Re: [PATCH 1/2] init: split get_fs_names

From: Al Viro
Date: Mon Jun 21 2021 - 10:59:15 EST


On Mon, Jun 21, 2021 at 04:53:09PM +0200, Christoph Hellwig wrote:
> On Mon, Jun 21, 2021 at 02:46:37PM +0000, Al Viro wrote:
> > TBH, I would rather take that one into fs/filesystems.c. Rationale:
> > get_filesystem_list(), for all its resemblance to /proc/filesystems
> > contents, is used only by init/*.c and it's not a big deal to make
> > it
>
> Yeah, unwinding this mess actually is a good idea. I didn't really
> look outside of do_mounts.c, but once doing that it becomes completely
> obvious.
>
> > int __init get_filesystem_list(char *buf, bool is_dev)
>
> As-is we don't even really need the is_dev argument, as the only
> callers wants block device file systems anyway.

*nod*

> In fact it would
> much rather have a cursor based iteration so that we can skip the
> allocation, but that is probaby overengineering the problem.

Very much so.

Sigh... I really wish we had more uniform syntax, though -
e.g. root=nfs(<options>) or root=xfs(sdb11,noatime), etc.
Oh, well...