[PATCH] HID: sony: disable descriptor fixup for FutureMax Dance Mat

From: Mikko Perttunen <mperttunen@nvidia.com>

The FutureMax Dance Mat claims to be a SixAxis controller
but breaks if descriptor fixups are applied. Detect the
device using its USB product string and disable fixups
when it is detected.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
---
 drivers/hid/hid-sony.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/drivers/hid/hid-sony.c b/drivers/hid/hid-sony.c
index 310436a..4c976f7 100644
--- a/drivers/hid/hid-sony.c
+++ b/drivers/hid/hid-sony.c

@@ -36,6 +36,9 @@
 #include <linux/list.h>
 #include <linux/idr.h>
 #include <linux/input/mt.h>
+#include <linux/usb.h>
+
+#include "usbhid/usbhid.h"
 
 #include "hid-ids.h"
 

@@ -51,6 +54,7 @@
 #define NAVIGATION_CONTROLLER_USB BIT(9)
 #define NAVIGATION_CONTROLLER_BT  BIT(10)
 #define SINO_LITE_CONTROLLER      BIT(11)
+#define FUTUREMAX_DANCE_MAT       BIT(12)
 
 #define SIXAXIS_CONTROLLER (SIXAXIS_CONTROLLER_USB | SIXAXIS_CONTROLLER_BT)
 #define MOTION_CONTROLLER (MOTION_CONTROLLER_USB | MOTION_CONTROLLER_BT)

@@ -1125,7 +1129,7 @@ static u8 *sony_report_fixup(struct hid_device *hdev, u8 *rdesc,
 {
 	struct sony_sc *sc = hid_get_drvdata(hdev);
 
-	if (sc->quirks & SINO_LITE_CONTROLLER)
+	if (sc->quirks & (SINO_LITE_CONTROLLER | FUTUREMAX_DANCE_MAT))
 		return rdesc;
 
 	/*

@@ -2288,6 +2292,15 @@ static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 	unsigned long quirks = id->driver_data;
 	struct sony_sc *sc;
 	unsigned int connect_mask = HID_CONNECT_DEFAULT;
+	struct usb_device *usb_dev;
+
+	if (quirks & SIXAXIS_CONTROLLER_USB) {
+		usb_dev = hid_to_usb_dev(hdev);
+		if (usb_dev && usb_dev->product &&
+		    !strcmp(usb_dev->product, "FutureMax Dance Mat")) {
+			quirks |= FUTUREMAX_DANCE_MAT;
+		}
+	}
 
 	sc = devm_kzalloc(&hdev->dev, sizeof(*sc), GFP_KERNEL);
 	if (sc == NULL) {

-- 
2.8.2

On Jul 17 2016 or thereabouts, Mikko Perttunen wrote:

I spent a lot of effort 2 years ago to remove the usb dependency, I'd
prefer not adding a strong deps on it again.

If they decided to reuse the same PID than one existing, and you have no
other choice but to rely on the name, you can simply have a match on
hdev->name instead of adding the USB dependency.

Also, I think it would be better to have a check on the existing report
descriptor instead of blindly fixing it. Other drivers make sure they
are fixing the correct region before overriding it, but I think using a
md5sum might be just easier (though that's not required for your patch I
think).

Cheers,
Benjamin

+			quirks |= FUTUREMAX_DANCE_MAT;
+		}
+	}
 
 	sc = devm_kzalloc(&hdev->dev, sizeof(*sc), GFP_KERNEL);
 	if (sc == NULL) {
-- 
2.8.2

On 07/18/16 17:28, Benjamin Tissoires wrote:

On Jul 17 2016 or thereabouts, Mikko Perttunen wrote:

quoted

From: Mikko Perttunen <mperttunen@nvidia.com>
...
 #include <linux/list.h>
 #include <linux/idr.h>
 #include <linux/input/mt.h>
+#include <linux/usb.h>
+
+#include "usbhid/usbhid.h"

I spent a lot of effort 2 years ago to remove the usb dependency, I'd
prefer not adding a strong deps on it again.

I see, I'll remove it.

quoted

 #include "hid-ids.h"

@@ -51,6 +54,7 @@
 #define NAVIGATION_CONTROLLER_USB BIT(9)
 #define NAVIGATION_CONTROLLER_BT  BIT(10)
 #define SINO_LITE_CONTROLLER      BIT(11)
+#define FUTUREMAX_DANCE_MAT       BIT(12)

 #define SIXAXIS_CONTROLLER (SIXAXIS_CONTROLLER_USB | SIXAXIS_CONTROLLER_BT)
 #define MOTION_CONTROLLER (MOTION_CONTROLLER_USB | MOTION_CONTROLLER_BT)

@@ -1125,7 +1129,7 @@ static u8 *sony_report_fixup(struct hid_device *hdev, u8 *rdesc,
 {
 	struct sony_sc *sc = hid_get_drvdata(hdev);

