Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1480207 > unrolled thread

[PATCH v3] leds: Introduce userspace leds driver

Back to article view | Back to linux.kernel

Contents

  [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-09 19:00 +0200
    Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-09 22:50 +0200
    Re: [PATCH v3] leds: Introduce userspace leds driver Jacek Anaszewski <j.anaszewski@samsung.com> - 2016-09-12 10:20 +0200
      Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-12 17:00 +0200
      Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-15 15:10 +0200
        Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-15 15:40 +0200
          Re: [PATCH v3] leds: Introduce userspace leds driver Jacek Anaszewski <j.anaszewski@samsung.com> - 2016-09-15 17:00 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 08:20 +0200
        Re: [PATCH v3] leds: Introduce userspace leds driver Jacek Anaszewski <j.anaszewski@samsung.com> - 2016-09-15 17:00 +0200
          Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-15 17:40 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-15 17:40 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 08:00 +0200
              Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-16 17:40 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 08:00 +0200
              Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-16 17:20 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 08:10 +0200
              Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-16 17:50 +0200
          Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-15 18:40 +0200
            Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 08:00 +0200
              Re: [PATCH v3] leds: Introduce userspace leds driver Jacek Anaszewski <j.anaszewski@samsung.com> - 2016-09-16 09:10 +0200
                Re: [PATCH v3] leds: Introduce userspace leds driver David Lechner <david@lechnology.com> - 2016-09-16 17:20 +0200
                  Re: [PATCH v3] leds: Introduce userspace leds driver Pavel Machek <pavel@ucw.cz> - 2016-09-16 21:40 +0200

Page 1 of 2  [1] 2  Next page →

#1480207 — [PATCH v3] leds: Introduce userspace leds driver

This driver creates a userspace leds driver similar to uinput.

New leds are created by opening /dev/uleds and writing a uleds_user_dev
struct. A new leds class device is registered with the name given in the
struct. Reading will return a single byte that is the current brightness.
The poll() syscall is also supported. It will be triggered whenever the
brightness changes. Closing the file handle to /dev/uleds will remove
the leds class device.

Signed-off-by: David Lechner <david@lechnology.com>
---

v2 changes:

* sort #includes
* fix typo in Kconfig
* Add Documentation text file

v3 changes:

* fix typos in docs
* rename "leds class"  to "LED class" in docs
* rebase on linux-leds/for-next


 Documentation/leds/uleds.txt | 105 ++++++++++++++++++++
 drivers/leds/Kconfig         |   8 ++
 drivers/leds/Makefile        |   3 +
 drivers/leds/uleds.c         | 224 +++++++++++++++++++++++++++++++++++++++++++
 include/uapi/linux/Kbuild    |   1 +
 include/uapi/linux/uleds.h   |  23 +++++
 6 files changed, 364 insertions(+)
 create mode 100644 Documentation/leds/uleds.txt
 create mode 100644 drivers/leds/uleds.c
 create mode 100644 include/uapi/linux/uleds.h

diff --git a/Documentation/leds/uleds.txt b/Documentation/leds/uleds.txt
new file mode 100644
index 0000000..de11dbb
--- /dev/null
+++ b/Documentation/leds/uleds.txt
@@ -0,0 +1,105 @@
+Userspace LEDs
+==============
+
+The uleds driver supports userspace LEDs. This can be useful for testing
+triggers and can also be used to implement virtual LEDs.
+
+
+Usage
+=====
+
+When the driver is loaded, a character device is created at /dev/uleds. To
+create a new LED class device, open /dev/uleds and write a uleds_user_dev
+structure to it (found in kernel public header file linux/uleds.h).
+
+    #define ULEDS_MAX_NAME_SIZE 80
+
+    struct uleds_user_dev {
+        char name[ULEDS_MAX_NAME_SIZE];
+    };
+
+A new LED class device will be created with the name given. The name can be
+any valid file name, but consider using the LED class naming convention of
+"devicename:color:function".
+
+The current brightness is found by reading a single byte from the character
+device. Values are unsigned: 0 to 255. Reading does not block and always returns
+the most recent brightness value. The device node can also be polled to notify
+when the brightness value changes.
+
+The LED class device will be removed when the open file handle to /dev/uleds
+is closed.
+
+
+Example
+=======
+
+/*
+ * uledmon.c
+ *
+ * This program creates a new userspace LED class device and monitors it. A
+ * timestamp and brightness value is printed each time the brightness changes.
+ *
+ * Usage: uledmon <device-name>
+ *
+ * <device-name> is the name of the LED class device to be created. Pressing
+ * CTRL+C will exit.
+ */
+
+#include <fcntl.h>
+#include <poll.h>
+#include <stdio.h>
+#include <string.h>
+#include <time.h>
+#include <unistd.h>
+
+#include <linux/uleds.h>
+
+int main(int argc, char const *argv[])
+{
+    struct uleds_user_dev uleds_dev;
+    int fd, ret;
+    struct pollfd pfd;
+    unsigned char brightness;
+    struct timespec ts;
+
+    if (argc != 2) {
+        fprintf(stderr, "Requires <device-name> argument\n");
+        return 1;
+    }
+
+    strncpy(uleds_dev.name, argv[1], ULEDS_MAX_NAME_SIZE);
+
+    fd = open("/dev/uleds", O_RDWR);
+    if (fd == -1) {
+        perror("Failed to open /dev/uleds");
+        return 1;
+    }
+
+    ret = write(fd, &uleds_dev, sizeof(uleds_dev));
+    if (ret == -1) {
+        perror("Failed to write to /dev/uleds");
+        close(fd);
+        return 1;
+    }
+
+    pfd.fd = fd;
+    pfd.events = POLLIN;
+    pfd.revents = 0;
+
+    while (!(pfd.revents & (POLLERR | POLLHUP | POLLNVAL))) {
+        ret = read(fd, &brightness, 1);
+        if (ret == -1) {
+            perror("Failed to read from /dev/uleds");
+            close(fd);
+            return 1;
+        }
+        clock_gettime(CLOCK_MONOTONIC, &ts);
+        printf("[%ld.%09ld] %u\n", ts.tv_sec, ts.tv_nsec, brightness);
+        poll(&pfd, 1, -1);
+    }
+
+    close(fd);
+
+    return 0;
+}
diff --git a/drivers/leds/Kconfig b/drivers/leds/Kconfig
index 7a628c6..5fd3f4c 100644
--- a/drivers/leds/Kconfig
+++ b/drivers/leds/Kconfig
@@ -659,6 +659,14 @@ config LEDS_MLXCPLD
 	  This option enabled support for the LEDs on the Mellanox
 	  boards. Say Y to enabled these.
 
