Re: Problem hotplugging USB flashdrive (race?)

[Patrick Mansfield <patmans@us.ibm.com>]

On Tue, Jan 06, 2004 at 12:47:12AM +0100, Olaf Hering wrote:
>  On Fri, Jan 02, Bertrik Sikken wrote:
> 
> > I don't see why doing something from usb.agent or scsi.agent is
> > obviously wrong. As far as I understand, plugging in the device
> > causes a USB hotplug event which triggers loading of the usb-storage
> > module. This in turn fires off a SCSI hotplug event which should
> > load the sd_mod module and create a block device. From then on,
> > its up to the user (me) to mount the device.

I can't find the email you're replying to (in my inbox or the archives,
anything dated Jan 02 from Bertrik).

> I see the same. The reason is probably in scsi_probe_and_add_lun(). It
> calls scsi_alloc_sdev() which seems to generate the event, and later it
> calls scsi_probe_lun() which seems to set the type.

scsi_add_lun() calls scsi_sysfs_add_sdev() for the scsi_device after it
has set the type, and there we have the race between the hotplug event
generated via the call to device_add() and the addition of the attributes
for multiple calls to attr_add().

The type is initialized to -1 in scsi_alloc_sdev, so at worst you might
see a -1 there, but that would be a bug.

Are you hitting the problem even after adding the sleep 1 hack?

-- Patrick Mansfield


-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel