HID: multitouch: do not blindly set EV_KEY or EV_ABS bits

PATCH In review Dmitry Torokhov 3 participants 3 messages Jul 11, 2017 – Jul 13, 2017 HID

Share & tools

Dmitry Torokhov

author

2017-07-11 17:02 UTC #0 lore raw

PATCH 1/1 HID: multitouch: do not blindly set EV_KEY or EV_ABS bits

Now that input core insists on having dev->absinfo when device claims to generate EV_ABS in its dev->evbit, we should not be blindly setting that bit.

The code in question might have been needed before input_set_abs_params() started setting EV_ABS in device's evbit, but not anymore, and is now breaking devices such as SMART SPNL-6075 Touchscreen.

Fixes 6ecfe51b4082 ("Input: refuse to register absolute devices ...")

Reported-by Matthias Fend <[email protected]>

Tested-by Matthias Fend <[email protected]>

Cc [email protected]

Signed-off-by Dmitry Torokhov <[email protected]>

---

Jiri, Benjamin, I marked it for stable as it causes regressions in 4.12 reported by users. Please take a look.

 drivers/hid/hid-multitouch.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

drivers/hid/hid-multitouch.c +4 ~~-12~~

index 24d5b6deb571..be5bea15ce52 100644
@@ -595,16 +595,6 @@ static int mt_touch_input_mapping(struct hid_device *hdev, struct hid_input *hi,
 	return 0;
 }
 
-static int mt_touch_input_mapped(struct hid_device *hdev, struct hid_input *hi,
-		struct hid_field *field, struct hid_usage *usage,
-		unsigned long **bit, int *max)
-{
-	if (usage->type == EV_KEY || usage->type == EV_ABS)
-		set_bit(usage->type, hi->input->evbit);
-
-	return -1;
-}
-
 static int mt_compute_slot(struct mt_device *td, struct input_dev *input)
 {
 	__s32 quirks = td->mtclass.quirks;

@@ -905,8 +895,10 @@ static int mt_input_mapped(struct hid_device *hdev, struct hid_input *hi,
 		return 0;
 
 	if (field->application == HID_DG_TOUCHSCREEN ||
-	    field->application == HID_DG_TOUCHPAD)
-		return mt_touch_input_mapped(hdev, hi, field, usage, bit, max);
+	    field->application == HID_DG_TOUCHPAD) {
+		/* We own these mappings, tell hid-input to ignore them */
+		return -1;
+	}
 
 	/* let hid-core decide for the others */
 	return 0;

Signature · 5 lines

2.13.2.725.g09c95d1e9-goog


-- 
Dmitry

Benjamin Tissoires

Reviewed-by on v1 1/1

2017-07-12 07:38 UTC #1 lore raw

Quoted 13 lines · Dmitry Torokhov

Now that input core insists on having dev->absinfo when device claims to generate EV_ABS in its dev->evbit, we should not be blindly setting that bit.
The code in question might have been needed before input_set_abs_params() started setting EV_ABS in device's evbit, but not anymore, and is now breaking devices such as SMART SPNL-6075 Touchscreen.
Fixes 6ecfe51b4082 ("Input: refuse to register absolute devices ...")
Reported-by Matthias Fend <[email protected]>
Tested-by Matthias Fend <[email protected]>
Cc [email protected]
Signed-off-by Dmitry Torokhov <[email protected]>

Quoted 4 lines

---
Jiri, Benjamin, I marked it for stable as it causes regressions in 4.12 reported by users. Please take a look.

set_bit(usage->type, input->evbit); is called by hid-input.c when the return code is positive, so we should be safe here.

Reviewed-by Benjamin Tissoires <[email protected]>

Cheers,
Benjamin

Quoted 44 lines

 drivers/hid/hid-multitouch.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

drivers/hid/hid-multitouch.c +4 ~~-12~~

index 24d5b6deb571..be5bea15ce52 100644
@@ -595,16 +595,6 @@ static int mt_touch_input_mapping(struct hid_device *hdev, struct hid_input *hi,
 	return 0;
 }
 
-static int mt_touch_input_mapped(struct hid_device *hdev, struct hid_input *hi,
-		struct hid_field *field, struct hid_usage *usage,
-		unsigned long **bit, int *max)
-{
-	if (usage->type == EV_KEY || usage->type == EV_ABS)
-		set_bit(usage->type, hi->input->evbit);
-
-	return -1;
-}
-
 static int mt_compute_slot(struct mt_device *td, struct input_dev *input)
 {
 	__s32 quirks = td->mtclass.quirks;

@@ -905,8 +895,10 @@ static int mt_input_mapped(struct hid_device *hdev, struct hid_input *hi,
 		return 0;
 
 	if (field->application == HID_DG_TOUCHSCREEN ||
-	    field->application == HID_DG_TOUCHPAD)
-		return mt_touch_input_mapped(hdev, hi, field, usage, bit, max);
+	    field->application == HID_DG_TOUCHPAD) {
+		/* We own these mappings, tell hid-input to ignore them */
+		return -1;
+	}
 
 	/* let hid-core decide for the others */
 	return 0;

Signature · 5 lines

2.13.2.725.g09c95d1e9-goog


-- 
Dmitry

Jiri Kosina

maintainer

2017-07-13 14:51 UTC #2 lore raw

In reply to Dmitry Torokhov

Quoted 13 lines · Dmitry Torokhov

Now that input core insists on having dev->absinfo when device claims to generate EV_ABS in its dev->evbit, we should not be blindly setting that bit.
The code in question might have been needed before input_set_abs_params() started setting EV_ABS in device's evbit, but not anymore, and is now breaking devices such as SMART SPNL-6075 Touchscreen.
Fixes 6ecfe51b4082 ("Input: refuse to register absolute devices ...")
Reported-by Matthias Fend <[email protected]>
Tested-by Matthias Fend <[email protected]>
Cc [email protected]
Signed-off-by Dmitry Torokhov <[email protected]>

Quoted 4 lines

---
Jiri, Benjamin, I marked it for stable as it causes regressions in 4.12 reported by users. Please take a look.

Thanks for a quick fix, Dmitry. Queued in for-4.13/upstream-fixes branch.

Signature · 2 lines

Jiri Kosina
SUSE Labs