+config LEDS_USER
+	tristate "Userspace LED support"
+	depends on LEDS_CLASS
+	help
+	  This option enables support for userspace LEDs. Say 'y' to enable this
+	  support in kernel. To compile this driver as a module, choose 'm' here:
+	  the module will be called uleds.
+
 comment "LED Triggers"
 source "drivers/leds/trigger/Kconfig"
 
diff --git a/drivers/leds/Makefile b/drivers/leds/Makefile
index 3965070..d5331ff 100644
--- a/drivers/leds/Makefile
+++ b/drivers/leds/Makefile
@@ -75,5 +75,8 @@ obj-$(CONFIG_LEDS_MLXCPLD)		+= leds-mlxcpld.o
 # LED SPI Drivers
 obj-$(CONFIG_LEDS_DAC124S085)		+= leds-dac124s085.o
 
+# LED Userspace Drivers
+obj-$(CONFIG_LEDS_USER)			+= uleds.o
+
 # LED Triggers
 obj-$(CONFIG_LEDS_TRIGGERS)		+= trigger/
diff --git a/drivers/leds/uleds.c b/drivers/leds/uleds.c
new file mode 100644
index 0000000..77a0bee
--- /dev/null
+++ b/drivers/leds/uleds.c
@@ -0,0 +1,224 @@
+/*
+ * Userspace driver for leds subsystem
+ *
+ * Copyright (C) 2016 David Lechner <david@lechnology.com>
+ *
+ * Based on uinput.c: Aristeu Sergio Rozanski Filho <aris@cathedrallabs.org>
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/leds.h>
+#include <linux/miscdevice.h>
+#include <linux/module.h>
+#include <linux/poll.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+
+#include <uapi/linux/uleds.h>
+
+#define ULEDS_NAME	"uleds"
+
+enum uleds_state {
+	ULEDS_STATE_UNKNOWN,
+	ULEDS_STATE_REGISTERED,
+};
+
+struct uleds_device {
+	struct uleds_user_dev	user_dev;
+	struct led_classdev	led_cdev;
+	struct mutex		mutex;
+	enum uleds_state	state;
+	wait_queue_head_t	waitq;
+	unsigned char		brightness;
+	unsigned char		new_data;
+};
+
+static void uleds_brightness_set(struct led_classdev *led_cdev,
+				 enum led_brightness brightness)
+{
+	struct uleds_device *udev = container_of(led_cdev, struct uleds_device,
+						 led_cdev);
+
+	if (udev->brightness != brightness) {
+		udev->brightness = brightness;
+		udev->new_data = 1;
+		wake_up_interruptible(&udev->waitq);
+	}
+}
+
+static int uleds_open(struct inode *inode, struct file *file)
+{
+	struct uleds_device *udev;
+
+	udev = kzalloc(sizeof(*udev), GFP_KERNEL);
+	if (!udev)
+		return -ENOMEM;
+
+	udev->led_cdev.name = udev->user_dev.name;
+	udev->led_cdev.max_brightness = LED_FULL;
+	udev->led_cdev.brightness_set = uleds_brightness_set;
+
+	mutex_init(&udev->mutex);
+	init_waitqueue_head(&udev->waitq);
+	udev->state = ULEDS_STATE_UNKNOWN;
+
+	file->private_data = udev;
+	nonseekable_open(inode, file);
+
+	return 0;
+}
+
+static ssize_t uleds_write(struct file *file, const char __user *buffer,
+			   size_t count, loff_t *ppos)
+{
+	struct uleds_device *udev = file->private_data;
+	int ret;
+
+	if (count == 0)
+		return 0;
+
+	ret = mutex_lock_interruptible(&udev->mutex);
+	if (ret)
+		return ret;
+
+	if (udev->state == ULEDS_STATE_REGISTERED) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	if (count != sizeof(struct uleds_user_dev)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (copy_from_user(&udev->user_dev, buffer,
+			   sizeof(struct uleds_user_dev))) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (!udev->user_dev.name[0]) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = led_classdev_register(NULL, &udev->led_cdev);
+	if (ret < 0)
+		goto out;
+
+	udev->state = ULEDS_STATE_REGISTERED;
+	ret = count;
+
+out:
+	mutex_unlock(&udev->mutex);
+
+	return ret;
+}
+
+static ssize_t uleds_read(struct file *file, char __user *buffer, size_t count,
+			  loff_t *ppos)
+{
+	struct uleds_device *udev = file->private_data;
+	ssize_t retval;
+
+	if (count == 0)
+		return 0;
+
+	if (count != 1)
+		return -EINVAL;
+
+	do {
+		retval = mutex_lock_interruptible(&udev->mutex);
+		if (retval)
+			return retval;
+
+		if (udev->state != ULEDS_STATE_REGISTERED) {
+			retval = -ENODEV;
+		} else if (!udev->new_data && (file->f_flags & O_NONBLOCK)) {
+			retval = -EAGAIN;
+		} else {
+			retval = copy_to_user(buffer, &udev->brightness, 1);
+			udev->new_data = 0;
+			retval = 1;
+		}
+
+		mutex_unlock(&udev->mutex);
+
+		if (retval || count == 0)
+			break;
+
+		if (!(file->f_flags & O_NONBLOCK))
+			retval = wait_event_interruptible(udev->waitq,
+					udev->new_data ||
+					udev->state != ULEDS_STATE_REGISTERED);
+	} while (retval == 0);
+
+	return retval;
+}
+
+static unsigned int uleds_poll(struct file *file, poll_table *wait)
+{
+	struct uleds_device *udev = file->private_data;
+
+	poll_wait(file, &udev->waitq, wait);
+
+	if (udev->new_data)
+		return POLLIN | POLLRDNORM;
+
+	return 0;
+}
+
+static int uleds_release(struct inode *inode, struct file *file)
+{
+	struct uleds_device *udev = file->private_data;
+
+	if (udev->state == ULEDS_STATE_REGISTERED) {
+		udev->state = ULEDS_STATE_UNKNOWN;
+		led_classdev_unregister(&udev->led_cdev);
+	}
+	kfree(udev);
+
+	return 0;
+}
+
+static const struct file_operations uleds_fops = {
+	.owner		= THIS_MODULE,
+	.open		= uleds_open,
+	.release	= uleds_release,
+	.read		= uleds_read,
+	.write		= uleds_write,
+	.poll		= uleds_poll,
+	.llseek		= no_llseek,
+};
+
+static struct miscdevice uleds_misc = {
+	.fops		= &uleds_fops,
+	.minor		= MISC_DYNAMIC_MINOR,
+	.name		= ULEDS_NAME,
+};
+
+static int __init uleds_init(void)
+{
+	return misc_register(&uleds_misc);
+}
+module_init(uleds_init);
+
+static void __exit uleds_exit(void)
+{
+	misc_deregister(&uleds_misc);
+}
+module_exit(uleds_exit);
+
+MODULE_AUTHOR("David Lechner <david@lechnology.com>");
+MODULE_DESCRIPTION("Userspace driver for leds subsystem");
+MODULE_LICENSE("GPL");
diff --git a/include/uapi/linux/Kbuild b/include/uapi/linux/Kbuild
index 185f8ea..416f5e6 100644
--- a/include/uapi/linux/Kbuild
+++ b/include/uapi/linux/Kbuild
@@ -421,6 +421,7 @@ header-y += udp.h
 header-y += uhid.h
 header-y += uinput.h
 header-y += uio.h
+header-y += uleds.h
 header-y += ultrasound.h
 header-y += un.h
 header-y += unistd.h
diff --git a/include/uapi/linux/uleds.h b/include/uapi/linux/uleds.h
new file mode 100644
index 0000000..e78ed46
--- /dev/null
+++ b/include/uapi/linux/uleds.h
@@ -0,0 +1,23 @@
+/*
+ * Userspace driver support for leds subsystem
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+#ifndef _UAPI__ULEDS_H_
+#define _UAPI__ULEDS_H_
+
+#define ULEDS_MAX_NAME_SIZE	80
+
+struct uleds_user_dev {
+	char name[ULEDS_MAX_NAME_SIZE];
+};
+
+#endif /* _UAPI__ULEDS_H_ */
-- 
2.7.4

