Message-Id: <20141117133518.ee306b3b35a0e0566a255b14@ao2.it>
Date:	Mon, 17 Nov 2014 13:35:18 +0100
From:	Antonio Ospite <ao2@....it>
To:	bri <bri@...ij.org>
Cc:	Frank Praznik <frank.praznik@...il.com>,
	Jiri Kosina <jkosina@...e.cz>,
	Henrik Rydberg <rydberg@...omail.se>,
	"open list:HID CORE LAYER" <linux-input@...r.kernel.org>,
	open list <linux-kernel@...r.kernel.org>,
	Michael Kerrisk <mtk.manpages@...il.com>
Subject: Re: [PATCH 001/001] hid-sony.c: add sysfs provisioning

On Sun, 16 Nov 2014 13:18:46 -0500
bri <bri@...ij.org> wrote:

Hi Brian,

> Add some sysfs interfaces to cover some corner use cases when provisioning:
> 
> 1) Allow read access to the IDA index through device/gamepad_number.  The
> IDA index is used to identify the device to the user as the "controller number"
> and is reflected via the initial LED states when plugged in.  This sysfs
> attribute could be used by userspace for a more seamless transition to
> userspace drivers without controller numbers changing.  Write access is not
> addressed as it raises the issue of a global IDA that works across different
> gamepad drivers.  Internally, remove the use of the term "device_id" for
> the gamepad number for code grepability purposes.
>

I am not sure about this, Frank can better comment on it.

>From how I always saw it, the IDA for sixaxis controller may be useful
in a scenario when you _only_ have Sixaxis controllers, but if you have
hooked up controllers relying on different drivers, then using the
joystick device numbers instead (i.e. the X in /dev/input/jsX) as
identifier is more reliable IMHO, so that's how the BlueZ sixaxis
plugin sets the LEDs, see:
http://git.kernel.org/cgit/bluetooth/bluez.git/tree/plugins/sixaxis.c

Imagine:
 1. plug in a sixaxis (IDA = 0, /dev/input/js0)
 2. plug in another joypad (/dev/input/js1)
 3. plug in a second sixaxis (IDA = 1, /dev/input/js2)

> 2) Allow manual provisioning of the Sixaxis/Dualshock3 USB "pairing" process
> through device/master_bdaddr.  The bluez plugin has removed the need for
> sixpair.c or other such utilities for the common use case of autoprovisioning.
> Under less usual situations, the bluez behavior may not be desired, e.g.
> if you need to pair a gamepad with a different machine, see what machine
> a pad is paired to, or restore old pairings after use.  This removes
> the need to use sixpair.c or roll-your-own libusb code for that sort
> of thing.
>

I had tried doing something similar in the past (the parsing was just a
sscanf): http://thread.gmane.org/gmane.linux.bluez.kernel/5261 but then
we deliberately decided against exposing specific sysfs interfaces for
device/master_bdaddr, you can just use generic HID feature reports from
userspace to get/set these, write a simple program reusing the code in
the BlueZ sixaxis plugin, using the ioclts
HIDIOCGFEATURE/HIDIOCSFEATURE, this way you don't depend on libusb. As
an historical note, the BlueZ sixaxis plugin was one of the first user
of these ioctls.

It's true that you still need some code, but when possible it's better
to have code for corner cases in userspace rather than in the kernel.

A more general note, when submitting kernel patches try to send one
patch per logical change, in this case I would have expected several
patches, one for the rename device_id/gamepad_number, one for the IDA
sysfs hook and one for the bdaddrs sysfs hooks. Self contained patches
are easier to review.

I'd also try to avoid custom parsing routines in kernel code as much as
possible.

That said I still don't think the changes you are proposing are strictly
necessary in the kernel driver, but let's see what the others have to
say about that.

Misc comments inlined below.

