Skip to contentSkip to conversation

HID: make raw output callback more flexible

PATCH In review Jiri Kosina 8 participants 66 messages Jan 29, 2010 – Feb 16, 2010 1 message not in the archive

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover HID: make raw output callback more flexible

Hi,

this is a followup to the thread

"Implement Bluetooth Wacom tablet's mode change in the kernel"

First patch alters the raw output API so that it is able to actually distinguish between feature and output reports (and modifies low-level drivers accordingly), and the two followup patches make use of this for in-kernel mode setting of Sony and Wacom bluetooth devices.

Marcel, could you please review the Bluetooth part?

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#1 lore raw

PATCH 1/3 HID: make raw reports possible for both feature and output reports

In commit 2da31939a42f7a676a0bc5155d6a0a39ed8451f2 ("Bluetooth: Implement raw output support for HIDP layer"), support for Bluetooth hid_output_raw_report was added, but it pushes the data to the intr socket instead of the ctrl one.

Plus it is necessary to distinguish whether the report in question should be either FEATURE or OUTPUT. For this, we have to extend the generic HID API, so that hid_output_raw_report() callback provides means to specify this value so that it can be passed down to lower level hardware drivers (currently Bluetooth and USB).

Based on original patch by Bastien Nocera <[email protected]>

---
 drivers/hid/hidraw.c          |    2 +-
 drivers/hid/usbhid/hid-core.c |    5 ++-
 include/linux/hid.h           |    2 +-
 net/bluetooth/hidp/core.c     |   79 +++++++++++++++++++++++------------------
 4 files changed, 49 insertions(+), 39 deletions(-)
drivers/hid/hidraw.c +1 -1
index cdd1369..d044767 100644
@@ -134,7 +134,7 @@ static ssize_t hidraw_write(struct file *file, const char __user *buffer, size_t
 		goto out;
 	}
 
-	ret = dev->hid_output_raw_report(dev, buf, count);
+	ret = dev->hid_output_raw_report(dev, buf, count, HID_OUTPUT_REPORT);
 out:
 	kfree(buf);
 	return ret;
drivers/hid/usbhid/hid-core.c +3 -2
index e2997a8..caa16c0 100644
@@ -774,7 +774,8 @@ static int hid_alloc_buffers(struct usb_device *dev, struct hid_device *hid)
 	return 0;
 }
 
-static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t count)
+static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t count,
+		unsigned char report_type)
 {
 	struct usbhid_device *usbhid = hid->driver_data;
 	struct usb_device *dev = hid_to_usb_dev(hid);
@@ -785,7 +786,7 @@ static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t co
 	ret = usb_control_msg(dev, usb_sndctrlpipe(dev, 0),
 		HID_REQ_SET_REPORT,
 		USB_DIR_OUT | USB_TYPE_CLASS | USB_RECIP_INTERFACE,
-		((HID_OUTPUT_REPORT + 1) << 8) | *buf,
+		((report_type + 1) << 8) | *buf,
 		interface->desc.bInterfaceNumber, buf + 1, count - 1,
 		USB_CTRL_SET_TIMEOUT);
 
include/linux/hid.h +1 -1
index 8709365..3661a62 100644
@@ -501,7 +501,7 @@ struct hid_device {							/* device report descriptor */
 	void (*hiddev_report_event) (struct hid_device *, struct hid_report *);
 
 	/* handler for raw output data, used by hidraw */
-	int (*hid_output_raw_report) (struct hid_device *, __u8 *, size_t);
+	int (*hid_output_raw_report) (struct hid_device *, __u8 *, size_t, unsigned char);
 
 	/* debugging support via debugfs */
 	unsigned short debug;
net/bluetooth/hidp/core.c +44 -35

98 lines not on this page · raw message

Signature · 1 line
1.6.4.2
#2 lore raw

PATCH 2/3 HID: Implement Wacom quirk in the kernel

In reply to Jiri Kosina

From: Bastien Nocera <[email protected]>

The hid-wacom driver required user-space to poke at the tablet to make it send data about the cursor location.

This patch makes it do the same thing but in the kernel.

---
 drivers/hid/hid-wacom.c |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)
drivers/hid/hid-wacom.c +27 -0
index 12dcda5..d7897b5 100644
@@ -156,7 +156,9 @@ static int wacom_probe(struct hid_device *hdev,
 	struct hid_input *hidinput;
 	struct input_dev *input;
 	struct wacom_data *wdata;
+	char rep_data[2];
 	int ret;
+	int limit;
 
 	wdata = kzalloc(sizeof(*wdata), GFP_KERNEL);
 	if (wdata == NULL) {
@@ -166,6 +168,7 @@ static int wacom_probe(struct hid_device *hdev,
 
 	hid_set_drvdata(hdev, wdata);
 
+	/* Parse the HID report now */
 	ret = hid_parse(hdev);
 	if (ret) {
 		dev_err(&hdev->dev, "parse failed\n");
@@ -178,6 +181,30 @@ static int wacom_probe(struct hid_device *hdev,
 		goto err_free;
 	}
 
+	/* Set Wacom mode2 */
+	rep_data[0] = 0x03; rep_data[1] = 0x00;
+	limit =3;
+	do {
+		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+				HID_FEATURE_REPORT);
+	} while (ret < 0 && limit-- > 0);
+	if (ret < 0) {
+		dev_err(&hdev->dev, "failed to poke device #1, %d\n", ret);
+		goto err_free;
+	}
+
+	/* 0x06 - high reporting speed, 0x05 - low speed */
+	rep_data[0] = 0x06; rep_data[1] = 0x00;
+	limit = 3;
+	do {
+		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+				HID_FEATURE_REPORT);
+	} while (ret < 0 && limit-- > 0);
+	if (ret < 0) {
+		dev_err(&hdev->dev, "failed to poke device #2, %d\n", ret);
+		goto err_free;
+	}
+
 	hidinput = list_entry(hdev->inputs.next, struct hid_input, list);
 	input = hidinput->input;
 
Signature · 1 line
1.6.4.2
#3 lore raw

Re: [PATCH 2/3] HID: Implement Wacom quirk in the kernel

Quoted 37 lines · Jiri Kosina

From: Bastien Nocera <[email protected]>

The hid-wacom driver required user-space to poke at the tablet to make it send data about the cursor location.

This patch makes it do the same thing but in the kernel.

---
 drivers/hid/hid-wacom.c |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)
drivers/hid/hid-wacom.c +3 -0
index 12dcda5..d7897b5 100644
@@ -156,7 +156,9 @@ static int wacom_probe(struct hid_device *hdev,
       struct hid_input *hidinput;
       struct input_dev *input;
       struct wacom_data *wdata;
+       char rep_data[2];
       int ret;
+       int limit;
 
       wdata = kzalloc(sizeof(*wdata), GFP_KERNEL);
       if (wdata == NULL) {
@@ -166,6 +168,7 @@ static int wacom_probe(struct hid_device *hdev,
 
       hid_set_drvdata(hdev, wdata);
 
+       /* Parse the HID report now */
       ret = hid_parse(hdev);
       if (ret) {
               dev_err(&hdev->dev, "parse failed\n");
@@ -178,6 +181,30 @@ static int wacom_probe(struct hid_device *hdev,
               goto err_free;

drivers/hid/hid-wacom.c static int wacom_probe(struct hid_device *hdev, · line 191

Quoted 10 lines
drivers/hid/hid-wacom.c +7 -0
       }
 
+       /* Set Wacom mode2 */
+       rep_data[0] = 0x03; rep_data[1] = 0x00;
+       limit =3;
+       do {
+               ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+                               HID_FEATURE_REPORT);
+       } while (ret < 0 && limit-- > 0);
 

hdev->hid_output_raw_report error is not a hard failure, i.e., the following four lines are unnecessay. We can continue with the rest of the code.

drivers/hid/hid-wacom.c static int wacom_probe(struct hid_device *hdev, · line 210

Quoted 13 lines
drivers/hid/hid-wacom.c +12 -0
+       if (ret < 0) {
+               dev_err(&hdev->dev, "failed to poke device #1, %d\n", ret);
+               goto err_free;
+       }
+
+       /* 0x06 - high reporting speed, 0x05 - low speed */
+       rep_data[0] = 0x06; rep_data[1] = 0x00;
+       limit = 3;
+       do {
+               ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+                               HID_FEATURE_REPORT);
+       } while (ret < 0 && limit-- > 0);
 

Same comments apply here. "/* Note that if query fails it is not a hard failure */" is the comments we use in the corresponding Wacom USB driver, wacom_sys.c

Quoted 7 lines
drivers/hid/hid-wacom.c +5 -0
+       if (ret < 0) {
+               dev_err(&hdev->dev, "failed to poke device #2, %d\n", ret);
+               goto err_free;
+       }
+
       hidinput = list_entry(hdev->inputs.next, struct hid_input, list);
       input = hidinput->input;

Everything else looks good to me.

Ping

#4 lore raw

Re: [PATCH 2/3] HID: Implement Wacom quirk in the kernel

Quoted 24 lines · Ping Cheng

+       rep_data[0] = 0x03; rep_data[1] = 0x00;
+       limit =3;
+       do {
+               ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+                               HID_FEATURE_REPORT);
+       } while (ret < 0 && limit-- > 0);


hdev->hid_output_raw_report error is not a hard failure, i.e., the following four lines are unnecessay. We can continue with the rest of the code.

+       if (ret < 0) {
+               dev_err(&hdev->dev, "failed to poke device #1, %d\n", ret);
+               goto err_free;
+       }
+
+       /* 0x06 - high reporting speed, 0x05 - low speed */
+       rep_data[0] = 0x06; rep_data[1] = 0x00;
+       limit = 3;
+       do {
+               ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+                               HID_FEATURE_REPORT);
+       } while (ret < 0 && limit-- > 0);

Quoted 3 lines


Same comments apply here. "/* Note that if query fails it is not a hard failure */" is the comments we use in the corresponding Wacom USB driver, wacom_sys.c

Thanks for the feedback. I have queued the patch below on top of all the other changes.

From: Jiri Kosina <[email protected]>
Subject: [PATCH] HID: make Wacom modesetting failures non-fatal

With Wacom tablet mode-setting moved from userspace into kernel, we don't have to consider failures of device queries through the _raw callback as hard failure, as the driver can safely continue anyway.

This is consistent with the current USB driver in wacom_sys.c

---
 drivers/hid/hid-wacom.c |   17 +++++++++--------
 1 files changed, 9 insertions(+), 8 deletions(-)
drivers/hid/hid-wacom.c +9 -8
index b8778db..8d3b46f 100644
@@ -181,6 +181,11 @@ static int wacom_probe(struct hid_device *hdev,
 		goto err_free;
 	}
 
+	/*
+	 * Note that if the raw queries fail, it's not a hard failure and it
+	 * is safe to continue
+	 */
+
 	/* Set Wacom mode2 */
 	rep_data[0] = 0x03; rep_data[1] = 0x00;
 	limit = 3;
@@ -188,10 +193,8 @@ static int wacom_probe(struct hid_device *hdev,
 		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
 				HID_FEATURE_REPORT);
 	} while (ret < 0 && limit-- > 0);
-	if (ret < 0) {
-		dev_err(&hdev->dev, "failed to poke device #1, %d\n", ret);
-		goto err_free;
-	}
+	if (ret < 0)
+		dev_warn(&hdev->dev, "failed to poke device #1, %d\n", ret);
 
 	/* 0x06 - high reporting speed, 0x05 - low speed */
 	rep_data[0] = 0x06; rep_data[1] = 0x00;
@@ -200,10 +203,8 @@ static int wacom_probe(struct hid_device *hdev,
 		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
 				HID_FEATURE_REPORT);
 	} while (ret < 0 && limit-- > 0);
-	if (ret < 0) {
-		dev_err(&hdev->dev, "failed to poke device #2, %d\n", ret);
-		goto err_free;
-	}
+	if (ret < 0)
+		dev_warn(&hdev->dev, "failed to poke device #2, %d\n", ret);
 
 	hidinput = list_entry(hdev->inputs.next, struct hid_input, list);
 	input = hidinput->input;
Signature · 1 line
1.6.4.2
#5 lore raw

PATCH 3/3 HID: Enable Sixaxis controller over Bluetooth

In reply to Jiri Kosina

From: Bastien Nocera <[email protected]>

Now that hid_output_raw_report works, port the PS3 Sixaxis Bluetooth quirk from user-space, into kernel-space.

---
 drivers/hid/hid-core.c |    1 +
 drivers/hid/hid-sony.c |   15 ++++++++++++++-
 2 files changed, 15 insertions(+), 1 deletions(-)
drivers/hid/hid-core.c +1 -0
index eabe5f8..f7f80e1 100644
@@ -1340,6 +1340,7 @@ static const struct hid_device_id hid_blacklist[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_PETALYNX, USB_DEVICE_ID_PETALYNX_MAXTER_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SAMSUNG, USB_DEVICE_ID_SAMSUNG_IR_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_VAIO_VGX_MOUSE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SUNPLUS, USB_DEVICE_ID_SUNPLUS_WDESKTOP) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_THRUSTMASTER, 0xb300) },
drivers/hid/hid-sony.c +14 -1
index 4e84502..08b7fa0 100644
@@ -73,6 +73,12 @@ static int sony_set_operational(struct hid_device *hdev)
 	return ret;
 }
 
+static int sony_set_operational_bt(struct hid_device *hdev)
+{
+	unsigned char buf[] = { 0x53, 0xf4,  0x42, 0x03, 0x00, 0x00 };
+	return hdev->hid_output_raw_report(hdev, buf, sizeof(buf), HID_FEATURE_REPORT);
+}
+
 static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 {
 	int ret;
@@ -101,7 +107,13 @@ static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 		goto err_free;
 	}
 
-	ret = sony_set_operational(hdev);
+	if (hdev->bus == BUS_USB)
+		ret = sony_set_operational(hdev);
+	else if (hdev->bus == BUS_BLUETOOTH)
+		ret = sony_set_operational_bt(hdev);
+	else
+		ret = 0;
+
 	if (ret < 0)
 		goto err_stop;
 
@@ -121,6 +133,7 @@ static void sony_remove(struct hid_device *hdev)
 
 static const struct hid_device_id sony_devices[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_VAIO_VGX_MOUSE),
 		.driver_data = VAIO_RDESC_CONSTANT },
 	{ }
Signature · 1 line
1.6.4.2
#6 lore raw
In reply to Jiri Kosina
Quoted 4 lines · Jiri Kosina

Hi,

this is a followup to the thread

Quoted 8 lines

"Implement Bluetooth Wacom tablet's mode change in the kernel"

First patch alters the raw output API so that it is able to actually distinguish between feature and output reports (and modifies low-level drivers accordingly), and the two followup patches make use of this for in-kernel mode setting of Sony and Wacom bluetooth devices.

Marcel, could you please review the Bluetooth part?

Patches 1 and 2 work as expected (minus the hidp parsing issue[1]). Couldn't test patch 3 for the Sony Sixaxis right now.

Cheers

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

#7 lore raw
Quoted 5 lines · Bastien Nocera
Quoted 4 lines · Jiri Kosina

Hi,

this is a followup to the thread

Quoted 8 lines
Quoted 8 lines

"Implement Bluetooth Wacom tablet's mode change in the kernel"

First patch alters the raw output API so that it is able to actually distinguish between feature and output reports (and modifies low-level drivers accordingly), and the two followup patches make use of this for in-kernel mode setting of Sony and Wacom bluetooth devices.

Marcel, could you please review the Bluetooth part?

Count this as another vote for review, please -- I have a similar patch[1] that I would prefer to drop.

[1]- http://www.spinics.net/lists/linux-input/msg06768.html

Quoted 8 lines

Patches 1 and 2 work as expected (minus the hidp parsing issue[1]). Couldn't test patch 3 for the Sony Sixaxis right now.

Cheers

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

This patch does not work for me. Before, the first time after each boot that I tried to connect to an Apple Magic Mouse, it failed with -14 (EFAULT). With this patch, it fails with -22 (EINVAL) instead. The -EFAULT *was* due to hidp_parse()'s copy_from_user(). I have not looked yet to see where the -EINVAL is coming from -- would that help? (Both with and without your patch, the second attempt to connect works.)

Michael Poole

#8 lore raw
Quoted 4 lines · Michael Poole
Quoted 3 lines

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

Quoted 8 lines

This patch does not work for me. Before, the first time after each boot
that I tried to connect to an Apple Magic Mouse, it failed with -14 (EFAULT). With this patch, it fails with -22 (EINVAL) instead. The -EFAULT *was* due to hidp_parse()'s copy_from_user(). I have not looked
yet to see where the -EINVAL is coming from -- would that help? (Both with and without your patch, the second attempt to connect works.)

I don't get -EFAULT anymore (it was failing to copy the rd_data from user-space), but I do get -EINVALs now. I haven't investigated it though. My guess is that the hid parser fails.

Could you compare the sizes of the data gathered in user-space?

#9 lore raw
Quoted 14 lines · Bastien Nocera
Quoted 4 lines · Michael Poole
Quoted 3 lines

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

Quoted 8 lines

This patch does not work for me. Before, the first time after each boot
that I tried to connect to an Apple Magic Mouse, it failed with -14 (EFAULT). With this patch, it fails with -22 (EINVAL) instead. The -EFAULT *was* due to hidp_parse()'s copy_from_user(). I have not looked
yet to see where the -EINVAL is coming from -- would that help? (Both with and without your patch, the second attempt to connect works.)

Quoted 5 lines

I don't get -EFAULT anymore (it was failing to copy the rd_data from user-space), but I do get -EINVALs now. I haven't investigated it though. My guess is that the hid parser fails.

Could you compare the sizes of the data gathered in user-space?

It looks like something else is overwriting parts of the hidp_session structure; I hope to find out what that is tomorrow. Based on a printk() in hidp_sock_ioctl(), user space consistently passes a 99-byte HID report descriptor. I added another printk() to hidp_parse() to characterize things:
printk("hidp_parse(%p): %u@%p\n", hid, req->rd_size, req->rd_data);

With your patch, the first time I connect to the mouse:

[  244.109619] hidp_parse(ffff880138f90000): 0@0000002800000000
[  244.109694] magicmouse 0005:05AC:030D.0001: magicmouse hid parse
  failed
[  244.109708] magicmouse: probe of 0005:05AC:030D.0001 failed with
  error -22

With your patch, the second time:

[  258.840542] hidp_parse(ffff880134e14000): 99@ffff88008fd8f880
[  258.841415] input: Apple Wireless Mouse as
  /devices/pci0000:00/0000:00:1d.2/usb8/8-1/8-1:1.0/bluetooth/hci0/hci0:12/input11

Without your patch, the first time:

[  157.569000] hidp_parse(ffff8801313cc000): 11859@ffffffff811397c0
[  157.569018] magicmouse 0005:05AC:030D.0001: magicmouse hid parse
  failed
[  157.569031] magicmouse: probe of 0005:05AC:030D.0001 failed with
  error -14

Without your patch, the second time:

[  178.983956] hidp_parse(ffff88012e660000): 99@00007f40d7d44880
[  178.984977] input: Apple Wireless Mouse as
  /devices/pci0000:00/0000:00:1d.2/usb8/8-1/8-1:1.0/bluetooth/hci0/hci0:12/input11

Michael Poole

#10 lore raw
In reply to Bastien Nocera
Quoted 14 lines · Bastien Nocera
Quoted 4 lines · Michael Poole
Quoted 3 lines

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

Quoted 8 lines

This patch does not work for me. Before, the first time after each boot
that I tried to connect to an Apple Magic Mouse, it failed with -14 (EFAULT). With this patch, it fails with -22 (EINVAL) instead. The -EFAULT *was* due to hidp_parse()'s copy_from_user(). I have not looked
yet to see where the -EINVAL is coming from -- would that help? (Both with and without your patch, the second attempt to connect works.)

Quoted 5 lines

I don't get -EFAULT anymore (it was failing to copy the rd_data from user-space), but I do get -EINVALs now. I haven't investigated it though. My guess is that the hid parser fails.

