| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20374432-d8d6-ceae-2f31-d154520288ee@huaweicloud.com>
Date: Tue, 27 Feb 2024 09:43:49 +0800
From: Yu Kuai <yukuai1@...weicloud.com>
To: Xiao Ni <xni@...hat.com>, Yu Kuai <yukuai1@...weicloud.com>
Cc: paul.e.luse@...ux.intel.com, song@...nel.org, neilb@...e.com,
shli@...com, linux-raid@...r.kernel.org, linux-kernel@...r.kernel.org,
yi.zhang@...wei.com, yangerkun@...wei.com, "yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH md-6.9 06/10] md/raid1: factor out read_first_rdev() from
read_balance()
Hi,
在 2024/02/27 9:23, Xiao Ni 写道:
> On Tue, Feb 27, 2024 at 9:06 AM Yu Kuai <yukuai1@...weicloud.com> wrote:
>>
>> Hi,
>>
>> 在 2024/02/26 22:16, Xiao Ni 写道:
>>> On Thu, Feb 22, 2024 at 4:04 PM Yu Kuai <yukuai1@...weicloud.com> wrote:
>>>>
>>>> From: Yu Kuai <yukuai3@...wei.com>
>>>>
>>>> read_balance() is hard to understand because there are too many status
>>>> and branches, and it's overlong.
>>>>
>>>> This patch factor out the case to read the first rdev from
>>>> read_balance(), there are no functional changes.
>>>>
>>>> Co-developed-by: Paul Luse <paul.e.luse@...ux.intel.com>
>>>> Signed-off-by: Paul Luse <paul.e.luse@...ux.intel.com>
>>>> Signed-off-by: Yu Kuai <yukuai3@...wei.com>
>>>> ---
>>>> drivers/md/raid1.c | 63 +++++++++++++++++++++++++++++++++-------------
>>>> 1 file changed, 46 insertions(+), 17 deletions(-)
>>>>
>>>> diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
>>>> index 8089c569e84f..08c45ca55a7e 100644
>>>> --- a/drivers/md/raid1.c
>>>> +++ b/drivers/md/raid1.c
>>>> @@ -579,6 +579,47 @@ static sector_t align_to_barrier_unit_end(sector_t start_sector,
>>>> return len;
>>>> }
>>>>
>>>> +static void update_read_sectors(struct r1conf *conf, int disk,
>>>> + sector_t this_sector, int len)
>>>> +{
>>>> + struct raid1_info *info = &conf->mirrors[disk];
>>>> +
>>>> + atomic_inc(&info->rdev->nr_pending);
>>>> + if (info->next_seq_sect != this_sector)
>>>> + info->seq_start = this_sector;
>>>> + info->next_seq_sect = this_sector + len;
>>>> +}
>>>> +
>>>> +static int choose_first_rdev(struct r1conf *conf, struct r1bio *r1_bio,
>>>> + int *max_sectors)
>>>> +{
>>>> + sector_t this_sector = r1_bio->sector;
>>>> + int len = r1_bio->sectors;
>>>> + int disk;
>>>> +
>>>> + for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
>>>> + struct md_rdev *rdev;
>>>> + int read_len;
>>>> +
>>>> + if (r1_bio->bios[disk] == IO_BLOCKED)
>>>> + continue;
>>>> +
>>>> + rdev = conf->mirrors[disk].rdev;
>>>> + if (!rdev || test_bit(Faulty, &rdev->flags))
>>>> + continue;
>>>> +
>>>> + /* choose the first disk even if it has some bad blocks. */
>>>> + read_len = raid1_check_read_range(rdev, this_sector, &len);
>>>> + if (read_len > 0) {
>>>> + update_read_sectors(conf, disk, this_sector, read_len);
>>>> + *max_sectors = read_len;
>>>> + return disk;
>>>> + }
>>>
>>> Hi Kuai
>>>
>>> It needs to update max_sectors even if the bad block starts before
>>> this_sector. Because it can't read more than bad_blocks from other
>>> member disks. If it reads more data than bad blocks, it will cause
>>> data corruption. One rule here is read from the primary disk (the
>>> first readable disk) if it has no bad block and read the
>>> badblock-data-length data from other disks.