Thanks,
   Antonio

> Signed-off-by: Brian S. Julin <bri@...ij.org>
> Tested-by: Brian S. Julin <bri@...ij.org>
> 
> ---
> 
> I've tested with both SixAxis and Dualshock 3 on stock Debian kernel 3.16.5.
> I've also tested a copy of the 3.18 git version inserted into the 3.16.5
> kernel.  The patch should apply to both (with fuzz on the newer version.)
> Tests are still needed on pure 3.18.
>

You know you can build Debian packages from the official git kernel
repository, don't you?
	make LOCALVERSION=-brian INSTALL_MOD_STRIP=1 deb-pkg

When sending kernel patches try to base them on the master branch from
the linus tree:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/

> I do not have access to the non-Sony pads handled by this driver, but did
> test the with a Dualshock4 to ensure that the sysfs files are harmless
> vestigials for such devices.
> 
> It has been a long time since I popped the hood on the kernel and much
> has changed.  I'm unclear as to under what conditions hid-core relieves
> the need to lock.  Some review is definitely needed there.
> 
> General comment on the driver: when a gamepad being used over
> BT is plugged back in to USB, it continues to use BT rather than
> switching to USB.  This has both benefits and disadvantages.
> 
> On the bright side, this prevents the evdev and js devices from being
> destroyed and recreated.  Some applications, many closed-source, may
> not respond well to that even when udev rules are used to pin the
> device names.
>

Exactly.

> On the other hand this is not the behavior gamers expect and will not
> help matters if the reason for plugging the pad in was to decrease
> latency due to RF contention.

This is an interesting point, can you show numbers about these latency
problems?

> I notice that sysfs-rules.txt mentions that a device's parentage in
> the tree may change, and teaching the hid core to allow graceful
> takeover of a device by a different host/bus driver for these
> devices might be a use case for that.
> 
> --- drivers/hid/hid-sony.c.orig	2014-11-05 20:32:47.387599486 -0500
> +++ drivers/hid/hid-sony.c	2014-11-14 20:35:28.327053036 -0500
> @@ -8,6 +8,7 @@
>   *  Copyright (c) 2012 David Dillow <dave@...dillows.org>
>   *  Copyright (c) 2006-2013 Jiri Kosina
>   *  Copyright (c) 2013 Colin Leitner <colin.leitner@...il.com>
> + *  Copyright (c) 2014 Brian S. Julin <bri@...ij.org>
>   */
>  
>  /*
> @@ -35,6 +36,9 @@
>  #include <linux/list.h>
>  #include <linux/idr.h>
>  #include <linux/input/mt.h>
> +#include <linux/ctype.h>
> +#include <net/bluetooth/bluetooth.h>
> +#include <net/bluetooth/hci_core.h>

A HID driver should not need to depend on bluetooth, no other HID
driver does. Layer violations should be avoided as much as possible.

>  #include "hid-ids.h"
>  
> @@ -750,7 +754,7 @@ union sixaxis_output_report_01 {
>  
>  static spinlock_t sony_dev_list_lock;
>  static LIST_HEAD(sony_device_list);
> -static DEFINE_IDA(sony_device_id_allocator);
> +static DEFINE_IDA(sony_gamepad_number_allocator);
>  
>  struct sony_sc {
>  	spinlock_t lock;
> @@ -760,7 +764,7 @@ struct sony_sc {
>  	unsigned long quirks;
>  	struct work_struct state_worker;
>  	struct power_supply battery;
> -	int device_id;
> +	int gamepad_number;

As I tried to explain above, the value of this variable won't
corresponds to the "gamepad number" if you have a mixed driver scenario.