Could you compare the sizes of the data gathered in user-space?

The bug mysteriously disappeared when I tried to apply kgdb to the problem. Your patch won't do the trick because the hidp_connadd_req structure has also gone out of scope by the time the hid-(whatever) probe function is called. The ioctl has returned to the user, but the new hid-(whatever) module is not yet loaded.

That is, the sequence looks like this:

- Application triggers hidp_sock_ioctl(socket, HIDPCONNADD, &connadd).
- This starts the connection, and returns 0...
- ... but also indicates that some other module needs to be loaded.

- User-space loads the appropriate module does init_module()...
- which calls the module_init() function...
- which calls hid_register_driver()...
- which eventually attaches the new driver to the device...
- triggering the module's probe() function to call hid_parse()...
- hid_parse() fails because its hidp_connadd_req is gone.

Marcel, I think this is a case where the subsystem maintainer should make the call on how to fix it. I can write and locally test a patch, but I don't want to assume that (for example) the desired solution is to keep a copy of the Report descriptor in the hidp_session. The USB HID core sends a GET_DESCRIPTOR request for the Report descriptor, which isn't practical here because that descriptor is only available through SDP.

Michael Poole

#11 lore raw
Quoted 20 lines · Michael Poole
Quoted 18 lines · Bastien Nocera
Quoted 4 lines · Michael Poole
Quoted 3 lines

[1]: Comments on the patch at
http://thread.gmane.org/gmane.linux.bluez.kernel/4279 would be appreciated

Quoted 8 lines

This patch does not work for me. Before, the first time after each boot
that I tried to connect to an Apple Magic Mouse, it failed with -14 (EFAULT). With this patch, it fails with -22 (EINVAL) instead. The -EFAULT *was* due to hidp_parse()'s copy_from_user(). I have not looked
yet to see where the -EINVAL is coming from -- would that help? (Both with and without your patch, the second attempt to connect works.)

I don't get -EFAULT anymore (it was failing to copy the rd_data from user-space), but I do get -EINVALs now. I haven't investigated it though. My guess is that the hid parser fails.

Quoted 7 lines
Quoted 1 line

Could you compare the sizes of the data gathered in user-space?

The bug mysteriously disappeared when I tried to apply kgdb to the problem. Your patch won't do the trick because the hidp_connadd_req structure has also gone out of scope by the time the hid-(whatever) probe function is called. The ioctl has returned to the user, but the new hid-(whatever) module is not yet loaded.

Right. I knew that rd_data would already be out of scope, didn't think about the hidp_connadd_req structure itself.

I guess we should make a copy of the rd_data/rd_size in hidp_setup_hid() instead.

Quoted 13 lines

That is, the sequence looks like this:

- Application triggers hidp_sock_ioctl(socket, HIDPCONNADD, &connadd).
- This starts the connection, and returns 0...
- ... but also indicates that some other module needs to be loaded.

- User-space loads the appropriate module does init_module()...
- which calls the module_init() function...
- which calls hid_register_driver()...
- which eventually attaches the new driver to the device...
- triggering the module's probe() function to call hid_parse()...
- hid_parse() fails because its hidp_connadd_req is gone.

Quoted 4 lines

Marcel, I think this is a case where the subsystem maintainer should make the call on how to fix it. I can write and locally test a patch, but I don't want to assume that (for example) the desired solution is to keep a copy of the Report descriptor in the hidp_session.

I would think it would be the cleanest way to do it. Quite a few members of the hidp_connadd_req structure are already copied into the session in hidp_setup_hid().

Quoted 4 lines

The USB HID
core sends a GET_DESCRIPTOR request for the Report descriptor, which isn't practical here because that descriptor is only available through SDP.

And we only do SDP queries in user-space...

If you're not going to work on it shortly, I can take a look at cooking a patch by the end of the week. Otherwise, I'd be happy testing your patches.

Cheers

A message that is not in the archive (sent off-list or not yet received): [email protected]

#13 lore raw
Quoted 9 lines · Jiri Kosina
Quoted 9 lines
Quoted 3 lines

The USB HID core sends a GET_DESCRIPTOR request for the Report descriptor, which isn't practical here because that descriptor is only available through SDP.

And we only do SDP queries in user-space...

If you're not going to work on it shortly, I can take a look at cooking a patch by the end of the week. Otherwise, I'd be happy testing your patches.

I have now rebased the three patches on top of 6bf8268f9a91f1 ("Bluetooth: Use the control channel for raw HID reports") which has already landed in Linus' tree. I will send them as a followup to this mail.

If you have any changes/acks/nacks, please let me know.

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#14 lore raw

PATCH 1/3 HID: make raw reports possible for both feature and output reports

In commit 2da31939a42 ("Bluetooth: Implement raw output support for HIDP layer"), support for Bluetooth hid_output_raw_report was added, but it pushes the data to the intr socket instead of the ctrl one. This has been fixed by 6bf8268f9a91f1 ("Bluetooth: Use the control channel for raw HID reports")

Still, it is necessary to distinguish whether the report in question should be either FEATURE or OUTPUT. For this, we have to extend the generic HID API, so that hid_output_raw_report() callback provides means to specify this value so that it can be passed down to lower level hardware drivers (currently Bluetooth and USB).

Based on original patch by Bastien Nocera <[email protected]>

---
 drivers/hid/hidraw.c          |    2 +-
 drivers/hid/usbhid/hid-core.c |    5 +++--
 include/linux/hid.h           |    2 +-
 net/bluetooth/hidp/core.c     |   13 ++++++++++---
 4 files changed, 15 insertions(+), 7 deletions(-)
drivers/hid/hidraw.c +1 -1
index cdd1369..d044767 100644
@@ -134,7 +134,7 @@ static ssize_t hidraw_write(struct file *file, const char __user *buffer, size_t
 		goto out;
 	}
 
-	ret = dev->hid_output_raw_report(dev, buf, count);
+	ret = dev->hid_output_raw_report(dev, buf, count, HID_OUTPUT_REPORT);
 out:
 	kfree(buf);
 	return ret;
drivers/hid/usbhid/hid-core.c +3 -2
index e2997a8..caa16c0 100644
@@ -774,7 +774,8 @@ static int hid_alloc_buffers(struct usb_device *dev, struct hid_device *hid)
 	return 0;
 }
 
-static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t count)
+static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t count,
+		unsigned char report_type)
 {
 	struct usbhid_device *usbhid = hid->driver_data;
 	struct usb_device *dev = hid_to_usb_dev(hid);
@@ -785,7 +786,7 @@ static int usbhid_output_raw_report(struct hid_device *hid, __u8 *buf, size_t co
 	ret = usb_control_msg(dev, usb_sndctrlpipe(dev, 0),
 		HID_REQ_SET_REPORT,
 		USB_DIR_OUT | USB_TYPE_CLASS | USB_RECIP_INTERFACE,
-		((HID_OUTPUT_REPORT + 1) << 8) | *buf,
+		((report_type + 1) << 8) | *buf,
 		interface->desc.bInterfaceNumber, buf + 1, count - 1,
 		USB_CTRL_SET_TIMEOUT);
 
include/linux/hid.h +1 -1
index 8709365..3661a62 100644
@@ -501,7 +501,7 @@ struct hid_device {							/* device report descriptor */
 	void (*hiddev_report_event) (struct hid_device *, struct hid_report *);
 
 	/* handler for raw output data, used by hidraw */
-	int (*hid_output_raw_report) (struct hid_device *, __u8 *, size_t);
+	int (*hid_output_raw_report) (struct hid_device *, __u8 *, size_t, unsigned char);
 
 	/* debugging support via debugfs */
 	unsigned short debug;
net/bluetooth/hidp/core.c +10 -3
index 6cf526d..9c88750 100644
@@ -313,10 +313,17 @@ static int hidp_send_report(struct hidp_session *session, struct hid_report *rep
 	return hidp_queue_report(session, buf, rsize);
 }
 
-static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count)
+static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count,
+		unsigned char report_type)
 {
-	if (hidp_send_ctrl_message(hid->driver_data,
-			HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE,
+	if (report_type == HID_FEATURE_REPORT)
+		report_type = HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE;
+	else if (report_type == HID_OUTPUT_REPORT)
+		report_type = HIDP_TRANS_DATA | HIDP_DATA_RTYPE_OUPUT;
+	else
+		return -EINVAL;
+
+	if (hidp_send_ctrl_message(hid->driver_data, report_type,
 			data, count))
 		return -ENOMEM;
 	return count;
Signature · 1 line
1.6.4.2
maintainer Acked-by on v1 1/3
#15 lore raw

Re: [PATCH 1/3] HID: make raw reports possible for both feature and output reports

Hi Jiri,

Quoted 11 lines

In commit 2da31939a42 ("Bluetooth: Implement raw output support for HIDP layer"), support for Bluetooth hid_output_raw_report was added, but it pushes the data to the intr socket instead of the ctrl one. This has been fixed by 6bf8268f9a91f1 ("Bluetooth: Use the control channel for raw HID reports")

Still, it is necessary to distinguish whether the report in question should be either FEATURE or OUTPUT. For this, we have to extend the generic HID API, so that hid_output_raw_report() callback provides means to specify this value so that it can be passed down to lower level hardware drivers (currently Bluetooth and USB).

Quoted 8 lines

Based on original patch by Bastien Nocera <[email protected]>

---
 drivers/hid/hidraw.c          |    2 +-
 drivers/hid/usbhid/hid-core.c |    5 +++--
 include/linux/hid.h           |    2 +-
 net/bluetooth/hidp/core.c     |   13 ++++++++++---
 4 files changed, 15 insertions(+), 7 deletions(-)

net/bluetooth/hidp/core.c static int hidp_send_report(struct hidp_session *session, struct hid_report *rep · line 324

Quoted 12 lines
net/bluetooth/hidp/core.c +8 -3
-static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count)
+static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count,
+		unsigned char report_type)
 {
-	if (hidp_send_ctrl_message(hid->driver_data,
-			HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE,
+	if (report_type == HID_FEATURE_REPORT)
+		report_type = HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE;
+	else if (report_type == HID_OUTPUT_REPORT)
+		report_type = HIDP_TRANS_DATA | HIDP_DATA_RTYPE_OUPUT;
+	else
+		return -EINVAL;

If you wanna be a little bit more readable, then using a switch statement helps instead of else if. Just my 2 cents.

Regards

Marcel

#16 lore raw

Re: [PATCH 1/3] HID: make raw reports possible for both feature and output reports

Quoted 22 lines · Marcel Holtmann
Quoted 11 lines

In commit 2da31939a42 ("Bluetooth: Implement raw output support for HIDP layer"), support for Bluetooth hid_output_raw_report was added, but it pushes the data to the intr socket instead of the ctrl one. This has been fixed by 6bf8268f9a91f1 ("Bluetooth: Use the control channel for raw HID reports")

Still, it is necessary to distinguish whether the report in question should be either FEATURE or OUTPUT. For this, we have to extend the generic HID API, so that hid_output_raw_report() callback provides means to specify this value so that it can be passed down to lower level hardware drivers (currently Bluetooth and USB).

Quoted 8 lines

Based on original patch by Bastien Nocera <[email protected]>

---
 drivers/hid/hidraw.c          |    2 +-
 drivers/hid/usbhid/hid-core.c |    5 +++--
 include/linux/hid.h           |    2 +-
 net/bluetooth/hidp/core.c     |   13 ++++++++++---
 4 files changed, 15 insertions(+), 7 deletions(-)
Quoted 15 lines

net/bluetooth/hidp/core.c static int hidp_send_report(struct hidp_session *session, struct hid_report *rep · line 324

Quoted 12 lines
net/bluetooth/hidp/core.c +8 -3
-static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count)
+static int hidp_output_raw_report(struct hid_device *hid, unsigned char *data, size_t count,
+		unsigned char report_type)
 {
-	if (hidp_send_ctrl_message(hid->driver_data,
-			HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE,
+	if (report_type == HID_FEATURE_REPORT)
+		report_type = HIDP_TRANS_SET_REPORT | HIDP_DATA_RTYPE_FEATURE;
+	else if (report_type == HID_OUTPUT_REPORT)
+		report_type = HIDP_TRANS_DATA | HIDP_DATA_RTYPE_OUPUT;
+	else
+		return -EINVAL;

If you wanna be a little bit more readable, then using a switch statement helps instead of else if. Just my 2 cents.

I have converted this to switch, added your Acked-by and queued in my tree.

Thanks Marcel.

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#17 lore raw

PATCH 2/3 HID: Implement Wacom quirk in the kernel

In reply to Jiri Kosina

From: Bastien Nocera <[email protected]>

The hid-wacom driver required user-space to poke at the tablet to make it send data about the cursor location.

This patch makes it do the same thing but in the kernel.

---
 drivers/hid/hid-wacom.c |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)
drivers/hid/hid-wacom.c +27 -0
index 12dcda5..d7897b5 100644
@@ -156,7 +156,9 @@ static int wacom_probe(struct hid_device *hdev,
 	struct hid_input *hidinput;
 	struct input_dev *input;
 	struct wacom_data *wdata;
+	char rep_data[2];
 	int ret;
+	int limit;
 
 	wdata = kzalloc(sizeof(*wdata), GFP_KERNEL);
 	if (wdata == NULL) {
@@ -166,6 +168,7 @@ static int wacom_probe(struct hid_device *hdev,
 
 	hid_set_drvdata(hdev, wdata);
 
+	/* Parse the HID report now */
 	ret = hid_parse(hdev);
 	if (ret) {
 		dev_err(&hdev->dev, "parse failed\n");
@@ -178,6 +181,30 @@ static int wacom_probe(struct hid_device *hdev,
 		goto err_free;
 	}
 
+	/* Set Wacom mode2 */
+	rep_data[0] = 0x03; rep_data[1] = 0x00;
+	limit =3;
+	do {
+		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+				HID_FEATURE_REPORT);
+	} while (ret < 0 && limit-- > 0);
+	if (ret < 0) {
+		dev_err(&hdev->dev, "failed to poke device #1, %d\n", ret);
+		goto err_free;
+	}
+
+	/* 0x06 - high reporting speed, 0x05 - low speed */
+	rep_data[0] = 0x06; rep_data[1] = 0x00;
+	limit = 3;
+	do {
+		ret = hdev->hid_output_raw_report(hdev, rep_data, 2,
+				HID_FEATURE_REPORT);
+	} while (ret < 0 && limit-- > 0);
+	if (ret < 0) {
+		dev_err(&hdev->dev, "failed to poke device #2, %d\n", ret);
+		goto err_free;
+	}
+
 	hidinput = list_entry(hdev->inputs.next, struct hid_input, list);
 	input = hidinput->input;
 
Signature · 1 line
1.6.4.2
maintainer Acked-by on v1 2/3
#18 lore raw

Re: [PATCH 2/3] HID: Implement Wacom quirk in the kernel

Hi Jiri,

Quoted 7 lines

The hid-wacom driver required user-space to poke at the tablet to make it send data about the cursor location.

This patch makes it do the same thing but in the kernel.

drivers/hid/hid-wacom.c static int wacom_probe(struct hid_device *hdev, · line 186

Quoted 4 lines
drivers/hid/hid-wacom.c +3 -0
+	/* Set Wacom mode2 */
+	rep_data[0] = 0x03; rep_data[1] = 0x00;
+	limit =3;

Small nitpick here. The space after = is missing ;)

Regards

Marcel

#19 lore raw

Re: [PATCH 2/3] HID: Implement Wacom quirk in the kernel

Quoted 10 lines · Marcel Holtmann

Hi Jiri,

Quoted 7 lines

The hid-wacom driver required user-space to poke at the tablet to make it send data about the cursor location.

This patch makes it do the same thing but in the kernel.

Quoted 8 lines

drivers/hid/hid-wacom.c static int wacom_probe(struct hid_device *hdev, · line 186

Quoted 4 lines
drivers/hid/hid-wacom.c +3 -0
+	/* Set Wacom mode2 */
+	rep_data[0] = 0x03; rep_data[1] = 0x00;
+	limit =3;

Small nitpick here. The space after = is missing ;)

Fixed and applied. Thank you.

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#20 lore raw

PATCH 3/3 HID: Enable Sixaxis controller over Bluetooth

In reply to Jiri Kosina

From: Bastien Nocera <[email protected]>

Now that hid_output_raw_report works, port the PS3 Sixaxis Bluetooth quirk from user-space, into kernel-space.

---
 drivers/hid/hid-core.c |    1 +
 drivers/hid/hid-sony.c |   15 ++++++++++++++-
 2 files changed, 15 insertions(+), 1 deletions(-)
drivers/hid/hid-core.c +1 -0
index eabe5f8..f7f80e1 100644
@@ -1340,6 +1340,7 @@ static const struct hid_device_id hid_blacklist[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_PETALYNX, USB_DEVICE_ID_PETALYNX_MAXTER_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SAMSUNG, USB_DEVICE_ID_SAMSUNG_IR_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_VAIO_VGX_MOUSE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SUNPLUS, USB_DEVICE_ID_SUNPLUS_WDESKTOP) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_THRUSTMASTER, 0xb300) },
drivers/hid/hid-sony.c +14 -1
index 4e84502..08b7fa0 100644
@@ -73,6 +73,12 @@ static int sony_set_operational(struct hid_device *hdev)
 	return ret;
 }
 
+static int sony_set_operational_bt(struct hid_device *hdev)
+{
+	unsigned char buf[] = { 0x53, 0xf4,  0x42, 0x03, 0x00, 0x00 };
+	return hdev->hid_output_raw_report(hdev, buf, sizeof(buf), HID_FEATURE_REPORT);
+}
+
 static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 {
 	int ret;
@@ -101,7 +107,13 @@ static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id)
 		goto err_free;
 	}
 
-	ret = sony_set_operational(hdev);
+	if (hdev->bus == BUS_USB)
+		ret = sony_set_operational(hdev);
+	else if (hdev->bus == BUS_BLUETOOTH)
+		ret = sony_set_operational_bt(hdev);
+	else
+		ret = 0;
+
 	if (ret < 0)
 		goto err_stop;
 
@@ -121,6 +133,7 @@ static void sony_remove(struct hid_device *hdev)
 
 static const struct hid_device_id sony_devices[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_PS3_CONTROLLER) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_SONY, USB_DEVICE_ID_SONY_VAIO_VGX_MOUSE),
 		.driver_data = VAIO_RDESC_CONSTANT },
 	{ }
Signature · 1 line
1.6.4.2
maintainer Acked-by on v1 3/3
#21 lore raw

Re: [PATCH 3/3] HID: Enable Sixaxis controller over Bluetooth

Hi Jiri,

Quoted 5 lines

Now that hid_output_raw_report works, port the PS3 Sixaxis Bluetooth quirk from user-space, into kernel-space.

drivers/hid/hid-sony.c static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id) · line 116

Quoted 8 lines
drivers/hid/hid-sony.c +7 -1
-	ret = sony_set_operational(hdev);
+	if (hdev->bus == BUS_USB)
+		ret = sony_set_operational(hdev);
+	else if (hdev->bus == BUS_BLUETOOTH)
+		ret = sony_set_operational_bt(hdev);
+	else
+		ret = 0;
+

I would do sony_set_operational_usb() and sony_set_operational_bt() and also turn this into a switch statement for better readability.

Regards

Marcel

#22 lore raw

Re: [PATCH 3/3] HID: Enable Sixaxis controller over Bluetooth

Quoted 10 lines · Marcel Holtmann

Hi Jiri,

Quoted 5 lines

Now that hid_output_raw_report works, port the PS3 Sixaxis Bluetooth quirk from user-space, into kernel-space.

Quoted 11 lines

drivers/hid/hid-sony.c static int sony_probe(struct hid_device *hdev, const struct hid_device_id *id) · line 116

Quoted 8 lines
drivers/hid/hid-sony.c +7 -1
-	ret = sony_set_operational(hdev);
+	if (hdev->bus == BUS_USB)
+		ret = sony_set_operational(hdev);
+	else if (hdev->bus == BUS_BLUETOOTH)
+		ret = sony_set_operational_bt(hdev);
+	else
+		ret = 0;
+

I would do sony_set_operational_usb() and sony_set_operational_bt() and also turn this into a switch statement for better readability.

