Recent patch frenzy, conflicts and Oopses

Recent patch frenzy, conflicts and Oopses

[Christian Trefzer]

Hi guys,

as I am heavily using reiser4 for some time now on my stable machine, I
keep a git tree with vanilla -stable and reiser4 since -mm has sometimes
been too unreliable. Lurking on this list I collected some patches, of
which two seem to collide. The first one fixes a nasty "out of memory?"
Oops for me and maybe others, and the second one was posted as a
response to other Oopses on the list.

Since I reversed the first one in order to apply the second one,
thinking (wrongly) that this one should be a better fix for the same issues and
maybe more, I have my old pal out of memory back. Maybe the two should
be merged in a compatible way, and a new patch against 2.6.16-stable be
put on the namesys ftp server?

Anyway, here come the patches in question:

From: Alexander Zarochentsev <zam@namesys.com>
To: reiserfs-list@namesys.com
Subject: Re: "warning("vs-44", "out of memory?")"
Date: Wed, 12 Apr 2006 14:26:59 +0400

On Wednesday 12 April 2006 13:06, Alexander Zarochentsev wrote:
> Hello
>
> On Wednesday 12 April 2006 11:42, Roy Lanek wrote:
> > I have managed to save a copy of my
> > /var/log/socklog-klog/others/current
> > before the next complete freezing happened.
>
> Please use the latest reiser4 patch reiser4-for-2.6.16-2 from
> ftp.namesys.com.

and apply the following patch:

Index: linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/extent_file_ops.c
===================================================================
--- linux-2.6.17-rc1-mm1.orig/fs/reiser4/plugin/item/extent_file_ops.c
+++ linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/extent_file_ops.c
@@ -807,7 +807,7 @@ extent_balance_dirty_pages(struct inode 
 		if (excl)
 			get_exclusive_access(uf_info);
 		else
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info, 1);
 	}
 	return 0;
 }
Index: linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/tail.c
===================================================================
--- linux-2.6.17-rc1-mm1.orig/fs/reiser4/plugin/item/tail.c
+++ linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/tail.c
@@ -510,7 +510,7 @@ tail_balance_dirty_pages(struct address_
 		if (excl)
 			get_exclusive_access(uf_info);
 		else
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info, 1);
 	}
 	return 0;
 }


This one helped me a lot. The second one brought older problems back,
but may have fixed others for some people on this list. It seems to
remove the second argument to get_nonexclusive_access(), though:

From: Alexander Zarochentsev <zam@namesys.com>
To: reiserfs-list@namesys.com
Subject: Re: Reproducible reiser4 bug with 2.6.16.2 patch on tail_conversion.c:80
Date: Thu, 11 May 2006 11:24:53 +0400

--Boundary-00=_FbuYEKWWhFAJ/IU
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Hello

[original message snipped]

please check whether the attached patch helps.

-- 
Alex.

--Boundary-00=_FbuYEKWWhFAJ/IU
Content-Type: text/x-diff;
  charset="iso-8859-1";
  name="reiser4-remove-atom_may_exists-from-get_nea.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
	filename="reiser4-remove-atom_may_exists-from-get_nea.diff"

 fs/reiser4/plugin/file/file.c            |   17 ++++++++---------
 fs/reiser4/plugin/file/file.h            |    2 +-
 fs/reiser4/plugin/file/funcs.h           |    5 -----
 fs/reiser4/plugin/file/tail_conversion.c |   12 +++---------
 fs/reiser4/plugin/item/extent_file_ops.c |    3 ++-
 fs/reiser4/plugin/item/tail.c            |    3 ++-
 6 files changed, 16 insertions(+), 26 deletions(-)

Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/item/extent_file_ops.c
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/item/extent_file_ops.c
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/item/extent_file_ops.c
@@ -804,10 +804,11 @@ extent_balance_dirty_pages(struct inode 
 					f->length > PAGE_CACHE_SIZE ?
 					PAGE_CACHE_SIZE : f->length);
 
