在 2025/3/4 20:26, Yu Kuai 写道:
> Hi,
>
> 在 2025/03/04 14:12, Wu Guanghao 写道:
>> When testing the raid1, I found that adding disk to raid1 fails.
>> Here's how to reproduce it:
>>
>> 1. modprobe brd rd_nr=3 rd_size=524288
>> 2. mdadm -Cv /dev/md0 -l1 -n2 -e1.0 /dev/ram0 /dev/ram1
>>
>> 3. mdadm /dev/md0 -f /dev/ram0
>> 4. mdadm /dev/md0 -r /dev/ram0
>>
>> 5. echo "10000 100" > /sys/block/md0/md/dev-ram1/bad_blocks
>> 6. echo "write_error" > /sys/block/md0/md/dev-ram1/state
>>
>> 7. mkfs.xfs /dev/md0
>> 8. mdadm --examine-badblocks /dev/ram1 # Bad block records can be seen
>> Bad-blocks on /dev/ram1:
>> 10000 for 100 sectors
>>
>> 9. mdadm /dev/md0 -a /dev/ram2
>> mdadm: add new device failed for /dev/ram2 as 2: Invalid argument
>
> Can you add a new regression test as well?
>
Yeah, I will add in v2.
>>
>> When adding a disk to a RAID1 array, the metadata is read from the existing
>> member disks for synchronization. However, only the bad_blocks flag are copied,
>> the bad_blocks records are not copied, so the bad_blocks records are all zeros.
>> The kernel function super_1_load() detects bad_blocks flag and reads the
>> bad_blocks record, then sets the bad block using badblocks_set().
>>
>> After the kernel commit 1726c7746("badblocks: improve badblocks_set() for
>> multiple ranges handling"), if the length of a bad_blocks record is 0, it will
>> return a failure. Therefore the device addition will fail.
>>
>> So, don't set the bad_blocks flag when initializing the metadata, kernel can
>> handle it.
>>
>> Signed-off-by: Wu Guanghao <wuguanghao3@xxxxxxxxxx>
>> ---
>> super1.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/super1.c b/super1.c
>> index fe3c4c64..03578e5b 100644
>> --- a/super1.c
>> +++ b/super1.c
>> @@ -2139,6 +2139,9 @@ static int write_init_super1(struct supertype *st)
>> if (raid0_need_layout)
>> sb->feature_map |= __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT);
>>
>> + if (sb->feature_map & MD_FEATURE_BAD_BLOCKS)
>> + sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BAD_BLOCKS);
>
> There are also other flags that is per rdev, like MD_FEATURE_REPLACEMENT
> and MD_FEATURE_JOURNAL, they should be excluded as well.
>
OK， These will be modified in v2.
> Thanks,
> Kuai
>
>> +
>> sb->sb_csum = calc_sb_1_csum(sb);
>> rv = store_super1(st, di->fd);
>>
>
> .