Fixed and applied. Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#23 lore raw

PATCH 1/1 Bluetooth: Keep a copy of each HID device's report descriptor.

In reply to Bastien Nocera

Bastien and Marcel,

This patch works for me. Bastien, does it also fix it for you?

Michael Poole

Quoted 1 line

From f8aeb64ac2d42f5fae080c00a287fdbd8304ffa2 Mon Sep 17 00:00:00 2001

From: Michael Poole <[email protected]>
Date: Wed, 3 Feb 2010 21:49:48 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

---
 net/bluetooth/hidp/core.c |   42 +++++++++++++++++++-----------------------
 net/bluetooth/hidp/hidp.h |    4 +++-
 2 files changed, 22 insertions(+), 24 deletions(-)
net/bluetooth/hidp/core.c +19 -23
index dde4c60..af4a448 100644
@@ -701,29 +701,9 @@ static void hidp_close(struct hid_device *hid)
 static int hidp_parse(struct hid_device *hid)
 {
 	struct hidp_session *session = hid->driver_data;
-	struct hidp_connadd_req *req = session->req;
-	unsigned char *buf;
-	int ret;
-
-	buf = kmalloc(req->rd_size, GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
-	if (copy_from_user(buf, req->rd_data, req->rd_size)) {
-		kfree(buf);
-		return -EFAULT;
-	}
-
-	ret = hid_parse_report(session->hid, buf, req->rd_size);
-
-	kfree(buf);
-
-	if (ret)
-		return ret;
 
-	session->req = NULL;
-
-	return 0;
+	return hid_parse_report(session->hid, session->rd_data,
+			session->rd_size);
 }
 
 static int hidp_start(struct hid_device *hid)
@@ -798,7 +778,17 @@ static int hidp_setup_hid(struct hidp_session *session,
 		return PTR_ERR(hid);
 
 	session->hid = hid;
-	session->req = req;
+	session->rd_size = req->rd_size;
+	session->rd_data = kzalloc(session->rd_size, GFP_KERNEL);
+	if (!session->rd_data) {
+		err = -ENOMEM;
+		goto failed;
+	}
+	if (copy_from_user(session->rd_data, req->rd_data, session->rd_size)) {
+		err = -EFAULT;
+		goto failed;
+	}
+
 	hid->driver_data = session;
 
 	baswap(&src, &bt_sk(session->ctrl_sock->sk)->src);
@@ -829,6 +819,9 @@ failed:
 	hid_destroy_device(hid);
 	session->hid = NULL;
 
+	kfree(session->rd_data);
+	session->rd_data = NULL;
+
 	return err;
 }
 
@@ -923,6 +916,9 @@ unlink:
 		session->hid = NULL;
 	}
 
+	kfree(session->rd_data);
+	session->rd_data = NULL;
+
 purge:
 	skb_queue_purge(&session->ctrl_transmit);
 	skb_queue_purge(&session->intr_transmit);
net/bluetooth/hidp/hidp.h +3 -1
index faf3d74..a4e215d 100644
@@ -154,7 +154,9 @@ struct hidp_session {
 	struct sk_buff_head ctrl_transmit;
 	struct sk_buff_head intr_transmit;
 
-	struct hidp_connadd_req *req;
+	/* Report descriptor */
+	__u8 *rd_data;
+	uint rd_size;
 };
 
 static inline void hidp_schedule(struct hidp_session *session)
Signature · 1 line
1.6.5.6
maintainer
#24 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Hi Michael,

Quoted 9 lines

This patch works for me. Bastien, does it also fix it for you?

Michael Poole

From f8aeb64ac2d42f5fae080c00a287fdbd8304ffa2 Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Wed, 3 Feb 2010 21:49:48 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

@@ -770,12 +770,24 @@ static int hidp_setup_hid(struct hidp_session *session,

        bdaddr_t src, dst;
        int err;

+ session->rd_data = kzalloc(req->rd_size, GFP_KERNEL);
+ if (!session->rd_data)
+ return -ENOMEM;
+
+ if (copy_from_user(session->rd_data, req->rd_data, req->rd_size)) {
+ err = -EFAULT;
+ goto fault;
+ }
+
+ session->rd_size = req->rd_size;
+

        hid = hid_allocate_device();

- if (IS_ERR(hid))
- return PTR_ERR(hid);
+ if (IS_ERR(hid)) {
+ err = PTR_ERR(hid);
+ goto fault;
+ }

        session->hid = hid;
-       session->req = req;
        hid->driver_data = session;
 
        baswap(&src, &bt_sk(session->ctrl_sock->sk)->src);
@@ -806,6 +818,10 @@ failed:
        hid_destroy_device(hid);
        session->hid = NULL;

+fault:
+ kfree(session->rd_data);
+ session->rd_data = NULL;
+

        return err;

Regards

Marcel

#25 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 2 lines · Marcel Holtmann

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

Quoted 1 line

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001

From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

---
 net/bluetooth/hidp/core.c |   49 ++++++++++++++++++++++-----------------------
 net/bluetooth/hidp/hidp.h |    4 ++-
 2 files changed, 27 insertions(+), 26 deletions(-)
net/bluetooth/hidp/core.c +24 -25
index dde4c60..2928e44 100644
@@ -701,29 +701,9 @@ static void hidp_close(struct hid_device *hid)
 static int hidp_parse(struct hid_device *hid)
 {
 	struct hidp_session *session = hid->driver_data;
-	struct hidp_connadd_req *req = session->req;
-	unsigned char *buf;
-	int ret;
-
-	buf = kmalloc(req->rd_size, GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
-	if (copy_from_user(buf, req->rd_data, req->rd_size)) {
-		kfree(buf);
-		return -EFAULT;
-	}
-
-	ret = hid_parse_report(session->hid, buf, req->rd_size);
-
-	kfree(buf);
-
-	if (ret)
-		return ret;
-
-	session->req = NULL;
 
-	return 0;
+	return hid_parse_report(session->hid, session->rd_data,
+			session->rd_size);
 }
 
 static int hidp_start(struct hid_device *hid)
@@ -793,12 +773,24 @@ static int hidp_setup_hid(struct hidp_session *session,
 	bdaddr_t src, dst;
 	int err;
 
+	session->rd_data = kzalloc(req->rd_size, GFP_KERNEL);
+	if (!session->rd_data)
+		return -ENOMEM;
+
+	if (copy_from_user(session->rd_data, req->rd_data, req->rd_size)) {
+		err = -EFAULT;
+		goto fault;
+	}
+	session->rd_size = req->rd_size;
+
 	hid = hid_allocate_device();
-	if (IS_ERR(hid))
-		return PTR_ERR(hid);
+	if (IS_ERR(hid)) {
+		err = PTR_ERR(hid);
+		goto fault;
+	}
 
 	session->hid = hid;
-	session->req = req;
+
 	hid->driver_data = session;
 
 	baswap(&src, &bt_sk(session->ctrl_sock->sk)->src);
@@ -829,6 +821,10 @@ failed:
 	hid_destroy_device(hid);
 	session->hid = NULL;
 
+fault:
+	kfree(session->rd_data);
+	session->rd_data = NULL;
+
 	return err;
 }
 
@@ -923,6 +919,9 @@ unlink:
 		session->hid = NULL;
 	}
 
+	kfree(session->rd_data);
+	session->rd_data = NULL;
+
 purge:
 	skb_queue_purge(&session->ctrl_transmit);
 	skb_queue_purge(&session->intr_transmit);
net/bluetooth/hidp/hidp.h +3 -1
index faf3d74..a4e215d 100644
@@ -154,7 +154,9 @@ struct hidp_session {
 	struct sk_buff_head ctrl_transmit;
 	struct sk_buff_head intr_transmit;
 
-	struct hidp_connadd_req *req;
+	/* Report descriptor */
+	__u8 *rd_data;
+	uint rd_size;
 };
 
 static inline void hidp_schedule(struct hidp_session *session)
Signature · 1 line
1.6.5.6
maintainer
#26 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Hi Michael,

Quoted 11 lines
Quoted 2 lines

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Regards

Marcel

#27 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Hi,

I have been trying to test Michael's magic mouse drivers against 32.7 (or .8-rc). Is there a tree
based on .32 with these patches? If not would it be possible to post the patches to lkml? I do not mind sorting out a few conflicts - this thread assume one knows a little more about the process being used here than is obvious.

TIA
Ed Tomlinson

Quoted 34 lines · On Friday 05 February 2010 12:51:43 Marcel Holtmann wrote:

Hi Michael,

Quoted 11 lines
Quoted 2 lines

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Regards

Marcel

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#28 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 7 lines

TIA
Ed Tomlinson

Quoted 3 lines · On Friday 05 February 2010 12:51:43 Marcel Holtmann wrote:

Hi Michael,

Quoted 1 line
Quoted 1 line

looks good to me. I just prefer that you do the allocation of the re=

port

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 1 line

descriptor before the HID object:

An updated patch is below. =A0Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report de=

scriptor.

Quoted 10 lines
Quoted 10 lines
Quoted 7 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. =A0However, the probe function that needs the descriptor might not be called until a specific module is loaded. =A0Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Quoted 7 lines
Quoted 7 lines

Regards

Marcel

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" =

in

Quoted 8 lines
Quoted 3 lines

the body of a message to [email protected]
More majordomo info at =A0http://vger.kernel.org/majordomo-info.html Please read the FAQ at =A0http://www.tux.org/lkml/

Hi,

I have been trying to test Michael's magic mouse drivers against 32.7 (or=

.8-rc). Is there a tree

Quoted 1 line

based on .32 with these patches? If not would it be possible to post the=

patches to lkml? I do not

Quoted 1 line

mind sorting out a few conflicts - this thread assume one knows a little =

more about the process being

Quoted 2 lines

used here than is obvious.

just got one of those magic things as well. I did apply the patchs to the current HEAD, but am getting a lifeless result i.g. here's what I did:

sudo /usr/bin/hcitool scan
sudo /usr/bin/simple-agent hci0 D4:9A:20:88:C7:48
sudo /usr/bin/dbus-send --system --dest=3Dorg.bluez --print-reply /org/bluez/1897/hci0/dev_D4_9A_20_88_C7_48 org.bluez.Input.Connect sudo /usr/bin/test-device trusted D4:9A:20:88:C7:48

the thing works after that(no scroll), fast cursor though. then reboot
once up and running nothing.
cat /var/log/daemon.log (shows this)

Feb  8 22:57:40 Linux-1 bluetoothd[1973]: link_key_request
(sba=3D00:25:00:C0:6C:4F, dba=3DD4:9A:20:88:C7:48)
Feb  8 22:58:21 Linux-1 bluetoothd[1973]: last message repeated 5 times
Feb  8 22:59:21 Linux-1 bluetoothd[1973]: last message repeated 3 times
Feb  8 23:03:10 Linux-1 bluetoothd[1973]: link_key_request

(sba=3D00:25:00:C0:6C:4F, dba=3DD4:9A:20:88:C7:48)

(could be missing something though(don't use gnome/kde-bluetooth)although mightymouse works).

bluez is v59(in the morning I can load up the latest and see). and then go from there if you need me to be a test subject let me know.

BTW: ed I can attach the patch(seperate e-mail) which is updated to rc6*(to the most part)
if you want.? this way you can give it a try, then once these are in
the mainline
you wont have todo so.

--=20
Justin P. Mattock

#29 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 62 lines · Justin Mattock
Quoted 38 lines

TIA
Ed Tomlinson

Quoted 24 lines · On Friday 05 February 2010 12:51:43 Marcel Holtmann wrote:

Hi Michael,

Quoted 11 lines
Quoted 2 lines

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Regards

Quoted 8 lines

Marcel

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Quoted 7 lines

Hi,

I have been trying to test Michael's magic mouse drivers against 32.7 (or .8-rc). Is there a tree
based on .32 with these patches? If not would it be possible to post the patches to lkml? I do not mind sorting out a few conflicts - this thread assume one knows a little more about the process being used here than is obvious.

just got one of those magic things as well. I did apply the patchs to the current HEAD, but am getting a lifeless result i.g. here's what I did:

sudo /usr/bin/hcitool scan
sudo /usr/bin/simple-agent hci0 D4:9A:20:88:C7:48
sudo /usr/bin/dbus-send --system --dest=org.bluez --print-reply /org/bluez/1897/hci0/dev_D4_9A_20_88_C7_48 org.bluez.Input.Connect sudo /usr/bin/test-device trusted D4:9A:20:88:C7:48

the thing works after that(no scroll), fast cursor though. then reboot
once up and running nothing.
cat /var/log/daemon.log (shows this)

Quoted 6 lines
Feb  8 22:57:40 Linux-1 bluetoothd[1973]: link_key_request
(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)
Feb  8 22:58:21 Linux-1 bluetoothd[1973]: last message repeated 5 times
Feb  8 22:59:21 Linux-1 bluetoothd[1973]: last message repeated 3 times
Feb  8 23:03:10 Linux-1 bluetoothd[1973]: link_key_request

(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)

That means that your computer and the mouse aren't paired. The pin code for Apple mice is usually "0000" (as opposed to most mice which don't need pairing, just a poke and setting as trusted).

Quoted 2 lines

(could be missing something though(don't use gnome/kde-bluetooth)although mightymouse works).

Cheers

#30 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 70 lines · On Tuesday 09 February 2010 05:14:20 Bastien Nocera wrote:
Quoted 62 lines · Justin Mattock
Quoted 38 lines

TIA
Ed Tomlinson

Quoted 24 lines · On Friday 05 February 2010 12:51:43 Marcel Holtmann wrote:

Hi Michael,

Quoted 11 lines
Quoted 2 lines

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Regards

Quoted 8 lines

Marcel

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Quoted 7 lines

Hi,

I have been trying to test Michael's magic mouse drivers against 32.7 (or .8-rc). Is there a tree
based on .32 with these patches? If not would it be possible to post the patches to lkml? I do not mind sorting out a few conflicts - this thread assume one knows a little more about the process being used here than is obvious.

just got one of those magic things as well. I did apply the patchs to the current HEAD, but am getting a lifeless result i.g. here's what I did:

sudo /usr/bin/hcitool scan
sudo /usr/bin/simple-agent hci0 D4:9A:20:88:C7:48
sudo /usr/bin/dbus-send --system --dest=org.bluez --print-reply /org/bluez/1897/hci0/dev_D4_9A_20_88_C7_48 org.bluez.Input.Connect sudo /usr/bin/test-device trusted D4:9A:20:88:C7:48

the thing works after that(no scroll), fast cursor though. then reboot
once up and running nothing.
cat /var/log/daemon.log (shows this)

Quoted 6 lines
Feb  8 22:57:40 Linux-1 bluetoothd[1973]: link_key_request
(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)
Feb  8 22:58:21 Linux-1 bluetoothd[1973]: last message repeated 5 times
Feb  8 22:59:21 Linux-1 bluetoothd[1973]: last message repeated 3 times
Feb  8 23:03:10 Linux-1 bluetoothd[1973]: link_key_request

(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)

Quoted 6 lines

That means that your computer and the mouse aren't paired. The pin code for Apple mice is usually "0000" (as opposed to most mice which don't need pairing, just a poke and setting as trusted).

Quoted 2 lines

(could be missing something though(don't use gnome/kde-bluetooth)although mightymouse works).

The magicmouse part is the easy piece. What I am looking for is the patches to the kernel that add the new function to bluetooth allowing the extra probing. eg the patches (or tree based on .32) that addthe initial version of "hidp_output_raw_report" and friends.

Thanks
Ed

#31 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 4 lines · Ed Tomlinson

What I am looking for is the patches to the kernel that add the new function to bluetooth allowing the extra probing. eg the patches (or tree based on .32) that addthe initial version of "hidp_output_raw_report" and friends.

It's commit 2da31939a4 (and fix 6bf8268f9a9 on top of that). You need quite recent 2.6.33-rc snapshot to have them both:

	$ git describe --contains 6bf8268f9a9
	v2.6.33-rc7~20^2~3^2

Plus there is additional patch queued in my tree, which allows user to specify type of the report that is to be sent (feature/output).

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#32 lore raw

cover Provide a driver for the Apple Magic Mouse

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

#33 lore raw

PATCH 1/2 Provide a driver for the Apple Magic Mouse

Quoted 1 line

From 6fd44770774cd672df6f10e4767f4c17bf90bc30 Mon Sep 17 00:00:00 2001

From: Michael Poole <[email protected]>
Date: Sun, 24 Jan 2010 22:32:29 -0500
Subject: [RFC PATCH 3/4] Export hid_register_report().

The Apple Magic Mouse (and probably other devices) publish reports that are not called out in their HID report descriptors -- they only send them when enabled through other writes to the device. This allows a driver to handle these unlisted reports.

---
 drivers/hid/hid-core.c |    3 ++-
 include/linux/hid.h    |    1 +
 2 files changed, 3 insertions(+), 1 deletions(-)
drivers/hid/hid-core.c +2 -1
index eabe5f8..e39055b 100644
@@ -51,7 +51,7 @@ EXPORT_SYMBOL_GPL(hid_debug);
  * Register a new report for a device.
  */
 
-static struct hid_report *hid_register_report(struct hid_device *device, unsigned type, unsigned id)
+struct hid_report *hid_register_report(struct hid_device *device, unsigned type, unsigned id)
 {
 	struct hid_report_enum *report_enum = device->report_enum + type;
 	struct hid_report *report;
@@ -75,6 +75,7 @@ static struct hid_report *hid_register_report(struct hid_device *device, unsigne
 
 	return report;
 }
+EXPORT_SYMBOL_GPL(hid_register_report);
 
 /*
  * Register a new field for this report.
include/linux/hid.h +1 -0
index 3661a62..456838c 100644
@@ -690,6 +690,7 @@ int hid_input_report(struct hid_device *, int type, u8 *, int, int);
 int hidinput_find_field(struct hid_device *hid, unsigned int type, unsigned int code, struct hid_field **field);
 void hid_output_report(struct hid_report *report, __u8 *data);
 struct hid_device *hid_allocate_device(void);
+struct hid_report *hid_register_report(struct hid_device *device, unsigned type, unsigned id);
 int hid_parse_report(struct hid_device *hid, __u8 *start, unsigned size);
 int hid_check_keys_pressed(struct hid_device *hid);
 int hid_connect(struct hid_device *hid, unsigned int connect_mask);
Signature · 1 line
1.6.5.6
#34 lore raw

PATCH 2/2 Add a device driver for the Apple Magic Mouse.

In reply to Michael Poole
Quoted 1 line

From da3e2b0330fccd584ad79495853e638e9652155b Mon Sep 17 00:00:00 2001

From: Michael Poole <[email protected]>
Date: Sat, 6 Feb 2010 12:24:36 -0500
Subject: Add a device driver for the Apple Magic Mouse.

---
 drivers/hid/Kconfig          |   10 +
 drivers/hid/Makefile         |    1 +
 drivers/hid/hid-core.c       |    1 +
 drivers/hid/hid-ids.h        |    1 +
 drivers/hid/hid-magicmouse.c |  469 ++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 482 insertions(+), 0 deletions(-)
 create mode 100644 drivers/hid/hid-magicmouse.c
drivers/hid/Kconfig +10 -0
index 24d90ea..ba14ec8 100644
@@ -183,6 +183,16 @@ config LOGIRUMBLEPAD2_FF
 	  Say Y here if you want to enable force feedback support for Logitech
 	  Rumblepad 2 devices.
 
