Re: Re: [PATCH 1/5] md: When resync aborted,using ->curr_resync as the last completed request rather than ->curr_resync_completed.

[majianpeng <majianpeng@gmail.com>]

>On Sat, 27 Oct 2012 10:28:58 +0800 kernelmail <kedacomkernel@gmail.com> wrote:
>
>> In commit(db91ff55bdf06736b849afc1b1fce5763bbb8d5d),it said:
>> >If a resync is aborted we should record how far we got
>> >(recovery_cp) the last request that we know has completed
>> >(->curr_resync_completed) rather than the last request that was
>> >submitted (->curr_resync).
>> But when resync aborted, it execed
>> >wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
>> After this, the ->curr_resync is the last completed request.
>> 
>> Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
>> ---
>>  drivers/md/md.c |    3 +--
>>  1 file changed, 1 insertion(+), 2 deletions(-)
>> 
>> diff --git a/drivers/md/md.c b/drivers/md/md.c
>> index 9ab768a..1d9940d 100644
>> --- a/drivers/md/md.c
>> +++ b/drivers/md/md.c
>> @@ -7558,8 +7558,7 @@ void md_do_sync(struct md_thread *thread)
>>  					printk(KERN_INFO
>>  					       "md: checkpointing %s of %s.\n",
>>  					       desc, mdname(mddev));
>> -					mddev->recovery_cp =
>> -						mddev->curr_resync_completed;
>> +					mddev->recovery_cp = mddev->curr_resync;
>>  				}
>>  			} else
>>  				mddev->recovery_cp = MaxSector;
>
>No.
>This reverts commit db91ff55bdf06736b849afc1b1fce5763bbb8d5d
>
>Maybe if we differentiated between an abort-due-to-error and an
>abort-due-to-request, then in the second card we could use ->curr_resync.
>
>NeilBrown
>
Yes, i missed the case which interrupted by io-error.
So i add a flag MD_RECOVERY_ERROR in md_done_sync.
 
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 9ab768a..cb9864c 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -7172,6 +7172,7 @@ void md_done_sync(struct mddev *mddev, int blocks, int ok)
        wake_up(&mddev->recovery_wait);
        if (!ok) {
                set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+               set_bit(MD_RECOVERY_ERROR, &mddev->recovery);
                md_wakeup_thread(mddev->thread);
                // stop recovery, signal do_sync ....
        }
@@ -7558,8 +7559,12 @@ void md_do_sync(struct md_thread *thread)
                                        printk(KERN_INFO
                                               "md: checkpointing %s of %s.\n",
                                               desc, mdname(mddev));
-                                       mddev->recovery_cp =
-                                               mddev->curr_resync_completed;
+                                       if (test_bit(MD_RECOVERY_ERROR, &mddev->recovery))
+                                               mddev->recovery_cp =
+                                                       mddev->curr_resync_completed;
+                                       else
+                                               mddev->recovery_cp =
+                                                       mddev->curr_resync;
                                }
                        } else
                                mddev->recovery_cp = MaxSector;
diff --git a/drivers/md/md.h b/drivers/md/md.h
index af443ab..3b1a35f 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -307,6 +307,7 @@ struct mddev {
         * REQUEST:  user-space has requested a sync (used with SYNC)
         * CHECK:    user-space request for check-only, no repair
         * RESHAPE:  A reshape is happening
+        * ERROR:    interrupted by IO error not by request
         *
         * If neither SYNC or RESHAPE are set, then it is a recovery.
         */
@@ -320,6 +321,7 @@ struct mddev {
 #define        MD_RECOVERY_CHECK       7
 #define MD_RECOVERY_RESHAPE    8
 #define        MD_RECOVERY_FROZEN      9
+#define MD_RECOVERY_ERROR      10
 
        unsigned long                   recovery;
        /* If a RAID personality determines that recovery (of a particular


How about this patch?

Jianpeng