From: Hans de Goede Date: Sun, 9 Sep 2012 09:30:02 +0000 (-0300) Subject: [media] gspca: Don't set gspca_dev->dev to NULL before stop0 X-Git-Tag: omap-for-v3.7-rc1/fixes-cpufreq-signed~3^2^2~296 X-Git-Url: http://git.openpandora.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=345321dc9c52b774f42c934339f9b3e2f0a39395;p=pandora-kernel.git [media] gspca: Don't set gspca_dev->dev to NULL before stop0 In commit a3d6e8cc0e6ddc8b3cfdeb3c979f07ed1aa528b3 gspca_dev->dev is set to NULL on disconnect, before calling stop0. The plan was to get rid of gspca_dev->present and instead simply check for gspca_dev->dev everywhere where we were checking for present. This should be race free since all users of gspca_dev->dev hold the usb_lock, or so I thought. But I was wrong, drivers which use a work-queue + synchronous bulk transfers to get the video data don't hold the usb_lock while doing so, their stop0 callbacks stop the workqueue, so they won't be using gspca_dev->dev anymore after the stop0 call, but they might be dereferincing it before, so we should not set gspca_dev->dev to NULL on disconnect before calling stop0. This also means that the workqueue functions in these drivers cannot use gspca_dev->dev to check if they need to stop because of disconnection, so we will need to keep gspca_dev->present around, and set that to 0 on disconnect, before calling stop0. Unfortunately as part of the plan to remove gspca_dev->present, these workqueues where already moved over to checking for gspca_dev->dev instead of gspca_dev->present as part of commit 254902b01d2acc6aced99ec17caa4c6cd890cdea, so this patch also reverts those parts of that commit. Signed-off-by: Hans de Goede Signed-off-by: Mauro Carvalho Chehab --- Reading git-diff-tree failed