Re: md/bitmap: fix GPF in write_page caused by resize race

[Yu Kuai]

Hi,

在 2026/3/6 0:41, Michał Mirosław 写道:
> Hi,
>
> Commit 5f73c8b33df9 ("md/bitmap: fix GPF in write_page caused by resize
> race") in stable 6.19 killed my machine. I confirmed that after reverting
> this commit the machine boots fine. The same commit was backported into
> 6.18 stable and shows the same symptoms. I tried the revert only with 6.19,
> though, as that revived my workflow.
>
> For context, I have several md devices, some with internal bitmaps and
> dm-crypt on top.  md3 and md4 belong to a single LVM VG. In the locked
> up state I was able to see `mdadm`, some `udev-trigger` and `mount` tasks
> all hanging in D state.

Thanks for the report. Do you have the log of task stack of these tasks?
This will be very helpful to locate the root cause.

>
> /proc/mdstat:
>
> md9 : active raid1 sdc1[0] sdh1[1]
>        547584 blocks super 1.0 [2/2] [UU]
>
> md3 : active raid1 sdg3[3] sdb3[2]
>        669920064 blocks super 1.2 [2/2] [UU]
>        bitmap: 2/5 pages [8KB], 65536KB chunk
>
> md0 : active raid1 sdb1[5] sda1[3]
>        123892 blocks super 1.2 [3/2] [UU_]
>
> md4 : active raid1 sde[0](W) nvme0n1[2] nvme1n1[3]
>        488254464 blocks super 1.2 [3/3] [UUU]
>        bitmap: 0/4 pages [0KB], 65536KB chunk
>
> md1 : active raid1 sda2[2] nvme4n1p1[4]
>        234305112 blocks super 1.2 [2/2] [UU]
>        bitmap: 2/2 pages [8KB], 65536KB chunk
>
> Best Regards
> Michał Mirosław

-- 
Thansk,
Kuai