>  
>  #ifdef CONFIG_SONY_FF
>  	__u8 left;
> @@ -1321,7 +1325,8 @@ static int sony_leds_init(struct sony_sc
>  		if (!hid_validate_values(hdev, HID_OUTPUT_REPORT, 0, 0, 7))
>  			return -ENODEV;
>  	} else if (sc->quirks & DUALSHOCK4_CONTROLLER) {
> -		dualshock4_set_leds_from_id(sc->device_id, initial_values);
> +		dualshock4_set_leds_from_id(sc->gamepad_number,
> +					    initial_values);
>  		initial_values[3] = 1;
>  		sc->led_count = 4;
>  		memset(max_brightness, 255, 3);
> @@ -1330,7 +1335,7 @@ static int sony_leds_init(struct sony_sc
>  		name_len = 0;
>  		name_fmt = "%s:%s";
>  	} else {
> -		sixaxis_set_leds_from_id(sc->device_id, initial_values);
> +		sixaxis_set_leds_from_id(sc->gamepad_number, initial_values);
>  		sc->led_count = 4;
>  		memset(use_hw_blink, 1, 4);
>  		use_ds4_names = 0;
> @@ -1758,38 +1763,209 @@ static int sony_check_add(struct sony_sc
>  	return sony_check_add_dev_list(sc);
>  }
>  
> -static int sony_set_device_id(struct sony_sc *sc)
> +/* PS3 pads do not pair over bluetooth; Instead a USB handshake is used. */
> +static ssize_t sony_master_bdaddr_show(struct device *dev,
> +				       struct device_attribute *attr,
> +				       char *buf)
> +{
> +	struct hid_device *hdev = container_of(dev, struct hid_device, dev);
> +	struct sony_sc *sc = hid_get_drvdata(hdev);
> +	size_t count;
> +	unsigned char msg[8];
> +	int ret;
> +
> +	if (!sc) {
> +		hid_err(hdev, "No device data\n");
> +		return 0;
> +	}
> +
> +	if (sc->quirks & SIXAXIS_CONTROLLER_BT) {
> +		struct device *parent;
> +
> +		parent = dev->parent;
> +		if (!parent)
> +			return 0;
> +		parent = parent->parent;
> +		if (!parent || !parent->type
> +		    || strcmp(parent->type->name, "host"))
> +			return 0;
> +
> +		count = scnprintf(buf, PAGE_SIZE, "%pMR\n",
> +				  &(to_hci_dev(parent)->bdaddr));
> +		return count;
> +	}
> +
> +	if (!(sc->quirks & SIXAXIS_CONTROLLER_USB))
> +		return 0;
> +
> +	ret = hid_hw_raw_request(hdev, 0xf5, msg, sizeof(msg),
> +				 HID_FEATURE_REPORT, HID_REQ_GET_REPORT);
> +	if (ret < 0) {
> +		/* This matches bluez error message */
> +		hid_err(hdev, "failed to read master address\n");
> +		return 0;
> +	}
> +
> +	count = scnprintf(buf, PAGE_SIZE, "%pM\n", msg + 2);
> +	return count;
> +}
> +
> +/*
> + * This could probably be useful in lib/.
> + *
> + * Convert a string to a MAC address, accepting multiple formats.
> + *
> + * Will take most commonly cut-and-pasted formats without being too liberal.
> + * Accepts any hex with delims [-:.] between any octets, no mixing delims,
> + * no splitting octets.
> + *
> + * *buf should have at least 6 bytes of space and will not be altered on fail
> + * Will halt at null terminator or before reading more than lim chars from *c
> + * Return code is sscanf-style: number of non-null bytes read, or 0 on failure
> + */
> +static int sony_kstrntomac(const char *c, u8 *buf, size_t lim)
> +{
> +	int cnt = 0;
> +	int dcnt = 0;
> +	char delim = '\0';
> +	u8 octet[6];