-	if (sc->quirks & SINO_LITE_CONTROLLER)
+	if (sc->quirks & (SINO_LITE_CONTROLLER | FUTUREMAX_DANCE_MAT))
 		return rdesc;

 	/*

@@ -2288,6 +2292,15 @@ static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 	unsigned long quirks = id->driver_data;
 	struct sony_sc *sc;
 	unsigned int connect_mask = HID_CONNECT_DEFAULT;
+	struct usb_device *usb_dev;
+
+	if (quirks & SIXAXIS_CONTROLLER_USB) {
+		usb_dev = hid_to_usb_dev(hdev);
+		if (usb_dev && usb_dev->product &&
+		    !strcmp(usb_dev->product, "FutureMax Dance Mat")) {i

If they decided to reuse the same PID than one existing, and you have no
other choice but to rely on the name, you can simply have a match on
hdev->name instead of adding the USB dependency.

Ah, didn't notice this!

Also, I think it would be better to have a check on the existing report
descriptor instead of blindly fixing it. Other drivers make sure they
are fixing the correct region before overriding it, but I think using a
md5sum might be just easier (though that's not required for your patch I
think).

I agree. In fact, the mat used to work back in 3.15 when the driver did 
some rudimentary checks before overwriting the descriptor. In any case, 
it would be difficult for me to add the checks since the mat is the only 
piece of hardware I own that is handled by this driver.

Cheers,
Benjamin

quoted

 ...

Cheers, and thanks for the review! I'll post a v2 once I get the chance.

Mikko

On Tue, 19 Jul 2016 09:08:27 +0300
Mikko Perttunen [off-list ref] wrote:

On 07/18/16 17:28, Benjamin Tissoires wrote:

[...]

quoted

Also, I think it would be better to have a check on the existing report
descriptor instead of blindly fixing it. Other drivers make sure they
are fixing the correct region before overriding it, but I think using a
md5sum might be just easier (though that's not required for your patch I
think).

I agree. In fact, the mat used to work back in 3.15 when the driver did 
some rudimentary checks before overwriting the descriptor. In any case, 
it would be difficult for me to add the checks since the mat is the only 
piece of hardware I own that is handled by this driver.

The fact that the report gets fixed up unconditionally is from
commit c607fb8 (HID: sony: Always override the Sixaxis descriptor).

The rationale behind this change was that compatible devices are
supposed to send exactly the same _raw_ data as the original ones, so
overriding the _whole_ descriptor unconditionally should keep them
working.

I would be interesting to understand why some (supposedly) compatible
devices break, maybe they rely on the fact that the PS3 would use the
original report?

BTW I'd slightly prefer the current way of having a straight path for
the common case (original and actually compatible devices) and handle
the exceptions explicitly using quirks, but I am open to changes and I
can test patches with some original controllers if needed.

Ciao ciao,
   Antonio

-- 
Antonio Ospite
http://ao2.it

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

On Jul 21 2016 or thereabouts, Antonio Ospite wrote:

On Tue, 19 Jul 2016 09:08:27 +0300
Mikko Perttunen [off-list ref] wrote:

quoted

On 07/18/16 17:28, Benjamin Tissoires wrote:

[...]

quoted

quoted

Also, I think it would be better to have a check on the existing report
descriptor instead of blindly fixing it. Other drivers make sure they
are fixing the correct region before overriding it, but I think using a
md5sum might be just easier (though that's not required for your patch I
think).

I agree. In fact, the mat used to work back in 3.15 when the driver did 
some rudimentary checks before overwriting the descriptor. In any case, 
it would be difficult for me to add the checks since the mat is the only 
piece of hardware I own that is handled by this driver.

The fact that the report gets fixed up unconditionally is from
commit c607fb8 (HID: sony: Always override the Sixaxis descriptor).

The rationale behind this change was that compatible devices are
supposed to send exactly the same _raw_ data as the original ones, so
overriding the _whole_ descriptor unconditionally should keep them
working.

I would be interesting to understand why some (supposedly) compatible
devices break, maybe they rely on the fact that the PS3 would use the
original report?

[just speculating]
I would be *really* surprised if the Sony software were to overwrite the
report descriptor of their own sony controllers. My guess is that their
HID stack or their Playstation driver is compatible with the original
report descriptors and so they can handle the differences between
original and weirdo controllers.

BTW I'd slightly prefer the current way of having a straight path for
the common case (original and actually compatible devices) and handle
the exceptions explicitly using quirks, but I am open to changes and I
can test patches with some original controllers if needed.

Would the v2 of Mikko's patch be OK with you?

Cheers,
Benjamin

Ciao ciao,
   Antonio

-- 
Antonio Ospite
http://ao2.it

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

On Mon, 25 Jul 2016 11:14:04 +0200
Benjamin Tissoires [off-list ref] wrote:

On Jul 21 2016 or thereabouts, Antonio Ospite wrote:

[...]

quoted

It would be interesting to understand why some (supposedly) compatible
devices break, maybe they rely on the fact that the PS3 would use the
original report?

