tty: synclink_gt: Fix potential deadlock on &info->lock
authorChengfeng Ye <dg573847474@gmail.com>
Fri, 28 Jul 2023 12:39:01 +0000 (12:39 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 4 Aug 2023 13:05:56 +0000 (15:05 +0200)
commit8b66eec1fd85ca9c4fd83f92e704de4fc0c583c1
tree1cba8c52fc5ee1c0700354dc40698a1d6e9114c0
parentf68279ca7f806c44fbbae0a631e5603ba426d9be
tty: synclink_gt: Fix potential deadlock on &info->lock

As &info->lock is acquired by slgt_interrupt() under irq context, other
process context code acquiring the lock should disable irq, otherwise
deadlock could happen if the irq preempt the execution while the
lock is held in process context on the same CPU.

Lock acquisition inside set_params32() does not disable irq, and this
function is called by slgt_compat_ioctl() from process context.

Possible deadlock scenario:
slgt_compat_ioctl()
    -> set_params32()
    -> spin_lock(&info->lock)
        <irq>
        -> slgt_interrupt()
        -> spin_lock(&info->lock); (deadlock here)

This flaw was found by an experimental static analysis tool I am developing
for irq-related deadlock. x86_64 allmodconfig using gcc shows no new
warning.

The patch fixes the potential deadlock by spin_lock_irqsave() like other
lock acquisition sites.

Signed-off-by: Chengfeng Ye <dg573847474@gmail.com>
Reviewed-by: Jiri Slaby <jirislaby@kernel.org>
Link: https://lore.kernel.org/r/20230728123901.64225-1-dg573847474@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/tty/synclink_gt.c