> +
> +	while (cnt < 6) {
> +		if (lim < 2 || !isxdigit(*c) || !isxdigit(*(c+1)))
> +			return 0;
> +
> +		lim -= 2;
> +
> +		if (isdigit(*c))
> +			octet[cnt] = (*c - '0') << 4;
> +		else
> +			octet[cnt] = (tolower(*c) - 'a' + 10) << 4;
> +		c++;
> +
> +		if (isdigit(*c))
> +			octet[cnt] |= *c - '0';
> +		else
> +			octet[cnt] |= tolower(*c) - 'a' + 10;
> +		c++;
> +
> +		if (cnt < 5) {
> +			if (!lim)
> +				return 0;
> +			if (!delim && (*c == '-' || *c == ':' || *c == '.'))
> +				delim = *c;
> +			if (delim && *c == delim) {
> +				dcnt++;
> +				c++;
> +				lim--;
> +			}
> +		}
> +		cnt++;
> +	}
> +
> +	memcpy(buf, octet, 6);
> +
> +	return cnt * 2 + dcnt;
> +}
> +
> +static ssize_t sony_master_bdaddr_store(struct device *dev,
> +					struct device_attribute *attr,
> +					const char *buf, size_t count)
> +{
> +	struct hid_device *hdev = container_of(dev, struct hid_device, dev);
> +	struct sony_sc *sc = hid_get_drvdata(hdev);
> +	u8 msg[8];
> +	int ret;
> +
> +	if (!sc) {
> +		hid_err(hdev, "No device data\n");
> +		return -EBADFD;
> +	}
> +
> +	if (!(sc->quirks & SIXAXIS_CONTROLLER_USB)) {
> +		if (sc->quirks & SIXAXIS_CONTROLLER_BT) {
> +			hid_err(hdev,
> +				"To set master disconnect BT then use USB\n");
> +			return -EOPNOTSUPP;
> +		}
> +		return -ENOTSUPP;
> +	}
> +
> +	msg[0] = 0x01;
> +	msg[1] = 0x00;
> +
> +	if (!sony_kstrntomac(buf, msg + 2, count))
> +		return -EINVAL;
> +
> +	ret = hid_hw_raw_request(hdev, 0xf5, msg, sizeof(msg),
> +				 HID_FEATURE_REPORT, HID_REQ_SET_REPORT);
> +	if (ret < 0) {
> +		/* This matches bluez error message */
> +		hid_err(hdev, "failed to write master address\n");
> +		return count;
> +	}
> +
> +	return count;
> +}
> +
> +static DEVICE_ATTR(master_bdaddr, S_IRUGO | S_IWUSR,
> +		   sony_master_bdaddr_show, sony_master_bdaddr_store);
> +
> +static int sony_set_gamepad_number(struct sony_sc *sc)
>  {
>  	int ret;
>  
>  	/*
> -	 * Only DualShock 4 or Sixaxis controllers get an id.
> +	 * Only DualShock 4 or Sixaxis controllers get a number.
>  	 * All others are set to -1.
>  	 */
>  	if ((sc->quirks & SIXAXIS_CONTROLLER) ||
>  	    (sc->quirks & DUALSHOCK4_CONTROLLER)) {
> -		ret = ida_simple_get(&sony_device_id_allocator, 0, 0,
> -					GFP_KERNEL);
> +		ret = ida_simple_get(&sony_gamepad_number_allocator, 0, 0,
> +				     GFP_KERNEL);
>  		if (ret < 0) {
> -			sc->device_id = -1;
> +			sc->gamepad_number = -1;
>  			return ret;
>  		}
> -		sc->device_id = ret;
> +		sc->gamepad_number = ret;
>  	} else {
> -		sc->device_id = -1;
> +		sc->gamepad_number = -1;
>  	}
>  
>  	return 0;
>  }
>  
> -static void sony_release_device_id(struct sony_sc *sc)
> +static void sony_release_gamepad_number(struct sony_sc *sc)
>  {
> -	if (sc->device_id >= 0) {
> -		ida_simple_remove(&sony_device_id_allocator, sc->device_id);
> -		sc->device_id = -1;
> +	if (sc->gamepad_number >= 0) {
> +		ida_simple_remove(&sony_gamepad_number_allocator,
> +				  sc->gamepad_number);