+		txn_restart_current();
 		if (excl)
 			get_exclusive_access(uf_info);
 		else
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info);
 	}
 	return 0;
 }
Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/item/tail.c
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/item/tail.c
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/item/tail.c
@@ -507,10 +507,11 @@ tail_balance_dirty_pages(struct address_
 		} else
 			drop_nonexclusive_access(uf_info);
 		reiser4_throttle_write(inode);
+		txn_restart_current();
 		if (excl)
 			get_exclusive_access(uf_info);
 		else
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info);
 	}
 	return 0;
 }
Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/file.c
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/file/file.c
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/file.c
@@ -1597,7 +1597,7 @@ writepages_unix_file(struct address_spac
 				break;
 			}
 		} else
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info);
 
 		while (to_capture > 0) {
 			pgoff_t start;
@@ -2023,7 +2023,7 @@ read_unix_file(struct file *file, char _
 	while (left > 0) {
 		txn_restart_current();
 
-		get_nonexclusive_access(uf_info, 0);
+		get_nonexclusive_access(uf_info);
 
 		size = i_size_read(inode);
 		if (*off >= size) {
@@ -2177,7 +2177,7 @@ append_and_or_overwrite(hint_t * hint, s
 				if (!exclusive) {
 					drop_exclusive_access(uf_info);
 					txn_restart_current();
-					get_nonexclusive_access(uf_info, 0);
+					get_nonexclusive_access(uf_info);
 				}
 				if (result)
 					return result;
@@ -2531,19 +2531,18 @@ ssize_t write_unix_file(struct file *fil
 	while (left > 0) {
 		int excl = 0;
 
-		/* getting exclusive or not exclusive access requires no
-		   transaction open */
-		txn_restart_current();
-
 		/* faultin user page */
 		fault_in_pages_readable(buf,
 					left > PAGE_CACHE_SIZE ? PAGE_CACHE_SIZE : left);
+		/* getting exclusive or not exclusive access requires no
+		   transaction open */
+		txn_restart_current();
 
 		if (inode->i_size == 0) {
 			get_exclusive_access(uf_info);
 			excl = 1;
 		} else {
-			get_nonexclusive_access(uf_info, 0);
+			get_nonexclusive_access(uf_info);
 			excl = 0;
 		}
 
@@ -3015,7 +3014,7 @@ sendfile_unix_file(struct file *file, lo
 	mutex_unlock(&inode->i_mutex);
 
 	uf_info = unix_file_inode_data(inode);
-	get_nonexclusive_access(uf_info, 0);
+	get_nonexclusive_access(uf_info);
 	result = generic_file_sendfile(file, ppos, count, actor, target);
 	drop_nonexclusive_access(uf_info);
  error:
Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/file.h
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/file/file.h
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/file.h
@@ -104,7 +104,7 @@ typedef struct unix_file_info {
 struct unix_file_info *unix_file_inode_data(const struct inode *inode);
 void get_exclusive_access(unix_file_info_t *);
 void drop_exclusive_access(unix_file_info_t *);
-void get_nonexclusive_access(unix_file_info_t *, int);
+void get_nonexclusive_access(unix_file_info_t *);
 void drop_nonexclusive_access(unix_file_info_t *);
 int try_to_get_nonexclusive_access(unix_file_info_t *);
 
Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/funcs.h
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/file/funcs.h
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/funcs.h
@@ -1,11 +1,6 @@
 /* Copyright 2001, 2002, 2003, 2004 by Hans Reiser, licensing governed by reiser4/README */
 
 /* this prototypes functions used by both file.c and tail_conversion.c */
-void get_exclusive_access(unix_file_info_t *);
-void drop_exclusive_access(unix_file_info_t *);
-void get_nonexclusive_access(unix_file_info_t *, int);
-void drop_nonexclusive_access(unix_file_info_t *);
-
 int tail2extent(unix_file_info_t *);
 int extent2tail(unix_file_info_t *);
 int finish_conversion(struct inode *inode);
Index: linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/tail_conversion.c
===================================================================
--- linux-2.6.17-rc3-mm1.orig/fs/reiser4/plugin/file/tail_conversion.c
+++ linux-2.6.17-rc3-mm1/fs/reiser4/plugin/file/tail_conversion.c
@@ -69,18 +69,12 @@ static void nea_grabbed(unix_file_info_t
  *
  * Nonexclusive access is obtained on a file before read, write, readpage.
  */
-void get_nonexclusive_access(unix_file_info_t *uf_info, int atom_may_exist)
+void get_nonexclusive_access(unix_file_info_t *uf_info)
 {
 	assert("nikita-3029", schedulable());
-	/* unix_file_filemap_nopage may call this when current atom exist already */
-	assert("nikita-3361",
-	       ergo(atom_may_exist == 0,
-		    get_current_context()->trans->atom == NULL));
-	BUG_ON(atom_may_exist == 0
-	       && get_current_context()->trans->atom != NULL);
-
+	assert("nikita-3361", get_current_context()->trans->atom == NULL);
+	BUG_ON(get_current_context()->trans->atom != NULL);
 	down_read(&uf_info->latch);
-
 	nea_grabbed(uf_info);
 }
 

