From: Richard Gooch <rgooch@ras.ucalgary.ca>
To: Alexander Viro <viro@math.psu.edu>
Cc: Linus Torvalds <torvalds@transmeta.com>,
Alan Cox <alan@lxorguk.ukuu.org.uk>,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH] one of $BIGNUM devfs races
Date: Mon, 6 Aug 2001 20:00:02 -0600 [thread overview]
Message-ID: <200108070200.f77202G27928@vindaloo.ras.ucalgary.ca> (raw)
In-Reply-To: <Pine.GSO.4.21.0108062129030.16817-100000@weyl.math.psu.edu>
In-Reply-To: <200108070127.f771RNe27524@vindaloo.ras.ucalgary.ca> <Pine.GSO.4.21.0108062129030.16817-100000@weyl.math.psu.edu>
Alexander Viro writes:
>
>
> On Mon, 6 Aug 2001, Richard Gooch wrote:
>
> > I'm referring specifically to this code:
> > new->inode.ino = fs_info.num_inodes + FIRST_INODE;
> > fs_info.table[fs_info.num_inodes++] = new;
> >
> > This is not SMP safe. Besides, even the allocation loop isn't SMP
> > safe. If two tasks both allocate a table, they each could end up
> > calling:
> > kfree (fs_info.table);
> > for the same value. Or for a different one (which is also bad).
>
> BKL. kfree() is non-blocking. IOW, critical area can be placed under a
> spinlock and BKL acts as such. We can trivially replace it with
> a spinlock (static in function).
>
> Actually, there is another problem with that code and it has nothing
> to SMP. You never shrink that table and AFAICS you never reuse the
> entries. IOW, you've got a leak there.
The devfs entries are reused *for the same name*. But yes, if "fred"
is registered and unregistered, and is never registered again, it will
indeed stick around forever. In general, this is not a significant
problem, since the same name tends to be re-registered later. Said
another way: there aren't many different temporary entries.
The reason for not freeing stuff is simplicity. Without proper
locking, I was able to avoid a lot of races. Now that I'm putting
locks in, I can consider freeing stuff (after the locks are in).
> Why on the Earth do you need it, in the first place? Just put the
> pointer to entry into inode->u.generic_ip and be done with that - it
> kills all that mess for good. AFAICS the only places where you
> really use that table is your get_devfs_entry_from_vfs_inode() and
> devfs_write_inode(). In both cases pointer would be obviously more
> convenient.
Again, historical reasons. When I wrote devfs, the pipe data trampled
the inode->u.generic_ip pointer. So that's no good. I see that the
pipe data has been moved away. Good. Hm. But there's still the
inode->u.socket_i structure. I'd need to check where that gets
trampled.
Regards,
Richard....
Permanent: rgooch@atnf.csiro.au
Current: rgooch@ras.ucalgary.ca
next prev parent reply other threads:[~2001-08-07 2:00 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-08-06 11:58 [PATCH] one of $BIGNUM devfs races Alexander Viro
2001-08-06 16:47 ` Richard Gooch
2001-08-06 16:56 ` Rik van Riel
2001-08-06 17:11 ` Richard Gooch
2001-08-06 23:50 ` Richard Gooch
2001-08-07 0:34 ` Alexander Viro
2001-08-07 0:51 ` Richard Gooch
2001-08-07 1:10 ` Alexander Viro
2001-08-07 1:27 ` Richard Gooch
2001-08-07 1:42 ` Alexander Viro
2001-08-07 2:00 ` Richard Gooch [this message]
2001-08-07 2:15 ` Alexander Viro
2001-08-07 5:17 ` Richard Gooch
2001-08-07 5:28 ` Alexander Viro
2001-08-07 10:21 ` Anton Altaparmakov
2001-08-07 17:09 ` Richard Gooch
2001-08-07 17:27 ` Arjan van de Ven
2001-08-07 17:28 ` Richard Gooch
2001-08-07 5:53 ` Richard Gooch
2001-08-07 6:23 ` Alexander Viro
2001-08-07 6:36 ` Richard Gooch
2001-08-07 6:40 ` Alexander Viro
2001-08-07 6:47 ` Richard Gooch
2001-08-07 7:31 ` Alexander Viro
2001-08-07 18:11 ` Richard Gooch
2001-08-07 18:23 ` Alexander Viro
2001-08-07 18:55 ` Richard Gooch
2001-08-07 19:10 ` Alexander Viro
2001-08-07 22:13 ` Richard Gooch
2001-08-06 23:59 Alan Cox
[not found] ` <no.id>
2001-08-06 23:54 ` Alan Cox
2001-08-06 23:55 ` Richard Gooch
2001-08-06 23:59 ` Richard Gooch
2001-08-07 16:22 ` Alan Cox
2001-08-07 19:04 ` Alan Cox
2001-08-07 19:16 ` Alexander Viro
2001-08-08 21:16 ` H. Peter Anvin
2001-08-08 21:47 ` Alexander Viro
2001-08-08 23:29 ` Richard Gooch
2001-08-07 19:09 ` Richard Gooch
2001-08-07 19:20 ` Alexander Viro
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200108070200.f77202G27928@vindaloo.ras.ucalgary.ca \
--to=rgooch@ras.ucalgary.ca \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@transmeta.com \
--cc=viro@math.psu.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).