Re: [OOps] 2.6.11-rc3 rmmod ide-scsi

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

Hi Florian,

Thanks for the report, addition of MRW support to cdrom.c exposed
bugs in ide-scsi locking.  ide-scsi is a "virtual" host driver which is a
bridge between IDE and SCSI subsystems so fixing ide-scsi requires
assuring that locking is correct both for IDE and SCSI "side".

Required changes consist of:
* adding reference counting to ide-scsi (now in ide-dev-2.6 tree)
* fixing various races in IDE device drivers to make conversion
  to driver model possible (now in ide-dev-2.6 tree)
* converting IDE device drivers to driver model (posted for review)
* fixing reference counting w.r.t. SCSI subsystem (in testing)

With all the above changes I can rmmod ide-scsi safely.
In short: this will hopefully get fixed in -mm tree soon.

Regards,
Bartlomiej

On Tue, 8 Feb 2005 18:55:29 +0100, Florian Lohoff <flo@rfc822.org> wrote:
> 
> Hi,
> got this oops while unloading ide-scsi (rmmod segfaulted)
> 
> UP P4 1.7Ghz non Preempt, No HT, IDE DVD/CD-RW, IDE Disk, Vanilla
> Kernel, (-chaos -> Debian make-kpkg --append-to-version=-chaos)
> 
> Linux chaos 2.6.11-rc3-chaos #1 Fri Feb 4 23:27:57 CET 2005 i686 GNU/Linux
> 
> Unable to handle kernel NULL pointer dereference at virtual address 000001e0
>  printing eip:
> e089e539
> *pde = 00000000
> Oops: 0000 [#1]
> Modules linked in: nls_iso8859_1 isofs autofs4 nfs lockd sunrpc snd_ens1371 snd_rawmidi snd_ac97_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd snd_page_alloc e100 i2c_i801 i2c_core evdev usbhid uhci_hcd usbcore intel_agp agpgart sr_mod cdrom dm_mod ide_scsi scsi_mod eepro100 mii
> CPU:    0
> EIP:    0060:[<e089e539>]    Not tainted VLI
> EFLAGS: 00010002   (2.6.11-rc3-chaos)
> EIP is at idescsi_queue+0x119/0x400 [ide_scsi]
> eax: 00000000   ebx: c0971a40   ecx: e08c4a30   edx: dedf8b00
> esi: 00001388   edi: c044e234   ebp: dedf8b54   esp: c910dc94
> ds: 007b   es: 007b   ss: 0068
> Process rmmod (pid: 11757, threadinfo=c910c000 task=c9eec510)
> Stack: dffef380 00000020 00000010 c0971a40 dedf8b00 d9546cb0 dedf8b68 deffcd80
>        000001d4 c044e234 00000293 dedf8b00 ded25400 00000000 e08c47cf dedf8b00
>        e08c4a30 e08c76d0 dedf8b88 dedf8b00 df692c00 ded25400 c1583030 e08caa9a
> Call Trace:
>  [<e08c47cf>] scsi_dispatch_cmd+0x16f/0x250 [scsi_mod]
>  [<e08c4a30>] scsi_done+0x0/0x30 [scsi_mod]
>  [<e08c76d0>] scsi_times_out+0x0/0xc0 [scsi_mod]
>  [<e08caa9a>] scsi_request_fn+0x1ca/0x380 [scsi_mod]
>  [<c0233858>] __elv_add_request+0x78/0xc0
>  [<c0236446>] blk_insert_request+0xa6/0xd0
>  [<e08c97b8>] scsi_insert_special_req+0x38/0x40 [scsi_mod]
>  [<e08c9a28>] scsi_wait_req+0x68/0xa0 [scsi_mod]
>  [<e08c9960>] scsi_wait_done+0x0/0x60 [scsi_mod]
>  [<e08a3462>] sr_do_ioctl+0x92/0x2a0 [sr_mod]
>  [<e08a3155>] sr_packet+0x25/0x40 [sr_mod]
>  [<e08e7a75>] cdrom_get_disc_info+0x65/0xb0 [cdrom]
>  [<e08e372b>] cdrom_mrw_exit+0x1b/0x70 [cdrom]
>  [<e08e3323>] unregister_cdrom+0x83/0xb0 [cdrom]
>  [<e08a319a>] sr_kref_release+0x2a/0x50 [sr_mod]
>  [<e08a3170>] sr_kref_release+0x0/0x50 [sr_mod]
>  [<c01bf769>] kref_put+0x39/0xa0
>  [<c01bed1f>] kobject_put+0x1f/0x30
>  [<e08a31f9>] sr_remove+0x39/0x4b [sr_mod]
>  [<e08a3170>] sr_kref_release+0x0/0x50 [sr_mod]
>  [<c022e946>] device_release_driver+0x86/0x90
>  [<c022ebc4>] bus_remove_device+0x64/0xb0
>  [<c022d8cd>] device_del+0x5d/0xa0
>  [<e08cda1f>] scsi_remove_device+0x4f/0xb0 [scsi_mod]
>  [<e08ccaca>] scsi_forget_host+0x2a/0x50 [scsi_mod]
>  [<e08c5319>] scsi_remove_host+0x19/0x80 [scsi_mod]
>  [<e089e2dd>] idescsi_cleanup+0x4d/0x60 [ide_scsi]
>  [<c024b95e>] ide_unregister_driver+0x5e/0x90
>  [<c012f7f8>] try_stop_module+0x28/0x30
>  [<e089edef>] exit_idescsi_module+0xf/0x11 [ide_scsi]
>  [<c012f9b4>] sys_delete_module+0x144/0x180
>  [<c0147500>] do_munmap+0xf0/0x160
>  [<c01475b4>] sys_munmap+0x44/0x70
>  [<c0102f33>] syscall_call+0x7/0xb
> Code: 8b 54 24 3c 8b 42 64 89 53 2c 89 43 14 89 43 0c 8b 4c 24 40 89 4b 30 a1 a0 0d 35 c0 8b 72 3c 01 f0 89 43 38 8b 7c 24 24 8b 47 20 <8b> 80 e0 01 00 00 a8 01 74 05 0f ba 6b 34 02 8b 43 1c 89 44 24
> 
> --
> Florian Lohoff                  flo@rfc822.org             +49-171-2280134
>                         Heisenberg may have been here.
> 
> 
>