linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: john stultz <johnstul@us.ibm.com>
To: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	Frank Mayhar <fmayhar@google.com>
Subject: Re: [patch 16/52] fs: dcache RCU for multi-step operaitons
Date: Thu, 24 Jun 2010 10:22:11 -0700	[thread overview]
Message-ID: <1277400131.15264.29.camel@work-vm> (raw)
In-Reply-To: <20100624150334.GE10441@laptop>

On Fri, 2010-06-25 at 01:03 +1000, Nick Piggin wrote:
> On Thu, Jun 24, 2010 at 09:58:10AM +0200, Peter Zijlstra wrote:
> > On Thu, 2010-06-24 at 13:02 +1000, npiggin@suse.de wrote:
> > > plain text document attachment (fs-dcache_lock-multi-step.patch)
> > > The remaining usages for dcache_lock is to allow atomic, multi-step read-side
> > > operations over the directory tree by excluding modifications to the tree.
> > > Also, to walk in the leaf->root direction in the tree where we don't have
> > > a natural d_lock ordering.
> > > 
> > > This could be accomplished by taking every d_lock, but this would mean a
> > > huge number of locks and actually gets very tricky.
> > > 
> > > Solve this instead by using the rename seqlock for multi-step read-side
> > > operations. Insert operations are not serialised. Delete operations are
> > > tricky when walking up the directory our parent might have been deleted
> > > when dropping locks so also need to check and retry for that.
> > > 
> > > XXX: hmm, we could of course just take the rename lock if there is any worry
> > > about livelock. Most of these are slow paths.
> > 
> > 
> > Ah, does this address John's issue? 
> 
> This is where John's issue is introduced. I actually again couldn't
> see the problem (thought I saw a problem, then lost it!).

Ok. So I need to review this new set in full, but the issue we tripped
with the patches in -rt was the following:

In select parent, when it does the dentry ascending, it has to release
the current dentry lock so it can aquire the parents dentry (for proper
ordering). At the point it released the lock, before it grabs the
parent's lock, there is nothing that is preventing dput from being
called on the next dentry, it grabbing the parent and dentry's d_lock
and killing it. Then back in select_parent, when we try to lock the next
entry, it no longer exists and we oops.

So I can't see anything that is protecting the dentry (or even the new
parent dentry, should everything be killed under it) at this point. The
dentries d_count might be zero, we don't hold the d_lock, and we don't
hold the parent's d_lock. What am I missing here?


> Got to think about it and test more... I couldn't reproduce the problem
> mind you, but I was testing mainline wheras bug was seen on -rt.

Yea, -rt may be allowing the race to more easily occur (I only found I
could trigger it on a UP machine) since we can be preempted right after
we released the dentry lock as we tried to grab the parent. Then dput
would jump in and wreck things. I had some pretty clear trace logs that
were easily repeated (well, took about an hour or two to trigger), that
showed about the same order of operations every time.

I don't remember if there's another lock being held at the point
select_parent is called, so on mainline it might be harder to trigger
(having to actually get the dput race on another cpu timed right).


thanks
-john



  reply	other threads:[~2010-06-24 17:22 UTC|newest]

