Re: [PATCH 01/11] drbd_nl: dont allow detating to be inttrupted in waiting D_DETACHING to DISKLESS

Hello Dongsheng,

First of all, thanks for contributing patches to us.
Please find my reply on the patch below the quote:

On Mon, Jun 24, 2024 at 7:52 AM zhengbing.huang
<[email protected]> wrote:
>
> From: Dongsheng Yang <[email protected]>
>
> In our network failure and drbd down testing, we found warning in dmesg and drbd down process into D state:
>
> "kernel: drbd /unregistered/ramtest3/0 drbd103: ASSERTION device->disk_state[NOW] == D_FAILED || device->disk_state[NOW] == D_DETACHING FAILED in go_diskless"
>
> the problem is the wait_event is inttruptable, it could be intrupted by signal and call drbd_cleanup_device before go_diskless()
>

In this case, I suggest improving the expression in the assertion.
Improving an assertion can also mean removing that assertion.

The wait_event_interruptible() is there for a reason. Think of a
backing disk that behaves like a tar pit—a backing device that no
longer finishes IO requests. You want a way to interrupt the drbdsetup
waiting in detach.

PS: A bit more elaborative commit messages are welcome.

best regards,
 Philipp