[toc] | [next] | [standalone]

#1480343

On 09/09/2016 02:30 PM, Kelly French wrote:
>
> Replying privately, as I'm not involved in the kernel community.
>

By offering your opinion, you are now involved in the community. No need 
to be shy. :-)

> I'm looking forward to this functionality.  I'm a little worried about
> the interface.  Is it possible to create the new uled devices in some
> other way?
>
> Maybe one insmod per uled device?  Maybe a new /proc/ entry (yuck) where
> you can echo the name into it, which will trigger udev to create the
> appropraite /dev/uled-name device?
>
> I'm hoping that the final interface is something I can just "cat
> /dev/uled-name" instead of having to write a binary struct into it at
> the beginning.  I'm even ok with just echoing text into the dev entry
> before reading.  It's the binary struct that's bugging me.
>


I originally considered using configfs to setup new LEDs[1], which would 
be very much like what you are suggesting, but I find the way my current 
patch works to be much simpler.

With this patch, you can create multiple LEDs. You just have to open a 
file handle to /dev/uleds for each LED that you want to create. Also, 
the LED device is automatically destroyed when the file handle is 
closed. I think this is nice because if the program that created it 
crashes, then it is automatically cleaned up.

I agree it would be much more convenient to not use a binary struct, but 
I think the complexity of implementing it outweighs any benefit to be 
gained. And it is easy enough to use an interactive python shell to 
handle the things that bash can't if the interactive part is what you 
really want.



[1]: https://lkml.org/lkml/2016/7/25/505

[toc] | [prev] | [next] | [standalone]

#1481024

Hi David,

Thanks for addressing the issues. I've applied the patch to the
for-next branch of linux-leds.git, after making few more
modifications indicated below.

Thanks,
Jacek Anaszewski

On 09/09/2016 06:49 PM, David Lechner wrote:
> This driver creates a userspace leds driver similar to uinput.

s/leds/LED class/

Also in the commit title:

s/leds/LED class/

>
> New leds are created by opening /dev/uleds and writing a uleds_user_dev

s/leds/LED class devices/

> struct. A new leds class device is registered with the name given in the

s/leds/LED/

> struct. Reading will return a single byte that is the current brightness.
> The poll() syscall is also supported. It will be triggered whenever the
> brightness changes. Closing the file handle to /dev/uleds will remove
> the leds class device.

