]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/commit
net: dcb: disable softirqs in dcbnl_flush_dev()
authorVladimir Oltean <vladimir.oltean@nxp.com>
Wed, 2 Mar 2022 19:39:39 +0000 (21:39 +0200)
committerPaolo Pisati <paolo.pisati@canonical.com>
Wed, 9 Mar 2022 14:17:58 +0000 (15:17 +0100)
commitd9f4f616105d99ce3cfa7a03cd1d6c380d6a809c
treebe3b4bd1325a74ef7a639ae0cbf18b6482169f51
parent8ae1cae8608a2ea7d9d566d605a49d9862f339e5
net: dcb: disable softirqs in dcbnl_flush_dev()

BugLink: https://bugs.launchpad.net/bugs/1964361
[ Upstream commit 10b6bb62ae1a49ee818fc479cf57b8900176773e ]

Ido Schimmel points out that since commit 52cff74eef5d ("dcbnl : Disable
software interrupts before taking dcb_lock"), the DCB API can be called
by drivers from softirq context.

One such in-tree example is the chelsio cxgb4 driver:
dcb_rpl
-> cxgb4_dcb_handle_fw_update
   -> dcb_ieee_setapp

If the firmware for this driver happened to send an event which resulted
in a call to dcb_ieee_setapp() at the exact same time as another
DCB-enabled interface was unregistering on the same CPU, the softirq
would deadlock, because the interrupted process was already holding the
dcb_lock in dcbnl_flush_dev().

Fix this unlikely event by using spin_lock_bh() in dcbnl_flush_dev() as
in the rest of the dcbnl code.

Fixes: 91b0383fef06 ("net: dcb: flush lingering app table entries for unregistered devices")
Reported-by: Ido Schimmel <idosch@idosch.org>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20220302193939.1368823-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
net/dcb/dcbnl.c