+config HID_MAGICMOUSE
+	tristate "Apple" if EMBEDDED
+	depends on BT_HIDP
+	default !EMBEDDED
+	---help---
+	Support for the Apple Magic Mouse.
+
+	Say Y here if you want support for the multi-touch features of the
+	Apple Wireless "Magic" Mouse.
+
 config HID_MICROSOFT
 	tristate "Microsoft" if EMBEDDED
 	depends on USB_HID
drivers/hid/Makefile +1 -0
index 0de2dff..45d81e9 100644
@@ -31,6 +31,7 @@ obj-$(CONFIG_HID_GYRATION)	+= hid-gyration.o
 obj-$(CONFIG_HID_KENSINGTON)	+= hid-kensington.o
 obj-$(CONFIG_HID_KYE)		+= hid-kye.o
 obj-$(CONFIG_HID_LOGITECH)	+= hid-logitech.o
+obj-$(CONFIG_HID_MAGICMOUSE)    += hid-magicmouse.o
 obj-$(CONFIG_HID_MICROSOFT)	+= hid-microsoft.o
 obj-$(CONFIG_HID_MONTEREY)	+= hid-monterey.o
 obj-$(CONFIG_HID_NTRIG)		+= hid-ntrig.o
drivers/hid/hid-core.c +1 -0
index e39055b..f23ca76 100644
@@ -1254,6 +1254,7 @@ static const struct hid_device_id hid_blacklist[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_ATV_IRCONTROL) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_IRCONTROL4) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MIGHTYMOUSE) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MAGICMOUSE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_FOUNTAIN_ANSI) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_FOUNTAIN_ISO) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_GEYSER_ANSI) },
drivers/hid/hid-ids.h +1 -0
index 010368e..11e5218 100644
@@ -56,6 +56,7 @@
 
 #define USB_VENDOR_ID_APPLE		0x05ac
 #define USB_DEVICE_ID_APPLE_MIGHTYMOUSE	0x0304
+#define USB_DEVICE_ID_APPLE_MAGICMOUSE	0x030d
 #define USB_DEVICE_ID_APPLE_FOUNTAIN_ANSI	0x020e
 #define USB_DEVICE_ID_APPLE_FOUNTAIN_ISO	0x020f
 #define USB_DEVICE_ID_APPLE_GEYSER_ANSI	0x0214
drivers/hid/hid-magicmouse.c new file +469 -0

469 lines not on this page · raw message

Signature · 1 line
1.6.5.6
#35 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

Quoted 20 lines · Michael Poole
Quoted 1 line

From da3e2b0330fccd584ad79495853e638e9652155b Mon Sep 17 00:00:00 2001

From: Michael Poole <[email protected]>
Date: Sat, 6 Feb 2010 12:24:36 -0500
Subject: Add a device driver for the Apple Magic Mouse.

---
 drivers/hid/Kconfig          |   10 +
 drivers/hid/Makefile         |    1 +
 drivers/hid/hid-core.c       |    1 +
 drivers/hid/hid-ids.h        |    1 +
 drivers/hid/hid-magicmouse.c |  469 ++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 482 insertions(+), 0 deletions(-)
 create mode 100644 drivers/hid/hid-magicmouse.c
drivers/hid/Kconfig +0 -0
index 24d90ea..ba14ec8 100644

drivers/hid/Kconfig config LOGIRUMBLEPAD2_FF · line 187

Quoted 5 lines
drivers/hid/Kconfig +2 -0
@@ -183,6 +183,16 @@ config LOGIRUMBLEPAD2_FF
 	  Say Y here if you want to enable force feedback support for Logitech
 	  Rumblepad 2 devices.
 
+config HID_MAGICMOUSE
+	tristate "Apple" if EMBEDDED

Could you please make this something else than "Apple"? This string is already used by HID_APPLE.

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#36 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

Quoted 5 lines · Jiri Kosina

drivers/hid/Kconfig config LOGIRUMBLEPAD2_FF · line 187

Quoted 2 lines
drivers/hid/Kconfig +2 -0
+config HID_MAGICMOUSE
+	tristate "Apple" if EMBEDDED

Could you please make this something else than "Apple"? This string is already used by HID_APPLE.

I have fixed this up on my side, no need for you to respin.

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#37 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

In reply to Michael Poole

Hi Michael,

Review

Quoted 2 lines · Michael Poole
(quoted diff) +2 -0
+
+static bool emulate_3button = 1;

If it is a bool then values should be true/false.

Quoted 5 lines
drivers/hid/hid-magicmouse.c new file +5 -0
+module_param(emulate_3button, bool, 0644);
+MODULE_PARM_DESC(emulate_3button, "Emulate a middle button");
+
+static int middle_button_start = -350;
+static int middle_button_stop = +350;

drivers/hid/hid-magicmouse.c line 36

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+
+static bool emulate_scroll_wheel = 1;
+module_param(emulate_scroll_wheel, bool, 0644);
+MODULE_PARM_DESC(emulate_scroll_wheel, "Emulate a scroll wheel");
+
+static bool report_touches = 1;
+module_param(report_touches, bool, 0644);
+MODULE_PARM_DESC(report_touches, "Emit touch records (otherwise, only use them for emulation)");
+
+static bool report_undeciphered = 0;

No need to initialize statics to 0/false.

Quoted 67 lines
drivers/hid/hid-magicmouse.c new file +67 -0
+module_param(report_undeciphered, bool, 0644);
+MODULE_PARM_DESC(report_undeciphered, "Report undeciphered multi-touch state field using a MSC_RAW event");
+
+#define TOUCH_REPORT_ID   0x29
+/* These definitions are not precise, but they're close enough.  (Bits
+ * 0x03 seem to indicate the aspect ratio of the touch, bits 0x70 seem
+ * to be some kind of bit mask -- 0x20 may be a near-field reading,
+ * and 0x40 is actual contact, and 0x10 may be a start/stop or change
+ * indication.)
+ */
+#define TOUCH_STATE_MASK  0xf0
+#define TOUCH_STATE_NONE  0x00
+#define TOUCH_STATE_START 0x30
+#define TOUCH_STATE_DRAG  0x40
+
+/**
+ * struct magicmouse_sc - Tracks Magic Mouse-specific data.
+ * @input: Input device through which we report events.
+ * @quirks: Currently unused.
+ * @last_timestamp: Timestamp from most recent (18-bit) touch report
+ *     (units of milliseconds over short windows, but seems to
+ *     increase faster when there are no touches).
+ * @delta_time: 18-bit difference between the two most recent touch
+ *     reports from the mouse.
+ * @ntouches: Number of touches in most recent touch report.
+ * @scroll_accel: Number of consecutive scroll motions.
+ * @scroll_jiffies: Time of last scroll motion.
+ * @touches: Most recent data for a touch, indexed by tracking ID.
+ * @tracking_ids: Mapping of current touch input data to @touches.
+ */
+struct magicmouse_sc {
+	struct input_dev *input;
+	unsigned long quirks;
+
+	int last_timestamp;
+	int delta_time;
+	int ntouches;
+	int scroll_accel;
+	unsigned long scroll_jiffies;
+
+	struct {
+		short x;
+		short y;
+		short scroll_y;
+		u8 size;
+	} touches[16];
+	int tracking_ids[16];
+};
+
+static int magicmouse_firm_touch(struct magicmouse_sc *msc)
+{
+	int touch = -1;
+	int ii;
+
+	/* If there is only one "firm" touch, set touch to its
+	 * tracking ID.
+	 */
+	for (ii = 0; ii < msc->ntouches; ii++) {
+		int idx = msc->tracking_ids[ii];
+		if (msc->touches[idx].size < 8) {
+			/* Ignore this touch. */
+		} else if (touch >= 0) {
+			touch = -1;
+			break;
+		} else {
+			touch = idx;
+		}

drivers/hid/hid-magicmouse.c line 113

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+	}
+
+	return touch;
+}
+
+static void magicmouse_emit_buttons(struct magicmouse_sc *msc, int state)
+{
+        int last_state = test_bit(BTN_LEFT, msc->input->key) << 0 |
+                test_bit(BTN_RIGHT, msc->input->key) << 1 |
+                test_bit(BTN_MIDDLE, msc->input->key) << 2;

Indented with spaces, we prefer tabs.

Quoted 84 lines
drivers/hid/hid-magicmouse.c new file +84 -0
+
+	if (emulate_3button) {
+		int id;
+
+		/* If some button was pressed before, keep it held
+		 * down.  Otherwise, if there's exactly one firm
+		 * touch, use that to override the mouse's guess.
+		 */
+		if (state == 0) {
+			/* The button was released. */
+		} else if (last_state != 0) {
+			state = last_state;
+		} else if ((id = magicmouse_firm_touch(msc)) >= 0) {
+			int x = msc->touches[id].x;
+			if (x < middle_button_start)
+				state = 1;
+			else if (x > middle_button_stop)
+				state = 2;
+			else
+				state = 4;
+		} /* else: we keep the mouse's guess */
+
+		input_report_key(msc->input, BTN_MIDDLE, state & 4);
+	}
+
+	input_report_key(msc->input, BTN_LEFT, state & 1);
+	input_report_key(msc->input, BTN_RIGHT, state & 2);
+
+	if (state != last_state)
+		msc->scroll_accel = 0;
+}
+
+static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tdata)
+{
+	struct input_dev *input = msc->input;
+	__s32 x_y = tdata[0] << 8 | tdata[1] << 16 | tdata[2] << 24;
+	int misc = tdata[5] | tdata[6] << 8;
+	int id = (misc >> 6) & 15;
+	int x = x_y << 12 >> 20;
+	int y = -(x_y >> 20);
+
+	/* Store tracking ID and other fields. */
+	msc->tracking_ids[raw_id] = id;
+	msc->touches[id].x = x;
+	msc->touches[id].y = y;
+	msc->touches[id].size = misc & 63;
+
+	/* If requested, emulate a scroll wheel by detecting small
+	 * vertical touch motions along the middle of the mouse.
+	 */
+	if (emulate_scroll_wheel &&
+	    middle_button_start < x && x < middle_button_stop) {
+		static const int accel_profile[] = {
+			256, 228, 192, 160, 128, 96, 64, 32,
+		};
+		unsigned long now = jiffies;
+		int step = msc->touches[id].scroll_y - y;
+
+		/* Reset acceleration after half a second. */
+		if (time_after(now, msc->scroll_jiffies + HZ / 2))
+			msc->scroll_accel = 0;
+
+		/* Calculate and apply the scroll motion. */
+		switch (tdata[7] & TOUCH_STATE_MASK) {
+		case TOUCH_STATE_START:
+			msc->touches[id].scroll_y = y;
+                        msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
+						ARRAY_SIZE(accel_profile) - 1);
+			break;
+		case TOUCH_STATE_DRAG:
+			step = step / accel_profile[msc->scroll_accel];
+			if (step != 0) {
+				msc->touches[id].scroll_y = y;
+				msc->scroll_jiffies = now;
+				input_report_rel(input, REL_WHEEL, step);
+			}
+			break;
+		}
+	}
+
+	/* Generate the input events for this touch. */
+	if (report_touches) {
+                int orientation = (misc >> 10) - 32;
+

drivers/hid/hid-magicmouse.c line 207

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+		input_report_abs(input, ABS_MT_TRACKING_ID, id);
+		input_report_abs(input, ABS_MT_TOUCH_MAJOR, tdata[3]);
+		input_report_abs(input, ABS_MT_TOUCH_MINOR, tdata[4]);
+		input_report_abs(input, ABS_MT_ORIENTATION, orientation);
+		input_report_abs(input, ABS_MT_POSITION_X, x);
+		input_report_abs(input, ABS_MT_POSITION_Y, y);
+
+		if (report_undeciphered) {
+			input_event(input, EV_MSC, MSC_RAW, tdata[7]);
+		}

No need for braces for single statement.

Quoted 88 lines
drivers/hid/hid-magicmouse.c new file +88 -0
+
+		input_mt_sync(input);
+	}
+}
+
+static int magicmouse_raw_event(struct hid_device *hdev,
+		struct hid_report *report, u8 *data, int size)
+{
+	struct magicmouse_sc *msc = hid_get_drvdata(hdev);
+	struct input_dev *input = msc->input;
+	int x, y, ts, ii, clicks;
+
+	switch (data[0]) {
+	case 0x10:
+		if (size != 6)
+			return 0;
+		x = (__s16)(data[2] | data[3] << 8);
+		y = (__s16)(data[4] | data[5] << 8);
+		clicks = data[1];
+		break;
+	case TOUCH_REPORT_ID:
+		/* Expect six bytes of prefix, and N*8 bytes of touch data. */
+		if (size < 6 || ((size - 6) % 8) != 0)
+			return 0;
+		ts = data[3] >> 6 | data[4] << 2 | data[5] << 10;
+		msc->delta_time = (ts - msc->last_timestamp) & 0x3ffff;
+		msc->last_timestamp = ts;
+		msc->ntouches = (size - 6) / 8;
+		for (ii = 0; ii < msc->ntouches; ii++)
+			magicmouse_emit_touch(msc, ii, data + ii * 8 + 6);
+		/* When emulating three-button mode, it is important
+		 * to have the current touch information before
+		 * generating a click event.
+		 */
+		x = (signed char)data[1];
+		y = (signed char)data[2];
+		clicks = data[3];
+		break;
+	case 0x20: /* Theoretically battery status (0-100), but I have
+		    * never seen it -- maybe it is only upon request.
+		    */
+	case 0x60: /* Unknown, maybe laser on/off. */
+	case 0x61: /* Laser reflection status change.
+		    * data[1]: 0 = spotted, 1 = lost
+		    */
+	default:
+		return 0;
+	}
+
+	magicmouse_emit_buttons(msc, clicks & 3);
+	input_report_rel(input, REL_X, x);
+	input_report_rel(input, REL_Y, y);
+	input_sync(input);
+	return 1;
+}
+
+static int magicmouse_input_open(struct input_dev *dev)
+{
+	struct hid_device *hid = input_get_drvdata(dev);
+
+	return hid->ll_driver->open(hid);
+}
+
+static void magicmouse_input_close(struct input_dev *dev)
+{
+	struct hid_device *hid = input_get_drvdata(dev);
+
+	hid->ll_driver->close(hid);
+}
+
+static void magicmouse_setup_input(struct input_dev *input, struct hid_device *hdev)
+{
+	input_set_drvdata(input, hdev);
+	input->event = hdev->ll_driver->hidinput_input_event;
+	input->open = magicmouse_input_open;
+	input->close = magicmouse_input_close;
+
+	input->name = hdev->name;
+	input->phys = hdev->phys;
+	input->uniq = hdev->uniq;
+	input->id.bustype = hdev->bus;
+	input->id.vendor = hdev->vendor;
+	input->id.product = hdev->product;
+	input->id.version = hdev->version;
+	input->dev.parent = hdev->dev.parent;
+
+	set_bit(EV_KEY, input->evbit);
+	set_bit(BTN_LEFT, input->keybit);

drivers/hid/hid-magicmouse.c line 305

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+	set_bit(BTN_RIGHT, input->keybit);
+	if (emulate_3button)
+		set_bit(BTN_MIDDLE, input->keybit);
+	set_bit(BTN_TOOL_FINGER, input->keybit);
+
+	set_bit(EV_REL, input->evbit);
+	set_bit(REL_X, input->relbit);
+	set_bit(REL_Y, input->relbit);
+	if (emulate_scroll_wheel)
+		set_bit(REL_WHEEL, input->relbit);

I'd use __set_bit() instead, no need to lock the bus.

drivers/hid/hid-magicmouse.c line 313

Quoted 8 lines
drivers/hid/hid-magicmouse.c new file +8 -0
+
+	if (report_touches) {
+		set_bit(EV_ABS, input->evbit);
+
+		set_bit(ABS_MT_TRACKING_ID, input->absbit);
+		input->absmin[ABS_MT_TRACKING_ID] = 0;
+		input->absmax[ABS_MT_TRACKING_ID] = 15;
+		input->absfuzz[ABS_MT_TRACKING_ID] = 0;

input_set_abs_params() is a bit more compact.

Quoted 70 lines
drivers/hid/hid-magicmouse.c new file +70 -0
+
+		set_bit(ABS_MT_TOUCH_MAJOR, input->absbit);
+		input->absmin[ABS_MT_TOUCH_MAJOR] = 0;
+		input->absmax[ABS_MT_TOUCH_MAJOR] = 255;
+		input->absfuzz[ABS_MT_TOUCH_MAJOR] = 4;
+
+		set_bit(ABS_MT_TOUCH_MINOR, input->absbit);
+		input->absmin[ABS_MT_TOUCH_MINOR] = 0;
+		input->absmax[ABS_MT_TOUCH_MINOR] = 255;
+		input->absfuzz[ABS_MT_TOUCH_MINOR] = 4;
+
+		set_bit(ABS_MT_ORIENTATION, input->absbit);
+		input->absmin[ABS_MT_ORIENTATION] = -32;
+		input->absmax[ABS_MT_ORIENTATION] = 31;
+		input->absfuzz[ABS_MT_ORIENTATION] = 1;
+
+		set_bit(ABS_MT_POSITION_X, input->absbit);
+		input->absmin[ABS_MT_POSITION_X] = -1100;
+		input->absmax[ABS_MT_POSITION_X] = 1358;
+		input->absfuzz[ABS_MT_POSITION_X] = 4;
+
+		/* Note: Touch Y position from the device is inverted relative
+		 * to how pointer motion is reported (and relative to how USB
+		 * HID recommends the coordinates work).  This driver keeps
+		 * the origin at the same position, and just uses the additive
+		 * inverse of the reported Y.
+		 */
+		set_bit(ABS_MT_POSITION_Y, input->absbit);
+		input->absmin[ABS_MT_POSITION_Y] = -1589;
+		input->absmax[ABS_MT_POSITION_Y] = 2047;
+		input->absfuzz[ABS_MT_POSITION_Y] = 4;
+	}
+
+	if (report_undeciphered) {
+		set_bit(EV_MSC, input->evbit);
+		set_bit(MSC_RAW, input->mscbit);
+	}
+}
+
+static int magicmouse_probe(struct hid_device *hdev,
+	const struct hid_device_id *id)
+{
+	__u8 feature_1[] = { 0xd7, 0x01 };
+	__u8 feature_2[] = { 0xf8, 0x01, 0x32 };
+	struct input_dev *input;
+	struct magicmouse_sc *msc;
+	struct hid_report *report;
+	int ret;
+
+	msc = kzalloc(sizeof(*msc), GFP_KERNEL);
+	if (msc == NULL) {
+		dev_err(&hdev->dev, "can't alloc magicmouse descriptor\n");
+		return -ENOMEM;
+	}
+
+	msc->quirks = id->driver_data;
+	hid_set_drvdata(hdev, msc);
+
+	ret = hid_parse(hdev);
+	if (ret) {
+		dev_err(&hdev->dev, "magicmouse hid parse failed\n");
+		goto err_free;
+	}
+
+	ret = hid_hw_start(hdev, HID_CONNECT_DEFAULT);
+	if (ret) {
+		dev_err(&hdev->dev, "magicmouse hw start failed\n");
+		goto err_free;
+	}
+

drivers/hid/hid-magicmouse.c line 393

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+	report = hid_register_report(hdev, HID_INPUT_REPORT, TOUCH_REPORT_ID);
+	if (!report) {
+		dev_err(&hdev->dev, "unable to register touch report\n");
+		ret = -ENOMEM;
+		goto err_free;
+	}
+	report->size = 6;
+
+	ret = hdev->hid_output_raw_report(hdev, feature_1, sizeof(feature_1),
+			HID_FEATURE_REPORT);

This will cause transmission of on-stack data through USB which it does not like.

drivers/hid/hid-magicmouse.c line 400

Quoted 7 lines
drivers/hid/hid-magicmouse.c new file +7 -0
+	if (ret != sizeof(feature_1)) {
+		dev_err(&hdev->dev, "unable to request touch data (1:%d)\n",
+				ret);
+		goto err_free;
+	}
+	ret = hdev->hid_output_raw_report(hdev, feature_2,
+			sizeof(feature_2), HID_FEATURE_REPORT);

Same here.