s/leds/LED/

>
> Signed-off-by: David Lechner <david@lechnology.com>
> ---
>
> v2 changes:
>
> * sort #includes
> * fix typo in Kconfig
> * Add Documentation text file
>
> v3 changes:
>
> * fix typos in docs
> * rename "leds class"  to "LED class" in docs
> * rebase on linux-leds/for-next
>
>
>  Documentation/leds/uleds.txt | 105 ++++++++++++++++++++
>  drivers/leds/Kconfig         |   8 ++
>  drivers/leds/Makefile        |   3 +
>  drivers/leds/uleds.c         | 224 +++++++++++++++++++++++++++++++++++++++++++
>  include/uapi/linux/Kbuild    |   1 +
>  include/uapi/linux/uleds.h   |  23 +++++
>  6 files changed, 364 insertions(+)
>  create mode 100644 Documentation/leds/uleds.txt
>  create mode 100644 drivers/leds/uleds.c
>  create mode 100644 include/uapi/linux/uleds.h
>
> diff --git a/Documentation/leds/uleds.txt b/Documentation/leds/uleds.txt
> new file mode 100644
> index 0000000..de11dbb
> --- /dev/null
> +++ b/Documentation/leds/uleds.txt
> @@ -0,0 +1,105 @@
> +Userspace LEDs
> +==============
> +
> +The uleds driver supports userspace LEDs. This can be useful for testing
> +triggers and can also be used to implement virtual LEDs.
> +
> +
> +Usage
> +=====
> +
> +When the driver is loaded, a character device is created at /dev/uleds. To
> +create a new LED class device, open /dev/uleds and write a uleds_user_dev
> +structure to it (found in kernel public header file linux/uleds.h).
> +
> +    #define ULEDS_MAX_NAME_SIZE 80
> +
> +    struct uleds_user_dev {
> +        char name[ULEDS_MAX_NAME_SIZE];
> +    };
> +
> +A new LED class device will be created with the name given. The name can be
> +any valid file name, but consider using the LED class naming convention of
> +"devicename:color:function".
> +
> +The current brightness is found by reading a single byte from the character
> +device. Values are unsigned: 0 to 255. Reading does not block and always returns
> +the most recent brightness value. The device node can also be polled to notify
> +when the brightness value changes.
> +
> +The LED class device will be removed when the open file handle to /dev/uleds
> +is closed.
> +
> +
> +Example
> +=======
> +
> +/*
> + * uledmon.c
> + *
> + * This program creates a new userspace LED class device and monitors it. A
> + * timestamp and brightness value is printed each time the brightness changes.
> + *
> + * Usage: uledmon <device-name>
> + *
> + * <device-name> is the name of the LED class device to be created. Pressing
> + * CTRL+C will exit.
> + */
> +
> +#include <fcntl.h>
> +#include <poll.h>
> +#include <stdio.h>
> +#include <string.h>
> +#include <time.h>
> +#include <unistd.h>
> +
> +#include <linux/uleds.h>
> +
> +int main(int argc, char const *argv[])
> +{
> +    struct uleds_user_dev uleds_dev;
> +    int fd, ret;
> +    struct pollfd pfd;
> +    unsigned char brightness;
> +    struct timespec ts;
> +
> +    if (argc != 2) {
> +        fprintf(stderr, "Requires <device-name> argument\n");
> +        return 1;
> +    }
> +
> +    strncpy(uleds_dev.name, argv[1], ULEDS_MAX_NAME_SIZE);
> +
> +    fd = open("/dev/uleds", O_RDWR);
> +    if (fd == -1) {
> +        perror("Failed to open /dev/uleds");
> +        return 1;
> +    }
> +
> +    ret = write(fd, &uleds_dev, sizeof(uleds_dev));
> +    if (ret == -1) {
> +        perror("Failed to write to /dev/uleds");
> +        close(fd);
> +        return 1;
> +    }
> +
> +    pfd.fd = fd;
> +    pfd.events = POLLIN;
> +    pfd.revents = 0;
> +
> +    while (!(pfd.revents & (POLLERR | POLLHUP | POLLNVAL))) {
> +        ret = read(fd, &brightness, 1);
> +        if (ret == -1) {
> +            perror("Failed to read from /dev/uleds");
> +            close(fd);
> +            return 1;
> +        }
> +        clock_gettime(CLOCK_MONOTONIC, &ts);
> +        printf("[%ld.%09ld] %u\n", ts.tv_sec, ts.tv_nsec, brightness);
> +        poll(&pfd, 1, -1);
> +    }
> +
> +    close(fd);
> +
> +    return 0;
> +}
> diff --git a/drivers/leds/Kconfig b/drivers/leds/Kconfig
> index 7a628c6..5fd3f4c 100644
> --- a/drivers/leds/Kconfig
> +++ b/drivers/leds/Kconfig
> @@ -659,6 +659,14 @@ config LEDS_MLXCPLD
>  	  This option enabled support for the LEDs on the Mellanox
>  	  boards. Say Y to enabled these.
>
> +config LEDS_USER
> +	tristate "Userspace LED support"
> +	depends on LEDS_CLASS
> +	help
> +	  This option enables support for userspace LEDs. Say 'y' to enable this
> +	  support in kernel. To compile this driver as a module, choose 'm' here:
> +	  the module will be called uleds.
> +
>  comment "LED Triggers"
>  source "drivers/leds/trigger/Kconfig"
>
> diff --git a/drivers/leds/Makefile b/drivers/leds/Makefile
> index 3965070..d5331ff 100644
> --- a/drivers/leds/Makefile
> +++ b/drivers/leds/Makefile
> @@ -75,5 +75,8 @@ obj-$(CONFIG_LEDS_MLXCPLD)		+= leds-mlxcpld.o
>  # LED SPI Drivers
>  obj-$(CONFIG_LEDS_DAC124S085)		+= leds-dac124s085.o
>
> +# LED Userspace Drivers
> +obj-$(CONFIG_LEDS_USER)			+= uleds.o
> +
>  # LED Triggers
>  obj-$(CONFIG_LEDS_TRIGGERS)		+= trigger/
> diff --git a/drivers/leds/uleds.c b/drivers/leds/uleds.c
> new file mode 100644
> index 0000000..77a0bee
> --- /dev/null
> +++ b/drivers/leds/uleds.c
> @@ -0,0 +1,224 @@
> +/*
> + * Userspace driver for leds subsystem

