Re: [PATCH 07/17] nbd: call blk_mark_disk_dead in nbd_clear_sock_ioctl

[Samuel Holland <samuel.holland@sifive.com>]

On 2023-10-02 1:21 AM, Christoph Hellwig wrote:
> On Sun, Oct 01, 2023 at 07:10:53PM +0200, Wouter Verhelst wrote:
>>> So what are the semantics of clearing the socket?
>>>
>>> The <= 6.5 behavior of invalidating fs caches, but not actually marking
>>> the fs shutdown is pretty broken, especially if this expects to resurrect
>>> the device and thus the file system later on.
>>
>> nbd-client -d calls
>>
>> ioctl(nbd, NBD_DISCONNECT);
>> ioctl(nbd, NBD_CLEAR_SOCK);
>>
>> (error handling removed for clarity)
>>
>> where "nbd" is the file handle to the nbd device. This expects that the
>> device is cleared and that then the device can be reused for a different
>> connection, much like "losetup -d". Expecting that the next connection
>> would talk to the same file system is wrong.
> 
> So a fs shutdown seems like a the right thing.  So I'm a little confused
> on what actualy broke here.

I'm not too familiar with the block subsystem, but my understanding is that
blk_mark_disk_dead(nbd->disk) is permanent -- there is no way to un-mark a disk
as dead. So this makes the device (e.g. /dev/nbd0) permanently unusable after
the call to ioctl(nbd, NBD_CLEAR_SOCK).

Like Wouter said, the semantics should be similar to a loop device, where the
same block device can be reused after being disconnected. That was why I
suggested disk_force_media_change() as called from __loop_clr_fd(). The loop
driver doesn't call blk_mark_disk_dead() anywhere.

Regards,
Samuel