--Boundary-00=_FbuYEKWWhFAJ/IU--


Kind regards,
Chris

[FYI]: Two compiler warnings in fs/reiser4/plugin/object.c with gcc-4.1.0

[Christian Trefzer]

[-- Attachment #1: Type: text/plain, Size: 1560 bytes --]

fs/reiser4/plugin/object.c:120: Warnung: Initialisierung von inkompatiblem Zeigertyp
fs/reiser4/plugin/object.c:320: Warnung: Initialisierung von inkompatiblem Zeigertyp

(in english: Warning: initialization from incompatible pointer type)

file_plugin file_plugins[LAST_FILE_PLUGIN_ID] = {
	[UNIX_FILE_PLUGIN_ID] = {
<snip>
		.as_ops = {
			.writepage = reiser4_writepage,
			.readpage = readpage_unix_file,
			.sync_page = block_sync_page,
			.writepages = writepages_unix_file,
			.set_page_dirty = reiser4_set_page_dirty,
			.readpages = reiser4_readpages,
			.prepare_write = prepare_write_unix_file,
			.commit_write =	commit_write_unix_file,
			.bmap = bmap_unix_file,
120:		.invalidatepage = reiser4_invalidatepage,
			.releasepage = reiser4_releasepage
		},
<snip>
	[CRC_FILE_PLUGIN_ID] = {
<snip>
		.as_ops = {
			.writepage = reiser4_writepage,
			.readpage = readpage_cryptcompress,
			.sync_page = block_sync_page,
			.writepages = writepages_cryptcompress,
			.set_page_dirty = reiser4_set_page_dirty,
			.readpages = reiser4_readpages,
			.prepare_write = prepare_write_common,
320:		.invalidatepage = reiser4_invalidatepage,
			.releasepage = reiser4_releasepage
		},

Another warning mentions _possibly_ uninitialized use of a variable:

fs/reiser4/plugin/space/bitmap.c: In Funktion »load_and_lock_bnode«:
fs/reiser4/plugin/space/bitmap.c:817: Warnung: »cjnode« könnte in dieser Funktion uninitialisiert verwendet werden
(»cjnode« could be used uninitialized in this function)

Kind regards,
Chris

[-- Attachment #2: Type: application/pgp-signature, Size: 829 bytes --]

Re: [FYI]: Two compiler warnings in fs/reiser4/plugin/object.c with gcc-4.1.0

[Vladimir V. Saveliev]

Hello

On Mon, 2006-05-15 at 16:53 +0200, Christian Trefzer wrote:
> fs/reiser4/plugin/object.c:120: Warnung: Initialisierung von inkompatiblem Zeigertyp
> fs/reiser4/plugin/object.c:320: Warnung: Initialisierung von inkompatiblem Zeigertyp
> 
> (in english: Warning: initialization from incompatible pointer type)
> 

What versions of linux and reiser4 do you compile?

