[PATCH] [input] Set up uinput device again, if it was previously closed.

[PATCH] [input] Set up uinput device again, if it was previously closed.

[Jeff Hansen]

If you connect a PS3 controller to bluetoothd as an input device, then
take the batteries out of a PS3 controller, then put them back in and
push a few buttons, it will eventually cause an error which causes the
uinput socket to be closed. It will then re-connect to bluetoothd, but
the uinput socket fd will be -1, so it needs to be set up again after
the re-connect.

Signed-off-by: Jeff Hansen <x@jeffhansen.com>
---
 input/fakehid.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/input/fakehid.c b/input/fakehid.c
index eebca05..a83f8c3 100644
--- a/input/fakehid.c
+++ b/input/fakehid.c
@@ -393,12 +393,15 @@ struct fake_input *fake_hid_connadd(struct fake_input *fake,
 
 	/* New device? Add it to the list of known devices,
 	 * and create the uinput necessary */
-	if (old == NULL) {
+	if (old == NULL || old->uinput == -1) {
 		if (fake_hid->setup_uinput(fake, fake_hid)) {
 			error("Error setting up uinput");
 			g_free(fake);
 			return NULL;
 		}
+	}
+
+	if (old == NULL) {
 		fake_hid->devices = g_list_append(fake_hid->devices, fake);
 	}
 
-- 
1.7.3.4

Re: [PATCH] [input] Set up uinput device again, if it was previously closed.

[Johan Hedberg]

Hi Jeff,

On Sun, Mar 04, 2012, Jeff Hansen wrote:
> If you connect a PS3 controller to bluetoothd as an input device, then
> take the batteries out of a PS3 controller, then put them back in and
> push a few buttons, it will eventually cause an error which causes the
> uinput socket to be closed. It will then re-connect to bluetoothd, but
> the uinput socket fd will be -1, so it needs to be set up again after
> the re-connect.
> 
> Signed-off-by: Jeff Hansen <x@jeffhansen.com>
> ---
>  input/fakehid.c |    5 ++++-
>  1 files changed, 4 insertions(+), 1 deletions(-)

Thanks, the patch has been pushed upstream but there were a few minor
fixes I had to do before that. Firstly, we don't use signed-off-by for
user space patches. Then, a few stylistic issues with the code:

> +	if (old == NULL || old->uinput == -1) {

We usually use < 0 for testing for an invalid fd.

> +	if (old == NULL) {
>  		fake_hid->devices = g_list_append(fake_hid->devices, fake);
>  	}

Single-line branches do not need { }

Johan