Bug#493184: aufs: Deadlocks in the rename() call.

2008-08-28 Thread Daniel Baumann
reopen 493184
found 493184 0+20080719-4
thanks

Hi,

this bug is back.. looks like -3 or -4 did break it again.
You can verify it with debian-live 5.0 beta1 images.

Regards,
Daniel

-- 
Address:Daniel Baumann, Burgunderstrasse 3, CH-4562 Biberist
Email:  [EMAIL PROTECTED]
Internet:   http://people.panthera-systems.net/~daniel-baumann/



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#493184: aufs: Deadlocks in the rename() call.

2008-08-28 Thread Daniel Baumann
notfound 493184 0+20080719-4
close 493184
thanks

sorry for the noise - I forgot that the lenny l-m-e was built against a
previous aufs version (and should have been requested to be binNUMed to
be able to close #493184).

Regards,
Daniel

-- 
Address:Daniel Baumann, Burgunderstrasse 3, CH-4562 Biberist
Email:  [EMAIL PROTECTED]
Internet:   http://people.panthera-systems.net/~daniel-baumann/



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Processed: Re: Bug#493184: aufs: Deadlocks in the rename() call.

2008-08-02 Thread Debian Bug Tracking System
Processing commands for [EMAIL PROTECTED]:

 tag 493184 pending
Bug#493184: aufs: Deadlocks in the rename() call.
Tags were: patch
Tags added: pending

 thanks
Stopping processing here.

Please contact me if you need assistance.

Debian bug tracking system administrator
(administrator, Debian Bugs database)


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#493184: aufs: Deadlocks in the rename() call.

2008-08-02 Thread Julian Andres Klode
tag 493184 pending
thanks

Cyril Brulebois wrote:
 Package: aufs
 Version: 0+20080719-1
 Severity: serious
 Tags: patch
 
 Hi jak,
 
 as you probably already know, there is a quite inconvenient bug in
 sid+lenny's aufs, which is blocking renames, caused by deadlocks. That
 means that one can't update the apt cache through apt-get/aptitude, at
 the very least (it's not like rename() is something very rare anyway),
 which IMHO makes it unfit for a stable release.
I know it, and I just waited until aufs migrated to testing. I just had no time
to upload it yet.
 
 For more details, initial upstream bugreport:
 http://article.gmane.org/gmane.linux.file-systems.aufs.user/1270
 
 The patch as provided by upstream:
 http://article.gmane.org/gmane.linux.file-systems.aufs.user/1291
 
 The thread contains various confirmations that the patch does its job
 and fixes the deadlocks. I personally tested it against a custom
 snapshot I prepared: 20080714 (I didn't know 20080719 would follow some
 days later, and I was checking whether the bug got fixed in the
 meanwhile), but also with 0+20080719-1, against which I attach a
 debdiff, so that you can eventually merge the patch quickly.
Actually, the package is 20080714, I just checked it out on 20080719.

 
 I kept the patch as is, only adapting the patch level. Feel free to
 drop/adapt attribution lines in the patch/the changelog, nothing fancy
 in preparing an upload. More precisely: I needed to do so to prepare
 some (better-working) LiveCD images.
 
 Cheers,
 
 


-- 
Julian Andres Klode, Fellow of the Free Software Foundation Europe
 Debian Maintainer | Developer | Ubuntu Member

try Debian: http://www.debian.org/ | my site: http://jak-linux.org/
jabber: [EMAIL PROTECTED] | IRC: juliank (FreeNode, OFTC)
languages: German  | English



signature.asc
Description: OpenPGP digital signature


Bug#493184: aufs: Deadlocks in the rename() call.

2008-07-31 Thread Cyril Brulebois
Package: aufs
Version: 0+20080719-1
Severity: serious
Tags: patch

Hi jak,

as you probably already know, there is a quite inconvenient bug in
sid+lenny's aufs, which is blocking renames, caused by deadlocks. That
means that one can't update the apt cache through apt-get/aptitude, at
the very least (it's not like rename() is something very rare anyway),
which IMHO makes it unfit for a stable release.

For more details, initial upstream bugreport:
http://article.gmane.org/gmane.linux.file-systems.aufs.user/1270

The patch as provided by upstream:
http://article.gmane.org/gmane.linux.file-systems.aufs.user/1291

The thread contains various confirmations that the patch does its job
and fixes the deadlocks. I personally tested it against a custom
snapshot I prepared: 20080714 (I didn't know 20080719 would follow some
days later, and I was checking whether the bug got fixed in the
meanwhile), but also with 0+20080719-1, against which I attach a
debdiff, so that you can eventually merge the patch quickly.

