2.4.17pre2: ieee1394 sd hotplug problems (Oops)

2.4.17pre2: ieee1394 sd hotplug problems (Oops)

[ki]

Hello all,

I am using an 80GB ide disk together with an 
ide <> firewire converter card. (Oxford chip)

usually I am doing the following: (no SCSI hd present but sd_mod loaded)

 * connect & power on hd
 * insmod ieee1394, ohci1394 and sbp2
 * look for the partition info in the syslog, looks like this:

Dec 13 20:24:29 ki_pc2 kernel: scsi0 : IEEE-1394 SBP-2 protocol driver
Dec 13 20:24:29 ki_pc2 kernel:   Vendor: MAXTOR 4  Model: K080H4            Rev:     
Dec 13 20:24:29 ki_pc2 kernel:   Type:   Direct-Access                      ANSI SCSI revision: 06
Dec 13 20:24:29 ki_pc2 kernel: Attached scsi disk sdd at scsi0, channel 0, id 0, lun 0
Dec 13 20:24:29 ki_pc2 kernel: SCSI device sdd: 156301487 512-byte hdwr sectors (80026 MB)
Dec 13 20:24:29 ki_pc2 kernel:  sdd: sdd1
Dec 13 20:24:58 ki_pc2 /sbin/hotplug: no runnable /etc/hotplug/ieee1394.agent is installed

note: sdd,  not sda as expected

 * use the disk - works ok, hdparm says  ~12MB/sec
 * rmmod sbp2 (and perhaps also the other 1394 modules) 
 * disconnect hard drive.
 etc..

The following happens: on  'insmod sbp2' the disk is found
but the minor device is incremented every time (now I am at /dev/sde :-( )

Trying to open one of the lower (non-present) minors results in an Oops:

in sd.c: at about line 470
        }
        /*
         * The following code can sleep.
         * Module unloading must be prevented
         */
        SDev = rscsi_disks[target].device;

	^^^^^ this becomes zero for the lower minor devices.

        if (SDev->host->hostt->module) ---- Oops

So it seems that the rmmod sbp2 does not properly clean up
the internal bookkeeping in sd.c when being unloaded.

I am not sure what happens if scsi-remove-single-device is done instead
of rmmod sbp2.

OTOH /proc/scsi/scsi correctly shows the present devices
 
there are a few /dev/sd[a-?]  left so I dont have to reboot every time but
the situation is a bit non-optimal for now.

Greetings,
karl

-- 
Karl Kiniger	    ki@kretz.co.at
Kretztechnik AG     OE5KVN
Tiefenbach 15
A-4871 Zipf	    Tel: (++43) 7682-3800-710  Fax (++43) 7682-3800-47

Re: 2.4.17pre2: ieee1394 sd hotplug problems (Oops)

[Olivier Daigle]

You should better ask your question on the linux1394 reflector.

http://sourceforge.net/mail/?group_id=2252

Someone already posted a question regarding the problem you have.

Regards,

   Olivier

On Thursday 13 December 2001 17:31, ki@kretz.co.at wrote:
> Hello all,
>
> I am using an 80GB ide disk together with an
> ide <> firewire converter card. (Oxford chip)
>
> usually I am doing the following: (no SCSI hd present but sd_mod loaded)
>
>  * connect & power on hd
>  * insmod ieee1394, ohci1394 and sbp2
>  * look for the partition info in the syslog, looks like this:
>
> Dec 13 20:24:29 ki_pc2 kernel: scsi0 : IEEE-1394 SBP-2 protocol driver
> Dec 13 20:24:29 ki_pc2 kernel:   Vendor: MAXTOR 4  Model: K080H4           
> Rev: Dec 13 20:24:29 ki_pc2 kernel:   Type:   Direct-Access                
>      ANSI SCSI revision: 06 Dec 13 20:24:29 ki_pc2 kernel: Attached scsi
> disk sdd at scsi0, channel 0, id 0, lun 0 Dec 13 20:24:29 ki_pc2 kernel:
> SCSI device sdd: 156301487 512-byte hdwr sectors (80026 MB) Dec 13 20:24:29
> ki_pc2 kernel:  sdd: sdd1
> Dec 13 20:24:58 ki_pc2 /sbin/hotplug: no runnable
> /etc/hotplug/ieee1394.agent is installed
>
> note: sdd,  not sda as expected
>
>  * use the disk - works ok, hdparm says  ~12MB/sec
>  * rmmod sbp2 (and perhaps also the other 1394 modules)
>  * disconnect hard drive.
>  etc..
>
> The following happens: on  'insmod sbp2' the disk is found
> but the minor device is incremented every time (now I am at /dev/sde :-( )
>
> Trying to open one of the lower (non-present) minors results in an Oops:
>
> in sd.c: at about line 470
>         }
>         /*
>          * The following code can sleep.
>          * Module unloading must be prevented
>          */
>         SDev = rscsi_disks[target].device;
>
> 	^^^^^ this becomes zero for the lower minor devices.
>
>         if (SDev->host->hostt->module) ---- Oops
>
> So it seems that the rmmod sbp2 does not properly clean up
> the internal bookkeeping in sd.c when being unloaded.
>
> I am not sure what happens if scsi-remove-single-device is done instead
> of rmmod sbp2.
>
> OTOH /proc/scsi/scsi correctly shows the present devices
>
> there are a few /dev/sd[a-?]  left so I dont have to reboot every time but
> the situation is a bit non-optimal for now.
>
> Greetings,
> karl

-- 
Olivier Daigle <odaigle@harfangmicro.com>

Re: 2.4.17pre2: ieee1394 sd hotplug problems (Oops)

[ki]

On Thu, Dec 13, 2001 at 05:43:58PM -0500, Olivier Daigle wrote:
> You should better ask your question on the linux1394 reflector.
> 
> http://sourceforge.net/mail/?group_id=2252
> 
> Someone already posted a question regarding the problem you have.
> 
> Regards,
> 
>    Olivier

Thanks for the pointer, it seems to be exactly the same problem.

No solution available there - unfortunately.

karl
-- 
Karl Kiniger	    ki@kretz.co.at
Kretztechnik AG     OE5KVN
Tiefenbach 15
A-4871 Zipf	    Tel: (++43) 7682-3800-710  Fax (++43) 7682-3800-47