Quoted 17 lines
drivers/hid/hid-magicmouse.c new file +17 -0
+	if (ret != sizeof(feature_2)) {
+		dev_err(&hdev->dev, "unable to request touch data (2:%d)\n",
+				ret);
+		goto err_free;
+	}
+
+	input = input_allocate_device();
+	if (!input) {
+		dev_err(&hdev->dev, "can't alloc input device\n");
+		ret = -ENOMEM;
+		goto err_free;
+	}
+	magicmouse_setup_input(input, hdev);
+
+	ret = input_register_device(input);
+	if (ret) {
+		dev_err(&hdev->dev, "input device registration failed\n");

drivers/hid/hid-magicmouse.c line 427

Quoted 10 lines
drivers/hid/hid-magicmouse.c new file +10 -0
+		goto err_both;
+	}
+	msc->input = input;
+
+	return 0;
+ err_both:
+	input_free_device(input);
+ err_free:
+	kfree(msc);
+	return ret;

hid_hw_stop() is missing in error path. Also see question about freeing report below.

drivers/hid/hid-magicmouse.c line 431

Quoted 4 lines
drivers/hid/hid-magicmouse.c new file +4 -0
+}
+
+static void magicmouse_remove(struct hid_device *hdev)
+{

Do we need to unregister report (I am not that familiar with HID, who is responsible for cleaning report lists?)?

Quoted 45 lines
drivers/hid/hid-magicmouse.c new file +38 -0
+	hid_hw_stop(hdev);
+	kfree(hid_get_drvdata(hdev));
+}
+
+static const struct hid_device_id magic_mice[] = {
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MAGICMOUSE),
+		.driver_data = 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(hid, magic_mice);
+
+static struct hid_driver magicmouse_driver = {
+	.name = "magicmouse",
+	.id_table = magic_mice,
+	.probe = magicmouse_probe,
+	.remove = magicmouse_remove,
+	.raw_event = magicmouse_raw_event,
+};
+
+static int __init magicmouse_init(void)
+{
+	int ret;
+
+	ret = hid_register_driver(&magicmouse_driver);
+	if (ret)
+		printk(KERN_ERR "can't register magicmouse driver\n");
+
+	return ret;
+}
+
+static void __exit magicmouse_exit(void)
+{
+	hid_unregister_driver(&magicmouse_driver);
+}
+
+module_init(magicmouse_init);
+module_exit(magicmouse_exit);
+MODULE_LICENSE("GPL");
Signature · 6 lines
1.6.5.6

--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Signature · 1 line
Dmitry
#38 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

I'll post a patch for these based on Jiri's for-next tree this evening.

Quoted 7 lines · Dmitry Torokhov

Hi Michael,

Review

Quoted 2 lines · Michael Poole
(quoted diff) +2 -0
+
+static bool emulate_3button = 1;

If it is a bool then values should be true/false.

Thanks.

Quoted 9 lines
Quoted 9 lines
drivers/hid/hid-magicmouse.c new file +9 -0
+module_param(emulate_3button, bool, 0644);
+MODULE_PARM_DESC(emulate_3button, "Emulate a middle button");
+
+static int middle_button_start = -350;
+static int middle_button_stop = +350;
+
+static bool emulate_scroll_wheel = 1;
+module_param(emulate_scroll_wheel, bool, 0644);
+MODULE_PARM_DESC(emulate_scroll_wheel, "Emulate a scroll wheel");
Quoted 8 lines

drivers/hid/hid-magicmouse.c line 36

Quoted 6 lines
drivers/hid/hid-magicmouse.c new file +6 -0
+
+static bool report_touches = 1;
+module_param(report_touches, bool, 0644);
+MODULE_PARM_DESC(report_touches, "Emit touch records (otherwise, only use them for emulation)");
+
+static bool report_undeciphered = 0;

No need to initialize statics to 0/false.

Okay, I'll remote that.

[snip]

Quoted 5 lines

drivers/hid/hid-magicmouse.c line 113

Quoted 3 lines
drivers/hid/hid-magicmouse.c new file +3 -0
+        int last_state = test_bit(BTN_LEFT, msc->input->key) << 0 |
+                test_bit(BTN_RIGHT, msc->input->key) << 1 |
+                test_bit(BTN_MIDDLE, msc->input->key) << 2;

Indented with spaces, we prefer tabs.

Whoops -- one of my machines must be misconfigured. I'll fix that.

[snip]

Quoted 5 lines

drivers/hid/hid-magicmouse.c line 207

Quoted 3 lines
drivers/hid/hid-magicmouse.c new file +3 -0
+		if (report_undeciphered) {
+			input_event(input, EV_MSC, MSC_RAW, tdata[7]);
+		}

No need for braces for single statement.

Okay.

[snip]

Quoted 4 lines

drivers/hid/hid-magicmouse.c line 305

Quoted 2 lines
drivers/hid/hid-magicmouse.c new file +2 -0
+	if (emulate_scroll_wheel)
+		set_bit(REL_WHEEL, input->relbit);

I'd use __set_bit() instead, no need to lock the bus.

Okay. I wasn't aware of that semantic difference.

Quoted 10 lines

drivers/hid/hid-magicmouse.c line 313

Quoted 8 lines
drivers/hid/hid-magicmouse.c new file +8 -0
+
+	if (report_touches) {
+		set_bit(EV_ABS, input->evbit);
+
+		set_bit(ABS_MT_TRACKING_ID, input->absbit);
+		input->absmin[ABS_MT_TRACKING_ID] = 0;
+		input->absmax[ABS_MT_TRACKING_ID] = 15;
+		input->absfuzz[ABS_MT_TRACKING_ID] = 0;

input_set_abs_params() is a bit more compact.

Thanks -- more idiomatic is always handy.

[snip]

Quoted 5 lines

drivers/hid/hid-magicmouse.c line 393

Quoted 2 lines
drivers/hid/hid-magicmouse.c new file +2 -0
+	ret = hdev->hid_output_raw_report(hdev, feature_1, sizeof(feature_1),
+			HID_FEATURE_REPORT);

This will cause transmission of on-stack data through USB which it does not like.

Current Magic Mouse devices are Bluetooth-only. Is this an issue for that stack as well? (hid-sony.c and hid-wacom.c also use hid_output_raw_report() with on-stack buffers for Bluetooth devices -- if they need it too, I can write a patch for those. I have a PS3 controller that I can test with, but don't have a Wacom tablet.) Would making those buffers static put them in an appropriate section, or does the memory need to be allocated from a heap?

Quoted 32 lines

drivers/hid/hid-magicmouse.c line 400

Quoted 7 lines
drivers/hid/hid-magicmouse.c new file +7 -0
+	if (ret != sizeof(feature_1)) {
+		dev_err(&hdev->dev, "unable to request touch data (1:%d)\n",
+				ret);
+		goto err_free;
+	}
+	ret = hdev->hid_output_raw_report(hdev, feature_2,
+			sizeof(feature_2), HID_FEATURE_REPORT);

Same here.

Quoted 22 lines
drivers/hid/hid-magicmouse.c new file +22 -0
+	if (ret != sizeof(feature_2)) {
+		dev_err(&hdev->dev, "unable to request touch data (2:%d)\n",
+				ret);
+		goto err_free;
+	}
+
+	input = input_allocate_device();
+	if (!input) {
+		dev_err(&hdev->dev, "can't alloc input device\n");
+		ret = -ENOMEM;
+		goto err_free;
+	}
+	magicmouse_setup_input(input, hdev);
+
+	ret = input_register_device(input);
+	if (ret) {
+		dev_err(&hdev->dev, "input device registration failed\n");
+		goto err_both;
+	}
+	msc->input = input;
+
+	return 0;
Quoted 8 lines

drivers/hid/hid-magicmouse.c line 427

Quoted 5 lines

+ err_both:

drivers/hid/hid-magicmouse.c new file +4 -0
+	input_free_device(input);
+ err_free:
+	kfree(msc);
+	return ret;

hid_hw_stop() is missing in error path. Also see question about freeing report below.

Thanks, I'll add the hid_hw_stop().

Quoted 7 lines

drivers/hid/hid-magicmouse.c line 431

Quoted 4 lines
drivers/hid/hid-magicmouse.c new file +4 -0
+}
+
+static void magicmouse_remove(struct hid_device *hdev)
+{

Do we need to unregister report (I am not that familiar with HID, who is responsible for cleaning report lists?)?

hid-core.c's hid_device_release() frees all the reports that registered with hid_register_report().

Michael Poole

#39 lore raw

PATCH 1/1 hid-magicmouse: Coding style and probe failure fixes.

In reply to Dmitry Torokhov
Quoted 1 line

From 04b395dbbd1ad2b836188e6f125940ae8fac6925 Mon Sep 17 00:00:00 2001

As suggested by Dmitry Torokhov on 10 Feb: Use proper values to initialize bool configuration variables, tabs rather than spaces, no braces for one-line else clause, __set_bit() when the operation doesn't have to be atomic, input_set_abs_params() rather than writing the fields directly, and call hid_hw_stop() when appropriate to handle failures in the probe.

---
Dmitry and Jiri,

I haven't had a chance to run-test these changes yet -- hid/for-next causes corrupt X display on my laptop, whereas v2.6.33-rc6 and -rc7 are fine; I'm still bisecting to figure out the cause -- but this patch is not complicated. (It does compile.)

I left the buffers for hdev->hid_output_raw_report() on the stack because the Bluetooth HIDP code memcpy's the contents into a freshly allocated skb.

Michael

 drivers/hid/hid-magicmouse.c |  100 +++++++++++++++++-------------------------
 1 files changed, 40 insertions(+), 60 deletions(-)
drivers/hid/hid-magicmouse.c +40 -60
index f94b3e4..4a3a94f 100644
@@ -18,22 +18,22 @@
 
 #include "hid-ids.h"
 
-static bool emulate_3button = 1;
+static bool emulate_3button = true;
 module_param(emulate_3button, bool, 0644);
 MODULE_PARM_DESC(emulate_3button, "Emulate a middle button");
 
 static int middle_button_start = -350;
 static int middle_button_stop = +350;
 
-static bool emulate_scroll_wheel = 1;
+static bool emulate_scroll_wheel = true;
 module_param(emulate_scroll_wheel, bool, 0644);
 MODULE_PARM_DESC(emulate_scroll_wheel, "Emulate a scroll wheel");
 
-static bool report_touches = 1;
+static bool report_touches = true;
 module_param(report_touches, bool, 0644);
 MODULE_PARM_DESC(report_touches, "Emit touch records (otherwise, only use them for emulation)");
 
-static bool report_undeciphered = 0;
+static bool report_undeciphered;
 module_param(report_undeciphered, bool, 0644);
 MODULE_PARM_DESC(report_undeciphered, "Report undeciphered multi-touch state field using a MSC_RAW event");
 
@@ -108,9 +108,9 @@ static int magicmouse_firm_touch(struct magicmouse_sc *msc)
 
 static void magicmouse_emit_buttons(struct magicmouse_sc *msc, int state)
 {
-        int last_state = test_bit(BTN_LEFT, msc->input->key) << 0 |
-                test_bit(BTN_RIGHT, msc->input->key) << 1 |
-                test_bit(BTN_MIDDLE, msc->input->key) << 2;
+	int last_state = test_bit(BTN_LEFT, msc->input->key) << 0 |
+		test_bit(BTN_RIGHT, msc->input->key) << 1 |
+		test_bit(BTN_MIDDLE, msc->input->key) << 2;
 
 	if (emulate_3button) {
 		int id;
@@ -177,7 +177,7 @@ static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tda
 		switch (tdata[7] & TOUCH_STATE_MASK) {
 		case TOUCH_STATE_START:
 			msc->touches[id].scroll_y = y;
-                        msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
+			msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
 						ARRAY_SIZE(accel_profile) - 1);
 			break;
 		case TOUCH_STATE_DRAG:
@@ -193,7 +193,7 @@ static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tda
 
 	/* Generate the input events for this touch. */
 	if (report_touches) {
-                int orientation = (misc >> 10) - 32;
+		int orientation = (misc >> 10) - 32;
 
 		input_report_abs(input, ABS_MT_TRACKING_ID, id);
 		input_report_abs(input, ABS_MT_TOUCH_MAJOR, tdata[3]);
@@ -202,9 +202,8 @@ static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tda
 		input_report_abs(input, ABS_MT_POSITION_X, x);
 		input_report_abs(input, ABS_MT_POSITION_Y, y);
 
-		if (report_undeciphered) {
+		if (report_undeciphered)
 			input_event(input, EV_MSC, MSC_RAW, tdata[7]);
-		}
 
 		input_mt_sync(input);
 	}
@@ -291,62 +290,41 @@ static void magicmouse_setup_input(struct input_dev *input, struct hid_device *h
 	input->id.version = hdev->version;
 	input->dev.parent = hdev->dev.parent;
 
-	set_bit(EV_KEY, input->evbit);
-	set_bit(BTN_LEFT, input->keybit);
-	set_bit(BTN_RIGHT, input->keybit);
+	__set_bit(EV_KEY, input->evbit);
+	__set_bit(BTN_LEFT, input->keybit);
+	__set_bit(BTN_RIGHT, input->keybit);
 	if (emulate_3button)
-		set_bit(BTN_MIDDLE, input->keybit);
-	set_bit(BTN_TOOL_FINGER, input->keybit);
+		__set_bit(BTN_MIDDLE, input->keybit);
+	__set_bit(BTN_TOOL_FINGER, input->keybit);
 
-	set_bit(EV_REL, input->evbit);
-	set_bit(REL_X, input->relbit);
-	set_bit(REL_Y, input->relbit);
+	__set_bit(EV_REL, input->evbit);
+	__set_bit(REL_X, input->relbit);
+	__set_bit(REL_Y, input->relbit);
 	if (emulate_scroll_wheel)
-		set_bit(REL_WHEEL, input->relbit);
+		__set_bit(REL_WHEEL, input->relbit);
 
 	if (report_touches) {
-		set_bit(EV_ABS, input->evbit);
-
-		set_bit(ABS_MT_TRACKING_ID, input->absbit);
-		input->absmin[ABS_MT_TRACKING_ID] = 0;
-		input->absmax[ABS_MT_TRACKING_ID] = 15;
-		input->absfuzz[ABS_MT_TRACKING_ID] = 0;
-
-		set_bit(ABS_MT_TOUCH_MAJOR, input->absbit);
-		input->absmin[ABS_MT_TOUCH_MAJOR] = 0;
-		input->absmax[ABS_MT_TOUCH_MAJOR] = 255;
-		input->absfuzz[ABS_MT_TOUCH_MAJOR] = 4;
-
-		set_bit(ABS_MT_TOUCH_MINOR, input->absbit);
-		input->absmin[ABS_MT_TOUCH_MINOR] = 0;
-		input->absmax[ABS_MT_TOUCH_MINOR] = 255;
-		input->absfuzz[ABS_MT_TOUCH_MINOR] = 4;
-
-		set_bit(ABS_MT_ORIENTATION, input->absbit);
-		input->absmin[ABS_MT_ORIENTATION] = -32;
-		input->absmax[ABS_MT_ORIENTATION] = 31;
-		input->absfuzz[ABS_MT_ORIENTATION] = 1;
-
-		set_bit(ABS_MT_POSITION_X, input->absbit);
-		input->absmin[ABS_MT_POSITION_X] = -1100;
-		input->absmax[ABS_MT_POSITION_X] = 1358;
-		input->absfuzz[ABS_MT_POSITION_X] = 4;
-
+		__set_bit(EV_ABS, input->evbit);
+
+		input_set_abs_params(input, ABS_MT_TRACKING_ID, 0, 15, 0, 0);
+		input_set_abs_params(input, ABS_MT_TOUCH_MAJOR, 0, 255, 4, 0);
+		input_set_abs_params(input, ABS_MT_TOUCH_MINOR, 0, 255, 4, 0);
+		input_set_abs_params(input, ABS_MT_ORIENTATION, -32, 31, 1, 0);
+		input_set_abs_params(input, ABS_MT_POSITION_X, -1100, 1358,
+				4, 0);
 		/* Note: Touch Y position from the device is inverted relative
 		 * to how pointer motion is reported (and relative to how USB
 		 * HID recommends the coordinates work).  This driver keeps
 		 * the origin at the same position, and just uses the additive
 		 * inverse of the reported Y.
 		 */
-		set_bit(ABS_MT_POSITION_Y, input->absbit);
-		input->absmin[ABS_MT_POSITION_Y] = -1589;
-		input->absmax[ABS_MT_POSITION_Y] = 2047;
-		input->absfuzz[ABS_MT_POSITION_Y] = 4;
+		input_set_abs_params(input, ABS_MT_POSITION_Y, -1589, 2047,
+				4, 0);
 	}
 
 	if (report_undeciphered) {
-		set_bit(EV_MSC, input->evbit);
-		set_bit(MSC_RAW, input->mscbit);
+		__set_bit(EV_MSC, input->evbit);
+		__set_bit(MSC_RAW, input->mscbit);
 	}
 }
 
@@ -385,7 +363,7 @@ static int magicmouse_probe(struct hid_device *hdev,
 	if (!report) {
 		dev_err(&hdev->dev, "unable to register touch report\n");
 		ret = -ENOMEM;
-		goto err_free;
+		goto err_stop_hw;
 	}
 	report->size = 6;
 
@@ -394,35 +372,37 @@ static int magicmouse_probe(struct hid_device *hdev,
 	if (ret != sizeof(feature_1)) {
 		dev_err(&hdev->dev, "unable to request touch data (1:%d)\n",
 				ret);
-		goto err_free;
+		goto err_stop_hw;
 	}
 	ret = hdev->hid_output_raw_report(hdev, feature_2,
 			sizeof(feature_2), HID_FEATURE_REPORT);
 	if (ret != sizeof(feature_2)) {
 		dev_err(&hdev->dev, "unable to request touch data (2:%d)\n",
 				ret);
-		goto err_free;
+		goto err_stop_hw;
 	}
 
 	input = input_allocate_device();
 	if (!input) {
 		dev_err(&hdev->dev, "can't alloc input device\n");
 		ret = -ENOMEM;
-		goto err_free;
+		goto err_stop_hw;
 	}
 	magicmouse_setup_input(input, hdev);
 
 	ret = input_register_device(input);
 	if (ret) {
 		dev_err(&hdev->dev, "input device registration failed\n");
-		goto err_both;
+		goto err_input;
 	}
 	msc->input = input;
 
 	return 0;
- err_both:
+err_input:
 	input_free_device(input);
- err_free:
+err_stop_hw:
+	hid_hw_stop(hdev);
+err_free:
 	kfree(msc);
 	return ret;
 }
Signature · 1 line
1.6.5.6
#40 lore raw

Re: [PATCH] hid-magicmouse: Coding style and probe failure fixes.

Quoted 13 lines · Michael Poole

From 04b395dbbd1ad2b836188e6f125940ae8fac6925 Mon Sep 17 00:00:00 2001

As suggested by Dmitry Torokhov on 10 Feb: Use proper values to initialize bool configuration variables, tabs rather than spaces, no braces for one-line else clause, __set_bit() when the operation doesn't have to be atomic, input_set_abs_params() rather than writing the fields directly, and call hid_hw_stop() when appropriate to handle failures in the probe.

---
Dmitry and Jiri,

Quoted 9 lines

I haven't had a chance to run-test these changes yet -- hid/for-next causes corrupt X display on my laptop, whereas v2.6.33-rc6 and -rc7 are fine; I'm still bisecting to figure out the cause -- but this patch is not complicated. (It does compile.)

I left the buffers for hdev->hid_output_raw_report() on the stack because the Bluetooth HIDP code memcpy's the contents into a freshly allocated skb.

Ah, OK, I missed the fact that itis BT only device.

The resul looks much better now, thank you for making the changes.

Signature · 1 line
Dmitry
#41 lore raw

Re: [PATCH] hid-magicmouse: Coding style and probe failure fixes.

In reply to Michael Poole
Quoted 3 lines · Michael Poole

I haven't had a chance to run-test these changes yet -- hid/for-next causes corrupt X display on my laptop, whereas v2.6.33-rc6 and -rc7 are fine;

That's indeed quite strange. I'd be curious to see the result of the bisection, thanks a lot for doing this.

I have applied your patch to the branch containing your driver.

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#42 lore raw

Re: [PATCH] hid-magicmouse: Coding style and probe failure fixes.