I kept the patch as is, only adapting the patch level. Feel free to
drop/adapt attribution lines in the patch/the changelog, nothing fancy
in preparing an upload. More precisely: I needed to do so to prepare
some (better-working) LiveCD images.

Cheers,

-- 
Cyril Brulebois
diff -u aufs-0+20080719/debian/changelog aufs-0+20080719/debian/changelog
--- aufs-0+20080719/debian/changelog
+++ aufs-0+20080719/debian/changelog
@@ -1,3 +1,11 @@
+aufs (0+20080719-2) unstable; urgency=low
+
+  * debian/patches/10_fix_deadlocks_in_rename.dpatch: Backport a fix for
+blocking rename calls, caused by deadlocks, which led to hanging
+package managers when trying to update the APT cache (Closes: #N).
+
+ -- Cyril Brulebois [EMAIL PROTECTED]  Fri, 01 Aug 2008 04:07:49 +
+
 aufs (0+20080719-1) unstable; urgency=low
 
   * New upstream snapshot, fixing many bugs
diff -u aufs-0+20080719/debian/patches/00list aufs-0+20080719/debian/patches/00list
--- aufs-0+20080719/debian/patches/00list
+++ aufs-0+20080719/debian/patches/00list
@@ -8,0 +9 @@
+10_fix_deadlocks_in_rename.dpatch
only in patch2:
unchanged:
--- aufs-0+20080719.orig/debian/patches/10_fix_deadlocks_in_rename.dpatch
+++ aufs-0+20080719/debian/patches/10_fix_deadlocks_in_rename.dpatch
@@ -0,0 +1,56 @@
+#! /bin/sh /usr/share/dpatch/dpatch-run
+## 10_fix_deadlocks_in_rename.dpatch by Cyril Brulebois [EMAIL PROTECTED]
+##
+## DP: Fix deadlocks as detected by running aptitude update.
+
[EMAIL PROTECTED]@
+Index: fs/aufs/cpup.c
+===
+RCS file: /cvsroot/aufs/aufs/fs/aufs/cpup.c,v
+retrieving revision 1.67
+diff -u -p -r1.67 cpup.c
+--- a/fs/aufs/cpup.c	21 Jul 2008 02:51:28 -	1.67
 b/fs/aufs/cpup.c	24 Jul 2008 04:01:23 -
+@@ -1004,7 +1004,7 @@ int au_cp_dirs(struct dentry *dentry, au
+ h_parent = au_h_dptr(parent, bdst);
+ 		} while (!h_parent);
+ 
+-		if (d != real_parent)
++		if (d != real_parent  d != locked)
+ 			di_write_lock_child3(d);
+ 
+ 		/* somebody else might create while we were sleeping */
+@@ -1041,7 +1041,7 @@ int au_cp_dirs(struct dentry *dentry, au
+ di_read_unlock(parent, AuLock_IR);
+ 		}
+ 
+-		if (d != real_parent)
++		if (d != real_parent  d != locked)
+ 			di_write_unlock(d);
+ 		if (unlikely(err))
+ 			break;
+Index: fs/aufs25/cpup.c
+===
+RCS file: /cvsroot/aufs/aufs/fs/aufs25/cpup.c,v
+retrieving revision 1.11
+diff -u -p -r1.11 cpup.c
+--- a/fs/aufs25/cpup.c	21 Jul 2008 02:53:36 -	1.11
 b/fs/aufs25/cpup.c	24 Jul 2008 04:01:23 -
+@@ -998,7 +998,7 @@ int au_cp_dirs(struct dentry *dentry, au
+ h_parent = au_h_dptr(parent, bdst);
+ 		} while (!h_parent);
+ 
+-		if (d != real_parent)
++		if (d != real_parent  d != locked)
+ 			di_write_lock_child3(d);
+ 
+ 		/* somebody else might create while we were sleeping */
+@@ -1036,7 +1036,7 @@ int au_cp_dirs(struct dentry *dentry, au
+ di_read_unlock(parent, AuLock_IR);
+ 		}
+ 
+-		if (d != real_parent)
++		if (d != real_parent  d != locked)
+ 			di_write_unlock(d);
+ 		if (unlikely(err))
+ 			break;