linux-kernel - Re: UML/hostfs - mount failure at tip of tree

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CANP3RGd7AQXPYQVrhjbgEN608Jo7hDUh7nc8VQ62gGQqW0iXMg@mail.gmail.com>
Date: Tue, 23 Jul 2024 20:59:30 -0700
From: Maciej Żenczykowski <maze@...gle.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Hongbo Li <lihongbo22@...wei.com>, Matthew Wilcox <willy@...radead.org>, 
	Kernel hackers <linux-kernel@...r.kernel.org>, Patrick Rohr <prohr@...gle.com>, 
	Christian Brauner <brauner@...nel.org>
Subject: Re: UML/hostfs - mount failure at tip of tree

On Tue, Jul 23, 2024 at 7:55 PM Maciej Żenczykowski <maze@...gle.com> wrote:
>
> On Tue, Jul 23, 2024 at 7:22 PM Linus Torvalds
> <torvalds@...ux-foundation.org> wrote:
> >
> > On Tue, 23 Jul 2024 at 18:35, Hongbo Li <lihongbo22@...wei.com> wrote:
> > >
> > > I apologize for causing this issue. I am currently tracking it down.  If
> > > reverting this can solve the problem, you can revert it first.
> >
> > I don't get the feeling that this is _so_ urgent that it needs to be
> > reverted immediately - let's give it at least a few days and see if
> > you (or somebody else) figures out the bug.
> >
> > Maciej - if you can verify that folio conversion fix suggestion of
> > mine (or alternatively report that it doesn't help and I was barking
> > up the wrong tree), that would be great.
>
> That appears to fix the folio patch indeed (ie. I no longer need to revert it).
>
> The tests are still super unhappy, but I've yet to fix our tests very
> broken netlink parser for changes that released in 6.10, so that may
> be unrelated ;-)

+++ fs/hostfs/hostfs_kern.c:
 static int hostfs_fill_super(struct super_block *sb, struct fs_context *fc)
 {
        struct hostfs_fs_info *fsi = sb->s_fs_info;
-       const char *host_root = fc->source;
+       const char *host_root = "/";

appears to fix the problem (when combined with Linus' folio fix).

I think fc->source is just the 'block device' passed to mount, and
thus for a virtual filesystem like hostfs, it is just garbage...

(and with the appropriate netlink fixes all the tests now pass at tip-of-tree:
87f3073c2871 (HEAD) hostfs_fill_super(): host_root := "/" (not fc->source)
2743a4aabac6 fs/hostfs/hostfs_kern.c:445 buffer =
folio_zero_tail(folio, bytes_read, buffer + bytes_read);
a2caf678d7e1 neighbour: add RTNL_FLAG_DUMP_SPLIT_NLM_DONE to RTM_GETNEIGH
3bb0c5772acf net: add RTNL_FLAG_DUMP_SPLIT_NLM_DONE to RTM_GET(RULE|ROUTE)
786c8248dbd3 (linux/master) Merge tag
'perf-tools-fixes-for-v6.11-2024-07-23' of
git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools
)

> > And perhaps remind me about this mount API thing too if it doesn't
> > seem to be resolved by the end of the week when I'm starting to get
> > ready to do the rc1?
> >
> >              Linus
>
> --
> Maciej Żenczykowski, Kernel Networking Developer @ Google

--
Maciej Żenczykowski, Kernel Networking Developer @ Google