Re: [PATCH 1/8] scsi: target: Modify core_tmr_abort_task()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2020-07-12 02:49, Mike Christie wrote:
On 7/10/20 5:48 AM, Bodo Stroesser wrote:

...

@@ -137,11 +138,16 @@ void core_tmr_abort_task(
          printk("ABORT_TASK: Found referenced %s task_tag: %llu\n",
              se_cmd->se_tfo->fabric_name, ref_tag);
-        if (!__target_check_io_state(se_cmd, se_sess,
-                         dev->dev_attrib.emulate_tas))
+        spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
+        rc = __target_check_io_state(se_cmd, se_sess, 0);
+        spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+        if (!rc)
              continue;
-        spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+        list_del_init(&se_cmd->state_list);

Do we only want to do this if in the next patch tmr_notify can successfully accept the tmr?

This change does not modify behavior of the core. Please see how
core_tmr_drain_state_list() uses state_list to queue commands after
successfully calling __target_check_io_state() for later call of target_put_cmd_and_wait().

This patch just unifies ABORT_TASK and LUN_RESET handling such,
that in next patch we can call tmr_notify handler with a list of
aborted commands in both cases. The list uses the state_list list_head
in the se_cmd.


If tmr_notify can't allocate memory for the abort case, we would delete it from the list. Later the initiator would send a LUN_RESET but core_tmr_drain_state_list won't see it and pass it to the tmr_notify call.


tmr_notify handler will not and must not modify the list of aborted
cmds. So if backend just does nothing for whatever reason we will end
up with the "old" behavior of the core as it was before the change.

The general idea is to send info about received TMRs together with the
list of aborted commands to the backend, to allow TMR tracing and
canceling of aborted commands. But backend must not interfere with TMR
processing in core.

+        se_cmd->state_active = false;
+
+        spin_unlock_irqrestore(&dev->execute_task_lock, flags);
          /*
           * Ensure that this ABORT request is visible to the LU RESET
@@ -159,7 +165,7 @@ void core_tmr_abort_task(
          atomic_long_inc(&dev->aborts_complete);
          return;
      }
-    spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+    spin_unlock_irqrestore(&dev->execute_task_lock, flags);
      printk("ABORT_TASK: Sending TMR_TASK_DOES_NOT_EXIST for ref_tag: %lld\n",
              tmr->ref_task_tag);






[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [SCSI Target Devel]     [Linux SCSI Target Infrastructure]     [Kernel Newbies]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Linux IIO]     [Samba]     [Device Mapper]

  Powered by Linux