diff options
| author | Nicholas Bellinger <nab@linux-iscsi.org> | 2011-11-02 03:33:16 -0700 |
|---|---|---|
| committer | Nicholas Bellinger <nab@linux-iscsi.org> | 2011-11-02 15:58:30 +0000 |
| commit | 88dd9e26d6d3e743f9c7e4562b94b2ad3c2994d3 (patch) | |
| tree | 452154b4137170de74d4445c5fcbbf62ce2f04fd /drivers/target/target_core_transport.c | |
| parent | 5a4c8666c6d576f076a7c6824589cdbb984c0f84 (diff) | |
target: Make TFO->check_stop_free return free status
This patch converts target_core_fabric_ops->check_stop_free() usage in
transport_cmd_check_stop() and associated fabric module usage to
return '1' when the passed se_cmd has been released directly within
->check_stop_free(), or return '0' when the passed se_cmd has not
been released.
This addresses an issue where transport_cmd_finish_abort() ->
transport_cmd_check_stop_to_fabric() was leaking descriptors during
LUN_RESET for modules using ->check_stop_free(), but not directly
releasing se_cmd in all cases.
Cc: stable@kernel.org
Signed-off-by: Nicholas Bellinger <nab@risingtidesystems.com>
Diffstat (limited to 'drivers/target/target_core_transport.c')
| -rw-r--r-- | drivers/target/target_core_transport.c | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c index 5dee44639f8f..bf8867f44021 100644 --- a/drivers/target/target_core_transport.c +++ b/drivers/target/target_core_transport.c @@ -514,13 +514,16 @@ static int transport_cmd_check_stop( * Some fabric modules like tcm_loop can release * their internally allocated I/O reference now and * struct se_cmd now. + * + * Fabric modules are expected to return '1' here if the + * se_cmd being passed is released at this point, + * or zero if not being released. */ if (cmd->se_tfo->check_stop_free != NULL) { spin_unlock_irqrestore( &cmd->t_state_lock, flags); - cmd->se_tfo->check_stop_free(cmd); - return 1; + return cmd->se_tfo->check_stop_free(cmd); } } spin_unlock_irqrestore(&cmd->t_state_lock, flags); |