Quoted 8 lines · Jiri Kosina
Quoted 3 lines · Michael Poole

I haven't had a chance to run-test these changes yet -- hid/for-next causes corrupt X display on my laptop, whereas v2.6.33-rc6 and -rc7 are fine;

That's indeed quite strange. I'd be curious to see the result of the bisection, thanks a lot for doing this.

The problem was introduced by 859ddf0 and reverted in 6f14a66 (both only touching lib/idr.c). You just happened to merge from master during an unfortunate window :)

Cherry-picking the revert (6f14a66) on top of commit 093657 (details shown below for reference) yields a working kernel with a working Magic Mouse driver. I expect that merging the rest of master would also work.

commit 09365737da8b3af76abd9d811d435d3ce379903a
Merge: 1b97e37 71b38bd
Author: Jiri Kosina <[email protected]>
Date: Thu Feb 11 11:22:52 2010 +0100

Merge branch 'apple_magic_mouse' into for-next

Michael Poole

#43 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

In reply to Michael Poole

Hi,

When I apply this to head, I get:

drivers/hid/hid-magicmouse.c: In function 'magicmouse_probe':
drivers/hid/hid-magicmouse.c:393: error: too many arguments to function 'hdev->hid_output_raw_report' drivers/hid/hid-magicmouse.c:400: error: too many arguments to function 'hdev->hid_output_raw_report'

I only need the second patch - the first seems to be in tree.

Any ideas as to what I am missing?

TIA
Ed

Quoted 548 lines · On Tuesday 09 February 2010 08:13:08 Michael Poole wrote:

From da3e2b0330fccd584ad79495853e638e9652155b Mon Sep 17 00:00:00 2001
From: Michael Poole <[email protected]>
Date: Sat, 6 Feb 2010 12:24:36 -0500
Subject: Add a device driver for the Apple Magic Mouse.

---
 drivers/hid/Kconfig          |   10 +
 drivers/hid/Makefile         |    1 +
 drivers/hid/hid-core.c       |    1 +
 drivers/hid/hid-ids.h        |    1 +
 drivers/hid/hid-magicmouse.c |  469 ++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 482 insertions(+), 0 deletions(-)
 create mode 100644 drivers/hid/hid-magicmouse.c
drivers/hid/Kconfig +10 -0
index 24d90ea..ba14ec8 100644
@@ -183,6 +183,16 @@ config LOGIRUMBLEPAD2_FF
 	  Say Y here if you want to enable force feedback support for Logitech
 	  Rumblepad 2 devices.
 
+config HID_MAGICMOUSE
+	tristate "Apple" if EMBEDDED
+	depends on BT_HIDP
+	default !EMBEDDED
+	---help---
+	Support for the Apple Magic Mouse.
+
+	Say Y here if you want support for the multi-touch features of the
+	Apple Wireless "Magic" Mouse.
+
 config HID_MICROSOFT
 	tristate "Microsoft" if EMBEDDED
 	depends on USB_HID
drivers/hid/Makefile +1 -0
index 0de2dff..45d81e9 100644
@@ -31,6 +31,7 @@ obj-$(CONFIG_HID_GYRATION)	+= hid-gyration.o
 obj-$(CONFIG_HID_KENSINGTON)	+= hid-kensington.o
 obj-$(CONFIG_HID_KYE)		+= hid-kye.o
 obj-$(CONFIG_HID_LOGITECH)	+= hid-logitech.o
+obj-$(CONFIG_HID_MAGICMOUSE)    += hid-magicmouse.o
 obj-$(CONFIG_HID_MICROSOFT)	+= hid-microsoft.o
 obj-$(CONFIG_HID_MONTEREY)	+= hid-monterey.o
 obj-$(CONFIG_HID_NTRIG)		+= hid-ntrig.o
drivers/hid/hid-core.c +1 -0
index e39055b..f23ca76 100644
@@ -1254,6 +1254,7 @@ static const struct hid_device_id hid_blacklist[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_ATV_IRCONTROL) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_IRCONTROL4) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MIGHTYMOUSE) },
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MAGICMOUSE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_FOUNTAIN_ANSI) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_FOUNTAIN_ISO) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_GEYSER_ANSI) },
drivers/hid/hid-ids.h +1 -0
index 010368e..11e5218 100644
@@ -56,6 +56,7 @@
 
 #define USB_VENDOR_ID_APPLE		0x05ac
 #define USB_DEVICE_ID_APPLE_MIGHTYMOUSE	0x0304
+#define USB_DEVICE_ID_APPLE_MAGICMOUSE	0x030d
 #define USB_DEVICE_ID_APPLE_FOUNTAIN_ANSI	0x020e
 #define USB_DEVICE_ID_APPLE_FOUNTAIN_ISO	0x020f
 #define USB_DEVICE_ID_APPLE_GEYSER_ANSI	0x0214
drivers/hid/hid-magicmouse.c new file +469 -0
new file mode 100644
index 0000000..f94b3e4
@@ -0,0 +1,469 @@
+/*
+ *   Apple "Magic" Wireless Mouse driver
+ *
+ *   Copyright (c) 2010 Michael Poole <[email protected]>
+ */
+
+/*
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ */
+
+#include <linux/device.h>
+#include <linux/hid.h>
+#include <linux/module.h>
+#include <linux/usb.h>
+
+#include "hid-ids.h"
+
+static bool emulate_3button = 1;
+module_param(emulate_3button, bool, 0644);
+MODULE_PARM_DESC(emulate_3button, "Emulate a middle button");
+
+static int middle_button_start = -350;
+static int middle_button_stop = +350;
+
+static bool emulate_scroll_wheel = 1;
+module_param(emulate_scroll_wheel, bool, 0644);
+MODULE_PARM_DESC(emulate_scroll_wheel, "Emulate a scroll wheel");
+
+static bool report_touches = 1;
+module_param(report_touches, bool, 0644);
+MODULE_PARM_DESC(report_touches, "Emit touch records (otherwise, only use them for emulation)");
+
+static bool report_undeciphered = 0;
+module_param(report_undeciphered, bool, 0644);
+MODULE_PARM_DESC(report_undeciphered, "Report undeciphered multi-touch state field using a MSC_RAW event");
+
+#define TOUCH_REPORT_ID   0x29
+/* These definitions are not precise, but they're close enough.  (Bits
+ * 0x03 seem to indicate the aspect ratio of the touch, bits 0x70 seem
+ * to be some kind of bit mask -- 0x20 may be a near-field reading,
+ * and 0x40 is actual contact, and 0x10 may be a start/stop or change
+ * indication.)
+ */
+#define TOUCH_STATE_MASK  0xf0
+#define TOUCH_STATE_NONE  0x00
+#define TOUCH_STATE_START 0x30
+#define TOUCH_STATE_DRAG  0x40
+
+/**
+ * struct magicmouse_sc - Tracks Magic Mouse-specific data.
+ * @input: Input device through which we report events.
+ * @quirks: Currently unused.
+ * @last_timestamp: Timestamp from most recent (18-bit) touch report
+ *     (units of milliseconds over short windows, but seems to
+ *     increase faster when there are no touches).
+ * @delta_time: 18-bit difference between the two most recent touch
+ *     reports from the mouse.
+ * @ntouches: Number of touches in most recent touch report.
+ * @scroll_accel: Number of consecutive scroll motions.
+ * @scroll_jiffies: Time of last scroll motion.
+ * @touches: Most recent data for a touch, indexed by tracking ID.
+ * @tracking_ids: Mapping of current touch input data to @touches.
+ */
+struct magicmouse_sc {
+	struct input_dev *input;
+	unsigned long quirks;
+
+	int last_timestamp;
+	int delta_time;
+	int ntouches;
+	int scroll_accel;
+	unsigned long scroll_jiffies;
+
+	struct {
+		short x;
+		short y;
+		short scroll_y;
+		u8 size;
+	} touches[16];
+	int tracking_ids[16];
+};
+
+static int magicmouse_firm_touch(struct magicmouse_sc *msc)
+{
+	int touch = -1;
+	int ii;
+
+	/* If there is only one "firm" touch, set touch to its
+	 * tracking ID.
+	 */
+	for (ii = 0; ii < msc->ntouches; ii++) {
+		int idx = msc->tracking_ids[ii];
+		if (msc->touches[idx].size < 8) {
+			/* Ignore this touch. */
+		} else if (touch >= 0) {
+			touch = -1;
+			break;
+		} else {
+			touch = idx;
+		}
+	}
+
+	return touch;
+}
+
+static void magicmouse_emit_buttons(struct magicmouse_sc *msc, int state)
+{
+        int last_state = test_bit(BTN_LEFT, msc->input->key) << 0 |
+                test_bit(BTN_RIGHT, msc->input->key) << 1 |
+                test_bit(BTN_MIDDLE, msc->input->key) << 2;
+
+	if (emulate_3button) {
+		int id;
+
+		/* If some button was pressed before, keep it held
+		 * down.  Otherwise, if there's exactly one firm
+		 * touch, use that to override the mouse's guess.
+		 */
+		if (state == 0) {
+			/* The button was released. */
+		} else if (last_state != 0) {
+			state = last_state;
+		} else if ((id = magicmouse_firm_touch(msc)) >= 0) {
+			int x = msc->touches[id].x;
+			if (x < middle_button_start)
+				state = 1;
+			else if (x > middle_button_stop)
+				state = 2;
+			else
+				state = 4;
+		} /* else: we keep the mouse's guess */
+
+		input_report_key(msc->input, BTN_MIDDLE, state & 4);
+	}
+
+	input_report_key(msc->input, BTN_LEFT, state & 1);
+	input_report_key(msc->input, BTN_RIGHT, state & 2);
+
+	if (state != last_state)
+		msc->scroll_accel = 0;
+}
+
+static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tdata)
+{
+	struct input_dev *input = msc->input;
+	__s32 x_y = tdata[0] << 8 | tdata[1] << 16 | tdata[2] << 24;
+	int misc = tdata[5] | tdata[6] << 8;
+	int id = (misc >> 6) & 15;
+	int x = x_y << 12 >> 20;
+	int y = -(x_y >> 20);
+
+	/* Store tracking ID and other fields. */
+	msc->tracking_ids[raw_id] = id;
+	msc->touches[id].x = x;
+	msc->touches[id].y = y;
+	msc->touches[id].size = misc & 63;
+
+	/* If requested, emulate a scroll wheel by detecting small
+	 * vertical touch motions along the middle of the mouse.
+	 */
+	if (emulate_scroll_wheel &&
+	    middle_button_start < x && x < middle_button_stop) {
+		static const int accel_profile[] = {
+			256, 228, 192, 160, 128, 96, 64, 32,
+		};
+		unsigned long now = jiffies;
+		int step = msc->touches[id].scroll_y - y;
+
+		/* Reset acceleration after half a second. */
+		if (time_after(now, msc->scroll_jiffies + HZ / 2))
+			msc->scroll_accel = 0;
+
+		/* Calculate and apply the scroll motion. */
+		switch (tdata[7] & TOUCH_STATE_MASK) {
+		case TOUCH_STATE_START:
+			msc->touches[id].scroll_y = y;
+                        msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
+						ARRAY_SIZE(accel_profile) - 1);
+			break;
+		case TOUCH_STATE_DRAG:
+			step = step / accel_profile[msc->scroll_accel];
+			if (step != 0) {
+				msc->touches[id].scroll_y = y;
+				msc->scroll_jiffies = now;
+				input_report_rel(input, REL_WHEEL, step);
+			}
+			break;
+		}
+	}
+
+	/* Generate the input events for this touch. */
+	if (report_touches) {
+                int orientation = (misc >> 10) - 32;
+
+		input_report_abs(input, ABS_MT_TRACKING_ID, id);
+		input_report_abs(input, ABS_MT_TOUCH_MAJOR, tdata[3]);
+		input_report_abs(input, ABS_MT_TOUCH_MINOR, tdata[4]);
+		input_report_abs(input, ABS_MT_ORIENTATION, orientation);
+		input_report_abs(input, ABS_MT_POSITION_X, x);
+		input_report_abs(input, ABS_MT_POSITION_Y, y);
+
+		if (report_undeciphered) {
+			input_event(input, EV_MSC, MSC_RAW, tdata[7]);
+		}
+
+		input_mt_sync(input);
+	}
+}
+
+static int magicmouse_raw_event(struct hid_device *hdev,
+		struct hid_report *report, u8 *data, int size)
+{
+	struct magicmouse_sc *msc = hid_get_drvdata(hdev);
+	struct input_dev *input = msc->input;
+	int x, y, ts, ii, clicks;
+
+	switch (data[0]) {
+	case 0x10:
+		if (size != 6)
+			return 0;
+		x = (__s16)(data[2] | data[3] << 8);
+		y = (__s16)(data[4] | data[5] << 8);
+		clicks = data[1];
+		break;
+	case TOUCH_REPORT_ID:
+		/* Expect six bytes of prefix, and N*8 bytes of touch data. */
+		if (size < 6 || ((size - 6) % 8) != 0)
+			return 0;
+		ts = data[3] >> 6 | data[4] << 2 | data[5] << 10;
+		msc->delta_time = (ts - msc->last_timestamp) & 0x3ffff;
+		msc->last_timestamp = ts;
+		msc->ntouches = (size - 6) / 8;
+		for (ii = 0; ii < msc->ntouches; ii++)
+			magicmouse_emit_touch(msc, ii, data + ii * 8 + 6);
+		/* When emulating three-button mode, it is important
+		 * to have the current touch information before
+		 * generating a click event.
+		 */
+		x = (signed char)data[1];
+		y = (signed char)data[2];
+		clicks = data[3];
+		break;
+	case 0x20: /* Theoretically battery status (0-100), but I have
+		    * never seen it -- maybe it is only upon request.
+		    */
+	case 0x60: /* Unknown, maybe laser on/off. */
+	case 0x61: /* Laser reflection status change.
+		    * data[1]: 0 = spotted, 1 = lost
+		    */
+	default:
+		return 0;
+	}
+
+	magicmouse_emit_buttons(msc, clicks & 3);
+	input_report_rel(input, REL_X, x);
+	input_report_rel(input, REL_Y, y);
+	input_sync(input);
+	return 1;
+}
+
+static int magicmouse_input_open(struct input_dev *dev)
+{
+	struct hid_device *hid = input_get_drvdata(dev);
+
+	return hid->ll_driver->open(hid);
+}
+
+static void magicmouse_input_close(struct input_dev *dev)
+{
+	struct hid_device *hid = input_get_drvdata(dev);
+
+	hid->ll_driver->close(hid);
+}
+
+static void magicmouse_setup_input(struct input_dev *input, struct hid_device *hdev)
+{
+	input_set_drvdata(input, hdev);
+	input->event = hdev->ll_driver->hidinput_input_event;
+	input->open = magicmouse_input_open;
+	input->close = magicmouse_input_close;
+
+	input->name = hdev->name;
+	input->phys = hdev->phys;
+	input->uniq = hdev->uniq;
+	input->id.bustype = hdev->bus;
+	input->id.vendor = hdev->vendor;
+	input->id.product = hdev->product;
+	input->id.version = hdev->version;
+	input->dev.parent = hdev->dev.parent;
+
+	set_bit(EV_KEY, input->evbit);
+	set_bit(BTN_LEFT, input->keybit);
+	set_bit(BTN_RIGHT, input->keybit);
+	if (emulate_3button)
+		set_bit(BTN_MIDDLE, input->keybit);
+	set_bit(BTN_TOOL_FINGER, input->keybit);
+
+	set_bit(EV_REL, input->evbit);
+	set_bit(REL_X, input->relbit);
+	set_bit(REL_Y, input->relbit);
+	if (emulate_scroll_wheel)
+		set_bit(REL_WHEEL, input->relbit);
+
+	if (report_touches) {
+		set_bit(EV_ABS, input->evbit);
+
+		set_bit(ABS_MT_TRACKING_ID, input->absbit);
+		input->absmin[ABS_MT_TRACKING_ID] = 0;
+		input->absmax[ABS_MT_TRACKING_ID] = 15;
+		input->absfuzz[ABS_MT_TRACKING_ID] = 0;
+
+		set_bit(ABS_MT_TOUCH_MAJOR, input->absbit);
+		input->absmin[ABS_MT_TOUCH_MAJOR] = 0;
+		input->absmax[ABS_MT_TOUCH_MAJOR] = 255;
+		input->absfuzz[ABS_MT_TOUCH_MAJOR] = 4;
+
+		set_bit(ABS_MT_TOUCH_MINOR, input->absbit);
+		input->absmin[ABS_MT_TOUCH_MINOR] = 0;
+		input->absmax[ABS_MT_TOUCH_MINOR] = 255;
+		input->absfuzz[ABS_MT_TOUCH_MINOR] = 4;
+
+		set_bit(ABS_MT_ORIENTATION, input->absbit);
+		input->absmin[ABS_MT_ORIENTATION] = -32;
+		input->absmax[ABS_MT_ORIENTATION] = 31;
+		input->absfuzz[ABS_MT_ORIENTATION] = 1;
+
+		set_bit(ABS_MT_POSITION_X, input->absbit);
+		input->absmin[ABS_MT_POSITION_X] = -1100;
+		input->absmax[ABS_MT_POSITION_X] = 1358;
+		input->absfuzz[ABS_MT_POSITION_X] = 4;
+
+		/* Note: Touch Y position from the device is inverted relative
+		 * to how pointer motion is reported (and relative to how USB
+		 * HID recommends the coordinates work).  This driver keeps
+		 * the origin at the same position, and just uses the additive
+		 * inverse of the reported Y.
+		 */
+		set_bit(ABS_MT_POSITION_Y, input->absbit);
+		input->absmin[ABS_MT_POSITION_Y] = -1589;
+		input->absmax[ABS_MT_POSITION_Y] = 2047;
+		input->absfuzz[ABS_MT_POSITION_Y] = 4;
+	}
+
+	if (report_undeciphered) {
+		set_bit(EV_MSC, input->evbit);
+		set_bit(MSC_RAW, input->mscbit);
+	}
+}
+
+static int magicmouse_probe(struct hid_device *hdev,
+	const struct hid_device_id *id)
+{
+	__u8 feature_1[] = { 0xd7, 0x01 };
+	__u8 feature_2[] = { 0xf8, 0x01, 0x32 };
+	struct input_dev *input;
+	struct magicmouse_sc *msc;
+	struct hid_report *report;
+	int ret;
+
+	msc = kzalloc(sizeof(*msc), GFP_KERNEL);
+	if (msc == NULL) {
+		dev_err(&hdev->dev, "can't alloc magicmouse descriptor\n");
+		return -ENOMEM;
+	}
+
+	msc->quirks = id->driver_data;
+	hid_set_drvdata(hdev, msc);
+
+	ret = hid_parse(hdev);
+	if (ret) {
+		dev_err(&hdev->dev, "magicmouse hid parse failed\n");
+		goto err_free;
+	}
+
+	ret = hid_hw_start(hdev, HID_CONNECT_DEFAULT);
+	if (ret) {
+		dev_err(&hdev->dev, "magicmouse hw start failed\n");
+		goto err_free;
+	}
+
+	report = hid_register_report(hdev, HID_INPUT_REPORT, TOUCH_REPORT_ID);
+	if (!report) {
+		dev_err(&hdev->dev, "unable to register touch report\n");
+		ret = -ENOMEM;
+		goto err_free;
+	}
+	report->size = 6;
+
+	ret = hdev->hid_output_raw_report(hdev, feature_1, sizeof(feature_1),
+			HID_FEATURE_REPORT);
+	if (ret != sizeof(feature_1)) {
+		dev_err(&hdev->dev, "unable to request touch data (1:%d)\n",
+				ret);
+		goto err_free;
+	}
+	ret = hdev->hid_output_raw_report(hdev, feature_2,
+			sizeof(feature_2), HID_FEATURE_REPORT);
+	if (ret != sizeof(feature_2)) {
+		dev_err(&hdev->dev, "unable to request touch data (2:%d)\n",
+				ret);
+		goto err_free;
+	}
+
+	input = input_allocate_device();
+	if (!input) {
+		dev_err(&hdev->dev, "can't alloc input device\n");
+		ret = -ENOMEM;
+		goto err_free;
+	}
+	magicmouse_setup_input(input, hdev);
+
+	ret = input_register_device(input);
+	if (ret) {
+		dev_err(&hdev->dev, "input device registration failed\n");
+		goto err_both;
+	}
+	msc->input = input;
+
+	return 0;
+ err_both:
+	input_free_device(input);
+ err_free:
+	kfree(msc);
+	return ret;
+}
+
+static void magicmouse_remove(struct hid_device *hdev)
+{
+	hid_hw_stop(hdev);
+	kfree(hid_get_drvdata(hdev));
+}
+
+static const struct hid_device_id magic_mice[] = {
+	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_APPLE, USB_DEVICE_ID_APPLE_MAGICMOUSE),
+		.driver_data = 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(hid, magic_mice);
+
+static struct hid_driver magicmouse_driver = {
+	.name = "magicmouse",
+	.id_table = magic_mice,
+	.probe = magicmouse_probe,
+	.remove = magicmouse_remove,
+	.raw_event = magicmouse_raw_event,
+};
+
+static int __init magicmouse_init(void)
+{
+	int ret;
+
+	ret = hid_register_driver(&magicmouse_driver);
+	if (ret)
+		printk(KERN_ERR "can't register magicmouse driver\n");
+
+	return ret;
+}
+
+static void __exit magicmouse_exit(void)
+{
+	hid_unregister_driver(&magicmouse_driver);
+}
+
+module_init(magicmouse_init);
+module_exit(magicmouse_exit);
+MODULE_LICENSE("GPL");
#44 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse.

