netdev - Re: [PATCH v3 6/7] 9p/trans_virtio: support larger msize values

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] [day] [month] [year] [list]

Date:   Wed, 24 Nov 2021 13:22:20 +0100
From:   Christian Schoenebeck <linux_oss@...debyte.com>
To:     Dominique Martinet <asmadeus@...ewreck.org>
Cc:     Nikolay Kichukov <nikolay@...um.net>,
        v9fs-developer@...ts.sourceforge.net, netdev@...r.kernel.org,
        Eric Van Hensbergen <ericvh@...il.com>,
        Latchesar Ionkov <lucho@...kov.net>,
        Greg Kurz <groug@...d.org>, Vivek Goyal <vgoyal@...hat.com>
Subject: Re: [PATCH v3 6/7] 9p/trans_virtio: support larger msize values

On Montag, 22. November 2021 23:35:18 CET Dominique Martinet wrote:
> Christian Schoenebeck wrote on Mon, Nov 22, 2021 at 02:32:23PM +0100:
> > I "think" this could be used for all 9p message types except for the
> > listed
> > former three, but I'll review the 9p specs more carefully before v4. For
> > Tread and Twrite we already received the requested size, which just
> > leaves Treaddir, which is however indeed tricky, because I don't think we
> > have any info how many directory entries we could expect.
> 
> count in Treaddir is a number of bytes, not a number of entries -- so
> it's perfect for this as well :)

Yes it is in bytes, but it's currently always simply msize - P9_READDIRHDRSZ:
https://github.com/torvalds/linux/blob/5d9f4cf36721aba199975a9be7863a3ff5cd4b59/fs/9p/vfs_dir.c#L159

As my planned milestone for this series is max. 4 MB msize, it might be okay
for now, but it is something to keep in mind and should be checked whether it
will slow down things.

On the long term, when msize >4MB is supported, this definitely must be
addressed.

> > A simple compile time constant (e.g. one macro) could be used instead of
> > this function. If you prefer a constant instead, I could go for it in v4
> > of course. For this 9p client I would recommend a function though, simply
> > because this code has already seen some authors come and go over the
> > years, so it might be worth the redundant code for future safety. But
> > I'll adapt to what others think.
> 
> In this case a fallback constant seems simpler than a big switch like
> you've done, but honestly I'm not fussy at this point -- if you work on
> this you have the right to decide this kind of things in my opinion.
> 
> My worry with the snippet you listed is that you need to enumerate all
> calls again, so if someday the protocol is extended it'll be a place to
> forget adding new calls (although compiler warnings help with that),
> whereas a fallback constant will always work as long as it's small
> messages.
> 
> But really, as long as it's not horrible I'll take it :)

Maybe I can find a compromise. :)

Best regards,
Christian Schoenebeck