> file_plugin file_plugins[LAST_FILE_PLUGIN_ID] = {
> 	[UNIX_FILE_PLUGIN_ID] = {
> <snip>
> 		.as_ops = {
> 			.writepage = reiser4_writepage,
> 			.readpage = readpage_unix_file,
> 			.sync_page = block_sync_page,
> 			.writepages = writepages_unix_file,
> 			.set_page_dirty = reiser4_set_page_dirty,
> 			.readpages = reiser4_readpages,
> 			.prepare_write = prepare_write_unix_file,
> 			.commit_write =	commit_write_unix_file,
> 			.bmap = bmap_unix_file,
> 120:		.invalidatepage = reiser4_invalidatepage,
> 			.releasepage = reiser4_releasepage
> 		},
> <snip>
> 	[CRC_FILE_PLUGIN_ID] = {
> <snip>
> 		.as_ops = {
> 			.writepage = reiser4_writepage,
> 			.readpage = readpage_cryptcompress,
> 			.sync_page = block_sync_page,
> 			.writepages = writepages_cryptcompress,
> 			.set_page_dirty = reiser4_set_page_dirty,
> 			.readpages = reiser4_readpages,
> 			.prepare_write = prepare_write_common,
> 320:		.invalidatepage = reiser4_invalidatepage,
> 			.releasepage = reiser4_releasepage
> 		},
> 
> Another warning mentions _possibly_ uninitialized use of a variable:
> 
> fs/reiser4/plugin/space/bitmap.c: In Funktion »load_and_lock_bnode«:
> fs/reiser4/plugin/space/bitmap.c:817: Warnung: »cjnode« könnte in dieser Funktion uninitialisiert verwendet werden
> (»cjnode« could be used uninitialized in this function)
> 
> Kind regards,
> Chris

Re: [FYI]: Two compiler warnings in fs/reiser4/plugin/object.c with gcc-4.1.0

[Christian Trefzer]

[-- Attachment #1: Type: text/plain, Size: 579 bytes --]

On Tue, May 16, 2006 at 01:35:24PM +0400, Vladimir V. Saveliev wrote:
> Hello
> 
> On Mon, 2006-05-15 at 16:53 +0200, Christian Trefzer wrote:
> > fs/reiser4/plugin/object.c:120: Warnung: Initialisierung von inkompatiblem Zeigertyp
> > fs/reiser4/plugin/object.c:320: Warnung: Initialisierung von inkompatiblem Zeigertyp
> > 
> > (in english: Warning: initialization from incompatible pointer type)
> > 
> 
> What versions of linux and reiser4 do you compile?
> 

That would be current git, which is mostly 2.6.17-rc4, with reiser4 on
top.

Kind regards,
Chris

[-- Attachment #2: Type: application/pgp-signature, Size: 829 bytes --]

Re: [FYI]: Two compiler warnings in fs/reiser4/plugin/object.c with gcc-4.1.0

[Christian Trefzer]

[-- Attachment #1: Type: text/plain, Size: 699 bytes --]

On Tue, May 16, 2006 at 01:35:24PM +0400, Vladimir V. Saveliev wrote:
> 
> What versions of linux and reiser4 do you compile?
> 
Oh, and reiser4 was reiser4-for-2.6.16-2 plus a few fixes:

01-reiser4-for-2.6.16-2.patch
02-reiser4-avoid-out-of-memory-question-bug.patch
03-reiser4-radix-tree-direct-data-fix.patch
04-reiser4-fix-block-alloc-assertions.patch

This works reliably so far for quite some time.

I might look into applying anything reiser4-related from current -mm and
checking the differences in fs/reiser4 - some of the more recent reiser4
patches are related to other changes in Andrew's tree, though, and only
mean to make reiser4 play along.

Kind regards,
Chris

[-- Attachment #2: Type: application/pgp-signature, Size: 829 bytes --]

Re: Recent patch frenzy, conflicts and Oopses

[Alexander Zarochentsev]