> +		sc->gamepad_number = -1;
>  	}
>  }
>  
> +static ssize_t sony_gamepad_number_show(struct device *dev,
> +					struct device_attribute *attr,
> +					char *buf)
> +{
> +	struct hid_device *hdev = container_of(dev, struct hid_device, dev);
> +	struct sony_sc *sc = hid_get_drvdata(hdev);
> +	size_t count;
> +
> +	if (!sc) {
> +		hid_err(hdev, "No device data\n");
> +		return 0;
> +	}
> +
> +	if (sc->gamepad_number < 0)
> +		return 0;
> +
> +	/* Add one to match console user expectations */
> +	count = scnprintf(buf, PAGE_SIZE, "%i\n", sc->gamepad_number + 1);
> +	return count;
> +}
> +
> +static DEVICE_ATTR(gamepad_number, S_IRUGO, sony_gamepad_number_show, NULL);
> +
>  static inline void sony_init_work(struct sony_sc *sc,
>  					void (*worker)(struct work_struct *))
>  {
> @@ -1841,10 +2017,22 @@ static int sony_probe(struct hid_device
>  		return ret;
>  	}
>  
> -	ret = sony_set_device_id(sc);
> +	ret = sony_set_gamepad_number(sc);
>  	if (ret < 0) {
> -		hid_err(hdev, "failed to allocate the device id\n");
> -		goto err_stop;
> +		hid_err(hdev, "failed to allocate the gamepad number\n");
> +		goto err_nosysfs2;
> +	}
> +
> +	ret = device_create_file(&hdev->dev, &dev_attr_gamepad_number);
> +	if (ret) {
> +		hid_err(hdev, "failed to create gamepad_number sysfs file\n");
> +		goto err_nosysfs2;
> +	}
> +
> +	ret = device_create_file(&hdev->dev, &dev_attr_master_bdaddr);
> +	if (ret) {
> +		hid_err(hdev, "failed to create master_bdaddr sysfs file\n");
> +		goto err_nosysfs1;
>  	}
>  
>  	if (sc->quirks & SIXAXIS_CONTROLLER_USB) {
> @@ -1932,13 +2120,17 @@ static int sony_probe(struct hid_device
>  err_close:
>  	hid_hw_close(hdev);
>  err_stop:
> +	device_remove_file(&hdev->dev, &dev_attr_master_bdaddr);
> +err_nosysfs1:
> +	device_remove_file(&hdev->dev, &dev_attr_gamepad_number);
> +err_nosysfs2:
>  	if (sc->quirks & SONY_LED_SUPPORT)
>  		sony_leds_remove(sc);
>  	if (sc->quirks & SONY_BATTERY_SUPPORT)
>  		sony_battery_remove(sc);
>  	sony_cancel_work_sync(sc);
>  	sony_remove_dev_list(sc);
> -	sony_release_device_id(sc);
> +	sony_release_gamepad_number(sc);
>  	hid_hw_stop(hdev);
>  	return ret;
>  }
> @@ -1959,7 +2151,11 @@ static void sony_remove(struct hid_devic
>  
>  	sony_remove_dev_list(sc);
>  
> -	sony_release_device_id(sc);
> +	device_remove_file(&hdev->dev, &dev_attr_master_bdaddr);
> +
> +	device_remove_file(&hdev->dev, &dev_attr_gamepad_number);
> +
> +	sony_release_gamepad_number(sc);
>  
>  	hid_hw_stop(hdev);
>  }
> @@ -2017,7 +2213,7 @@ static void __exit sony_exit(void)
>  {
>  	dbg_hid("Sony:%s\n", __func__);
>  
> -	ida_destroy(&sony_device_id_allocator);
> +	ida_destroy(&sony_gamepad_number_allocator);
>  	hid_unregister_driver(&sony_driver);
>  }
>  module_init(sony_init);
> --
> To unsubscribe from this list: send the line "unsubscribe linux-input" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


-- 
Antonio Ospite
http://ao2.it

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists