Re: [RFC PATCH 5/5] md/raid1: introduce rectify action to repair badblocks

在 2026/1/14 11:11, Li Nan 写道:
>
>
> 在 2025/12/31 15:09, Zheng Qixing 写道:
>> From: Zheng Qixing <[email protected]>
>>
>> Add support for repairing known badblocks in RAID1. When disks
>> have known badblocks (shown in sysfs bad_blocks), data can be
>> read from other healthy disks in the array and written to repair
>> the badblock areas and clear it in bad_blocks.
>>
>> echo rectify > sync_action can trigger this action.
>>
>> Signed-off-by: Zheng Qixing <[email protected]>
>
>> +static void end_rectify_read(struct bio *bio)
>> +{
>> +    struct r1bio *r1_bio = get_resync_r1bio(bio);
>> +    struct r1conf *conf = r1_bio->mddev->private;
>> +    struct md_rdev *rdev;
>> +    struct bio *next_bio;
>> +    bool all_fail = true;
>> +    int i;
>> +
>> +    update_head_pos(r1_bio->read_disk, r1_bio);
>> +
>> +    if (!bio->bi_status) {
>> +        set_bit(R1BIO_Uptodate, &r1_bio->state);
>> +        goto out;
>> +    }
>> +
>> +    for (i = r1_bio->read_disk + 1; i < conf->raid_disks; i++) {
>> +        rdev = conf->mirrors[i].rdev;
>> +        if (!rdev || test_bit(Faulty, &rdev->flags))
>> +            continue;
>> +
>> +        next_bio = r1_bio->bios[i];
>> +        if (next_bio->bi_end_io == end_rectify_read) {
>> +            next_bio->bi_opf &= ~MD_FAILFAST;
>
> Why set MD_FAILFAST and clear it soon?
> And submit_rectify_read() will clear it again.


Indeed.

>
>> +static void rectify_request_write(struct mddev *mddev, struct r1bio *r1_bio)
>> +{
>> +    struct r1conf *conf = mddev->private;
>> +    struct bio *wbio = NULL;
>> +    int wcnt = 0;
>> +    int i;
>> +
>> +    if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
>> +        submit_rectify_read(r1_bio);
>> +        return;
>> +    }
>> +
>> +    atomic_set(&r1_bio->remaining, 0);
>> +    for (i = 0; i < conf->raid_disks; i++) {
>> +        wbio = r1_bio->bios[i];
>> +        if (wbio->bi_end_io == end_rectify_write) {
>> +            atomic_inc(&r1_bio->remaining);
>> +            wcnt++;
>> +            submit_bio_noacct(wbio);
>> +        }
>> +    }
>> +
>> +    if (unlikely(!wcnt)) {
>> +        md_done_sync(r1_bio->mddev, r1_bio->sectors);
>> +        put_buf(r1_bio);
>> +    }
>
> How can 'wcnt' be 0?


Oh, I forgot to check the faulty state of rdev.😳

>
>> +}
>> +
>> +static void handle_rectify_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
>> +{
>> +    struct md_rdev *rdev;
>> +    struct bio *bio;
>> +    int i;
>> +
>> +    for (i = 0; i < conf->raid_disks; i++) {
>> +        rdev = conf->mirrors[i].rdev;
>> +        bio = r1_bio->bios[i];
>> +        if (bio->bi_end_io == NULL)
>> +            continue;
>> +
>> +        if (!bio->bi_status && bio->bi_end_io == end_rectify_write &&
>> +            test_bit(R1BIO_MadeGood, &r1_bio->state)) {
>> +            rdev_clear_badblocks(rdev, r1_bio->sector,
>> +                         r1_bio->sectors, 0);
>> +        }
>
> Reuse handle_sync_write_finished() seems better.

Good point.


Thanks,

Qixing