Thread overview: 152+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-06-24  3:02 [patch 00/52] vfs scalability patches updated npiggin
2010-06-24  3:02 ` [patch 01/52] kernel: add bl_list npiggin
2010-06-24  6:04   ` Eric Dumazet
2010-06-24 14:42     ` Nick Piggin
2010-06-24 16:01       ` Eric Dumazet
2010-06-28 21:37   ` Paul E. McKenney
2010-06-29  6:30     ` Nick Piggin
2010-06-24  3:02 ` [patch 02/52] fs: fix superblock iteration race npiggin
2010-06-29 13:02   ` Christoph Hellwig
2010-06-29 14:56     ` Nick Piggin
2010-06-29 17:35       ` Linus Torvalds
2010-06-29 17:41         ` Nick Piggin
2010-06-29 17:52           ` Linus Torvalds
2010-06-29 17:58             ` Linus Torvalds
2010-06-29 20:04               ` Chris Clayton
2010-06-29 20:14                 ` Nick Piggin
2010-06-29 20:38                   ` Chris Clayton
2010-06-30  7:13                     ` Chris Clayton
2010-06-30 12:51               ` Al Viro
2010-06-24  3:02 ` [patch 03/52] fs: fs_struct rwlock to spinlock npiggin
2010-06-24  3:02 ` [patch 04/52] fs: cleanup files_lock npiggin
2010-06-24  3:02 ` [patch 05/52] lglock: introduce special lglock and brlock spin locks npiggin
2010-06-24 18:15   ` Thomas Gleixner
2010-06-25  6:22     ` Nick Piggin
2010-06-25  9:50       ` Thomas Gleixner
2010-06-25 10:11         ` Nick Piggin
2010-06-24  3:02 ` [patch 06/52] fs: scale files_lock npiggin
2010-06-24  7:52   ` Peter Zijlstra
2010-06-24 15:00     ` Nick Piggin
2010-06-24  3:02 ` [patch 07/52] fs: brlock vfsmount_lock npiggin
2010-06-24  3:02 ` [patch 08/52] fs: scale mntget/mntput npiggin
2010-06-24  3:02 ` [patch 09/52] fs: dcache scale hash npiggin
2010-06-24  3:02 ` [patch 10/52] fs: dcache scale lru npiggin
2010-06-24  3:02 ` [patch 11/52] fs: dcache scale nr_dentry npiggin
2010-06-24  3:02 ` [patch 12/52] fs: dcache scale dentry refcount npiggin
2010-06-24  3:02 ` [patch 13/52] fs: dcache scale d_unhashed npiggin
2010-06-24  3:02 ` [patch 14/52] fs: dcache scale subdirs npiggin
2010-06-24  7:56   ` Peter Zijlstra
2010-06-24  9:50   ` Andi Kleen
2010-06-24 15:53     ` Nick Piggin
2010-06-24  3:02 ` [patch 15/52] fs: dcache scale inode alias list npiggin
2010-06-24  3:02 ` [patch 16/52] fs: dcache RCU for multi-step operaitons npiggin
2010-06-24  7:58   ` Peter Zijlstra
2010-06-24 15:03     ` Nick Piggin
2010-06-24 17:22       ` john stultz [this message]
2010-06-24 17:26   ` john stultz
2010-06-25  6:45     ` Nick Piggin
2010-06-24  3:02 ` [patch 17/52] fs: dcache remove dcache_lock npiggin
2010-06-24  3:02 ` [patch 18/52] fs: dcache reduce dput locking npiggin
2010-06-24  3:02 ` [patch 19/52] fs: dcache per-bucket dcache hash locking npiggin
2010-06-24  3:02 ` [patch 20/52] fs: dcache reduce dcache_inode_lock npiggin
2010-06-24  3:02 ` [patch 21/52] fs: dcache per-inode inode alias locking npiggin
2010-06-24  3:02 ` [patch 22/52] fs: dcache rationalise dget variants npiggin
2010-06-24  3:02 ` [patch 23/52] fs: dcache percpu nr_dentry npiggin
2010-06-24  3:02 ` [patch 24/52] fs: dcache reduce d_parent locking npiggin
2010-06-24  8:44   ` Peter Zijlstra
2010-06-24 15:07     ` Nick Piggin
2010-06-24 15:32       ` Paul E. McKenney
2010-06-24 16:05         ` Nick Piggin
2010-06-24 16:41           ` Paul E. McKenney
2010-06-28 21:50   ` Paul E. McKenney
2010-07-07 14:35     ` Nick Piggin
2010-06-24  3:02 ` [patch 25/52] fs: dcache DCACHE_REFERENCED improve npiggin
2010-06-24  3:02 ` [patch 26/52] fs: icache lock s_inodes list npiggin
2010-06-24  3:02 ` [patch 27/52] fs: icache lock inode hash npiggin
2010-06-24  3:02 ` [patch 28/52] fs: icache lock i_state npiggin
2010-06-24  3:02 ` [patch 29/52] fs: icache lock i_count npiggin
2010-06-30  7:27   ` Dave Chinner
2010-06-30 12:05     ` Nick Piggin
2010-07-01  2:36       ` Dave Chinner
2010-07-01  7:54         ` Nick Piggin
2010-07-01  9:36           ` Nick Piggin
2010-07-01 16:21           ` Frank Mayhar
2010-07-03  2:03       ` Andrew Morton
2010-07-03  3:41         ` Nick Piggin
2010-07-03  4:31           ` Andrew Morton
2010-07-03  5:06             ` Nick Piggin
