target: Avoid mappedlun symlink creation during lun shutdown
authorNicholas Bellinger <nab@linux-iscsi.org>
Mon, 27 Mar 2017 23:12:43 +0000 (16:12 -0700)
committerWilly Tarreau <w@1wt.eu>
Wed, 1 Nov 2017 21:12:44 +0000 (22:12 +0100)
commit003bea4a702279fc339a0d2f1262e907364a2c73
treefaf38527025ede4e8073010f82f2853a8f6c01c2
parenta6a256f6125b80f980e8add0d72153c2426adad3
target: Avoid mappedlun symlink creation during lun shutdown

commit 49cb77e297dc611a1b795cfeb79452b3002bd331 upstream.

This patch closes a race between se_lun deletion during configfs
unlink in target_fabric_port_unlink() -> core_dev_del_lun()
-> core_tpg_remove_lun(), when transport_clear_lun_ref() blocks
waiting for percpu_ref RCU grace period to finish, but a new
NodeACL mappedlun is added before the RCU grace period has
completed.

This can happen in target_fabric_mappedlun_link() because it
only checks for se_lun->lun_se_dev, which is not cleared until
after transport_clear_lun_ref() percpu_ref RCU grace period
finishes.

This bug originally manifested as NULL pointer dereference
OOPsen in target_stat_scsi_att_intr_port_show_attr_dev() on
v4.1.y code, because it dereferences lun->lun_se_dev without
a explicit NULL pointer check.

In post v4.1 code with target-core RCU conversion, the code
in target_stat_scsi_att_intr_port_show_attr_dev() no longer
uses se_lun->lun_se_dev, but the same race still exists.

To address the bug, go ahead and set se_lun>lun_shutdown as
early as possible in core_tpg_remove_lun(), and ensure new
NodeACL mappedlun creation in target_fabric_mappedlun_link()
fails during se_lun shutdown.

Reported-by: James Shen <jcs@datera.io>
Cc: James Shen <jcs@datera.io>
Tested-by: James Shen <jcs@datera.io>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Willy Tarreau <w@1wt.eu>
drivers/target/target_core_fabric_configfs.c
drivers/target/target_core_tpg.c
include/target/target_core_base.h