s/leds subsystem/the LED subsystem/

> + *
> + * Copyright (C) 2016 David Lechner <david@lechnology.com>
> + *
> + * Based on uinput.c: Aristeu Sergio Rozanski Filho <aris@cathedrallabs.org>
> + *
> + * 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.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + */
> +#include <linux/fs.h>
> +#include <linux/init.h>
> +#include <linux/leds.h>
> +#include <linux/miscdevice.h>
> +#include <linux/module.h>
> +#include <linux/poll.h>
> +#include <linux/sched.h>
> +#include <linux/slab.h>
> +
> +#include <uapi/linux/uleds.h>
> +
> +#define ULEDS_NAME	"uleds"
> +
> +enum uleds_state {
> +	ULEDS_STATE_UNKNOWN,
> +	ULEDS_STATE_REGISTERED,
> +};
> +
> +struct uleds_device {
> +	struct uleds_user_dev	user_dev;
> +	struct led_classdev	led_cdev;
> +	struct mutex		mutex;
> +	enum uleds_state	state;
> +	wait_queue_head_t	waitq;
> +	unsigned char		brightness;
> +	unsigned char		new_data;
> +};
> +
> +static void uleds_brightness_set(struct led_classdev *led_cdev,
> +				 enum led_brightness brightness)
> +{
> +	struct uleds_device *udev = container_of(led_cdev, struct uleds_device,
> +						 led_cdev);
> +
> +	if (udev->brightness != brightness) {
> +		udev->brightness = brightness;
> +		udev->new_data = 1;
> +		wake_up_interruptible(&udev->waitq);
> +	}
> +}
> +
> +static int uleds_open(struct inode *inode, struct file *file)
> +{
> +	struct uleds_device *udev;
> +
> +	udev = kzalloc(sizeof(*udev), GFP_KERNEL);
> +	if (!udev)
> +		return -ENOMEM;
> +
> +	udev->led_cdev.name = udev->user_dev.name;
> +	udev->led_cdev.max_brightness = LED_FULL;
> +	udev->led_cdev.brightness_set = uleds_brightness_set;
> +
> +	mutex_init(&udev->mutex);
> +	init_waitqueue_head(&udev->waitq);
> +	udev->state = ULEDS_STATE_UNKNOWN;
> +
> +	file->private_data = udev;
> +	nonseekable_open(inode, file);
> +
> +	return 0;
> +}
> +
> +static ssize_t uleds_write(struct file *file, const char __user *buffer,
> +			   size_t count, loff_t *ppos)
> +{
> +	struct uleds_device *udev = file->private_data;
> +	int ret;
> +
> +	if (count == 0)
> +		return 0;
> +
> +	ret = mutex_lock_interruptible(&udev->mutex);
> +	if (ret)
> +		return ret;
> +
> +	if (udev->state == ULEDS_STATE_REGISTERED) {
> +		ret = -EBUSY;
> +		goto out;
> +	}
> +
> +	if (count != sizeof(struct uleds_user_dev)) {
> +		ret = -EINVAL;
> +		goto out;
> +	}
> +
> +	if (copy_from_user(&udev->user_dev, buffer,
> +			   sizeof(struct uleds_user_dev))) {
> +		ret = -EFAULT;
> +		goto out;
> +	}
> +
> +	if (!udev->user_dev.name[0]) {
> +		ret = -EINVAL;
> +		goto out;
> +	}
> +
> +	ret = led_classdev_register(NULL, &udev->led_cdev);
> +	if (ret < 0)
> +		goto out;
> +
> +	udev->state = ULEDS_STATE_REGISTERED;
> +	ret = count;
> +
> +out:
> +	mutex_unlock(&udev->mutex);
> +
> +	return ret;
> +}
> +
> +static ssize_t uleds_read(struct file *file, char __user *buffer, size_t count,
> +			  loff_t *ppos)
> +{
> +	struct uleds_device *udev = file->private_data;
> +	ssize_t retval;
> +
> +	if (count == 0)
> +		return 0;
> +
> +	if (count != 1)
> +		return -EINVAL;
> +
> +	do {
> +		retval = mutex_lock_interruptible(&udev->mutex);
> +		if (retval)
> +			return retval;
> +
> +		if (udev->state != ULEDS_STATE_REGISTERED) {
> +			retval = -ENODEV;
> +		} else if (!udev->new_data && (file->f_flags & O_NONBLOCK)) {
> +			retval = -EAGAIN;
> +		} else {
> +			retval = copy_to_user(buffer, &udev->brightness, 1);
> +			udev->new_data = 0;
> +			retval = 1;
> +		}
> +
> +		mutex_unlock(&udev->mutex);
> +
> +		if (retval || count == 0)
> +			break;
> +
> +		if (!(file->f_flags & O_NONBLOCK))
> +			retval = wait_event_interruptible(udev->waitq,
> +					udev->new_data ||
> +					udev->state != ULEDS_STATE_REGISTERED);
> +	} while (retval == 0);
> +
> +	return retval;
> +}
> +
> +static unsigned int uleds_poll(struct file *file, poll_table *wait)
> +{
> +	struct uleds_device *udev = file->private_data;
> +
> +	poll_wait(file, &udev->waitq, wait);
> +
> +	if (udev->new_data)
> +		return POLLIN | POLLRDNORM;
> +
> +	return 0;
> +}
> +
> +static int uleds_release(struct inode *inode, struct file *file)
> +{
> +	struct uleds_device *udev = file->private_data;
> +
> +	if (udev->state == ULEDS_STATE_REGISTERED) {
> +		udev->state = ULEDS_STATE_UNKNOWN;
> +		led_classdev_unregister(&udev->led_cdev);
> +	}
> +	kfree(udev);
> +
> +	return 0;
> +}
> +
> +static const struct file_operations uleds_fops = {
> +	.owner		= THIS_MODULE,
> +	.open		= uleds_open,
> +	.release	= uleds_release,
> +	.read		= uleds_read,
> +	.write		= uleds_write,
> +	.poll		= uleds_poll,
> +	.llseek		= no_llseek,
> +};
> +
> +static struct miscdevice uleds_misc = {
> +	.fops		= &uleds_fops,
> +	.minor		= MISC_DYNAMIC_MINOR,
> +	.name		= ULEDS_NAME,
> +};
> +
> +static int __init uleds_init(void)
> +{
> +	return misc_register(&uleds_misc);
> +}
> +module_init(uleds_init);
> +
> +static void __exit uleds_exit(void)
> +{
> +	misc_deregister(&uleds_misc);
> +}
> +module_exit(uleds_exit);
> +
> +MODULE_AUTHOR("David Lechner <david@lechnology.com>");
> +MODULE_DESCRIPTION("Userspace driver for leds subsystem");

s/leds subsystem/the LED subsystem/

> +MODULE_LICENSE("GPL");
> diff --git a/include/uapi/linux/Kbuild b/include/uapi/linux/Kbuild
> index 185f8ea..416f5e6 100644
> --- a/include/uapi/linux/Kbuild
> +++ b/include/uapi/linux/Kbuild
> @@ -421,6 +421,7 @@ header-y += udp.h
>  header-y += uhid.h
>  header-y += uinput.h
>  header-y += uio.h
> +header-y += uleds.h
>  header-y += ultrasound.h
>  header-y += un.h
>  header-y += unistd.h
> diff --git a/include/uapi/linux/uleds.h b/include/uapi/linux/uleds.h
> new file mode 100644
> index 0000000..e78ed46
> --- /dev/null
> +++ b/include/uapi/linux/uleds.h
> @@ -0,0 +1,23 @@
> +/*
> + * Userspace driver support for leds subsystem

s/leds subsystem/the LED subsystem/

> + *
> + * 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.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + */
> +#ifndef _UAPI__ULEDS_H_
> +#define _UAPI__ULEDS_H_
> +
> +#define ULEDS_MAX_NAME_SIZE	80
> +
> +struct uleds_user_dev {
> +	char name[ULEDS_MAX_NAME_SIZE];
> +};
> +
> +#endif /* _UAPI__ULEDS_H_ */
>

[toc] | [prev] | [next] | [standalone]

#1481387

On 09/12/2016 03:18 AM, Jacek Anaszewski wrote:
> Hi David,
>
> Thanks for addressing the issues. I've applied the patch to the
> for-next branch of linux-leds.git, after making few more
> modifications indicated below.

