Skip to contentSkip to conversation

Input: Add userio module

PATCH Superseded [email protected] 4 participants 19 messages Sep 17, 2015 – Oct 28, 2015

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH v5 1/1 Input: Add userio module

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes

* Correct "VIRTUAL PS/2 DEVICE" to "VIRTUAL SERIO DEVICE" in MAINTAINERS
* Remove uneccessary forward declaration in userio
* s/repeat repeat/repeat/
* Drop userio->serio checks, this struct is allocated right when we register the serio port so we don't need to worry about this ever being NULL
* Allocate a static minor and use that
* Add MODULE_ALIAS_MISCDEV() and MODULOE_ALIAS() lines to the module
* Improve the module description
* Change description for userio.h to "userio: virtual serio device support"
* Fix double space in comments in userio.h
* Change userio command types to enumerators

Responses
* Re: if (!userio) in userio_device_write(): it happens if we close the file descriptor while the input driver is trying to talk to the device, since we can't immediately bring down the driver. Removing the condition breaks the driver
* Re: ignoring trailing bytes: The main reason for that is if we need to add something else to the struct in the future. I doubt this will ever be done, but I figured I might as well let that behavior go since it can't harm anyone.
* Re: ignoring larger payloads + adding larger payloads: As of right now there really isn't one. serio_send() only allows you to send a single character at a time, and we can add the ability to have flags using this same packet format. I don't think userio is ever going to get much more complex then this, since I don't think serio is going to be changing very much in the future. This being said, if some unlikely circumstances happen and it does happen to change in such a way we need to extend the payload, we're currently ignoring the rest of the payload so we could safely add additional data onto the end of the struct.

 Documentation/input/userio.txt |  70 +++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 277 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 410 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 7ba7ab7..0a0fe08 100644
@@ -11086,6 +11086,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/ps2emu.c
+F:	include/uapi/linux/ps2emu.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	[email protected]
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +277 -0

277 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
maintainer
#1 lore raw

Hi Stephen,

Quoted 5 lines · On Thu, Sep 17, 2015 at 07:00:10PM -0400, [email protected] wrote:

Responses
* Re: if (!userio) in userio_device_write(): it happens if we close the file descriptor while the input driver is trying to talk to the device, since we can't immediately bring down the driver. Removing the condition breaks the driver

...

drivers/input/serio/userio.c line 58

Quoted 11 lines
drivers/input/serio/userio.c new file +11 -0
+/**
+ * userio_device_write - Write data from serio to a userio device in userspace
+ * @id: The serio port for the userio device
+ * @val: The data to write to the device
+ */
+static int userio_device_write(struct serio *id, unsigned char val)
+{
+	struct userio_device *userio = id->port_data;
+
+	if (!userio)
+		return -1;

1. I do not see where we reset port data.
2. I do not see what prevents object to which you now have a pointer to from disappearing at any moment past your check.

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+	mutex_lock(&userio->lock);

serio_write() must allow to be called from interrupt context, so you can't use mutex here.

Thanks.

Signature · 1 line
Dmitry
#2 lore raw
In reply to [email protected]

Hi

Quoted 38 lines · On Fri, Sep 18, 2015 at 1:00 AM, <[email protected]> wrote:

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes

* Correct "VIRTUAL PS/2 DEVICE" to "VIRTUAL SERIO DEVICE" in MAINTAINERS
* Remove uneccessary forward declaration in userio
* s/repeat repeat/repeat/
* Drop userio->serio checks, this struct is allocated right when we register the serio port so we don't need to worry about this ever being NULL
* Allocate a static minor and use that
* Add MODULE_ALIAS_MISCDEV() and MODULOE_ALIAS() lines to the module
* Improve the module description
* Change description for userio.h to "userio: virtual serio device support"
* Fix double space in comments in userio.h
* Change userio command types to enumerators

Responses
* Re: if (!userio) in userio_device_write(): it happens if we close the file descriptor while the input driver is trying to talk to the device, since we can't immediately bring down the driver. Removing the condition breaks the driver
* Re: ignoring trailing bytes: The main reason for that is if we need to add something else to the struct in the future. I doubt this will ever be done, but I figured I might as well let that behavior go since it can't harm anyone.
* Re: ignoring larger payloads + adding larger payloads: As of right now there

Quoted 8 lines

really isn't one. serio_send() only allows you to send a single character at a time, and we can add the ability to have flags using this same packet format. I don't think userio is ever going to get much more complex then this, since I don't think serio is going to be changing very much in the future. This being said, if some unlikely circumstances happen and it does happen to change in such a way we need to extend the payload, we're currently ignoring the rest of the payload so we could safely add additional data onto the end of the struct.

Please always respond inline to the previous emails. Otherwise, you break the thread and make it hard to track your responses. Nobody minds multiple emails.

Quoted 11 lines
 Documentation/input/userio.txt |  70 +++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 277 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 410 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h

[snip]

Quoted 51 lines
drivers/input/serio/userio.c new file +48 -0
@@ -0,0 +1,277 @@
+/*
+ * userio kernel serio device emulation module
+ * Copyright (C) 2015 Red Hat
+ * Copyright (C) 2015 Stephen Chandler Paul <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU Lesser 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 Lesser General Public License for more
+ * details.
+ */
+#include <linux/circ_buf.h>
+#include <linux/mutex.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/serio.h>
+#include <linux/slab.h>
+#include <linux/fs.h>
+#include <linux/miscdevice.h>
+#include <linux/sched.h>
+#include <linux/poll.h>
+#include <uapi/linux/userio.h>
+
+#define USERIO_NAME "userio"
+#define USERIO_BUFSIZE 16
+
+static struct miscdevice userio_misc;
+
+struct userio_device {
+       struct serio *serio;
+       struct mutex lock;
+
+       bool running;
+
+       u8 head;
+       u8 tail;
+
+       unsigned char buf[USERIO_BUFSIZE];
+
+       wait_queue_head_t waitq;
+};
+
+/**

drivers/input/serio/userio.c line 58

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+ * userio_device_write - Write data from serio to a userio device in userspace
+ * @id: The serio port for the userio device
+ * @val: The data to write to the device
+ */
+static int userio_device_write(struct serio *id, unsigned char val)
+{
+       struct userio_device *userio = id->port_data;
+
+       if (!userio)
+               return -1;

I still cannot see why this is necessary. Care to elaborate?

drivers/input/serio/userio.c line 69

Quoted 11 lines
drivers/input/serio/userio.c new file +11 -0
+
+       mutex_lock(&userio->lock);
+
+       userio->buf[userio->head] = val;
+       userio->head = (userio->head + 1) % USERIO_BUFSIZE;
+
+       if (userio->head == userio->tail)
+               dev_warn(userio_misc.this_device,
+                        "Buffer overflowed, userio client isn't keeping up");
+
+       mutex_unlock(&userio->lock);

As Dmitry mentioned, just add a spinlock to "struct userio_device" (supplementary to the mutex) which just protects buf, head and tail.