Quoted 8 lines · Ed Tomlinson

Hi,

When I apply this to head, I get:

drivers/hid/hid-magicmouse.c: In function 'magicmouse_probe':
drivers/hid/hid-magicmouse.c:393: error: too many arguments to function 'hdev->hid_output_raw_report' drivers/hid/hid-magicmouse.c:400: error: too many arguments to function 'hdev->hid_output_raw_report'

Quoted 6 lines

I only need the second patch - the first seems to be in tree.

Any ideas as to what I am missing?

TIA
Ed

You're missing the patch that makes hid_output_raw_report() more generic. It is in the HID tree[1] as commit d4bfa033.

[1]- http://git.kernel.org/?p=linux/kernel/git/jikos/hid.git

Michael Poole

#45 lore raw

Re: [PATCH 2/2] Add a device driver for the Apple Magic Mouse (2.6.32.8)

Quoted 14 lines · On Wednesday 10 February 2010 22:20:13 Michael Poole wrote:
Quoted 12 lines · Ed Tomlinson

Hi,

When I apply this to head, I get:

drivers/hid/hid-magicmouse.c: In function 'magicmouse_probe':
drivers/hid/hid-magicmouse.c:393: error: too many arguments to function 'hdev->hid_output_raw_report' drivers/hid/hid-magicmouse.c:400: error: too many arguments to function 'hdev->hid_output_raw_report'

I only need the second patch - the first seems to be in tree.

Any ideas as to what I am missing?

Quoted 7 lines
Quoted 2 lines

TIA
Ed

You're missing the patch that makes hid_output_raw_report() more generic. It is in the HID tree[1] as commit d4bfa033.

[1]- http://git.kernel.org/?p=linux/kernel/git/jikos/hid.git

Ok. With this I have a patchset that applies to 2.6.32.8. If anyone wants it just email.

Of course, Michael's cleanups came just after I booted with the new kernel. I'll apply it and update & post my changes that allow both x and y scrolling and almost work for a few multitouch gestures.

Ed

#46 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse

In reply to Michael Poole
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 11 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

First, is I made a mistake with the procedure(early post) i.g. sudo /usr/bin/test-device trusted D4:9A:20:88:C7:48 forgot to add yes at the end.
(so rebooting works and so forth).

As to this module I pulled the latest mainstream
added:
http://lkml.org/lkml/2010/1/29/146
plus the two other patches you had submitted with this thread and everything compiled good.

Glad to see the mouse is working with these patches(thank you vary much gentlemen).

I can scroll up and down(need to get used to this). copy/past works
right/left click works.
(need to figure out forward/reverse with firefox probably).

suspend works. I'm getting a resumbit urb, but that's been there
prior to this module. Although there was a moment after wakeup
I turned off the mouse, then reconnecting didnt, but then on another go(reboot)it worked(maybe hit some odd thing).

Other than not seeing any menu entry under hid with make menuconfig, I would have to say vary nice job with this, all functions I need to function are there(scroll,right/left click, copy/past).

nice job, and thanks for this..
(without this I would of had a non
functional magic mouse).

Justin P. Mattock

#47 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse

In reply to Michael Poole
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#48 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

[ 5955.908380] input: Apple Wireless Mouse as /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7
[ 5955.921165] magicmouse 0005:05AC:030D.0004: input,hidraw3: BLUETOOTH HID v0.84 Mouse [Apple Wireless Mouse] on 00:0A:3A:55:07:5A
[ 5955.934120] input: Apple Wireless Mouse as /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8
[ 6180.899332] input: Apple Wireless Mouse as /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input9
[ 6180.911914] magicmouse 0005:05AC:030D.0005: input,hidraw3: BLUETOOTH HID v0.84 Mouse [Apple Wireless Mouse] on 00:0A:3A:55:07:5A
[ 6180.923988] input: Apple Wireless Mouse as /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input10
[ 6391.991295] BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
[ 6391.995818] IP: [<ffffffffa05a604f>] magicmouse_input_open+0x1f/0x30 [hid_magicmouse]
[ 6392.009801] PGD 16c3b6067 PUD 16b139067 PMD 0 
[ 6392.009801] Oops: 0000 [#1] PREEMPT SMP 
[ 6392.009801] last sysfs file: /sys/devices/pci0000:00/0000:00:18.3/temp1_input
[ 6392.009801] CPU 1 
[ 6392.009801] Pid: 2763, comm: gpm Not tainted 2.6.33-rc8-crc #106 M3A78-T/System Product Name
[ 6392.064520] RIP: 0010:[<ffffffffa05a604f>]  [<ffffffffa05a604f>] magicmouse_input_open+0x1f/0x30 [hid_magicmouse]
[ 6392.064520] RSP: 0018:ffff88016c2ddba8  EFLAGS: 00010282
[ 6392.064520] RAX: ffff88016dd90000 RBX: ffff88016bc9b000 RCX: 0000000000000001
[ 6392.107009] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88016dd90000
[ 6392.107009] RBP: ffff88016c2ddba8 R08: 2222222222222222 R09: 2222222222222222
[ 6392.107009] R10: 0000000000000000 R11: 0000000000000001 R12: ffff880133b5a810
[ 6392.107009] R13: ffff88016bc9b820 R14: ffff88016dd60150 R15: ffff88016dd600a8
[ 6392.107009] FS:  00007f5ee305b700(0000) GS:ffff880028280000(0000) knlGS:0000000000000000
[ 6392.107009] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[ 6392.107009] CR2: 0000000000000010 CR3: 000000016c3b7000 CR4: 00000000000006e0
[ 6392.107009] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 6392.107009] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 6392.107009] Process gpm (pid: 2763, threadinfo ffff88016c2dc000, task ffff88016dde0000)
[ 6392.107009] Stack:
[ 6392.107009]  ffff88016c2ddbe8 ffffffff814010e9 ffff88016dd60150 ffff88016dd600a8
[ 6392.107009] <0> ffff88016c2ddbe8 ffff880133b5a800 ffff880133b5a8d8 0000000000000000
[ 6392.107009] <0> ffff88016c2ddc18 ffffffff81404236 ffff88016c2ddc18 ffff880133b5a800
[ 6392.107009] Call Trace:
[ 6392.107009]  [<ffffffff814010e9>] input_open_device+0x89/0xb0
[ 6392.107009]  [<ffffffff81404236>] mousedev_open_device+0x76/0x100
[ 6392.107009]  [<ffffffff814042a9>] mousedev_open_device+0xe9/0x100
[ 6392.107009]  [<ffffffff8140523f>] mousedev_open+0x19f/0x260
[ 6392.107009]  [<ffffffff814bc243>] ? _lock_kernel+0x143/0x1e0
[ 6392.107009]  [<ffffffff81402ba7>] input_open_file+0x227/0x410
[ 6392.107009]  [<ffffffff814bf981>] ? sub_preempt_count+0x51/0x60
[ 6392.107009]  [<ffffffff8112384d>] chrdev_open+0x17d/0x320
[ 6392.107009]  [<ffffffff814bbb7c>] ? _raw_spin_unlock+0x5c/0x70
[ 6392.107009]  [<ffffffff8111d358>] __dentry_open+0x1a8/0x400
[ 6392.107009]  [<ffffffff811236d0>] ? chrdev_open+0x0/0x320
[ 6392.107009]  [<ffffffff8111d6b4>] nameidata_to_filp+0x54/0x70
[ 6392.107009]  [<ffffffff8112e7f1>] do_filp_open+0x841/0xc00
[ 6392.107009]  [<ffffffff814bf981>] ? sub_preempt_count+0x51/0x60
[ 6392.107009]  [<ffffffff814bbb7c>] ? _raw_spin_unlock+0x5c/0x70
[ 6392.107009]  [<ffffffff81002b0c>] ? sysret_check+0x27/0x62
[ 6392.107009]  [<ffffffff8111e954>] do_sys_open+0xa4/0x180
[ 6392.107009]  [<ffffffff8111ea70>] sys_open+0x20/0x30
[ 6392.107009]  [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b
[ 6392.107009] Code: 66 66 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 0f 1f 44 00 00 48 81 c7 a0 08 00 00 e8 3b f0 de e0 48 8b 90 38 1b 00 00 48 89 c7 <ff> 52 10 c9 c3 66 66 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 
[ 6392.107009] RIP  [<ffffffffa05a604f>] magicmouse_input_open+0x1f/0x30 [hid_magicmouse]
[ 6392.107009]  RSP <ffff88016c2ddba8>
[ 6392.107009] CR2: 0000000000000010
[ 6392.107321] ---[ end trace c83e80c68826df09 ]---

grover ~ # udevadm monitor
monitor will print the received events for:
UDEV - the event which udev sends out after rule processing KERNEL - the kernel uevent

KERNEL[1266085238.085802] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41 (bluetooth)
UDEV  [1266085238.113475] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41 (bluetooth)
KERNEL[1266085238.412016] add      /module/hidp (module)
UDEV  [1266085238.412247] add      /module/hidp (module)
KERNEL[1266085238.418193] add      /bus/hid/drivers/generic-bluetooth (drivers)
UDEV  [1266085238.418365] add      /bus/hid/drivers/generic-bluetooth (drivers)
KERNEL[1266085238.418507] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004 (hid)
KERNEL[1266085238.460711] add      /module/hid_magicmouse (module)
UDEV  [1266085238.460906] add      /module/hid_magicmouse (module)
KERNEL[1266085238.460977] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7 (input)
UDEV  [1266085238.462772] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7 (input)
KERNEL[1266085238.473232] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/mouse1 (input)
KERNEL[1266085238.473323] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/event7 (input)
KERNEL[1266085238.473662] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004/hidraw/hidraw3 (hidraw)
UDEV  [1266085238.479691] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/event7 (input)
KERNEL[1266085238.486090] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8 (input)
UDEV  [1266085238.488291] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/mouse1 (input)
UDEV  [1266085238.493195] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8 (input)
KERNEL[1266085238.498954] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8/mouse2 (input)
KERNEL[1266085238.499066] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8/event8 (input)
KERNEL[1266085238.499154] add      /bus/hid/drivers/magicmouse (drivers)
UDEV  [1266085238.499624] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004 (hid)
UDEV  [1266085238.501108] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004/hidraw/hidraw3 (hidraw)
UDEV  [1266085238.501311] add      /bus/hid/drivers/magicmouse (drivers)
UDEV  [1266085238.509221] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8/event8 (input)
UDEV  [1266085238.511459] add      /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input8/mouse2 (input)
KERNEL[1266085311.467279] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/mouse1 (input)
UDEV  [1266085311.468189] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/mouse1 (input)
KERNEL[1266085311.471961] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/event7 (input)
UDEV  [1266085311.472632] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7/event7 (input)
KERNEL[1266085311.475921] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7 (input)
KERNEL[1266085311.475985] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004/hidraw/hidraw3 (hidraw)
KERNEL[1266085311.476153] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004 (hid)
KERNEL[1266085311.476214] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41 (bluetooth)
UDEV  [1266085311.476397] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/input7 (input)
UDEV  [1266085311.476734] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004/hidraw/hidraw3 (hidraw)
UDEV  [1266085311.477000] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41/0005:05AC:030D.0004 (hid)
UDEV  [1266085311.503056] remove   /devices/pci0000:00/0000:00:13.2/usb7/7-4/7-4.4/7-4.4:1.0/bluetooth/hci0/hci0:41 (bluetooth)

Ideas?
Ed Tomlinson

#49 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 25 lines · Ed Tomlinson
Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

Quoted 9 lines

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

Indeed, we seem to be missing call to input_unregister_device() in magicmouse_remove().

Signature · 1 line
Dmitry
#50 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 36 lines · On Sunday 14 February 2010 03:03:44 Dmitry Torokhov wrote:
Quoted 25 lines · Ed Tomlinson
Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

Quoted 9 lines

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

Quoted 2 lines

Indeed, we seem to be missing call to input_unregister_device() in magicmouse_remove().

How does this look? With this udevadm shows input8 being removed and there is no more opps.

----

drivers/hid/hid-magicmouse.c +4 -1
index f94b3e4..71a8669 100644
@@ -429,8 +429,11 @@ static int magicmouse_probe(struct hid_device *hdev,
 
 static void magicmouse_remove(struct hid_device *hdev)
 {
+	struct magicmouse_sc *msc;
+	msc = hid_get_drvdata(hdev);
+	input_unregister_device(msc->input);
 	hid_hw_stop(hdev);
-	kfree(hid_get_drvdata(hdev));
+	kfree(msc);
 }
 
 static const struct hid_device_id magic_mice[] = {

----

Thanks
Ed

#51 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 37 lines · Ed Tomlinson
Quoted 36 lines · On Sunday 14 February 2010 03:03:44 Dmitry Torokhov wrote:
Quoted 34 lines · Ed Tomlinson
Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

Quoted 5 lines
Quoted 2 lines

Indeed, we seem to be missing call to input_unregister_device() in magicmouse_remove().

How does this look? With this udevadm shows input8 being removed and there is no more opps.

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

Thanks.

Signature · 1 line
Dmitry
#52 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 47 lines · On Monday 15 February 2010 02:11:46 Dmitry Torokhov wrote:
Quoted 37 lines · Ed Tomlinson
Quoted 36 lines · On Sunday 14 February 2010 03:03:44 Dmitry Torokhov wrote:
Quoted 34 lines · Ed Tomlinson
Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

Quoted 5 lines
Quoted 2 lines

Indeed, we seem to be missing call to input_unregister_device() in magicmouse_remove().

How does this look? With this udevadm shows input8 being removed and there is no more opps.

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

#53 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

In reply to Dmitry Torokhov
Quoted 41 lines · On Monday 15 February 2010 02:11:46 Dmitry Torokhov wrote:
Quoted 40 lines · Ed Tomlinson
Quoted 38 lines · On Sunday 14 February 2010 03:03:44 Dmitry Torokhov wrote:
Quoted 25 lines · Ed Tomlinson
Quoted 14 lines · On Wednesday 10 February 2010 08:57:37 Jiri Kosina wrote:
Quoted 12 lines · Michael Poole

I think this patch is ready for real review. The Magic Mouse requires that a driver send an unlock Report(Feature) command, similar to the Wacom wireless tablet and Sixaxis controller quirks. This turns on an Input Report that isn't published in the input Report descriptor that contains touch data (and usually overrides the normal motion and click Report).

Because the mouse has only one switch and no scroll wheel, the driver (under control of parameters) emulates a middle button and scroll wheel. User space could also ignore and/or re-synthesize those events based on the reported events.

Quoted 8 lines
Quoted 5 lines

The first patch exports hid_register_report() so the driver can turn on the multitouch report. The second patch adds the device ID and the driver. Some user-space tools to talk to the mouse directly (that is, when it is not associated with the host's HIDP stack) are at
http://github.com/entrope/linux-magicmouse .

I have applied the driver into apple_magic_mouse branch and merged this branch into for-next, so it should appear in the upcoming linux-next.

Quoted 9 lines

This driver (or the hid changes) can triggers an opps. What I did was start X. Turn on the magic mouse. It connected on input7&8. Then I powered it off and on. This time it conneced on input9&10. Then I exited X and got the opps. Note my X does not hotplug the magic mouse. I've also included a trace of the udev events that generated the log below (if there was a remove after X stopped it would not be included). To my eyes it looks like we leak an input device (there is not a remove event for input8).

Indeed, we seem to be missing call to input_unregister_device() in magicmouse_remove().

Quoted 6 lines
Quoted 2 lines

How does this look? With this udevadm shows input8 being removed and there is no more opps.

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

How about this? It applies on top of yesterdays patch.

Ed

---

drivers/hid/hid-magicmouse.c +1 -1
index 7d252d2..46fdeee 100644
@@ -430,8 +430,8 @@ static void magicmouse_remove(struct hid_device *hdev)
 {
 	struct magicmouse_sc *msc;
 	msc = hid_get_drvdata(hdev);
-	input_unregister_device(msc->input);
 	hid_hw_stop(hdev);
+	input_unregister_device(msc->input);
 	kfree(msc);
 }
 
#54 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 14 lines · Ed Tomlinson
Quoted 3 lines

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

How about this? It applies on top of yesterdays patch.

Ed

---

drivers/hid/hid-magicmouse.c +0 -0
index 7d252d2..46fdeee 100644

drivers/hid/hid-magicmouse.c static void magicmouse_remove(struct hid_device *hdev) · line 436

Quoted 8 lines
drivers/hid/hid-magicmouse.c +1 -1
@@ -430,8 +430,8 @@ static void magicmouse_remove(struct hid_device *hdev)
 {
 	struct magicmouse_sc *msc;
 	msc = hid_get_drvdata(hdev);
-	input_unregister_device(msc->input);
 	hid_hw_stop(hdev);
+	input_unregister_device(msc->input);
 	kfree(msc);
 }

This looks OK.

Could you please send it altogether with short changelog entry and Signed-off-by line so that I could queue that up?

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#55 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

In reply to Dmitry Torokhov
Quoted 3 lines · On Monday 15 February 2010 02:11:46 Dmitry Torokhov wrote:

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

The exit routing for the module also has a hid_hw_stop. Is it going to cause problems when it gets called twice?

TIA
Ed

#56 lore raw

Re: [PATCH 0/2] Provide a driver for the Apple Magic Mouse - opps

Quoted 6 lines · Ed Tomlinson
Quoted 3 lines

Almost... you need to do hid_hw_stop() first and only then unregister input device, Otherwise if you unload the module while moving the mouse it is likely to still oops.

The exit routing for the module also has a hid_hw_stop. Is it going to cause problems when it gets called twice?

The routine wasn't probably meant/designed with re-entrancy in mind, but looking quickly at all the subsequent callpaths, I don't see why it should cause any problem, as skb_queue_purge(), usb_kill_urb() and usb_free_urb() should be safe.

Maybe we should even add test_bit() for HID_STARTED and HID_DISCONNECTED at the beginning of the low-level drivers' ->stop callbacks, so that we don't do all the magic when all the work has been already done.

