HID: wacom: Allow wacom_{get,set}_report to retry on -EAGAIN

Hardware may not respond to a request for the HID_DG_CONTACTMAX feature and we should be tolerant of such a failure. This is especially true when using hid-replay where the hardware doesn't exist, but also for devices attached to a flaky bus. This patch increases the number of allowable retries to match other calls to 'wacom_get_report' and also provides a fallback which forces 'touch_max = 16' (enough for any Wacom device seen so far).

Signed-off-by Jason Gerecke <[email protected]>

---

 drivers/hid/wacom_sys.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

drivers/hid/wacom_sys.c +0 -0

index 3cd74d9..c9a7e30 100644

drivers/hid/wacom_sys.c +4 -0

+					 "could not get HID_DG_CONTACTMAX, "
+					 "defaulting to %d\n",
+					  features->touch_max);
+			}
 			kfree(data);
 		}
 		break;

Signature · 1 line

2.4.0

If we get an -EAGAIN error within either of the wacom_{get,set}_report functions, we should use one of our retries if available. Also, log an error if the functions fail so that we can be aware of what's going on.

Signed-off-by Jason Gerecke <[email protected]>

---
 drivers/hid/wacom_sys.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

drivers/hid/wacom_sys.c +0 -0

index 7abf52c..3cd74d9 100644

drivers/hid/wacom_sys.c +1 -0

+			"(last error = %d)\n", retval);
 
 	return retval;
 }

drivers/hid/wacom_sys.c +0 -0

 	return retval;
 }

Signature · 1 line

2.4.0

Retrying on -EPIPE makes very little sense since this typically indicates a problem that will not just disappear on its own. For instance, the USB documentation states that it will be sent if the endpoint is stalled or the device has disconnected. Instead, we should retry if -EAGAIN is received since this indicates a temporary error condition such as a busy bus.

In addition to adjusting the conditions we retry under, we also log an error on failure so that we can be aware of what's going on.

Signed-off-by Jason Gerecke <[email protected]>

---

Changed in v2:
- Renamed from "HID: wacom: Allow wacom_{get,set}_report to retry on -EAGAIN"
- Changed retry conditions so that we no longer retry on -EPIPE
- Use 'hid_err' instead of 'dev_err'

 drivers/hid/wacom_sys.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

drivers/hid/wacom_sys.c +10 -2

index 7abf52c..109312f 100644
@@ -35,7 +35,11 @@ static int wacom_get_report(struct hid_device *hdev, u8 type, u8 *buf,
 	do {
 		retval = hid_hw_raw_request(hdev, buf[0], buf, size, type,
 				HID_REQ_GET_REPORT);
-	} while ((retval == -ETIMEDOUT || retval == -EPIPE) && --retries);
+	} while ((retval == -ETIMEDOUT || retval == -EAGAIN) && --retries);
+
+	if (retval < 0)
+		hid_err(hdev, "wacom_get_report: ran out of retries "
+			"(last error = %d)\n", retval);
 
 	return retval;
 }

@@ -48,7 +52,11 @@ static int wacom_set_report(struct hid_device *hdev, u8 type, u8 *buf,
 	do {
 		retval = hid_hw_raw_request(hdev, buf[0], buf, size, type,
 				HID_REQ_SET_REPORT);
-	} while ((retval == -ETIMEDOUT || retval == -EPIPE) && --retries);
+	} while ((retval == -ETIMEDOUT || retval == -EAGAIN) && --retries);
+
+	if (retval < 0)
+		hid_err(hdev, "wacom_set_report: ran out of retries "
+			"(last error = %d)\n", retval);
 
 	return retval;
 }

Signature · 1 line

2.4.1

Retrying on -EPIPE makes very little sense since this typically indicates a problem that will not just disappear on its own. For instance, the USB documentation states that it will be sent if the endpoint is stalled or the device has disconnected. Instead, we should retry if -EAGAIN is received since this indicates a temporary error condition such as a busy bus.

In addition to adjusting the conditions we retry under, we also log an error on failure so that we can be aware of what's going on.

Signed-off-by Jason Gerecke <[email protected]>