[Bluez-devel] [PATCH] Deleted acl-handles on failed hci_create_connection

[Bluez-devel] [PATCH] Deleted acl-handles on failed hci_create_connection

[Andreas Gaufer]

[-- Attachment #1: Type: text/plain, Size: 569 bytes --]

Hi,

Heres a patch for hci_event.c to fix the problem that acl-handles are removed 
from the kernels connection hash when a hci_create_connection returnes a
command status thats not 0x00.

One of the problems that was caused by this bug was that cc could be issued 
even if the peer was already connected. This case leads to the error message 
"Too many links" with sdptool and rfcomm for example. This connection could 
only be removed if the acl handle was known otherwise or by removing power 
from the usb-dongle (unplugg).

Have Fun!

Greetings

Andreas Gaufer
 


[-- Attachment #2: patch-2.4.26-ag1.gz --]
[-- Type: application/octet-stream, Size: 388 bytes --]

Re: [Bluez-devel] [PATCH] Deleted acl-handles on failed hci_create_connection

[Marcel Holtmann]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]

Hi Andreas,

> Heres a patch for hci_event.c to fix the problem that acl-handles are removed 
> from the kernels connection hash when a hci_create_connection returnes a
> command status thats not 0x00.
> 
> One of the problems that was caused by this bug was that cc could be issued 
> even if the peer was already connected. This case leads to the error message 
> "Too many links" with sdptool and rfcomm for example. This connection could 
> only be removed if the acl handle was known otherwise or by removing power 
> from the usb-dongle (unplugg).

I think your patch is wrong, because in case of the first try through
L2CAP etc. a negative return value must result in a drop of the hci_conn
structure. Check if the attached patch is also working (againt 2.6).

Regards

Marcel


[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 411 bytes --]

===== net/bluetooth/hci_event.c 1.13 vs edited =====
--- 1.13/net/bluetooth/hci_event.c	2004-05-31 22:15:12 +02:00
+++ edited/net/bluetooth/hci_event.c	2004-06-10 20:48:55 +02:00
@@ -358,7 +358,7 @@
 			status, batostr(&cp->bdaddr), conn);
 
 	if (status) {
-		if (conn) {
+		if (conn && conn->state == BT_CONNECT) {
 			conn->state = BT_CLOSED;
 			hci_proto_connect_cfm(conn, status);
 			hci_conn_del(conn);

Re: [Bluez-devel] [PATCH] Deleted acl-handles on failed hci_create_connection

[Andreas Gaufer]

[-- Attachment #1: Type: text/plain, Size: 1152 bytes --]

Hi Marcel,

Your right (like always ;). I assumed that in case af a negative return in 
first try there is no entry created but that was wrong. 

Your change works for me, a patch for 2.4.26 is attached

Greetings & Regards

Andreas Gaufer



On Thu, 10 Jun 2004 20:53:02 +0200
Marcel Holtmann <marcel@holtmann.org> wrote:

> Hi Andreas,
> 
> > Heres a patch for hci_event.c to fix the problem that acl-handles are removed 
> > from the kernels connection hash when a hci_create_connection returnes a
> > command status thats not 0x00.
> > 
> > One of the problems that was caused by this bug was that cc could be issued 
> > even if the peer was already connected. This case leads to the error message 
> > "Too many links" with sdptool and rfcomm for example. This connection could 
> > only be removed if the acl handle was known otherwise or by removing power 
> > from the usb-dongle (unplugg).
> 
> I think your patch is wrong, because in case of the first try through
> L2CAP etc. a negative return value must result in a drop of the hci_conn
> structure. Check if the attached patch is also working (againt 2.6).
> 
> Regards
> 
> Marcel
> 
> 

[-- Attachment #2: patch-2.4.26 --]
[-- Type: application/octet-stream, Size: 504 bytes --]

diff --unified --recursive -I /rootgauf/ --new-file linux-2.4.26/net/bluetooth/hci_event.c linux-2.4.26-mh/net/bluetooth/hci_event.c
--- linux-2.4.26/net/bluetooth/hci_event.c	Wed Feb 18 14:36:32 2004
+++ linux-2.4.26-mh/net/bluetooth/hci_event.c	Thu Jun 10 22:13:07 2004
@@ -320,7 +320,7 @@
 			status, batostr(&cc->bdaddr), conn);
 
 	if (status) {
-		if (conn) {
+		if (conn && conn->state == BT_CONNECT) {
 			conn->state = BT_CLOSED;
 			hci_proto_connect_cfm(conn, status);
 			hci_conn_del(conn);