Signature · 2 lines
Jiri Kosina
SUSE Labs, Novell Inc.
#57 lore raw

PATCH 1/1 Enable xy scrolling for Apple Magic Mouse

In reply to Jiri Kosina

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Comments
Ed Tomlinson

---

drivers/hid/hid-magicmouse.c +36 -18
index b20484a..3075d78 100644
@@ -59,7 +59,7 @@ MODULE_PARM_DESC(report_undeciphered, "Report undeciphered multi-touch state fie
  * @delta_time: 18-bit difference between the two most recent touch
  *     reports from the mouse.
  * @ntouches: Number of touches in most recent touch report.
- * @scroll_accel: Number of consecutive scroll motions.
+ * @scroll_accely: Number of consecutive scroll motions.
  * @scroll_jiffies: Time of last scroll motion.
  * @touches: Most recent data for a touch, indexed by tracking ID.
  * @tracking_ids: Mapping of current touch input data to @touches.
@@ -71,11 +71,13 @@ struct magicmouse_sc {
 	int last_timestamp;
 	int delta_time;
 	int ntouches;
-	int scroll_accel;
+	int scroll_accely;
+	int scroll_accelx;
 	unsigned long scroll_jiffies;
 
 	struct {
 		short x;
+		short scroll_x;
 		short y;
 		short scroll_y;
 		u8 size;
@@ -139,8 +141,10 @@ static void magicmouse_emit_buttons(struct magicmouse_sc *msc, int state)
 	input_report_key(msc->input, BTN_LEFT, state & 1);
 	input_report_key(msc->input, BTN_RIGHT, state & 2);
 
-	if (state != last_state)
-		msc->scroll_accel = 0;
+	if (state != last_state) {
+		msc->scroll_accely = 0;
+		msc->scroll_accelx = 0;
+	}
 }
 
 static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tdata)
@@ -159,34 +163,46 @@ static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tda
 	msc->touches[id].size = misc & 63;
 
 	/* If requested, emulate a scroll wheel by detecting small
-	 * vertical touch motions along the middle of the mouse.
+	 * touch motions on the mouse.
 	 */
 	if (emulate_scroll_wheel &&
-	    middle_button_start < x && x < middle_button_stop) {
+	    msc->ntouches == 1) {
 		static const int accel_profile[] = {
-			256, 228, 192, 160, 128, 96, 64, 32,
+			192, 160, 128, 96, 64, 48, 32, 24,
 		};
 		unsigned long now = jiffies;
-		int step = msc->touches[id].scroll_y - y;
+		int stepx, stepy;
 
 		/* Reset acceleration after half a second. */
-		if (time_after(now, msc->scroll_jiffies + HZ / 2))
-			msc->scroll_accel = 0;
+		if (time_after(now, msc->scroll_jiffies + HZ / 2)) {
+			msc->scroll_accely = 0;
+			msc->scroll_accelx = 0;
+		}
 
-		/* Calculate and apply the scroll motion. */
 		switch (tdata[7] & TOUCH_STATE_MASK) {
 		case TOUCH_STATE_START:
 			msc->touches[id].scroll_y = y;
-			msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
-						ARRAY_SIZE(accel_profile) - 1);
+			msc->touches[id].scroll_x = x;
 			break;
 		case TOUCH_STATE_DRAG:
-			step = step / accel_profile[msc->scroll_accel];
-			if (step != 0) {
+			/* Calculate and apply the scroll motion. */
+			stepy = (msc->touches[id].scroll_y - y)/accel_profile[msc->scroll_accely];
+			stepx = (msc->touches[id].scroll_x - x)/accel_profile[msc->scroll_accelx];
+
+			/* tell input about any motions */
+			if (stepy != 0) {
 				msc->touches[id].scroll_y = y;
-				msc->scroll_jiffies = now;
-				input_report_rel(input, REL_WHEEL, step);
+				input_report_rel(input, REL_WHEEL, stepy);
+				msc->scroll_accely = min_t(int, msc->scroll_accely + 1,
+							        ARRAY_SIZE(accel_profile) - 1);
 			}
+			if (stepx != 0) {
+				msc->touches[id].scroll_x = x;
+				input_report_rel(input, REL_HWHEEL, stepx);
+				msc->scroll_accelx = min_t(int, msc->scroll_accelx + 1,
+							        ARRAY_SIZE(accel_profile) - 1);
+			}
+			msc->scroll_jiffies = now;
 			break;
 		}
 	}
@@ -300,8 +316,10 @@ static void magicmouse_setup_input(struct input_dev *input, struct hid_device *h
 	__set_bit(EV_REL, input->evbit);
 	__set_bit(REL_X, input->relbit);
 	__set_bit(REL_Y, input->relbit);
-	if (emulate_scroll_wheel)
+	if (emulate_scroll_wheel) {
 		__set_bit(REL_WHEEL, input->relbit);
+		__set_bit(REL_HWHEEL, input->relbit);
+	}
 
 	if (report_touches) {
 		__set_bit(EV_ABS, input->evbit);
#58 lore raw

Re: [PATCH 1/1] Enable xy scrolling for Apple Magic Mouse

Quoted 4 lines · Ed Tomlinson

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Hi Ed,

Your other patch to call input_unregister_device() looks good -- thanks!

I've never used a horizontal scroll wheel -- what are the common uses for it? Why should the acceleration be separate for the two directions rather than using the same factor? Why does the kernel need to emulate this rather than having user-space implement the emulation?

(If the answers to the last two questions justify putting this feature in the kernel, my only suggestions on the patch are style-related:
scroll_accel{x,y} should probably be spelled scroll_accel_{x,y}, and there are missing spaces around the / operator when calculating stepx and stepy.)

Michael Poole

#59 lore raw

Re: [PATCH 1/1] Enable xy scrolling for Apple Magic Mouse

Quoted 10 lines · On Sunday 14 February 2010 17:51:15 Michael Poole wrote:
Quoted 4 lines · Ed Tomlinson

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Hi Ed,

Your other patch to call input_unregister_device() looks good -- thanks!

Thanks. One question about it though. Do we have to check if msc is null before the unregister?

Quoted 4 lines

I've never used a horizontal scroll wheel -- what are the common uses for it? Why should the acceleration be separate for the two directions rather than using the same factor? Why does the kernel need to emulate this rather than having user-space implement the emulation?

Its usefull for scrolling left and right while browsing. If you use kde it can
be used to scroll between applications on the taskbar. Here most applications
with a horizontial scrollbar work as expected.

I first tried with a single acceleration value for both axies. It leads to confusing
things happening. For example. I quickly scroll down, then nudge the
window to the left. This works as expect with two values. With one the nudge is accelerated and moves too far.

My personal goal is to have the basic, apple defined, gestures working from kernel space. This way the device works as expected without needing to fiddle with X or other managers (think wayland and/or chromeOS).

Quoted 5 lines

(If the answers to the last two questions justify putting this feature in the kernel, my only suggestions on the patch are style-related:
scroll_accel{x,y} should probably be spelled scroll_accel_{x,y}, and there are missing spaces around the / operator when calculating stepx and stepy.)

I'll fix up the style and resubmit.

Thanks
Ed

#60 lore raw

Re: [PATCH 1/1] Enable xy scrolling for Apple Magic Mouse

Quoted 8 lines · Ed Tomlinson
Quoted 7 lines · On Sunday 14 February 2010 17:51:15 Michael Poole wrote:
Quoted 4 lines · Ed Tomlinson

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Quoted 6 lines
Quoted 3 lines

Hi Ed,

Your other patch to call input_unregister_device() looks good -- thanks!

Thanks. One question about it though. Do we have to check if msc is null before the unregister?

If it is NULL whan remove() is running there are much bigger problems with the driver/HID subsystem.

Quoted 10 lines
Quoted 5 lines

I've never used a horizontal scroll wheel -- what are the common uses for it? Why should the acceleration be separate for the two directions rather than using the same factor? Why does the kernel need to emulate this rather than having user-space implement the emulation?

Its usefull for scrolling left and right while browsing. If you use kde it can
be used to scroll between applications on the taskbar. Here most applications
with a horizontial scrollbar work as expected.

Quoted 8 lines

I first tried with a single acceleration value for both axies. It leads to confusing
things happening. For example. I quickly scroll down, then nudge the
window to the left. This works as expect with two values. With one the nudge is accelerated and moves too far.

My personal goal is to have the basic, apple defined, gestures working from kernel space. This way the device works as expected without needing to fiddle with X or other managers (think wayland and/or chromeOS).

I am not sure if this is the desired approach. The current idea is to export useable but minimally processed events to userspace and let them be turned into gestures there (by evdev, synaptics driver and so forth).

Thanks.

Signature · 1 line
Dmitry
#61 lore raw

Re: [PATCH 1/1] Enable xy scrolling for Apple Magic Mouse

Quoted 34 lines · On Monday 15 February 2010 02:18:50 Dmitry Torokhov wrote:
Quoted 8 lines · Ed Tomlinson
Quoted 7 lines · On Sunday 14 February 2010 17:51:15 Michael Poole wrote:
Quoted 4 lines · Ed Tomlinson

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Quoted 6 lines
Quoted 3 lines

Hi Ed,

Your other patch to call input_unregister_device() looks good -- thanks!

Thanks. One question about it though. Do we have to check if msc is null before the unregister?

If it is NULL whan remove() is running there are much bigger problems with the driver/HID subsystem.

Quoted 15 lines
Quoted 5 lines

I've never used a horizontal scroll wheel -- what are the common uses for it? Why should the acceleration be separate for the two directions rather than using the same factor? Why does the kernel need to emulate this rather than having user-space implement the emulation?

Its usefull for scrolling left and right while browsing. If you use kde it can
be used to scroll between applications on the taskbar. Here most applications
with a horizontial scrollbar work as expected.

I first tried with a single acceleration value for both axies. It leads to confusing
things happening. For example. I quickly scroll down, then nudge the
window to the left. This works as expect with two values. With one the nudge is accelerated and moves too far.

Quoted 7 lines
Quoted 3 lines

My personal goal is to have the basic, apple defined, gestures working from kernel space. This way the device works as expected without needing to fiddle with X or other managers (think wayland and/or chromeOS).

I am not sure if this is the desired approach. The current idea is to export useable but minimally processed events to userspace and let them be turned into gestures there (by evdev, synaptics driver and so forth).

I would go along with this if the code to add the emulation in kernel was large or complex. Its not. About 30 more lines should add the pgup/pgdn code.

Thanks
Ed

#62 lore raw

Re: [PATCH 1/1] Enable xy scrolling for Apple Magic Mouse

In reply to Michael Poole
Quoted 7 lines · On Sunday 14 February 2010 17:51:15 Michael Poole wrote:
Quoted 4 lines · Ed Tomlinson

Hi,

Here is a patch that enables xy scrolling with the magic mouse. I have also changed the accelleration logic to work better with xy scrolling.

Quoted 8 lines

Hi Ed,

Your other patch to call input_unregister_device() looks good -- thanks!

I've never used a horizontal scroll wheel -- what are the common uses for it? Why should the acceleration be separate for the two directions rather than using the same factor? Why does the kernel need to emulate this rather than having user-space implement the emulation?

With style fixups.

Ed

--

drivers/hid/hid-magicmouse.c +37 -18
index b20484a..7d252d2 100644
@@ -59,7 +59,8 @@ MODULE_PARM_DESC(report_undeciphered, "Report undeciphered multi-touch state fie
  * @delta_time: 18-bit difference between the two most recent touch
  *     reports from the mouse.
  * @ntouches: Number of touches in most recent touch report.
- * @scroll_accel: Number of consecutive scroll motions.
+ * @scroll_accel_y: Number of consecutive scroll motions.
+ * @scroll_accel_x: Number of consecutive scroll motions.
  * @scroll_jiffies: Time of last scroll motion.
  * @touches: Most recent data for a touch, indexed by tracking ID.
  * @tracking_ids: Mapping of current touch input data to @touches.
@@ -71,11 +72,13 @@ struct magicmouse_sc {
 	int last_timestamp;
 	int delta_time;
 	int ntouches;
-	int scroll_accel;
+	int scroll_accel_y;
+	int scroll_accel_x;
 	unsigned long scroll_jiffies;
 
 	struct {
 		short x;
+		short scroll_x;
 		short y;
 		short scroll_y;
 		u8 size;
@@ -139,8 +142,10 @@ static void magicmouse_emit_buttons(struct magicmouse_sc *msc, int state)
 	input_report_key(msc->input, BTN_LEFT, state & 1);
 	input_report_key(msc->input, BTN_RIGHT, state & 2);
 
-	if (state != last_state)
-		msc->scroll_accel = 0;
+	if (state != last_state) {
+		msc->scroll_accel_y = 0;
+		msc->scroll_accel_x = 0;
+	}
 }
 
 static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tdata)
@@ -159,34 +164,46 @@ static void magicmouse_emit_touch(struct magicmouse_sc *msc, int raw_id, u8 *tda
 	msc->touches[id].size = misc & 63;
 
 	/* If requested, emulate a scroll wheel by detecting small
-	 * vertical touch motions along the middle of the mouse.
+	 * touch motions on the mouse.
 	 */
 	if (emulate_scroll_wheel &&
-	    middle_button_start < x && x < middle_button_stop) {
+	    msc->ntouches == 1) {
 		static const int accel_profile[] = {
-			256, 228, 192, 160, 128, 96, 64, 32,
+			192, 160, 128, 96, 64, 48, 32, 24,
 		};
 		unsigned long now = jiffies;
-		int step = msc->touches[id].scroll_y - y;
+		int stepx, stepy;
 
 		/* Reset acceleration after half a second. */
-		if (time_after(now, msc->scroll_jiffies + HZ / 2))
-			msc->scroll_accel = 0;
+		if (time_after(now, msc->scroll_jiffies + HZ / 2)) {
+			msc->scroll_accel_y = 0;
+			msc->scroll_accel_x = 0;
+		}
 
-		/* Calculate and apply the scroll motion. */
 		switch (tdata[7] & TOUCH_STATE_MASK) {
 		case TOUCH_STATE_START:
 			msc->touches[id].scroll_y = y;
-			msc->scroll_accel = min_t(int, msc->scroll_accel + 1,
-						ARRAY_SIZE(accel_profile) - 1);
+			msc->touches[id].scroll_x = x;
 			break;
 		case TOUCH_STATE_DRAG:
-			step = step / accel_profile[msc->scroll_accel];
-			if (step != 0) {
+			/* Calculate the scroll motion. */
+			stepy = (msc->touches[id].scroll_y - y) / accel_profile[msc->scroll_accel_y];
+			stepx = (msc->touches[id].scroll_x - x) / accel_profile[msc->scroll_accel_x];
+
+			/* Tell input about any motion. */
+			if (stepy != 0) {
 				msc->touches[id].scroll_y = y;
-				msc->scroll_jiffies = now;
-				input_report_rel(input, REL_WHEEL, step);
+				input_report_rel(input, REL_WHEEL, stepy);
+				msc->scroll_accel_y = min_t(int, msc->scroll_accel_y + 1,
+							        ARRAY_SIZE(accel_profile) - 1);
 			}
+			if (stepx != 0) {
+				msc->touches[id].scroll_x = x;
+				input_report_rel(input, REL_HWHEEL, stepx);
+				msc->scroll_accel_x = min_t(int, msc->scroll_accel_x + 1,
+							        ARRAY_SIZE(accel_profile) - 1);
+			}
+			msc->scroll_jiffies = now;
 			break;
 		}
 	}
@@ -300,8 +317,10 @@ static void magicmouse_setup_input(struct input_dev *input, struct hid_device *h
 	__set_bit(EV_REL, input->evbit);
 	__set_bit(REL_X, input->relbit);
 	__set_bit(REL_Y, input->relbit);
-	if (emulate_scroll_wheel)
+	if (emulate_scroll_wheel) {
 		__set_bit(REL_WHEEL, input->relbit);
+		__set_bit(REL_HWHEEL, input->relbit);
+	}
 
 	if (report_touches) {
 		__set_bit(EV_ABS, input->evbit);
#63 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

In reply to Bastien Nocera
Quoted 48 lines · Bastien Nocera
Quoted 47 lines · Justin Mattock
Quoted 45 lines

TIA
Ed Tomlinson

Quoted 24 lines · On Friday 05 February 2010 12:51:43 Marcel Holtmann wrote:

Hi Michael,

Quoted 11 lines
Quoted 2 lines

looks good to me. I just prefer that you do the allocation of the report descriptor before the HID object:

An updated patch is below. Sorry for the delay -- inclement weather here got in the way of testing this as quickly as I would have liked.

From e245ef87247f5e257db40c412af7991c9af375ab Mon Sep 17 00:00:00 2001
From: Michael Poole<[email protected]>
Date: Fri, 5 Feb 2010 12:21:38 -0500
Subject: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Quoted 6 lines

The report descriptor is read by user space (via the Service Discovery Protocol), so it is only available during the ioctl to connect. However, the probe function that needs the descriptor might not be called until a specific module is loaded. Keep a copy of the descriptor so it is available for later use.

patch has been applied. Thanks.

Regards

Quoted 8 lines

Marcel

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Hi,

I have been trying to test Michael's magic mouse drivers against 32.7 (or .8-rc). Is there a tree
based on .32 with these patches? If not would it be possible to post the patches to lkml? I do not mind sorting out a few conflicts - this thread assume one knows a little more about the process being used here than is obvious.

Quoted 6 lines
Quoted 6 lines

just got one of those magic things as well. I did apply the patchs to the current HEAD, but am getting a lifeless result i.g. here's what I did:

sudo /usr/bin/hcitool scan
sudo /usr/bin/simple-agent hci0 D4:9A:20:88:C7:48

pin was already done at this point i.g. after 0000 was able to use the mouse(no scroll function), but was able to work.

Quoted 4 lines
Quoted 4 lines

sudo /usr/bin/dbus-send --system --dest=org.bluez --print-reply /org/bluez/1897/hci0/dev_D4_9A_20_88_C7_48 org.bluez.Input.Connect sudo /usr/bin/test-device trusted D4:9A:20:88:C7:48

this is to verify that you want to trust this device. after this you should always be connected even after a reboot. in this case magic mouse was not functioning after a reboot.

Quoted 12 lines
Quoted 11 lines

the thing works after that(no scroll), fast cursor though. then reboot
once up and running nothing.
cat /var/log/daemon.log (shows this)

Feb  8 22:57:40 Linux-1 bluetoothd[1973]: link_key_request
(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)
Feb  8 22:58:21 Linux-1 bluetoothd[1973]: last message repeated 5 times
Feb  8 22:59:21 Linux-1 bluetoothd[1973]: last message repeated 3 times
Feb  8 23:03:10 Linux-1 bluetoothd[1973]: link_key_request

(sba=00:25:00:C0:6C:4F, dba=D4:9A:20:88:C7:48)

Quoted 4 lines

That means that your computer and the mouse aren't paired. The pin code for Apple mice is usually "0000" (as opposed to most mice which don't need pairing, just a poke and setting as trusted).

I'll spend some time today(need a break from looking at another problem), to see if I can get this thing functioning.

Quoted 6 lines
Quoted 2 lines

(could be missing something though(don't use gnome/kde-bluetooth)although mightymouse works).

Cheers

Justin P. Mattock

#64 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

In reply to Michael Poole
Quoted 3 lines · Michael Poole

Bastien and Marcel,

This patch works for me. Bastien, does it also fix it for you?

Patch works great. Thanks for hunting it down.

maintainer
#65 lore raw

Re: [PATCH] Bluetooth: Keep a copy of each HID device's report descriptor.

Hi Bastien,

Quoted 3 lines
Quoted 1 line

This patch works for me. Bastien, does it also fix it for you?

Patch works great. Thanks for hunting it down.

now I only need a modified version as requested and I can push that one upstream with all the other fixes from the UnPlugFest testing.

Regards

Marcel

maintainer
#66 lore raw
In reply to Jiri Kosina

Hi Jiri,

Quoted 10 lines

this is a followup to the thread

"Implement Bluetooth Wacom tablet's mode change in the kernel"

First patch alters the raw output API so that it is able to actually distinguish between feature and output reports (and modifies low-level drivers accordingly), and the two followup patches make use of this for in-kernel mode setting of Sony and Wacom bluetooth devices.

Marcel, could you please review the Bluetooth part?

I pushed the change for control vs. interrupt channel to the bluetooth-2.6 tree and asked Dave to pull it. This was clear mistake from in the first place.

When this lands, you need to redo the raw report patch.

Regards

Marcel