Quoted 57 lines
drivers/input/serio/userio.c new file +57 -0
+
+       wake_up_interruptible(&userio->waitq);
+
+       return 0;
+}
+
+static int userio_char_open(struct inode *inode, struct file *file)
+{
+       struct userio_device *userio;
+
+       userio = kzalloc(sizeof(struct userio_device), GFP_KERNEL);
+       if (!userio)
+               return -ENOMEM;
+
+       mutex_init(&userio->lock);
+       init_waitqueue_head(&userio->waitq);
+
+       userio->serio = kzalloc(sizeof(struct serio), GFP_KERNEL);
+       if (!userio->serio) {
+               kfree(userio);
+               return -ENOMEM;
+       }
+
+       userio->serio->write = userio_device_write;
+       userio->serio->port_data = userio;
+
+       file->private_data = userio;
+
+       return 0;
+}
+
+static int userio_char_release(struct inode *inode, struct file *file)
+{
+       struct userio_device *userio = file->private_data;
+
+       /* Don't free the serio port here, serio_unregister_port() does
+        * this for us */
+       if (userio->running)
+               serio_unregister_port(userio->serio);
+       else
+               kfree(userio->serio);
+
+       kfree(userio);
+
+       return 0;
+}
+
+static ssize_t userio_char_read(struct file *file, char __user *buffer,
+                               size_t count, loff_t *ppos)
+{
+       struct userio_device *userio = file->private_data;
+       int ret;
+       size_t nonwrap_len, copylen;
+
+       if (!count)
+               return 0;
+

drivers/input/serio/userio.c line 136

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+       /*
+        * By the time we get here, the data that was waiting might have been
+        * taken by another thread. Grab the mutex and check if there's still
+        * any data waiting, otherwise repeat repeat this process until we have
+        * data (unless the file descriptor is non-blocking of course)
+        */
+       for (;;) {
+               ret = mutex_lock_interruptible(&userio->lock);
+               if (ret)
+                       return ret;

Here it will be enough to take the spinlock rather than the mutex. The buffer is empty by default, so no reason to synchronize against write() calls.

Quoted 189 lines
drivers/input/serio/userio.c new file +141 -0
+
+               if (userio->head != userio->tail)
+                       break;
+
+               mutex_unlock(&userio->lock);
+
+               if (file->f_flags & O_NONBLOCK)
+                       return -EAGAIN;
+
+               ret = wait_event_interruptible(userio->waitq,
+                                              userio->head != userio->tail);
+               if (ret)
+                       return ret;
+       }
+
+       nonwrap_len = CIRC_CNT_TO_END(userio->head, userio->tail,
+                                     USERIO_BUFSIZE);
+       copylen = min(nonwrap_len, count);
+
+       if (copy_to_user(buffer, &userio->buf[userio->tail], copylen)) {
+               mutex_unlock(&userio->lock);
+               return -EFAULT;
+       }
+
+       userio->tail = (userio->tail + copylen) % USERIO_BUFSIZE;
+
+       mutex_unlock(&userio->lock);
+
+       return copylen;
+}
+
+static ssize_t userio_char_write(struct file *file, const char __user *buffer,
+                                size_t count, loff_t *ppos)
+{
+       struct userio_device *userio = file->private_data;
+       struct userio_cmd cmd;
+       int ret;
+
+       if (count < sizeof(cmd))
+               return -EINVAL;
+
+       if (copy_from_user(&cmd, buffer, sizeof(cmd)))
+               return -EFAULT;
+
+       ret = mutex_lock_interruptible(&userio->lock);
+       if (ret)
+               return ret;
+
+       switch (cmd.type) {
+       case USERIO_CMD_REGISTER:
+               if (!userio->serio->id.type) {
+                       dev_warn(userio_misc.this_device,
+                                "No port type given on /dev/userio\n");
+
+                       ret = -EINVAL;
+                       goto out;
+               }
+               if (userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "Begin command sent, but we're already running\n");
+
+                       ret = -EBUSY;
+                       goto out;
+               }
+
+               userio->running = true;
+               serio_register_port(userio->serio);
+               break;
+
+       case USERIO_CMD_SET_PORT_TYPE:
+               if (userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "Can't change port type on an already running userio instance\n");
+
+                       ret = -EBUSY;
+                       goto out;
+               }
+
+               userio->serio->id.type = cmd.data;
+               break;
+
+       case USERIO_CMD_SEND_INTERRUPT:
+               if (!userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "The device must be registered before sending interrupts\n");
+
+                       ret = -ENODEV;
+                       goto out;
+               }
+
+               serio_interrupt(userio->serio, cmd.data, 0);
+               break;
+
+       default:
+               ret = -EOPNOTSUPP;
+               goto out;
+       }
+
+       ret = sizeof(cmd);
+
+out:
+       mutex_unlock(&userio->lock);
+       return ret;
+}
+
+static unsigned int userio_char_poll(struct file *file, poll_table *wait)
+{
+       struct userio_device *userio = file->private_data;
+
+       poll_wait(file, &userio->waitq, wait);
+
+       if (userio->head != userio->tail)
+               return POLLIN | POLLRDNORM;
+
+       return 0;
+}
+
+static const struct file_operations userio_fops = {
+       .owner          = THIS_MODULE,
+       .open           = userio_char_open,
+       .release        = userio_char_release,
+       .read           = userio_char_read,
+       .write          = userio_char_write,
+       .poll           = userio_char_poll,
+       .llseek         = no_llseek,
+};
+
+static struct miscdevice userio_misc = {
+       .fops   = &userio_fops,
+       .minor  = USERIO_MINOR,
+       .name   = USERIO_NAME,
+};
+
+MODULE_ALIAS_MISCDEV(USERIO_MINOR);
+MODULE_ALIAS("devname:" USERIO_NAME);
+
+MODULE_AUTHOR("Stephen Chandler Paul <[email protected]>");
+MODULE_DESCRIPTION("Virtual Serio Device Support");
+MODULE_LICENSE("GPL");
+
+module_driver(userio_misc, misc_register, misc_deregister);
include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR                237
 #define VHOST_NET_MINOR                238
 #define UHID_MINOR             239
+#define USERIO_MINOR           240
 #define MISC_DYNAMIC_MINOR     255
 