On Monday 15 May 2006 17:04, Christian Trefzer wrote:
> Hi guys,
>
> as I am heavily using reiser4 for some time now on my stable machine,
> I keep a git tree with vanilla -stable and reiser4 since -mm has
> sometimes been too unreliable. Lurking on this list I collected some
> patches, of which two seem to collide. The first one fixes a nasty
> "out of memory?" Oops for me and maybe others, and the second one was
> posted as a response to other Oopses on the list.
>
> Since I reversed the first one in order to apply the second one,
> thinking (wrongly) that this one should be a better fix for the same
> issues and maybe more, I have my old pal out of memory back. Maybe
> the two should be merged in a compatible way, and a new patch against
> 2.6.16-stable be put on the namesys ftp server?
>
> Anyway, here come the patches in question:
>
> From: Alexander Zarochentsev <zam@namesys.com>
> To: reiserfs-list@namesys.com
> Subject: Re: "warning("vs-44", "out of memory?")"
> Date: Wed, 12 Apr 2006 14:26:59 +0400
>
> On Wednesday 12 April 2006 13:06, Alexander Zarochentsev wrote:
> > Hello
> >
> > On Wednesday 12 April 2006 11:42, Roy Lanek wrote:
> > > I have managed to save a copy of my
> > > /var/log/socklog-klog/others/current
> > > before the next complete freezing happened.
> >
> > Please use the latest reiser4 patch reiser4-for-2.6.16-2 from
> > ftp.namesys.com.
>
> and apply the following patch:

The patch disables the BUG() but reiser4 can deadlock.  However the 
deadlock requires 3 processes to use the same EA/NEA semaphore.  

better patches are there: 

ftp.kernel.org:/pub/linux/kernel/people/akpm/patches/2.6/2.6.17-rc4/2.6.17-rc4-mm1/broken-out/
reiser4-have-get_exclusive_access-restart-transaction.patch
reiser4-add-missing-txn_restart-before-get_nonexclusive_access-calls.patch


> Index: linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/extent_file_ops.c
> ===================================================================
> ---
> linux-2.6.17-rc1-mm1.orig/fs/reiser4/plugin/item/extent_file_ops.c
> +++ linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/extent_file_ops.c @@
> -807,7 +807,7 @@ extent_balance_dirty_pages(struct inode
>  		if (excl)
>  			get_exclusive_access(uf_info);
>  		else
> -			get_nonexclusive_access(uf_info, 0);
> +			get_nonexclusive_access(uf_info, 1);
>  	}
>  	return 0;
>  }
> Index: linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/tail.c
> ===================================================================
> --- linux-2.6.17-rc1-mm1.orig/fs/reiser4/plugin/item/tail.c
> +++ linux-2.6.17-rc1-mm1/fs/reiser4/plugin/item/tail.c
> @@ -510,7 +510,7 @@ tail_balance_dirty_pages(struct address_
>  		if (excl)
>  			get_exclusive_access(uf_info);
>  		else
> -			get_nonexclusive_access(uf_info, 0);
> +			get_nonexclusive_access(uf_info, 1);
>  	}
>  	return 0;
>  }
>
>
> This one helped me a lot. The second one brought older problems back,
> but may have fixed others for some people on this list. It seems to
> remove the second argument to get_nonexclusive_access(), though:
>
> From: Alexander Zarochentsev <zam@namesys.com>
> To: reiserfs-list@namesys.com
> Subject: Re: Reproducible reiser4 bug with 2.6.16.2 patch on
> tail_conversion.c:80 Date: Thu, 11 May 2006 11:24:53 +0400
>
> --Boundary-00=_FbuYEKWWhFAJ/IU
> Content-Type: text/plain;
>   charset="iso-8859-1"
> Content-Transfer-Encoding: 7bit
> Content-Disposition: inline
>
> Hello
>
> [original message snipped]
>
> please check whether the attached patch helps.

-- 
Alex.

Re: Recent patch frenzy, conflicts and Oopses

[Christian Trefzer]

[-- Attachment #1: Type: text/plain, Size: 677 bytes --]

On Tue, May 16, 2006 at 02:32:20PM +0400, Alexander Zarochentsev wrote:
> 
> The patch disables the BUG() but reiser4 can deadlock.  However the 
> deadlock requires 3 processes to use the same EA/NEA semaphore.  
> 
> better patches are there: 
> 
> ftp.kernel.org:/pub/linux/kernel/people/akpm/patches/2.6/2.6.17-rc4/2.6.17-rc4-mm1/broken-out/
> reiser4-have-get_exclusive_access-restart-transaction.patch
> reiser4-add-missing-txn_restart-before-get_nonexclusive_access-calls.patch
> 

Thanks a bunch for clarification! It seems I am not the only one on this
list who got confused by bug reports and different approaches to fix
those.

Kind regards,
Chris

[-- Attachment #2: Type: application/pgp-signature, Size: 829 bytes --]