On Wed, Mar 16, 2022 at 07:59:15PM +0900, Tetsuo Handa wrote:
> But I can see that due to no longer waiting for lo->lo_mutex from lo_open(),
> there are occasional I/O errors. What is your plan to avoid this?
There is no plan to avoid that. We'll get this areas due to have the
remove handling works. We could play tricks with RQF_QUIET, but I'm
not sure that is worth it.
> By the way, if CONFIG_BLOCK_LEGACY_AUTOLOAD=n,
>
> # mount -o loop,ro isofs.iso isofs/
>
> unconditionally fails with
>
> mount: isofs/: failed to setup loop device for isofs.iso.
>
> message. Commit 451f0b6f4c44d7b6 ("block: default BLOCK_LEGACY_AUTOLOAD to y")
> says "if the device node already exists because old scripts created it manually".
> But it is not always manual creation of loop devices; I think it is
> ioctl(LOOP_CTL_GET_FREE) in my case.
I'll take a look, thanks!