 struct device;
include/uapi/linux/userio.h new file +30 -0
new file mode 100644
index 0000000..37935b6
@@ -0,0 +1,44 @@
+/*
+ * userio: virtual serio device support
+ * Copyright (C) 2015 Red Hat
+ * Copyright (C) 2015 Lyude (Stephen Chandler Paul) <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU Lesser 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 Lesser General Public License for more
+ * details.
+ *
+ * This is the public header used for user-space communication with the userio
+ * driver. __attribute__((__packed__)) is used for all structs to keep ABI
+ * compatibility between all architectures.
+ */
+
+#ifndef _USERIO_H
+#define _USERIO_H
+
+#include <linux/types.h>
+
+enum userio_cmd_type {
+       USERIO_CMD_REGISTER = 0,
+       USERIO_CMD_SET_PORT_TYPE = 1,
+       USERIO_CMD_SEND_INTERRUPT = 2
+};

include/uapi/linux/userio.h line 40

Quoted 10 lines
include/uapi/linux/userio.h new file +10 -0
+
+/*
+ * userio Commands
+ * All commands sent to /dev/userio are encoded using this structure. The type
+ * field should contain a USERIO_CMD* value that indicates what kind of command
+ * is being sent to userio. The data field should contain the accompanying
+ * argument for the command, if there is one.
+ */
+struct userio_cmd {
+       enum userio_cmd_type type;

Here I'd stick to "__u8". Sorry, that wasn't clear in my previous response. I just meant changing the definition to an enum.

Otherwise looks good to me.

Thanks
David

Quoted 7 lines
include/uapi/linux/userio.h new file +4 -1
+       __u8 data;
+} __attribute__((__packed__));
+
+#endif /* !_USERIO_H */
--

2.4.3

#3 lore raw

PATCH v6 1/1 Input: Add userio module

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes
* Change type in userio.h struct to __u8
* Add a spin lock that is used in addition to the mutex, no more completion struct
* Fix "repeat repeat" typo again (not sure how this got here?)
* Use USERIO_BUFSIZE for size of buf array in userio_char_read()
* Rename buffer in userio_char_read() to user_buffer (having both buffer and buf could be confusing)

 Documentation/input/userio.txt |  70 ++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 289 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 422 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 274f854..44740de 100644
@@ -11086,6 +11086,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/ps2emu.c
+F:	include/uapi/linux/ps2emu.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	[email protected]
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +289 -0

289 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
#4 lore raw

PATCH 1/1 Input: Add userio module

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
 Documentation/input/userio.txt |  70 ++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 289 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 422 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h

Changes
* Fixed the files listed in MAINTAINERS, surprised no one (myself included) noticed this earlier.

Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 274f854..7613654 100644
@@ -11086,6 +11086,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/userio.c
+F:	include/uapi/linux/userio.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	virtualization-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +289 -0

289 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
maintainer
#5 lore raw

drivers/input/serio/userio.c line 60

Quoted 7 lines · On Wed, Sep 23, 2015 at 01:54:59PM -0400, [email protected] wrote:
drivers/input/serio/userio.c new file +7 -0
+static int userio_device_write(struct serio *id, unsigned char val)
+{
+	struct userio_device *userio = id->port_data;
+	unsigned long flags;
+
+	if (!userio)
+		return -1;

I still have the same question: how can this happen? Where do we reset port data to NULL? What happens if we set to NULL after checking?

Quoted 6 lines
drivers/input/serio/userio.c new file +6 -0
+
+	spin_lock_irqsave(&userio->buf_lock, flags);
+
+	userio->buf[userio->head] = val;
+	userio->head = (userio->head + 1) % USERIO_BUFSIZE;
+

drivers/input/serio/userio.c line 76

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+	if (userio->head == userio->tail)
+		dev_warn(userio_misc.this_device,
+			 "Buffer overflowed, userio client isn't keeping up");
+
+	spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+	wake_up_interruptible(&userio->waitq);
+
+	return 0;
+}

Thanks.

Signature · 1 line
Dmitry
#6 lore raw
Quoted 14 lines · Dmitry Torokhov

drivers/input/serio/userio.c line 60

Quoted 8 lines · On Wed, Sep 23, 2015 at 01:54:59PM -0400…

+static int userio_device_write(struct serio *id, unsigned char val)

drivers/input/serio/userio.c new file +6 -0
+{
+	struct userio_device *userio = id->port_data;
+	unsigned long flags;
+
+	if (!userio)
+		return -1;

I still have the same question: how can this happen? Where do we reset
port data to NULL? What happens if we set to NULL after checking?

Actually David pointed out this line was not needed, somehow it still got left in the patch. I'll send an updated version in just a minute with the changes.

Cheers,
Lyude

Quoted 20 lines
Quoted 10 lines
drivers/input/serio/userio.c new file +9 -0
+
+	spin_lock_irqsave(&userio->buf_lock, flags);
+
+	userio->buf[userio->head] = val;
+	userio->head = (userio->head + 1) % USERIO_BUFSIZE;
+
+	if (userio->head == userio->tail)
+		dev_warn(userio_misc.this_device,
+			 "Buffer overflowed, userio client isn't

keeping up");

drivers/input/serio/userio.c line 76

Quoted 7 lines
drivers/input/serio/userio.c new file +7 -0
+
+	spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+	wake_up_interruptible(&userio->waitq);
+
+	return 0;
+}

Thanks.

#7 lore raw

PATCH 1/1 Input: Add userio module

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes
* Remove the if (!userio) { return -1; } check that somehow got left in.

Sorry this took so long! I was wondering why you hadn't replied yet, only to notice I only made this change on my own tree and never sent out a response patch. Oops.

