[SCSI] ch: fix ch_remove oops
authorFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Sat, 26 Jul 2008 14:25:43 +0000 (23:25 +0900)
committerJames Bottomley <James.Bottomley@HansenPartnership.com>
Sat, 26 Jul 2008 19:17:47 +0000 (15:17 -0400)
commit3d164fb09bb5cb8a223eddf634fc0d355714fcfe
treee0e52cbc911dc8602a1f2d17d994a19e478fcc8e
parent3dabec7175bc6d49e88748cf03951357e74496ca
[SCSI] ch: fix ch_remove oops

The following commit causes ch_remove oops:

commit 24b42566c3fcbb5a9011d1446783d0f5844ccd45
Author: Greg Kroah-Hartman <gregkh@suse.de>
Date:   Fri May 16 17:55:12 2008 -0700

    SCSI: fix race in device_create

    There is a race from when a device is created with device_create() and
    then the drvdata is set with a call to dev_set_drvdata() in which a
    sysfs file could be open, yet the drvdata will be NULL, causing all
    sorts of bad things to happen.

    This patch fixes the problem by using the new function,
    device_create_drvdata().  It fixes the problem in all of the scsi
    drivers that need it.

Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Doug Gilbert <dgilbert@interlog.com>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The problem is ch_probe stores ch's private data at a wrong place.

We need to store it at scsi_device->sdev_gendev but the above patch
stores it at device struct that device_create_drvdata returns. So we
hit an oops when ch_remove accesses
scsi_device->sdev_gendev->driver_data, which is NULL.

Actually, there wasn't a race because ch doesn't create sysfs files
with device struct that device_create returns. This patch puts back
dev_set_drvdata() to set ch's private data properly.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
drivers/scsi/ch.c