Great!

[toc] | [prev] | [next] | [standalone]

#1484146

Hi!

> >+	if (copy_from_user(&udev->user_dev, buffer,
> >+			   sizeof(struct uleds_user_dev))) {
> >+		ret = -EFAULT;
> >+		goto out;
> >+	}
> >+
> >+	if (!udev->user_dev.name[0]) {
> >+		ret = -EINVAL;
> >+		goto out;
> >+	}
> >+
> >+	ret = led_classdev_register(NULL, &udev->led_cdev);
> >+	if (ret < 0)
> >+		goto out;

No sanity checking on the name -> probably a security hole. Do not
push this upstream before this is fixed.

Thanks,
								Pavel
								
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1484173

Hi!

> > >+	if (copy_from_user(&udev->user_dev, buffer,
> > >+			   sizeof(struct uleds_user_dev))) {
> > >+		ret = -EFAULT;
> > >+		goto out;
> > >+	}
> > >+
> > >+	if (!udev->user_dev.name[0]) {
> > >+		ret = -EINVAL;
> > >+		goto out;
> > >+	}
> > >+
> > >+	ret = led_classdev_register(NULL, &udev->led_cdev);
> > >+	if (ret < 0)
> > >+		goto out;
> 
> No sanity checking on the name -> probably a security hole. Do not
> push this upstream before this is fixed.

And actually... is it possible to have more then one userspace LED
with this interface? I do have RGB LED connect on /dev/ttyUSB0, and I
guess userspace driver would be appropriate, but it needs 3
channels...

Best regards,
								Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1484273

Hi Pavel,

On 09/15/2016 03:35 PM, Pavel Machek wrote:
> Hi!
>
>>>> +	if (copy_from_user(&udev->user_dev, buffer,
>>>> +			   sizeof(struct uleds_user_dev))) {
>>>> +		ret = -EFAULT;
>>>> +		goto out;
>>>> +	}
>>>> +
>>>> +	if (!udev->user_dev.name[0]) {
>>>> +		ret = -EINVAL;
>>>> +		goto out;
>>>> +	}
>>>> +
>>>> +	ret = led_classdev_register(NULL, &udev->led_cdev);
>>>> +	if (ret < 0)
>>>> +		goto out;
>>
>> No sanity checking on the name -> probably a security hole. Do not
>> push this upstream before this is fixed.
>
> And actually... is it possible to have more then one userspace LED
> with this interface? I do have RGB LED connect on /dev/ttyUSB0, and I
> guess userspace driver would be appropriate, but it needs 3
> channels...

Each new successful write to /dev/uleds creates a new LED class device.

See Documentation/leds/uleds.txt.

-- 
Best regards,
Jacek Anaszewski

[toc] | [prev] | [next] | [standalone]

#1484699

On Thu 2016-09-15 16:54:35, Jacek Anaszewski wrote:
> Hi Pavel,
> 
> On 09/15/2016 03:35 PM, Pavel Machek wrote:
> >Hi!
> >
> >>>>+	if (copy_from_user(&udev->user_dev, buffer,
> >>>>+			   sizeof(struct uleds_user_dev))) {
> >>>>+		ret = -EFAULT;
> >>>>+		goto out;
> >>>>+	}
> >>>>+
> >>>>+	if (!udev->user_dev.name[0]) {
> >>>>+		ret = -EINVAL;
> >>>>+		goto out;
> >>>>+	}
> >>>>+
> >>>>+	ret = led_classdev_register(NULL, &udev->led_cdev);
> >>>>+	if (ret < 0)
> >>>>+		goto out;
> >>
> >>No sanity checking on the name -> probably a security hole. Do not
> >>push this upstream before this is fixed.
> >
> >And actually... is it possible to have more then one userspace LED
> >with this interface? I do have RGB LED connect on /dev/ttyUSB0, and I
> >guess userspace driver would be appropriate, but it needs 3
> >channels...
> 
> Each new successful write to /dev/uleds creates a new LED class device.
> 
> See Documentation/leds/uleds.txt.

Umm. No, only one write is permitted to /dev/uleds according to the
code. In the email thread, it says:

# With this patch, you can create multiple LEDs. You just have to open a
# file handle to /dev/uleds for each LED that you want to create. Also,
# the LED device is automatically destroyed when the file handle is
# closed. I think this is nice because if the program that created it
# crashes, then it is automatically cleaned up.

...which should be actually ok. Sorry for the noise.

Best regards,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1484275

Hi Pavel,

On 09/15/2016 03:08 PM, Pavel Machek wrote:
> Hi!
>
>>> +	if (copy_from_user(&udev->user_dev, buffer,
>>> +			   sizeof(struct uleds_user_dev))) {
>>> +		ret = -EFAULT;
>>> +		goto out;
>>> +	}
>>> +
>>> +	if (!udev->user_dev.name[0]) {
>>> +		ret = -EINVAL;
>>> +		goto out;
>>> +	}
>>> +
>>> +	ret = led_classdev_register(NULL, &udev->led_cdev);
>>> +	if (ret < 0)
>>> +		goto out;
>
> No sanity checking on the name -> probably a security hole. Do not
> push this upstream before this is fixed.

Thanks for catching this.

David, please check if the LED name sticks to the LED class
device naming convention.

And one thing that caught my eye only now - please use
devm_led_classdev_register().

For now I'm dropping the patch.

-- 
Best regards,
Jacek Anaszewski

[toc] | [prev] | [next] | [standalone]

#1484326

On 09/15/2016 09:54 AM, Jacek Anaszewski wrote:
> Hi Pavel,
>
> On 09/15/2016 03:08 PM, Pavel Machek wrote:
>> Hi!
>>
>>>> +    if (copy_from_user(&udev->user_dev, buffer,
>>>> +               sizeof(struct uleds_user_dev))) {
>>>> +        ret = -EFAULT;
>>>> +        goto out;
>>>> +    }
>>>> +
>>>> +    if (!udev->user_dev.name[0]) {
>>>> +        ret = -EINVAL;
>>>> +        goto out;
>>>> +    }
>>>> +
>>>> +    ret = led_classdev_register(NULL, &udev->led_cdev);
>>>> +    if (ret < 0)
>>>> +        goto out;
>>
>> No sanity checking on the name -> probably a security hole. Do not
>> push this upstream before this is fixed.
>
> Thanks for catching this.
>
> David, please check if the LED name sticks to the LED class
> device naming convention.