2010-07-03  5:18               ` Nick Piggin
2010-07-05 22:41               ` Dave Chinner
2010-07-06  4:34                 ` Nick Piggin
2010-07-06 10:38                   ` Theodore Tso
2010-07-06 13:04                     ` Nick Piggin
2010-07-07 17:00                     ` Frank Mayhar
2010-06-24  3:02 ` [patch 30/52] fs: icache lock lru/writeback lists npiggin
2010-06-24  8:58   ` Peter Zijlstra
2010-06-24 15:09     ` Nick Piggin
2010-06-24 15:13       ` Peter Zijlstra
2010-06-24  3:02 ` [patch 31/52] fs: icache atomic inodes_stat npiggin
2010-06-24  3:02 ` [patch 32/52] fs: icache protect inode state npiggin
2010-06-24  3:02 ` [patch 33/52] fs: icache atomic last_ino, iunique lock npiggin
2010-06-24  3:02 ` [patch 34/52] fs: icache remove inode_lock npiggin
2010-06-24  3:02 ` [patch 35/52] fs: icache factor hash lock into functions npiggin
2010-06-24  3:02 ` [patch 36/52] fs: icache per-bucket inode hash locks npiggin
2010-06-24  3:02 ` [patch 37/52] fs: icache lazy lru npiggin
2010-06-24  9:52   ` Andi Kleen
2010-06-24 15:59     ` Nick Piggin
2010-06-30  8:38   ` Dave Chinner
2010-06-30 12:06     ` Nick Piggin
2010-07-01  2:46       ` Dave Chinner
2010-07-01  7:57         ` Nick Piggin
2010-06-24  3:02 ` [patch 38/52] fs: icache RCU free inodes npiggin
2010-06-30  8:57   ` Dave Chinner
2010-06-30 12:07     ` Nick Piggin
2010-06-24  3:02 ` [patch 39/52] fs: icache rcu walk for i_sb_list npiggin
2010-06-24  3:02 ` [patch 40/52] fs: dcache improve scalability of pseudo filesystems npiggin
2010-06-24  3:02 ` [patch 41/52] fs: icache reduce atomics npiggin
2010-06-24  3:02 ` [patch 42/52] fs: icache per-cpu last_ino allocator npiggin
2010-06-24  9:48   ` Andi Kleen
2010-06-24 15:52     ` Nick Piggin
2010-06-24 16:19       ` Andi Kleen
2010-06-24 16:38         ` Nick Piggin
2010-06-24  3:02 ` [patch 43/52] fs: icache per-cpu nr_inodes counter npiggin
2010-06-24  3:02 ` [patch 44/52] fs: icache per-CPU sb inode lists and locks npiggin
2010-06-30  9:26   ` Dave Chinner
2010-06-30 12:08     ` Nick Piggin
2010-07-01  3:12       ` Dave Chinner
2010-07-01  8:00         ` Nick Piggin
2010-06-24  3:02 ` [patch 45/52] fs: icache RCU hash lookups npiggin
2010-06-24  3:02 ` [patch 46/52] fs: icache reduce locking npiggin
2010-06-24  3:02 ` [patch 47/52] fs: keep inode with backing-dev npiggin
2010-06-24  3:03 ` [patch 48/52] fs: icache split IO and LRU lists npiggin
2010-06-24  3:03 ` [patch 49/52] fs: icache scale writeback list locking npiggin
2010-06-24  3:03 ` [patch 50/52] mm: implement per-zone shrinker npiggin
2010-06-24 10:06   ` Andi Kleen
2010-06-24 16:00     ` Nick Piggin
2010-06-24 16:27       ` Andi Kleen
2010-06-24 16:32         ` Andi Kleen
2010-06-24 16:37         ` Andi Kleen
2010-06-30  6:28   ` Dave Chinner
2010-06-30 12:03     ` Nick Piggin
2010-06-24  3:03 ` [patch 51/52] fs: per-zone dentry and inode LRU npiggin
2010-06-30 10:09   ` Dave Chinner
2010-06-30 12:13     ` Nick Piggin
2010-06-24  3:03 ` [patch 52/52] fs: icache less I_FREEING time npiggin
2010-06-30 10:13   ` Dave Chinner
2010-06-30 12:14     ` Nick Piggin
2010-07-01  3:33       ` Dave Chinner
2010-07-01  8:06         ` Nick Piggin
2010-06-25  7:12 ` [patch 00/52] vfs scalability patches updated Christoph Hellwig
2010-06-25  8:05   ` Nick Piggin
2010-06-30 11:30 ` Dave Chinner
2010-06-30 12:40   ` Nick Piggin
2010-07-01  3:56     ` Dave Chinner
2010-07-01  8:20       ` Nick Piggin
2010-07-01 17:36       ` Andi Kleen
2010-07-01 17:23     ` Nick Piggin
2010-07-01 17:28       ` Andi Kleen
2010-07-06 17:49       ` Nick Piggin
2010-07-01 17:35     ` Linus Torvalds
2010-07-01 17:52       ` Nick Piggin
2010-07-02  4:01       ` Paul E. McKenney
2010-06-30 17:08   ` Frank Mayhar

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=1277400131.15264.29.camel@work-vm \
    --to=johnstul@us.ibm.com \
    --cc=fmayhar@google.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=npiggin@suse.de \
    --cc=peterz@infradead.org \
    /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).