[PATCH]: set state to XFRM_STATE_DEAD before calling xfrm_state_put in pfkey_msg2xfrm_state

[PATCH]: set state to XFRM_STATE_DEAD before calling xfrm_state_put in pfkey_msg2xfrm_state

[Patrick McHardy]

[-- Attachment #1: Type: text/plain, Size: 234 bytes --]

This patch sets x->state to XFRM_STATE_DEAD before calling
xfrm_state_put in pfkey_msg2xfrm_state to avoid triggering
the BUG_TRAP in __xfrm_state_destroy. The patch applies to both
2.5 and the 2.4 backport.

Best regards,
Patrick




[-- Attachment #2: af_key-set-xfrm-dead.diff --]
[-- Type: text/plain, Size: 844 bytes --]

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
#	           ChangeSet	1.1113  -> 1.1114 
#	    net/key/af_key.c	1.35    -> 1.36   
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/05/14	kaber@trash.net	1.1114
# [IPSEC]: set state to XFRM_STATE_DEAD before calling xfrm_state_put in pfkey_msg2xfrm_state
# --------------------------------------------
#
diff -Nru a/net/key/af_key.c b/net/key/af_key.c
--- a/net/key/af_key.c	Wed May 14 16:56:52 2003
+++ b/net/key/af_key.c	Wed May 14 16:56:52 2003
@@ -1090,6 +1090,7 @@
 	return x;
 
 out:
+	x->type = XFRM_STATE_DEAD;
 	xfrm_state_put(x);
 	return ERR_PTR(-ENOBUFS);
 }

Re: [PATCH]: set state to XFRM_STATE_DEAD before calling xfrm_state_put in pfkey_msg2xfrm_state

[David S. Miller]

   From: Patrick McHardy <kaber@trash.net>
   Date: Wed, 14 May 2003 17:00:13 +0200

   This patch sets x->state to XFRM_STATE_DEAD before calling
   xfrm_state_put in pfkey_msg2xfrm_state to avoid triggering
   the BUG_TRAP in __xfrm_state_destroy. The patch applies to both
   2.5 and the 2.4 backport.
   
You didn't test this change.  x->type is a pointer, not a place where
you put XFRM_STATE_DEAD, a simple compile test would have alerted
this to you.

This also means you couldn't possibly have tested if this even
makes the assertion go away, it couldn't possibly have fixed this..

The correct fix, of course, is to set x->km.state to this value.
This is what I've done in my tree.

Please be a LOT more careful with your changes.

Re: [PATCH]: set state to XFRM_STATE_DEAD before calling xfrm_state_put in pfkey_msg2xfrm_state

[Patrick McHardy]

David S. Miller wrote:

>You didn't test this change.  x->type is a pointer, not a place where
>you put XFRM_STATE_DEAD, a simple compile test would have alerted
>this to you.
>  
>

Sorry about that, i recreated the fix in a different tree and broke it.

>This also means you couldn't possibly have tested if this even
>makes the assertion go away, it couldn't possibly have fixed this..
>
>The correct fix, of course, is to set x->km.state to this value.
>This is what I've done in my tree.
>
>Please be a LOT more careful with your changes.
>

Promised.