I don't think it is a good idea to enforce the LED class naming 
convention. Someone may have a userspace application they want to test 
that has a hard-coded name that does not follow the convention.


>
> And one thing that caught my eye only now - please use
> devm_led_classdev_register().

How do I use devm_* when there is no parent device?


>
> For now I'm dropping the patch.
>

[toc] | [prev] | [next] | [standalone]

#1484327

On 09/15/2016 10:31 AM, David Lechner wrote:

>
> How do I use devm_* when there is no parent device?
>
>

Answering my own question. I seen now that struct miscdevice has struct 
device that should be used as the parent.

[toc] | [prev] | [next] | [standalone]

#1484686

Hi!

+The current brightness is found by reading a single byte from the
character
+device. Values are unsigned: 0 to 255. Reading does not block and
always returns
+the most recent brightness value. The device node can also be polled
to notify
+when the brightness value changes.

What is going on there? We have O_NONBLOCK, user should be able to
select if he wants blocking behaviour or not.

And yes, there's interface for creating a LED, but not one for
deleting it?

And... how is it going to work with multiple LEDs? Userspace gets
single bytes with brightness. How does it know to which LED the
brightness belongs?

Thanks,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1485129

On 09/16/2016 12:59 AM, Pavel Machek wrote:
> Hi!
>
> +The current brightness is found by reading a single byte from the
> character
> +device. Values are unsigned: 0 to 255. Reading does not block and
> always returns
> +the most recent brightness value. The device node can also be polled
> to notify
> +when the brightness value changes.
>
> What is going on there? We have O_NONBLOCK, user should be able to
> select if he wants blocking behaviour or not.

I will look into this.

>
> And yes, there's interface for creating a LED, but not one for
> deleting it?

uleds.txt says "The LED class device will be removed when the open file 
handle to /dev/uleds is closed."

Is this not clear?

>
> And... how is it going to work with multiple LEDs? Userspace gets
> single bytes with brightness. How does it know to which LED the
> brightness belongs?

It looks like this is missing from uleds.txt. By opening multiple file 
handles to /dev/uleds, you can create multiple devices. Each file handle 
will return a different value when read that corresponds to the LEDs 
class device that belongs to it.

[toc] | [prev] | [next] | [standalone]

#1484691

On Thu 2016-09-15 10:31:50, David Lechner wrote:
> On 09/15/2016 09:54 AM, Jacek Anaszewski wrote:
> >Hi Pavel,
> >
> >On 09/15/2016 03:08 PM, Pavel Machek wrote:
> >>Hi!
> >>
> >>>>+    if (copy_from_user(&udev->user_dev, buffer,
> >>>>+               sizeof(struct uleds_user_dev))) {
> >>>>+        ret = -EFAULT;
> >>>>+        goto out;
> >>>>+    }
> >>>>+
> >>>>+    if (!udev->user_dev.name[0]) {
> >>>>+        ret = -EINVAL;
> >>>>+        goto out;
> >>>>+    }
> >>>>+
> >>>>+    ret = led_classdev_register(NULL, &udev->led_cdev);
> >>>>+    if (ret < 0)
> >>>>+        goto out;
> >>
> >>No sanity checking on the name -> probably a security hole. Do not
> >>push this upstream before this is fixed.
> >
> >Thanks for catching this.
> >
> >David, please check if the LED name sticks to the LED class
> >device naming convention.
> 
> I don't think it is a good idea to enforce the LED class naming convention.
> Someone may have a userspace application they want to test that has a
> hard-coded name that does not follow the convention.

Umm.

Noone has applications with hardcoded names that are not possible
today, right?

And better not encourage crazy names.

Best regards,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1485111

On 09/16/2016 12:51 AM, Pavel Machek wrote:
> On Thu 2016-09-15 10:31:50, David Lechner wrote:
>> On 09/15/2016 09:54 AM, Jacek Anaszewski wrote:
>>> Hi Pavel,
>>>
>>> On 09/15/2016 03:08 PM, Pavel Machek wrote:
>>>> Hi!
>>>>
>>>>>> +    if (copy_from_user(&udev->user_dev, buffer,
>>>>>> +               sizeof(struct uleds_user_dev))) {
>>>>>> +        ret = -EFAULT;
>>>>>> +        goto out;
>>>>>> +    }
>>>>>> +
>>>>>> +    if (!udev->user_dev.name[0]) {
>>>>>> +        ret = -EINVAL;
>>>>>> +        goto out;
>>>>>> +    }
>>>>>> +
>>>>>> +    ret = led_classdev_register(NULL, &udev->led_cdev);
>>>>>> +    if (ret < 0)
>>>>>> +        goto out;
>>>>
>>>> No sanity checking on the name -> probably a security hole. Do not
>>>> push this upstream before this is fixed.
>>>
>>> Thanks for catching this.
>>>
>>> David, please check if the LED name sticks to the LED class
>>> device naming convention.
>>
>> I don't think it is a good idea to enforce the LED class naming convention.
>> Someone may have a userspace application they want to test that has a
>> hard-coded name that does not follow the convention.
>
> Umm.
>
> Noone has applications with hardcoded names that are not possible
> today, right?
>
> And better not encourage crazy names.
>
> Best regards,
> 									Pavel
>

Here is the actual `ls /sys/class/leds` from my Raspberry Pi:

led0                      pistorms:BA:red:ev3dev    pistorms:BB:red:ev3dev
pistorms:BA:blue:ev3dev   pistorms:BB:blue:ev3dev
pistorms:BA:green:ev3dev  pistorms:BB:green:ev3dev



Suppose I want to use uleds on my desktop to simulate my Raspberry Pi. 
If we restrict the name to the LEDs class convention of 
device:color:function, then I can't do this. led0 does not follow the 
convention at all. The other do follow the convention, but only if we 
allow that the device portion of the name can also include ':'.

It's too late, the crazy names already exist.

[toc] | [prev] | [next] | [standalone]

#1484697

Hi!

+static ssize_t uleds_read(struct file *file, char __user *buffer,
size_t count,
+ loff_t *ppos)
+{
+ struct uleds_device *udev = file->private_data;
+ ssize_t retval;
+
+ if (count == 0)
+ return 0;
+
+ if (count != 1)
+ return -EINVAL;

This is quite anti-social. You are free to return 1 byte on any read
(short read), but please allow reads with bigger buffers.

