]> git.baikalelectronics.ru Git - kernel.git/commitdiff
target: Don't override EXTENDED_COPY xcopy_pt_cmd SCSI status code
authorDinesh Israni <ddi@datera.io>
Tue, 11 Oct 2016 03:22:03 +0000 (20:22 -0700)
committerNicholas Bellinger <nab@linux-iscsi.org>
Thu, 20 Oct 2016 04:22:33 +0000 (21:22 -0700)
This patch addresses a bug where a local EXTENDED_COPY WRITE or READ
backend I/O request would always return SAM_STAT_CHECK_CONDITION,
even if underlying xcopy_pt_cmd->se_cmd generated a different
SCSI status code.

ESX host environments expect to hit SAM_STAT_RESERVATION_CONFLICT
for certain scenarios, and SAM_STAT_CHECK_CONDITION results in
non-retriable status for these cases.

Tested on v4.1.y with ESX v5.5u2+ with local IBLOCK backend copy.

Reported-by: Nixon Vincent <nixon.vincent@calsoftinc.com>
Tested-by: Nixon Vincent <nixon.vincent@calsoftinc.com>
Cc: Nixon Vincent <nixon.vincent@calsoftinc.com>
Tested-by: Dinesh Israni <ddi@datera.io>
Signed-off-by: Dinesh Israni <ddi@datera.io>
Cc: Dinesh Israni <ddi@datera.io>
Cc: stable@vger.kernel.org # 3.14+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
drivers/target/target_core_xcopy.c

index f2a1443ecb33f8983de9c73ed2d9e308f67e34c7..094a1440eacb3dccdd9c35a678a2940c3e03216d 100644 (file)
@@ -662,6 +662,7 @@ static int target_xcopy_read_source(
        rc = target_xcopy_setup_pt_cmd(xpt_cmd, xop, src_dev, &cdb[0],
                                remote_port, true);
        if (rc < 0) {
+               ec_cmd->scsi_status = xpt_cmd->se_cmd.scsi_status;
                transport_generic_free_cmd(se_cmd, 0);
                return rc;
        }
@@ -673,6 +674,7 @@ static int target_xcopy_read_source(
 
        rc = target_xcopy_issue_pt_cmd(xpt_cmd);
        if (rc < 0) {
+               ec_cmd->scsi_status = xpt_cmd->se_cmd.scsi_status;
                transport_generic_free_cmd(se_cmd, 0);
                return rc;
        }
@@ -723,6 +725,7 @@ static int target_xcopy_write_destination(
                                remote_port, false);
        if (rc < 0) {
                struct se_cmd *src_cmd = &xop->src_pt_cmd->se_cmd;
+               ec_cmd->scsi_status = xpt_cmd->se_cmd.scsi_status;
                /*
                 * If the failure happened before the t_mem_list hand-off in
                 * target_xcopy_setup_pt_cmd(), Reset memory + clear flag so that
@@ -738,6 +741,7 @@ static int target_xcopy_write_destination(
 
        rc = target_xcopy_issue_pt_cmd(xpt_cmd);
        if (rc < 0) {
+               ec_cmd->scsi_status = xpt_cmd->se_cmd.scsi_status;
                se_cmd->se_cmd_flags &= ~SCF_PASSTHROUGH_SG_TO_MEM_NOALLOC;
                transport_generic_free_cmd(se_cmd, 0);
                return rc;
@@ -824,10 +828,14 @@ static void target_xcopy_do_work(struct work_struct *work)
 out:
        xcopy_pt_undepend_remotedev(xop);
        kfree(xop);
-
-       pr_warn_ratelimited("target_xcopy_do_work: rc: %d, Setting X-COPY CHECK_CONDITION"
-                           " -> sending response\n", rc);
-       ec_cmd->scsi_status = SAM_STAT_CHECK_CONDITION;
+       /*
+        * Don't override an error scsi status if it has already been set
+        */
+       if (ec_cmd->scsi_status == SAM_STAT_GOOD) {
+               pr_warn_ratelimited("target_xcopy_do_work: rc: %d, Setting X-COPY"
+                       " CHECK_CONDITION -> sending response\n", rc);
+               ec_cmd->scsi_status = SAM_STAT_CHECK_CONDITION;
+       }
        target_complete_cmd(ec_cmd, SAM_STAT_CHECK_CONDITION);
 }