[just speculating]
I would be *really* surprised if the Sony software were to overwrite the
report descriptor of their own sony controllers. My guess is that their
HID stack or their Playstation driver is compatible with the original
report descriptors and so they can handle the differences between
original and weirdo controllers.

That's basically what I was thinking too, in my last sentence I meant
to write report _descriptor_ :)

Mikko, if you have some time, could you post the original report
descriptor for your device?

$ hexdump /sys/bus/hid/devices/XXXX\:YYYY\:ZZZZ.0001/report_descriptor

or

$ hidrd-convert -i natv -o code /sys/bus/hid/devices/XXXX\:YYYY\:ZZZZ.0001/report_descriptor

You can also find out the length of the raw output report with trial and
error, start with a line like this:

$ sudo hexdump -v -e '49/1 "%02x " "\n"' /dev/hidraw0

and increase/decrease the value (49 works with original controllers)
until you see a regular pattern when comparing lines.

Anyways this is just for curiosity, we are not going to spend too much
time on that... or are we? :)

Thanks,
   Antonio

-- 
Antonio Ospite
http://ao2.it

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

On 07/25/16 14:05, Antonio Ospite wrote:

On Mon, 25 Jul 2016 11:14:04 +0200
Benjamin Tissoires [off-list ref] wrote:

quoted

On Jul 21 2016 or thereabouts, Antonio Ospite wrote:

[...]

quoted

quoted

It would be interesting to understand why some (supposedly) compatible
devices break, maybe they rely on the fact that the PS3 would use the
original report?

[just speculating]
I would be *really* surprised if the Sony software were to overwrite the
report descriptor of their own sony controllers. My guess is that their
HID stack or their Playstation driver is compatible with the original
report descriptors and so they can handle the differences between
original and weirdo controllers.

That's basically what I was thinking too, in my last sentence I meant
to write report _descriptor_ :)

Mikko, if you have some time, could you post the original report
descriptor for your device?

Sure, here's the output from hidrd-convert:

0x05, 0x01, /*  Usage Page (Desktop),               */
0x09, 0x04, /*  Usage (Joystick),                   */
0xA1, 0x01, /*  Collection (Application),           */
0x09, 0x01, /*      Usage (Pointer),                */
0xA1, 0x00, /*      Collection (Physical),          */
0x05, 0x09, /*          Usage Page (Button),        */
0x19, 0x01, /*          Usage Minimum (01h),        */
0x29, 0x0A, /*          Usage Maximum (0Ah),        */
0x15, 0x00, /*          Logical Minimum (0),        */
0x25, 0x01, /*          Logical Maximum (1),        */
0x75, 0x01, /*          Report Size (1),            */
0x95, 0x0A, /*          Report Count (10),          */
0x81, 0x02, /*          Input (Variable),           */
0x95, 0x06, /*          Report Count (6),           */
0x81, 0x03, /*          Input (Constant, Variable), */
0xC0,       /*      End Collection,                 */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x15, 0x00, /*      Logical Minimum (0),            */
0x25, 0x01, /*      Logical Maximum (1),            */
0x25, 0x01, /*      Logical Maximum (1),            */
0x25, 0x01, /*      Logical Maximum (1),            */
0xC0        /*  End Collection                      */

$ hexdump /sys/bus/hid/devices/XXXX\:YYYY\:ZZZZ.0001/report_descriptor

or

$ hidrd-convert -i natv -o code /sys/bus/hid/devices/XXXX\:YYYY\:ZZZZ.0001/report_descriptor

You can also find out the length of the raw output report with trial and
error, start with a line like this:

$ sudo hexdump -v -e '49/1 "%02x " "\n"' /dev/hidraw0

and increase/decrease the value (49 works with original controllers)
until you see a regular pattern when comparing lines.

/dev/hidraw0 seems to be all zeroes straight from the beginning

Anyways this is just for curiosity, we are not going to spend too much
time on that... or are we? :)

At least it's some reference information.. :)

Thanks,
   Antonio

Cheers,
Mikko.

On 07/25/16 18:23, Mikko Perttunen wrote:

On 07/25/16 14:05, Antonio Ospite wrote:

quoted

You can also find out the length of the raw output report with trial and
error, start with a line like this:

$ sudo hexdump -v -e '49/1 "%02x " "\n"' /dev/hidraw0

and increase/decrease the value (49 works with original controllers)
until you see a regular pattern when comparing lines.

/dev/hidraw0 seems to be all zeroes straight from the beginning

Actually, pressing some buttons on the pad, it seems that the report 
length is two bytes

quoted

Anyways this is just for curiosity, we are not going to spend too much
time on that... or are we? :)

At least it's some reference information.. :)

quoted

Thanks,
   Antonio

Cheers,
Mikko.