In the Linux kernel, the following vulnerability has been resolved: scsi: qedi: Fix use after free bug in qedi_remove() In qedi_probe() we call __qedi_probe() which initializes &qedi->recovery_work with qedi_recovery_handler() and &qedi->board_disable_work with qedi_board_disable_work(). When qedi_schedule_recovery_handler() is called, schedule_delayed_work() will finally start the work. In qedi_remove(), which is called to remove the driver, the following sequence may be observed: Fix this by finishing the work before cleanup in qedi_remove(). CPU0 CPU1 |qedi_recovery_handler qedi_remove | __qedi_remove | iscsi_host_free | scsi_host_put | //free shost | |iscsi_host_for_each_session |//use qedi->shost Cancel recovery_work and board_disable_work in __qedi_remove().
https://git.kernel.org/stable/c/fa19c533ab19161298f0780bcc6523af88f6fd20
https://git.kernel.org/stable/c/c5749639f2d0a1f6cbe187d05f70c2e7c544d748
https://git.kernel.org/stable/c/89f6023fc321c958a0fb11f143a6eb4544ae3940
https://git.kernel.org/stable/c/5e756a59cee6a8a79b9059c5bdf0ecbf5bb8d151
https://git.kernel.org/stable/c/3738a230831e861503119ee2691c4a7dc56ed60a
https://git.kernel.org/stable/c/124027cd1a624ce0347adcd59241a9966a726b22