I think I looked at the comments to that function and then didn't read
the code too closely. A few extra tests and jumps, but in the grand
scheme of things when dealing with disks, not noticeable like I
thought before. Thanks for pointing it out.
----------------
Robert LeBlanc
PGP Fingerprint 79A2 9CA4 6CC4 45DD A904 C70E E654 3BB2 FA62 B9F1
On Mon, Nov 28, 2016 at 4:53 PM, Doug Dumitru <doug@xxxxxxxxxx> wrote:
> Not the best code organization, but no real harm done. Look at the
> front of md_write_start:
>
> void md_write_start(struct mddev *mddev, struct bio *bi)
> {
> int did_change = 0;
> if (bio_data_dir(bi) != WRITE)
> return;
>
> So writes fall thru really quickly.
>
> Doug
>
>
> On Mon, Nov 28, 2016 at 3:41 PM, Robert LeBlanc <robert@xxxxxxxxxxxxx> wrote:
>> I'm looking through the code and am wondering why this is at the start
>> of raid1_make_request. It seems that it is only needed for WRITE and
>> would unnecessarily delay READ requests waiting for the superblock.
>> Would it make more sense to put this at the start of the WRITE branch?
>> What am I missing?
>>
>> 1057 /*
>> 1058 * Register the new request and wait if the
>> reconstruction
>> 1059 * thread has put up a bar for new requests.
>> 1060 * Continue immediately if no resync is active currently.
>> 1061 */
>> 1062
>> 1063 md_write_start(mddev, bio); /* wait on superblock update
>> early */
>> 1064
>> 1065 if (bio_data_dir(bio) == WRITE &&
>> 1066 ((bio_end_sector(bio) > mddev->suspend_lo &&
>> 1067 bio->bi_iter.bi_sector < mddev->suspend_hi) ||
>> 1068 (mddev_is_clustered(mddev) &&
>> 1069 md_cluster_ops->area_resyncing(mddev, WRITE,
>> 1070 bio->bi_iter.bi_sector,
>> bio_end_sector(bio))))) {
>> 1071 /* As the suspend_* range is controlled by
>> 1072 * userspace, we want an interruptible
>> 1073 * wait.
>> 1074 */
>> 1075 DEFINE_WAIT(w);
>> 1076 for (;;) {
>> 1077 flush_signals(current);
>> 1078 prepare_to_wait(&conf->wait_barrier,
>> 1079 &w, TASK_INTERRUPTIBLE);
>> 1080 if (bio_end_sector(bio) <=
>> mddev->suspend_lo ||
>> 1081 bio->bi_iter.bi_sector >=
>> mddev->suspend_hi ||
>> 1082 (mddev_is_clustered(mddev) &&
>> 1083
>> !md_cluster_ops->area_resyncing(mddev, WRITE,
>> 1084 bio->bi_iter.bi_sector,
>> bio_end_sector(bio))))
>> 1085 break;
>> 1086 schedule();
>> 1087 }
>> 1088 finish_wait(&conf->wait_barrier, &w);
>> 1089 }
>> 1090
>> 1091 start_next_window = wait_barrier(conf, bio);
>> 1092
>> 1093 bitmap = mddev->bitmap;
>> 1094
>>
>> Thanks
>>
>> ----------------
>> Robert LeBlanc
>> PGP Fingerprint 79A2 9CA4 6CC4 45DD A904 C70E E654 3BB2 FA62 B9F1
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
>
> --
> Doug Dumitru
> EasyCo LLC
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