 Documentation/input/userio.txt |  70 +++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 279 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 412 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 797236b..ba59bd7 100644
@@ -11100,6 +11100,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/userio.c
+F:	include/uapi/linux/userio.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	[email protected]
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +279 -0

279 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
maintainer
#8 lore raw

Hi Stephen,

Quoted 14 lines · On Fri, Oct 23, 2015 at 04:47:46PM -0400, [email protected] wrote:

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

Quoted 7 lines

---
Changes
* Remove the if (!userio) { return -1; } check that somehow got left in.

Sorry this took so long! I was wondering why you hadn't replied yet, only to notice I only made this change on my own tree and never sent out a response patch. Oops.

Thank you for making all the changes.

Review

Quoted 12 lines
(quoted diff) +12 -0
+
+static ssize_t userio_char_read(struct file *file, char __user *user_buffer,
+				size_t count, loff_t *ppos)
+{
+	struct userio_device *userio = file->private_data;
+	int ret;
+	size_t nonwrap_len, copylen;
+	unsigned char buf[USERIO_BUFSIZE];
+	unsigned long flags;
+
+	if (!count)
+		return 0;

This is not quite right: read of size 0 should still check if there is data in the buffer and return -EAGAIN for non-blocking descriptors.

I cooked a patch (below) that should adjust the read behavior (_+ a coupe of formatting changes), please take a look.

Thanks!

Signature · 11 lines
Dmitry

userio - misc fixups

From: Dmitry Torokhov <[email protected]>

Signed-off-by: Dmitry Torokhov <[email protected]>
---
 drivers/input/serio/Kconfig  |    9 ++-
 drivers/input/serio/userio.c |  122 ++++++++++++++++++++++--------------------
 2 files changed, 70 insertions(+), 61 deletions(-)
drivers/input/serio/Kconfig +6 -3
index 22b8b58..c3d05b4 100644
@@ -293,10 +293,13 @@ config SERIO_SUN4I_PS2
 	  module will be called sun4i-ps2.
 
 config USERIO
-	tristate "Virtual serio device support"
+	tristate "User space serio port driver support"
 	help
-	  Say Y here if you want to emulate serio devices using the userio
-	  kernel module.
+	  Say Y here if you want to support user level drivers for serio
+	  subsystem accessible under char device 10:240 - /dev/userio. Using
+	  this facility userspace programs can implement serio ports that
+	  will be used by the standard in-kernel serio consumer drivers,
+	  such as psmouse and atkbd.
 
 	  To compile this driver as a module, choose M here: the module will be
 	  called userio.
drivers/input/serio/userio.c +64 -58

231 lines not on this page · raw message

author
#9 lore raw

Hi! So, I'm guessing you want me to merge this with the patch I've got right now and post the new version? (Just making sure since I'm a bit new to this :)

Quoted 60 lines · Dmitry Torokhov

Hi Stephen,

Quoted 26 lines · On Fri, Oct 23, 2015 at 04:47:46PM -0400, [email protected] wrote:

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky
without having the physical device handy. Here we try to remedy that
with userio. This module allows an application to connect to a character
device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and
replay them through the /dev/userio device, this allows developers to
debug driver issues on the PS/2 level with devices simply by requesting
a recording from the user experiencing the issue without having to have
the physical hardware in front of them.

---
Changes
* Remove the if (!userio) { return -1; } check that somehow got left in.

Quoted 5 lines

Sorry this took so long! I was wondering why you hadn't replied yet, only to
notice I only made this change on my own tree and never sent out a response
patch. Oops.

Thank you for making all the changes.

Review

Quoted 13 lines
(quoted diff) +2 -0
+
+static ssize_t userio_char_read(struct file *file, char __user

*user_buffer,

(quoted diff) +10 -0
+				size_t count, loff_t *ppos)
+{
+	struct userio_device *userio = file->private_data;
+	int ret;
+	size_t nonwrap_len, copylen;
+	unsigned char buf[USERIO_BUFSIZE];
+	unsigned long flags;
+
+	if (!count)
+		return 0;

This is not quite right: read of size 0 should still check if there is
data in the buffer and return -EAGAIN for non-blocking descriptors.

I cooked a patch (below) that should adjust the read behavior (_+ a coupe of formatting changes), please take a look.

Thanks!

Signature · 2 lines
Cheers,
	Lyude
maintainer
#10 lore raw

Hi,

Quoted 4 lines · Lyude

Hi! So, I'm guessing you want me to merge this with the patch I've got right now and post the new version? (Just making sure since I'm a bit new to this :)

If you apply my patch on top of yours and it still works you do not need to resend anything. If I messed up and you need to fix something ideally just send me an incremental patch so that I can see where exactly I screwed up and I'll fold it all together.

Thanks.

Signature · 1 line
Dmitry
author
#11 lore raw

Gotcha, tested it and took a look at the changes, looks good to me.

Quoted 15 lines · Dmitry Torokhov

Hi,

Quoted 4 lines · Lyude

Hi! So, I'm guessing you want me to merge this with the patch I've got right now and post the new version? (Just making sure since I'm a bit new to this :)

If you apply my patch on top of yours and it still works you do not need to resend anything. If I messed up and you need to fix something ideally just send me an incremental patch so that I can see where exactly I screwed up and I'll fold it all together.

Thanks.

Signature · 2 lines
Cheers,
	Lyude
maintainer
#12 lore raw
Quoted 1 line · Lyude

Gotcha, tested it and took a look at the changes, looks good to me.

Excellent, thank you, applied.

Quoted 21 lines · Dmitry Torokhov
Quoted 15 lines

Hi,

Quoted 4 lines · Lyude

Hi! So, I'm guessing you want me to merge this with the patch I've got right now and post the new version? (Just making sure since I'm a bit new to this :)

If you apply my patch on top of yours and it still works you do not need to resend anything. If I messed up and you need to fix something ideally just send me an incremental patch so that I can see where exactly I screwed up and I'll fold it all together.

Thanks.

Signature · 2 lines
Cheers,
	Lyude
Signature · 1 line
Dmitry
#13 lore raw

PATCH v7 1/1 Input: Add userio module

In reply to Dmitry Torokhov

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes
* Removed useless if (!userio) check in userio_device_write()

 Documentation/input/userio.txt |  70 ++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 286 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 419 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 797236b..ba59bd7 100644
@@ -11100,6 +11100,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/userio.c
+F:	include/uapi/linux/userio.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	[email protected]
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +286 -0

286 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
Reviewed-by on v7 1/1
#14 lore raw

Hi

Quoted 12 lines · On Mon, Oct 5, 2015 at 5:55 PM, <[email protected]> wrote:

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

Quoted 1 line

Apart from the serio-unregistration, all I have left is bike-shedding. Regardless of those changes, this is:

See below for some details..

