Hello,
- currently, when performing a CreateDevice, input service connects
then
disconnects from remote device. This makes some tests failed.
What do you think if this does not disconnect (and create classic
device
connection) ?
does using CreateSecureDevice helps? Can you verify with bluez-
utils-3.29 and see if it is still a problem?
I tried with CreateSecureDevice and latest 3.29 version but PTS
tests failed.
I think this is related to the disconnection in Create*Device process.
What do you think if disconnection is removed from this process ? Or
add a new API, something like CreateAndConnectSecureDevice ?
can you please check the service records of the PTS. Does it set the
reconnect value to TRUE. Please check hcidump and see if CreateDevice
actually connects the control and interrupt PSM.
If I do not make any mistake it seems to me that reconnect is set to
false (see attached record - PTS_HID_record.txt).
Create Device connects control and interrupt PSMs (see
PTS_HID_createdevice_dump.txt), but not HID kernel part.
Does it help if you call CreateDevice() and then Connect() to pass
the test.
Yes, for most of the tests. But for one test (TC_EXT_HOS_CON_BV_0201_I)
that check if pairing and connection are available, this makes it failed
(CreateSecureDevice then Connect).
I create a patch that allows to create a device and keep it connected (non-secure and secure version) in input manager hierarchy.
It also add a method in device hierarchy to unplug virtual cable without removing device from registered input device list.
This patch allows to pass PTS tests.
Is it OK ?
- when kernel (2.6.22) receives a VIRTUAL_CABLE_UNPLUG from PTS
(tests
TC_HOS_HCR_BV_02_I and TC_HOS_HCR_BV_04_I) it does not close the
L2CAP
channels as PTS waits for.
I take a look in 2.6.22 sources, and it seems to me this
disconnection
should happen, am I right ?
Is it fixed in later kernel ?
I recall that there was a fix for this.
Do you remember if this has been integrated in kernel ?
I take a look to latest kernel (2.6.24) and do not see any change
that seems to be related to this problem.
Please include the hcidump -X -V output for this case.
See PTS_HID_virtualcableunplug_dump.txt.
I also add some traces in kernel (see
PTS_HID_virtualcableunplug_kernel.patch) and attached results (see
PTS_HID_virtualcableunplug_kernellog.txt).
Did traces help ?
Is it possible that hidp code in kernel send data on the control socket when it receives a Virtual Cable Unplug, so input service can be informed of this and
perform the l2cap sockets disconnection ?
What is your opinion about that ?
Regards
Fred
--
-----------------------------------------------
It is not by improving the oil lamp that one invents the electric bulb!
-----------------------------------------------
Danis Frederic Access Company
Software engineer
Mail : mailto:frederic.danis@xxxxxxxxxxxxxxxxxx
-----------------------------------------------
Index: device.c
===================================================================
RCS file: /cvsroot/bluez/utils/input/device.c,v
retrieving revision 1.71
diff -a -u -r1.71 device.c
--- device.c 14 Mar 2008 18:44:00 -0000 1.71
+++ device.c 31 Mar 2008 08:36:28 -0000
@@ -1039,6 +1039,19 @@
return send_message_and_unref(conn, reply);
}
+static DBusHandlerResult device_unplug(DBusConnection *conn,
+ DBusMessage *msg, void *data)
+{
+ struct device *idev = data;
+
+ if (disconnect(idev, (1 << HIDP_VIRTUAL_CABLE_UNPLUG)) < 0)
+ return error_failed_errno(conn, msg, errno);
+
+ /* Replying to the requestor */
+ return send_message_and_unref(conn,
+ dbus_message_new_method_return(msg));
+}
+
static void device_unregister(DBusConnection *conn, void *data)
{
struct device *idev = data;
@@ -1057,6 +1070,7 @@
{ "GetName", device_get_name, "", "s" },
{ "GetProductId", device_get_product_id, "", "q" },
{ "GetVendorId", device_get_vendor_id, "", "q" },
+ { "Unplug", device_unplug, "", "" },
{ NULL, NULL, NULL, NULL }
};
Index: input-api.txt
===================================================================
RCS file: /cvsroot/bluez/utils/input/input-api.txt,v
retrieving revision 1.12
diff -a -u -r1.12 input-api.txt
--- input-api.txt 26 Nov 2007 13:41:38 -0000 1.12
+++ input-api.txt 31 Mar 2008 08:36:28 -0000
@@ -39,6 +39,32 @@
org.bluez.Error.ConnectionAttemptFailed
org.bluez.Error.Failed
+ string CreateAndConnectDevice(string address)
+
+ Create an input device object.