>>
>> Noted that raid1_check_read_range() will return readable length from
>> this rdev, hence if bad block starts before this_sector, 0 is returned,
>> and 'len' is updated to the length of badblocks(if not exceed read
>> range), and following iteration will find the first disk to read updated
>> 'len' data and update max_sectors.
>
> Hi Kuai
>
> The problem is that choose_first_rdev doesn't return 'len' from
> max_sectors when bad blocks start before this_sector. In the following
> iteration, it can't read more than 'len' from other disks to avoid
> data corruption. I haven't read all the patches. To this patch, it
> resets best_good_sectors to sectors when it encounters a good member
> disk without bad blocks.
In this case, 'len' is not supposed to be returned, caller will split
orignal IO based on 'max_sectors', for example:
IO: 2, 4 | ----
rdev1: BB: 0, 4 |xxxx
rdev2: no BB
Then choose_first_rdev() will set max_sectors to 2, and return rdev2,
then caller will split and issue new IO:
orignal IO: 4, 2 | --
splited IO: 2, 2 | --
Finally, issue splited IO to rdev2. Later orignal IO will be handled by
read_balance() again, and rdev1 will be returned.
Is this case what you concerned?
Thanks,
Kuai
>
> Regards
> Xiao
>>
>> Thanks,
>> Kuai
>>
>>>
>>> Best Regards
>>> Xiao
>>>
>>>> + }
>>>> +
>>>> + return -1;
>>>> +}
>>>> +
>>>> /*
>>>> * This routine returns the disk from which the requested read should
>>>> * be done. There is a per-array 'next expected sequential IO' sector
>>>> @@ -603,7 +644,6 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>> sector_t best_dist;
>>>> unsigned int min_pending;
>>>> struct md_rdev *rdev;
>>>> - int choose_first;
>>>>
>>>> retry:
>>>> sectors = r1_bio->sectors;
>>>> @@ -613,10 +653,11 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>> best_pending_disk = -1;
>>>> min_pending = UINT_MAX;
>>>> best_good_sectors = 0;
>>>> - choose_first = raid1_should_read_first(conf->mddev, this_sector,
>>>> - sectors);
>>>> clear_bit(R1BIO_FailFast, &r1_bio->state);
>>>>
>>>> + if (raid1_should_read_first(conf->mddev, this_sector, sectors))
>>>> + return choose_first_rdev(conf, r1_bio, max_sectors);
>>>> +
>>>> for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
>>>> sector_t dist;
>>>> sector_t first_bad;
>>>> @@ -662,8 +703,6 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>> * bad_sectors from another device..
>>>> */
>>>> bad_sectors -= (this_sector - first_bad);
>>>> - if (choose_first && sectors > bad_sectors)
>>>> - sectors = bad_sectors;
>>>> if (best_good_sectors > sectors)
>>>> best_good_sectors = sectors;
>>>>
>>>> @@ -673,8 +712,6 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>> best_good_sectors = good_sectors;
>>>> best_disk = disk;
>>>> }
>>>> - if (choose_first)
>>>> - break;
>>>> }
>>>> continue;
>>>> } else {
>>>> @@ -689,10 +726,6 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>>
>>>> pending = atomic_read(&rdev->nr_pending);
>>>> dist = abs(this_sector - conf->mirrors[disk].head_position);
>>>> - if (choose_first) {
>>>> - best_disk = disk;
>>>> - break;
>>>> - }
>>>> /* Don't change to another disk for sequential reads */
>>>> if (conf->mirrors[disk].next_seq_sect == this_sector
>>>> || dist == 0) {
>>>> @@ -760,13 +793,9 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
>>>> rdev = conf->mirrors[best_disk].rdev;
>>>> if (!rdev)
>>>> goto retry;
>>>> - atomic_inc(&rdev->nr_pending);
>>>> - sectors = best_good_sectors;
>>>> -
>>>> - if (conf->mirrors[best_disk].next_seq_sect != this_sector)
>>>> - conf->mirrors[best_disk].seq_start = this_sector;
>>>>
>>>> - conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
>>>> + sectors = best_good_sectors;
>>>> + update_read_sectors(conf, disk, this_sector, sectors);
>>>> }
>>>> *max_sectors = sectors;
>>>>
>>>> --
>>>> 2.39.2
>>>>
>>>>
>>>
>>> .
>>>
>>
>
> .
>
Powered by blists - more mailing lists