Thanks,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1485133

On 09/16/2016 01:07 AM, Pavel Machek wrote:
> Hi!
>
> +static ssize_t uleds_read(struct file *file, char __user *buffer,
> size_t count,
> + loff_t *ppos)
> +{
> + struct uleds_device *udev = file->private_data;
> + ssize_t retval;
> +
> + if (count == 0)
> + return 0;
> +
> + if (count != 1)
> + return -EINVAL;
>
> This is quite anti-social. You are free to return 1 byte on any read
> (short read), but please allow reads with bigger buffers.
>
> Thanks,
> 									Pavel
>

Sure.

[toc] | [prev] | [next] | [standalone]

#1484400

On 09/15/2016 09:54 AM, Jacek Anaszewski wrote:
> Hi Pavel,
>
> On 09/15/2016 03:08 PM, Pavel Machek wrote:
>> Hi!
>>
>>>> +    if (copy_from_user(&udev->user_dev, buffer,
>>>> +               sizeof(struct uleds_user_dev))) {
>>>> +        ret = -EFAULT;
>>>> +        goto out;
>>>> +    }
>>>> +
>>>> +    if (!udev->user_dev.name[0]) {
>>>> +        ret = -EINVAL;
>>>> +        goto out;
>>>> +    }
>>>> +
>>>> +    ret = led_classdev_register(NULL, &udev->led_cdev);
>>>> +    if (ret < 0)
>>>> +        goto out;
>>
>> No sanity checking on the name -> probably a security hole. Do not
>> push this upstream before this is fixed.
>

If this is a serious security issue, then you should also raise an issue 
with input maintainers because this is the extent of sanity checking for 
uinput device names as well.

I must confess that I am no security expert, so unless you can give 
specific examples of what potential threats are, I will not be able to 
guess what I need to do to fix it.

After some digging around the kernel, I don't see many instances of 
validating device node names. The best I have found so far comes from 
create_entry() in binfmt_misc.c

	if (!e->name[0] ||
	    !strcmp(e->name, ".") ||
	    !strcmp(e->name, "..") ||
	    strchr(e->name, '/'))
		goto einval;

Would something like this be a sufficient sanity check? I suppose we 
could also check for non-printing characters, but I don't think ignoring 
them would be a security issue.


> Thanks for catching this.
>
> David, please check if the LED name sticks to the LED class
> device naming convention.
>
> And one thing that caught my eye only now - please use
> devm_led_classdev_register().
>
> For now I'm dropping the patch.
>

[toc] | [prev] | [next] | [standalone]

#1484692

Hi!

> >>>>+    if (copy_from_user(&udev->user_dev, buffer,
> >>>>+               sizeof(struct uleds_user_dev))) {
> >>>>+        ret = -EFAULT;
> >>>>+        goto out;
> >>>>+    }
> >>>>+
> >>>>+    if (!udev->user_dev.name[0]) {
> >>>>+        ret = -EINVAL;
> >>>>+        goto out;
> >>>>+    }
> >>>>+
> >>>>+    ret = led_classdev_register(NULL, &udev->led_cdev);
> >>>>+    if (ret < 0)
> >>>>+        goto out;
> >>
> >>No sanity checking on the name -> probably a security hole. Do not
> >>push this upstream before this is fixed.
> >
> 
> If this is a serious security issue, then you should also raise an issue
> with input maintainers because this is the extent of sanity checking for
> uinput device names as well.

I guess that should be fixed. But lets not add new ones.

> I must confess that I am no security expert, so unless you can give specific
> examples of what potential threats are, I will not be able to guess what I
> need to do to fix it.
> 
> After some digging around the kernel, I don't see many instances of
> validating device node names. The best I have found so far comes from
> create_entry() in binfmt_misc.c
> 
> 	if (!e->name[0] ||
> 	    !strcmp(e->name, ".") ||
> 	    !strcmp(e->name, "..") ||
> 	    strchr(e->name, '/'))
> 		goto einval;
> 
> Would something like this be a sufficient sanity check? I suppose we could
> also check for non-printing characters, but I don't think ignoring them
> would be a security issue.

That would be minimum, yes. I guess it would be better/easier to just
limit the names to [a-zA-Z:-_0-9]*?

Thanks,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1484715

On 09/16/2016 07:50 AM, Pavel Machek wrote:
> Hi!
>
>>>>>> +    if (copy_from_user(&udev->user_dev, buffer,
>>>>>> +               sizeof(struct uleds_user_dev))) {
>>>>>> +        ret = -EFAULT;
>>>>>> +        goto out;
>>>>>> +    }
>>>>>> +
>>>>>> +    if (!udev->user_dev.name[0]) {
>>>>>> +        ret = -EINVAL;
>>>>>> +        goto out;
>>>>>> +    }
>>>>>> +
>>>>>> +    ret = led_classdev_register(NULL, &udev->led_cdev);
>>>>>> +    if (ret < 0)
>>>>>> +        goto out;
>>>>
>>>> No sanity checking on the name -> probably a security hole. Do not
>>>> push this upstream before this is fixed.
>>>
>>
>> If this is a serious security issue, then you should also raise an issue
>> with input maintainers because this is the extent of sanity checking for
>> uinput device names as well.
>
> I guess that should be fixed. But lets not add new ones.
>
>> I must confess that I am no security expert, so unless you can give specific
>> examples of what potential threats are, I will not be able to guess what I
>> need to do to fix it.
>>
>> After some digging around the kernel, I don't see many instances of
>> validating device node names. The best I have found so far comes from
>> create_entry() in binfmt_misc.c
>>
>> 	if (!e->name[0] ||
>> 	    !strcmp(e->name, ".") ||
>> 	    !strcmp(e->name, "..") ||
>> 	    strchr(e->name, '/'))
>> 		goto einval;
>>
>> Would something like this be a sufficient sanity check? I suppose we could
>> also check for non-printing characters, but I don't think ignoring them
>> would be a security issue.
>
> That would be minimum, yes. I guess it would be better/easier to just
> limit the names to [a-zA-Z:-_0-9]*?

Right, and we also could check if there are no more then two ":"
characters in the name.

-- 
Best regards,
Jacek Anaszewski

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | linux.kernel

csiph-web