+
+ On success it will return the path of the
+ newly created device object.
+
+ Possible errors: org.bluez.Error.AlreadyExists
+ org.bluez.Error.NotSupported
+ org.bluez.Error.ConnectionAttemptFailed
+ org.bluez.Error.Failed
+
+ string CreateAndConnectSecureDevice(string address)
+
+ Create an input device object. Pairing will
+ be initiated if needed(keyboard/combo devices).
+
+ On success it will return the path of the
+ newly created device object.
+
+ Possible errors: org.bluez.Error.AlreadyExists
+ org.bluez.Error.NotSupported
+ org.bluez.Error.ConnectionAttemptFailed
+ org.bluez.Error.AuthenticationFailed
+ org.bluez.Error.Failed
+
void RemoveDevice(string path)
Remove the input device object for a given path.
@@ -98,6 +124,12 @@
Possible errors: org.bluez.Error.Failed
+ void Unplug()
+
+ Unplug virtual cable of the input device.
+
+ Possible errors: org.bluez.Error.Failed
+
Signals void Connected()
void Disconnected()
Index: manager.c
===================================================================
RCS file: /cvsroot/bluez/utils/input/manager.c,v
retrieving revision 1.39
diff -a -u -r1.39 manager.c
--- manager.c 27 Mar 2008 14:05:55 -0000 1.39
+++ manager.c 31 Mar 2008 08:36:28 -0000
@@ -337,6 +337,33 @@
device_paths = g_slist_append(device_paths, g_strdup(path));
+ if ((strcmp("CreateAndConnectDevice", dbus_message_get_member(pr->msg)) == 0) ||
+ (strcmp("CreateAndConnectSecureDevice", dbus_message_get_member(pr->msg)) == 0)) {
+ if ((err = input_device_set_channel(&pr->src, &pr->dst, L2CAP_PSM_HIDP_CTRL, pr->ctrl_sock)) < 0) {
+ /* Send unplug virtual cable to unknown devices */
+ unsigned char unplug[] = { 0x15 };
+ write(pr->ctrl_sock, unplug, sizeof(unplug));
+
+ goto failed;
+ }
+
+ pr->ctrl_sock = -1;
+
+ if ((err = input_device_set_channel(&pr->src, &pr->dst, L2CAP_PSM_HIDP_INTR, isk)) < 0) {
+ input_device_close_channels(&pr->src, &pr->dst);
+
+ goto failed;
+ }
+
+ isk = -1;
+
+ if ((err = input_device_connadd(&pr->src, &pr->dst)) < 0) {
+ input_device_close_channels(&pr->src, &pr->dst);
+
+ goto failed;
+ }
+ }
+
/* Replying to the requestor */
reply = dbus_message_new_method_return(pr->msg);
@@ -354,7 +381,8 @@
if (isk >= 0)
close(isk);
- close(pr->ctrl_sock);
+ if (pr->ctrl_sock >= 0)
+ close(pr->ctrl_sock);
pending_req_free(pr);
if (hidp.rd_data)
@@ -544,7 +572,8 @@
dbus_message_unref(reply);
- if (strcmp("CreateSecureDevice", dbus_message_get_member(pr->msg)) == 0) {
+ if ((strcmp("CreateSecureDevice", dbus_message_get_member(pr->msg)) == 0) ||
+ (strcmp("CreateAndConnectSecureDevice", dbus_message_get_member(pr->msg)) == 0)) {
sdp_data_t *d;
/* Pairing mandatory for keyboard and combo */
@@ -1136,6 +1165,8 @@
{ "ListDevices", list_devices, "", "as" },
{ "CreateDevice", create_device, "s", "s" },
{ "CreateSecureDevice", create_device, "s", "s" },
+ { "CreateAndConnectDevice", create_device, "s", "s" },
+ { "CreateAndConnectSecureDevice", create_device, "s", "s" },
{ "RemoveDevice", remove_device, "s", "" },
{ NULL, NULL, NULL, NULL },
};
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
Bluez-devel mailing list
Bluez-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/bluez-devel
