problems with JFFS2 assert failure on mutex

problems with JFFS2 assert failure on mutex

[Øyvind Harboe]

[-- Attachment #1: Type: text/plain, Size: 238 bytes --]

I was working on a different problem when I came across a mutex that
was being obtained twice(discovered via eCos asserts).

On the surface of it, this patch looks correct.

Comments?


-- 
Øyvind Harboe
http://www.zylin.com


[-- Attachment #2: mutex.txt --]
[-- Type: text/x-patch, Size: 546 bytes --]

Index: gc.c
===================================================================
RCS file: /cvs/ecos/ecos/packages/fs/jffs2/current/src/gc.c,v
retrieving revision 1.7
diff -w -u -r1.7 gc.c
--- gc.c	1 Apr 2004 03:17:57 -0000	1.7
+++ gc.c	19 Jul 2004 12:54:18 -0000
@@ -358,10 +358,10 @@
 	spin_unlock(&c->inocache_lock);
 
 	f = jffs2_gc_fetch_inode(c, inum, nlink);
-	if (IS_ERR(f))
-		return PTR_ERR(f);
-	if (!f)
-		return 0;
+	if (!f||IS_ERR(f)) {
+	  up(&c->alloc_sem);
+		return f;
+	}
 
 	ret = jffs2_garbage_collect_live(c, jeb, raw, f);
 

Re: problems with JFFS2 assert failure on mutex

[David Woodhouse]

On Mon, 2004-07-19 at 17:17 +0200, Øyvind Harboe wrote:
> I was working on a different problem when I came across a mutex that
> was being obtained twice(discovered via eCos asserts).
> 
> On the surface of it, this patch looks correct.

Looks correct; I've committed an equivalent to CVS. Thanks.

> Comments?

Only that this error path should never happen. The call to
jffs2_gc_fetch_inode() for a valid inode should _never_ fail.

-- 
dwmw2

Re: problems with JFFS2 assert failure on mutex

[Øyvind Harboe]

tir, 20.07.2004 kl. 15.41 skrev David Woodhouse:
> On Mon, 2004-07-19 at 17:17 +0200, Øyvind Harboe wrote:
> > I was working on a different problem when I came across a mutex that
> > was being obtained twice(discovered via eCos asserts).
> > 
> > On the surface of it, this patch looks correct.
> 
> Looks correct; I've committed an equivalent to CVS. Thanks.
> 
> > Comments?
> 
> Only that this error path should never happen. The call to
> jffs2_gc_fetch_inode() for a valid inode should _never_ fail.

I believe this was with a corrupt JFFS2 disk.

-- 
Øyvind Harboe
http://www.zylin.com