Quoted 149 lines

---
Changes
* Removed useless if (!userio) check in userio_device_write()

 Documentation/input/userio.txt |  70 ++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 286 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 419 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+                             The userio Protocol
+            (c) 2015 Stephen Chandler Paul <[email protected]>
+                             Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+       struct userio_cmd {
+               __u8 type;
+               __u8 data;
+       };
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+       https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 797236b..ba59bd7 100644
@@ -11100,6 +11100,12 @@ S:     Maintained
 F:     drivers/media/v4l2-core/videobuf2-*
 F:     include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:     Stephen Chandler Paul <[email protected]>
+S:     Maintained
+F:     drivers/input/serio/userio.c
+F:     include/uapi/linux/userio.h
+
 VIRTIO CONSOLE DRIVER
 M:     Amit Shah <[email protected]>
 L:     [email protected]
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
          To compile this driver as a module, choose M here: the
          module will be called sun4i-ps2.
 
+config USERIO
+       tristate "Virtual serio device support"
+       help
+         Say Y here if you want to emulate serio devices using the userio
+         kernel module.
+
+         To compile this driver as a module, choose M here: the module will be
+         called userio.
+
+         If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)    += apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)  += olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)  += hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)  += sun4i-ps2.o
+obj-$(CONFIG_USERIO)           += userio.o
drivers/input/serio/userio.c new file +5 -0
new file mode 100644
index 0000000..0ada7cf
@@ -0,0 +1,286 @@
+/*
+ * userio kernel serio device emulation module
+ * Copyright (C) 2015 Red Hat
+ * Copyright (C) 2015 Stephen Chandler Paul <[email protected]>
+ *

drivers/input/serio/userio.c line 15

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU Lesser 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 Lesser General Public License for more
+ * details.
+ */

We put an empty line here, usually.

Quoted 87 lines
drivers/input/serio/userio.c new file +87 -0
+#include <linux/circ_buf.h>
+#include <linux/mutex.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/serio.h>
+#include <linux/slab.h>
+#include <linux/fs.h>
+#include <linux/miscdevice.h>
+#include <linux/sched.h>
+#include <linux/poll.h>
+#include <uapi/linux/userio.h>
+
+#define USERIO_NAME "userio"
+#define USERIO_BUFSIZE 16
+
+static struct miscdevice userio_misc;
+
+struct userio_device {
+       struct serio *serio;
+       struct mutex lock;
+
+       bool running;
+
+       u8 head;
+       u8 tail;
+
+       spinlock_t buf_lock;
+       unsigned char buf[USERIO_BUFSIZE];
+
+       wait_queue_head_t waitq;
+};
+
+/**
+ * userio_device_write - Write data from serio to a userio device in userspace
+ * @id: The serio port for the userio device
+ * @val: The data to write to the device
+ */
+static int userio_device_write(struct serio *id, unsigned char val)
+{
+       struct userio_device *userio = id->port_data;
+       unsigned long flags;
+
+       spin_lock_irqsave(&userio->buf_lock, flags);
+
+       userio->buf[userio->head] = val;
+       userio->head = (userio->head + 1) % USERIO_BUFSIZE;
+
+       if (userio->head == userio->tail)
+               dev_warn(userio_misc.this_device,
+                        "Buffer overflowed, userio client isn't keeping up");
+
+       spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+       wake_up_interruptible(&userio->waitq);
+
+       return 0;
+}
+
+static int userio_char_open(struct inode *inode, struct file *file)
+{
+       struct userio_device *userio;
+
+       userio = kzalloc(sizeof(struct userio_device), GFP_KERNEL);
+       if (!userio)
+               return -ENOMEM;
+
+       mutex_init(&userio->lock);
+       spin_lock_init(&userio->buf_lock);
+       init_waitqueue_head(&userio->waitq);
+
+       userio->serio = kzalloc(sizeof(struct serio), GFP_KERNEL);
+       if (!userio->serio) {
+               kfree(userio);
+               return -ENOMEM;
+       }
+
+       userio->serio->write = userio_device_write;
+       userio->serio->port_data = userio;
+
+       file->private_data = userio;
+
+       return 0;
+}
+
+static int userio_char_release(struct inode *inode, struct file *file)
+{

drivers/input/serio/userio.c line 112

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+       struct userio_device *userio = file->private_data;
+
+       /* Don't free the serio port here, serio_unregister_port() does
+        * this for us */
+       if (userio->running)
+               serio_unregister_port(userio->serio);
+       else
+               kfree(userio->serio);
+
+       kfree(userio);

I'm not very familiar with the serio-subsystem, but is there a guarantee that .write() will not be called once unregistered? I can see that it's not scheduled by userio anymore, but maybe there's an async path that can trigger .write() callbacks. Dunno.. I'll leave that to Dmitry.

Quoted 9 lines
drivers/input/serio/userio.c new file +9 -0
+
+       return 0;
+}
+
+static ssize_t userio_char_read(struct file *file, char __user *user_buffer,
+                               size_t count, loff_t *ppos)
+{
+       struct userio_device *userio = file->private_data;
+       int ret;

drivers/input/serio/userio.c line 131

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+       size_t nonwrap_len, copylen;
+       unsigned char buf[USERIO_BUFSIZE];
+       unsigned long flags;
+
+       if (!count)
+               return 0;
+
+       ret = mutex_lock_interruptible(&userio->lock);
+       if (ret)
+               return ret;

I cannot see why you need to lock the mutex here. The spin-lock should be perfectly fine, given that you don't check the device-state, anyway. Also, given that you don't have a shutdown-commands, except for close(), I don't see why we'd ever need the mutex-lock here, ever.

