write_2copy+0x121/0x320
[ 1320.478856] =======================
[ 1320.478859] ---[ end trace a049394c432b0fd9 ]---

this happens on pktsetup, even if invoked on the command line rather than with the udftools service.

3) Finally, pktcdvd module often complains on kernel syslog that it could not do optimal power calibration.

I do not know if 1) 2) 3) and 4) are different bugs and/or if they should be reported separately. 1) 2) and 3) appear strongly related.

Indeed all points look serious. The possibility of data loss and the suspect that the drive might get damaged by the continuous seek is worring. I wonder if the data loss may be induced by writing with no or wrong power calibration.

[Sergio Callegari (callegar)]

Ok...:

char device 252:0 is hidraw0, which I guess is the system keyboard. And that is why hal gets confused.
And that is why the dvdwriter probably received crazy commands that cook the disk that is inside (and maybe the drive too).

There is a long thread on this on the linux kernel mailing list.
It seems to affect 2.6.27.

Cannot see if it was fixed later on, but if it was, the ubuntu kernel doesn't have the fix.

This looks serious:
SURE DATA LOSS (after this treatement the DVD+RW that is in drive not only cannot be read, but cannot be formatted anymore) and possible hardware damage (drive spins and seeks in a crazy way).

Please adopt one of the following in order of priority:

1) Provide a true fix for the kernel or pktsetup if it is available.

2) Make sure that the udftools package contains HUGE warnings in the documentation.

[Sergio Callegari (callegar)]

In fact a kernel bug and not a bug in the udftools package.

Seems to be (partially) fixed in jaunty.
I say partially, because when udftools is restarted, the kernel still complains:

[ 978.277135] ------------[ cut here ]------------
[ 978.277139] WARNING: at /home/kernel-ppa/mainline/build/fs/sysfs/dir.c:462 sysfs_add_one+0x49/0x50()
[ 978.277144] Hardware name: System Name
[ 978.277146] sysfs: duplicate filename 'pktcdvd0' can not be created
[ 978.277149] Modules linked in: binfmt_misc bridge stp bnep radeon drm vboxnetflt vboxdrv video output ipt_REJECT ipt_LOG xt_recent xt_limit xt_tcpudp xt_state ipt_addrtype ip6table_filter ip6_tables nf_nat_irc nf_conntrack_irc nf_nat_ftp nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_conntrack_ftp nf_conntrack iptable_filter ip_tables x_tables dm_crypt psmouse lp rsrc_nonstatic pcmcia_core snd_cmipci gameport snd_pcm_oss snd_mixer_oss snd_pcm snd_page_alloc snd_opl3_lib snd_hwdep snd_mpu401_uart i2c_viapro snd_seq_dummy snd_seq_oss via_ircc ppdev snd_seq_midi pcspkr snd_rawmidi snd_seq_midi_event irda snd_seq crc_ccitt usblp snd_timer snd_seq_device parport_pc via_agp snd parport shpchp agpgart soundcore 8139too usbhid 8139cp mii aic7xxx scsi_transport_spi raid10 raid456 async_xor async_memcpy async_tx xor raid1 raid0 usb_storage multipath linear
[ 978.277230] Pid: 5878, comm: pktsetup Not tainted 2.6.29-02062901-generic #02062901
[ 978.277233] Call Trace:
[ 978.277246] [<c013a736>] warn_slowpath+0x86/0xa0
[ 978.277256] [<c02c1200>] ? sub_alloc+0x0/0x150
[ 978.277260] [<c02c1261>] ? sub_alloc+0x61/0x150
[ 978.277266] [<c02c13f0>] ? idr_get_empty_slot+0xa0/0x140
[ 978.277270] [<c02c150b>] ? ida_get_new_above+0x7b/0x180
[ 978.277276] [<c01d0bed>] ? find_inode+0x3d/0x70
[ 978.277280] [<c020bc00>] ? sysfs_ilookup_test+0x0/0x20
[ 978.277284] [<c020bf21>] ? sysfs_find_dirent+0x21/0x30
[ 978.277288] [<c020bf4d>] ? __sysfs_add_one+0x1d/0x90
[ 978.277292] [<c01d22b9>] ? ilookup5+0x39/0x50
[ 978.277296] [<c020c009>] sysfs_add_one+0x49/0x50
[ 978.277300] [<c020c929>] create_dir+0x49/0x90
[ 978.277304] [<c020c99b>] sysfs_create_dir+0x2b/0x50
[ 978.277310] [<c02c1eee>] create_dir+0x1e/0x50
[ 978.277314] [<c02c2296>] kobject_add_internal+0x86/0x100
[ 978.277318] [<c02c243c>] kobject_add_varg+0x4c/0x50
[ 978.277324] [<c02c674a>] ? string+0x2a/0xa0
[ 978.277327] [<c02c250f>] kobject_add+0x2f/0x60
[ 978.277334] [<c034ad33>] device_add+0x93/0x280
...

[pirx67 (pirx67)]

Still similar bug present in Karmic. Filed that as bug https://bugs.launchpad.net/ubuntu/+source/linux/+bug/526774 against the kernel.

I hope it will be fixed in Lucid Lynx.