mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Stephen C. Tweedie" <sct@redhat.com>
To: linux-kernel <linux-kernel@vger.kernel.org>,
	"ext2-devel@lists.sourceforge.net"
	<ext2-devel@lists.sourceforge.net>
Cc: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>,
	Stephen Tweedie <sct@redhat.com>, Andrew Morton <akpm@osdl.org>
Subject: ext2/3: Incomplete/buggy use of s_debts in Orlov
Date: 10 Sep 2004 17:49:49 +0100	[thread overview]
Message-ID: <1094834989.2047.171.camel@sisko.scot.redhat.com> (raw)

Hi all,

I've been tidying up the online resize code for -mm and come across one
nasty item.  The s_debts array is the only dynamic, fs-sized array in
the entirety of ext2/3, so it's the only dynamic array we need to grow
when we resize.

But to get the locking for that right, we need to know what the locking
for s_debts is in the first place.  Looking at that I found two rather
odd things:

On ext2, s_debts is modified with only the per-group lock:

	spin_lock(sb_bgl_lock(sbi, group));
	if (S_ISDIR(mode)) {
		if (sbi->s_debts[group] < 255)
			sbi->s_debts[group]++;

where s_debts is defined as

	u8 *s_debts;

in the superblock.  That was OK when we had a global lock_super(sb)
round every allocation, but it's extremely dangerous to perform byte
accesses like this when the rest of the memory word is shared with other
CPUs.  It certainly risks word-tearing on machines like Alpha; I think
it's theoretically unsafe even on x86.

On ext3, the situation is actually much simpler:

s_debts is never modified at all.

Ever.

It's allocated, carefully checked, then freed.  But never modified.  

Now, that's fine with me --- it makes the locking for s_debts *really*
easy to get right. :-)   And in my current resize patches, I just nuke
s_debts entirely.  

But if anybody thinks it really needs to stay, then a Plan B is
required, and we probably need to start off by fixing the locking in
ext2 itself.

--Stephen


                 reply	other threads:[~2004-09-10 16:58 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=1094834989.2047.171.camel@sisko.scot.redhat.com \
    --to=sct@redhat.com \
    --cc=akpm@osdl.org \
    --cc=ext2-devel@lists.sourceforge.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=viro@parcelfarce.linux.theplanet.co.uk \
    /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

all inboxes | Powered by JetHome®