Quoted 26 lines
drivers/input/serio/userio.c new file +26 -0
+
+       /*
+        * By the time we get here, the data that was waiting might have been
+        * taken by another thread. Grab the mutex and check if there's still
+        * any data waiting, otherwise repeat this process until we have data
+        * (unless the file descriptor is non-blocking of course)
+        */
+       for (;;) {
+               spin_lock_irqsave(&userio->buf_lock, flags);
+
+               if (userio->head != userio->tail)
+                       break;
+
+               spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+               if (file->f_flags & O_NONBLOCK) {
+                       ret = -EAGAIN;
+                       goto out;
+               }
+
+               ret = wait_event_interruptible(userio->waitq,
+                                              userio->head != userio->tail);
+               if (ret)
+                       goto out;
+       }
+

drivers/input/serio/userio.c line 167

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+       nonwrap_len = CIRC_CNT_TO_END(userio->head, userio->tail,
+                                     USERIO_BUFSIZE);
+       copylen = min(nonwrap_len, count);
+       memcpy(&buf, &userio->buf[userio->tail], copylen);
+
+       userio->tail = (userio->tail + copylen) % USERIO_BUFSIZE;
+
+       spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+       if (copy_to_user(user_buffer, &buf, copylen))

No need for '&' before 'buf'.

Quoted 11 lines
drivers/input/serio/userio.c new file +11 -0
+               ret = -EFAULT;
+       else
+               ret = copylen;
+
+out:
+       mutex_unlock(&userio->lock);
+       return ret;
+}
+
+static ssize_t userio_char_write(struct file *file, const char __user *buffer,
+                                size_t count, loff_t *ppos)

drivers/input/serio/userio.c line 188

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+{
+       struct userio_device *userio = file->private_data;
+       struct userio_cmd cmd;
+       int ret;
+
+       if (count < sizeof(cmd))
+               return -EINVAL;
+
+       if (copy_from_user(&cmd, buffer, sizeof(cmd)))
+               return -EFAULT;

I know we discussed that before, but I'm still slightly uncomfortable that we ignore trailing bytes. How about adding a check to each case below? (see below for an example)

drivers/input/serio/userio.c line 195

Quoted 7 lines
drivers/input/serio/userio.c new file +7 -0
+
+       ret = mutex_lock_interruptible(&userio->lock);
+       if (ret)
+               return ret;
+
+       switch (cmd.type) {
+       case USERIO_CMD_REGISTER:
                if (count != sizeof(cmd)) {
                        ret = -EINVAL;
                        goto out;
                }

And same for the other 2 commands. This guarantees that each cmd gets the correct payload size.

Quoted 143 lines
drivers/input/serio/userio.c new file +91 -0
+               if (!userio->serio->id.type) {
+                       dev_warn(userio_misc.this_device,
+                                "No port type given on /dev/userio\n");
+
+                       ret = -EINVAL;
+                       goto out;
+               }
+               if (userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "Begin command sent, but we're already running\n");
+
+                       ret = -EBUSY;
+                       goto out;
+               }
+
+               userio->running = true;
+               serio_register_port(userio->serio);
+               break;
+
+       case USERIO_CMD_SET_PORT_TYPE:
+               if (userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "Can't change port type on an already running userio instance\n");
+
+                       ret = -EBUSY;
+                       goto out;
+               }
+
+               userio->serio->id.type = cmd.data;
+               break;
+
+       case USERIO_CMD_SEND_INTERRUPT:
+               if (!userio->running) {
+                       dev_warn(userio_misc.this_device,
+                                "The device must be registered before sending interrupts\n");
+
+                       ret = -ENODEV;
+                       goto out;
+               }
+
+               serio_interrupt(userio->serio, cmd.data, 0);
+               break;
+
+       default:
+               ret = -EOPNOTSUPP;
+               goto out;
+       }
+
+       ret = sizeof(cmd);
+
+out:
+       mutex_unlock(&userio->lock);
+       return ret;
+}
+
+static unsigned int userio_char_poll(struct file *file, poll_table *wait)
+{
+       struct userio_device *userio = file->private_data;
+
+       poll_wait(file, &userio->waitq, wait);
+
+       if (userio->head != userio->tail)
+               return POLLIN | POLLRDNORM;
+
+       return 0;
+}
+
+static const struct file_operations userio_fops = {
+       .owner          = THIS_MODULE,
+       .open           = userio_char_open,
+       .release        = userio_char_release,
+       .read           = userio_char_read,
+       .write          = userio_char_write,
+       .poll           = userio_char_poll,
+       .llseek         = no_llseek,
+};
+
+static struct miscdevice userio_misc = {
+       .fops   = &userio_fops,
+       .minor  = USERIO_MINOR,
+       .name   = USERIO_NAME,
+};
+
+MODULE_ALIAS_MISCDEV(USERIO_MINOR);
+MODULE_ALIAS("devname:" USERIO_NAME);
+
+MODULE_AUTHOR("Stephen Chandler Paul <[email protected]>");
+MODULE_DESCRIPTION("Virtual Serio Device Support");
+MODULE_LICENSE("GPL");
+
+module_driver(userio_misc, misc_register, misc_deregister);
include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR                237
 #define VHOST_NET_MINOR                238
 #define UHID_MINOR             239
+#define USERIO_MINOR           240
 #define MISC_DYNAMIC_MINOR     255
 
 struct device;
include/uapi/linux/userio.h new file +34 -0
new file mode 100644
index 0000000..37d147f
@@ -0,0 +1,44 @@
+/*
+ * userio: virtual serio device support
+ * Copyright (C) 2015 Red Hat
+ * Copyright (C) 2015 Lyude (Stephen Chandler Paul) <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU Lesser 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 Lesser General Public License for more
+ * details.
+ *
+ * This is the public header used for user-space communication with the userio
+ * driver. __attribute__((__packed__)) is used for all structs to keep ABI
+ * compatibility between all architectures.
+ */
+
+#ifndef _USERIO_H
+#define _USERIO_H
+
+#include <linux/types.h>
+
+enum userio_cmd_type {
+       USERIO_CMD_REGISTER = 0,
+       USERIO_CMD_SET_PORT_TYPE = 1,
+       USERIO_CMD_SEND_INTERRUPT = 2
+};
+
+/*
+ * userio Commands
+ * All commands sent to /dev/userio are encoded using this structure. The type

include/uapi/linux/userio.h line 44

Quoted 10 lines
include/uapi/linux/userio.h new file +10 -0
+ * field should contain a USERIO_CMD* value that indicates what kind of command
+ * is being sent to userio. The data field should contain the accompanying
+ * argument for the command, if there is one.
+ */
+struct userio_cmd {
+       __u8 type;
+       __u8 data;
+} __attribute__((__packed__));
+
+#endif /* !_USERIO_H */

Otherwise, looks all good.

