[dm-devel] Patchset 2.4.20-dm-8

[Kevin Corry]

On Thursday 13 February 2003 05:32, Joe Thornber wrote:
>
> Changes since 2.4.20-dm-7
> -------------------------
>
> Revision 41:
>   Hook the io before we call the targets map function, this allows us to
>   implement little things like the zero target.

Wahoo! :)  Mike Tran and I noticed this yesterday, and were about to send you 
a nearly identical patch. Looks like you beat us to it.

> Revision 42:
>   Silly mistake in error path when an unknown target type is requested.

--- diff/drivers/md/dm-table.c	2003-01-10 10:20:07.000000000 +0000
+++ source/drivers/md/dm-table.c	2003-02-13 10:42:19.000000000 +0000
@@ -523,7 +523,7 @@
 	tgt->type = dm_get_target_type(type);
 	if (!tgt->type) {
 		tgt->error = "unknown target type";
-		goto bad;
+		return r;
 	}
 
 	tgt->table = t;


Should this patch actually be "return -EINVAL;" (or something similar)? Just 
above this, the call to check_space() sets r to 0, so this patch is going to 
return success to the caller of dm_table_add_target() in this error path. 
Along the same lines, the error path for the call to adjoin() will also end 
up returning success from this function.

Patch 5 in the new 2.5.60 patchset has the same issue. The patch below ought 
to apply to both 2.4 and 2.5.

-- 
Kevin Corry
corryk at us.ibm.com
http://evms.sourceforge.net/



--- linux-2.4.20a/drivers/md/dm-table.c	2003/02/13 16:18:01
+++ linux-2.4.20b/drivers/md/dm-table.c	2003/02/13 16:19:39
@@ -523,7 +523,7 @@
 	tgt->type = dm_get_target_type(type);
 	if (!tgt->type) {
 		tgt->error = "unknown target type";
-		return r;
+		return -EINVAL;
 	}
 
 	tgt->table = t;
@@ -536,6 +536,7 @@
 	 */
 	if (!adjoin(t, tgt)) {
 		tgt->error = "Gap in table";
+		r = -EINVAL;
 		goto bad;
 	}