Thanks
David

#15 lore raw

PATCH v8 1/1 Input: Add userio module

From: Stephen Chandler Paul <[email protected]>

Debugging input devices, specifically laptop touchpads, can be tricky without having the physical device handy. Here we try to remedy that with userio. This module allows an application to connect to a character device provided by the kernel, and emulate any serio device. In combination with userspace programs that can record PS/2 devices and replay them through the /dev/userio device, this allows developers to debug driver issues on the PS/2 level with devices simply by requesting a recording from the user experiencing the issue without having to have the physical hardware in front of them.

---
Changes
* Add empty line under GPL notice
* Remove mutex lock in userio_char_read()
* Also remove the goto in userio_char_read(), we only needed it before to ensure the mutex got unlocked at the end of the function
* Remove & in front of buf in userio_char_read()
* Return error if count != sizeof(cmd)
(changed the top of the function instead of the change you suggested, does
the same thing with fewer lines)
* Added a warning message if we get an invalid payload header. We do this for everything else but segfaults so I figured it would be appropriate

Re: write() after serio; originally I thought that there was no guarantee of this (hence the ineffective if (!userio) checks before), however going off Dmitry I think serio_unregister_port() gives us the guarantee this will never happen. Of course it's up to him to clarify this.

 Documentation/input/userio.txt |  70 +++++++++++
 MAINTAINERS                    |   6 +
 drivers/input/serio/Kconfig    |  11 ++
 drivers/input/serio/Makefile   |   1 +
 drivers/input/serio/userio.c   | 279 +++++++++++++++++++++++++++++++++++++++++
 include/linux/miscdevice.h     |   1 +
 include/uapi/linux/userio.h    |  44 +++++++
 7 files changed, 412 insertions(+)
 create mode 100644 Documentation/input/userio.txt
 create mode 100644 drivers/input/serio/userio.c
 create mode 100644 include/uapi/linux/userio.h
Documentation/input/userio.txt new file +70 -0
new file mode 100644
index 0000000..0880c0f
@@ -0,0 +1,70 @@
+			      The userio Protocol
+	     (c) 2015 Stephen Chandler Paul <[email protected]>
+			      Sponsored by Red Hat
+--------------------------------------------------------------------------------
+
+1. Introduction
+~~~~~~~~~~~~~~~
+  This module is intended to try to make the lives of input driver developers
+easier by allowing them to test various serio devices (mainly the various
+touchpads found on laptops) without having to have the physical device in front
+of them. userio accomplishes this by allowing any privileged userspace program
+to directly interact with the kernel's serio driver and control a virtual serio
+port from there.
+
+2. Usage overview
+~~~~~~~~~~~~~~~~~
+  In order to interact with the userio kernel module, one simply opens the
+/dev/userio character device in their applications. Commands are sent to the
+kernel module by writing to the device, and any data received from the serio
+driver is read as-is from the /dev/userio device. All of the structures and
+macros you need to interact with the device are defined in <linux/userio.h> and
+<linux/serio.h>.
+
+3. Command Structure
+~~~~~~~~~~~~~~~~~~~~
+  The struct used for sending commands to /dev/userio is as follows:
+
+	struct userio_cmd {
+		__u8 type;
+		__u8 data;
+	};
+
+  "type" describes the type of command that is being sent. This can be any one
+of the USERIO_CMD macros defined in <linux/userio.h>. "data" is the argument
+that goes along with the command. In the event that the command doesn't have an
+argument, this field can be left untouched and will be ignored by the kernel.
+Each command should be sent by writing the struct directly to the character
+device. In the event that the command you send is invalid, an error will be
+returned by the character device and a more descriptive error will be printed
+to the kernel log. Only one command can be sent at a time, any additional data
+written to the character device after the initial command will be ignored.
+  To close the virtual serio port, just close /dev/userio.
+
+4. Commands
+~~~~~~~~~~~
+
+4.1 USERIO_CMD_REGISTER
+~~~~~~~~~~~~~~~~~~~~~~~
+  Registers the port with the serio driver and begins transmitting data back and
+forth. Registration can only be performed once a port type is set with
+USERIO_CMD_SET_PORT_TYPE. Has no argument.
+
+4.2 USERIO_CMD_SET_PORT_TYPE
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sets the type of port we're emulating, where "data" is the port type being
+set. Can be any of the macros from <linux/serio.h>. For example: SERIO_8042
+would set the port type to be a normal PS/2 port.
+
+4.3 USERIO_CMD_SEND_INTERRUPT
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+  Sends an interrupt through the virtual serio port to the serio driver, where
+"data" is the interrupt data being sent.
+
+5. Userspace tools
+~~~~~~~~~~~~~~~~~~
+  The userio userspace tools are able to record PS/2 devices using some of the
+debugging information from i8042, and play back the devices on /dev/userio. The
+latest version of these tools can be found at:
+
+	https://github.com/Lyude/ps2emu
MAINTAINERS +6 -0
index 797236b..ba59bd7 100644
@@ -11100,6 +11100,12 @@ S:	Maintained
 F:	drivers/media/v4l2-core/videobuf2-*
 F:	include/media/videobuf2-*
 
+VIRTUAL SERIO DEVICE DRIVER
+M:	Stephen Chandler Paul <[email protected]>
+S:	Maintained
+F:	drivers/input/serio/userio.c
+F:	include/uapi/linux/userio.h
+
 VIRTIO CONSOLE DRIVER
 M:	Amit Shah <[email protected]>
 L:	virtualization-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org
drivers/input/serio/Kconfig +11 -0
index 200841b..22b8b58 100644
@@ -292,4 +292,15 @@ config SERIO_SUN4I_PS2
 	  To compile this driver as a module, choose M here: the
 	  module will be called sun4i-ps2.
 
+config USERIO
+	tristate "Virtual serio device support"
+	help
+	  Say Y here if you want to emulate serio devices using the userio
+	  kernel module.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called userio.
+
+	  If you are unsure, say N.
+
 endif
drivers/input/serio/Makefile +1 -0
index c600089..2374ef9 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_SERIO_APBPS2)	+= apbps2.o
 obj-$(CONFIG_SERIO_OLPC_APSP)	+= olpc_apsp.o
 obj-$(CONFIG_HYPERV_KEYBOARD)	+= hyperv-keyboard.o
 obj-$(CONFIG_SERIO_SUN4I_PS2)	+= sun4i-ps2.o
+obj-$(CONFIG_USERIO)		+= userio.o
drivers/input/serio/userio.c new file +279 -0

279 lines not on this page · raw message

include/linux/miscdevice.h +1 -0
index 81f6e42..5430374 100644
@@ -49,6 +49,7 @@
 #define LOOP_CTRL_MINOR		237
 #define VHOST_NET_MINOR		238
 #define UHID_MINOR		239
+#define USERIO_MINOR		240
 #define MISC_DYNAMIC_MINOR	255
 
 struct device;
include/uapi/linux/userio.h new file +44 -0

44 lines not on this page · raw message

Signature · 1 line
2.4.3
#16 lore raw
Quoted 41 lines · On Fri, Oct 09, 2015 at 10:30:53AM -0400, [email protected] wrote:

From: Stephen Chandler Paul <[email protected]>

drivers/input/serio/userio.c new file +40 -0
+
+static ssize_t userio_char_read(struct file *file, char __user *user_buffer,
+				size_t count, loff_t *ppos)
+{
+	struct userio_device *userio = file->private_data;
+	int ret;
+	size_t nonwrap_len, copylen;
+	unsigned char buf[USERIO_BUFSIZE];
+	unsigned long flags;
+
+	if (!count)
+		return 0;
+
+	/*
+	 * By the time we get here, the data that was waiting might have been
+	 * taken by another thread. Grab the mutex and check if there's still
+	 * any data waiting, otherwise repeat this process until we have data
+	 * (unless the file descriptor is non-blocking of course)
+	 */
+	for (;;) {
+		spin_lock_irqsave(&userio->buf_lock, flags);
+
+		if (userio->head != userio->tail)
+			break;
+
+		spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+		if (file->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		ret = wait_event_interruptible(userio->waitq,
+					       userio->head != userio->tail);
+		if (ret)
+			return ret;
+	}
+
+	nonwrap_len = CIRC_CNT_TO_END(userio->head, userio->tail,
+				      USERIO_BUFSIZE);
+	copylen = min(nonwrap_len, count);
+	memcpy(buf, &userio->buf[userio->tail], copylen);

drivers/input/serio/userio.c line 166

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+
+	userio->tail = (userio->tail + copylen) % USERIO_BUFSIZE;
+
+	spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+	if (copy_to_user(user_buffer, buf, copylen))
+		return -EFAULT;
+
+	return copylen;
+}

Hello

I think you could simplify by just return copy_to_user() since it return copylen in case of success.

Regards

#17 lore raw
Quoted 60 lines · LABBE Corentin
Quoted 46 lines · On Fri, Oct 09, 2015 at 10:30:53AM -0400…

From: Stephen Chandler Paul <[email protected]>

(quoted diff) +2 -0
+
+static ssize_t userio_char_read(struct file *file, char __user

*user_buffer,

drivers/input/serio/userio.c new file +13 -0
+				size_t count, loff_t *ppos)
+{
+	struct userio_device *userio = file->private_data;
+	int ret;
+	size_t nonwrap_len, copylen;
+	unsigned char buf[USERIO_BUFSIZE];
+	unsigned long flags;
+
+	if (!count)
+		return 0;
+
+	/*
+	 * By the time we get here, the data that was waiting

might have been

(quoted diff) +1 -0
+	 * taken by another thread. Grab the mutex and check if

there's still

(quoted diff) +1 -0
+	 * any data waiting, otherwise repeat this process until

we have data

drivers/input/serio/userio.c new file +15 -0
+	 * (unless the file descriptor is non-blocking of course)
+	 */
+	for (;;) {
+		spin_lock_irqsave(&userio->buf_lock, flags);
+
+		if (userio->head != userio->tail)
+			break;
+
+		spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+		if (file->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		ret = wait_event_interruptible(userio->waitq,
+					       userio->head !=

userio->tail);

drivers/input/serio/userio.c new file +8 -0
+		if (ret)
+			return ret;
+	}
+
+	nonwrap_len = CIRC_CNT_TO_END(userio->head, userio->tail,
+				      USERIO_BUFSIZE);
+	copylen = min(nonwrap_len, count);
+	memcpy(buf, &userio->buf[userio->tail], copylen);

drivers/input/serio/userio.c line 166

Quoted 10 lines
drivers/input/serio/userio.c new file +10 -0
+
+	userio->tail = (userio->tail + copylen) % USERIO_BUFSIZE;
+
+	spin_unlock_irqrestore(&userio->buf_lock, flags);
+
+	if (copy_to_user(user_buffer, buf, copylen))
+		return -EFAULT;
+
+	return copylen;
+}

Hello

Quoted 4 lines

I think you could simplify by just return copy_to_user() since it return copylen in case of success.

Regards

Hi! You're mistaken here actually:
https://www.fsl.cs.sunysb.edu/kernel-api/re256.html
"Returns number of bytes that could not be copied. On success, this will be zero". Plus, we need to check if the copy failed or not since if it did we have to indicate -EFAULT to signal that the memory address the user's buffer is pointed to isn't actually accessible.

Cheers,
Lyude

maintainer
#18 lore raw
In reply to David Herrmann
Quoted 11 lines · David Herrmann
Quoted 10 lines · On Mon, Oct 5, 2015 at 5:55 PM, <[email protected]> wrote:
drivers/input/serio/userio.c new file +10 -0
+static int userio_char_release(struct inode *inode, struct file *file)
+{
+       struct userio_device *userio = file->private_data;
+
+       /* Don't free the serio port here, serio_unregister_port() does
+        * this for us */
+       if (userio->running)
+               serio_unregister_port(userio->serio);
+       else
+               kfree(userio->serio);
Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+       kfree(userio);

I'm not very familiar with the serio-subsystem, but is there a guarantee that .write() will not be called once unregistered? I can see that it's not scheduled by userio anymore, but maybe there's an async path that can trigger .write() callbacks. Dunno.. I'll leave that to Dmitry.

I think this is OK. When port is unregistered we the driver will be disconnected from it, and is not supposed to call serio_write() anymore. In fact, it should not access port after calling serio_close(). If it keeps the pointer and uses it is a bug in the other driver; we can't really do much about it.

Thanks.

Signature · 1 line
Dmitry