From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: Andrey Konovalov <andreyknvl@google.com>
Cc: linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org,
Alan Stern <stern@rowland.harvard.edu>,
Jonathan Corbet <corbet@lwn.net>, Felipe Balbi <balbi@kernel.org>,
Dmitry Vyukov <dvyukov@google.com>,
Alexander Potapenko <glider@google.com>,
Marco Elver <elver@google.com>
Subject: Re: [PATCH 1/1] usb: gadget: add raw-gadget interface
Date: Fri, 8 Nov 2019 22:17:45 +0100 [thread overview]
Message-ID: <20191108211745.GA1282512@kroah.com> (raw)
In-Reply-To: <282c5da077ad53ce4e5ff9b4350bbf62b33bb6a9.1573236684.git.andreyknvl@google.com>
On Fri, Nov 08, 2019 at 07:26:55PM +0100, Andrey Konovalov wrote:
> USB Raw Gadget is a kernel module that provides a userspace interface for
> the USB Gadget subsystem. Essentially it allows to emulate USB devices
> from userspace. Enabled with CONFIG_USB_RAW_GADGET. Raw Gadget is
> currently a strictly debugging feature and shouldn't be used in
> production.
>
> Raw Gadget is similar to GadgetFS, but provides a more low-level and
> direct access to the USB Gadget layer for the userspace. The key
> differences are:
>
> 1. Every USB request is passed to the userspace to get a response, while
> GadgetFS responds to some USB requests internally based on the provided
> descriptors. However note, that the UDC driver might respond to some
> requests on its own and never forward them to the Gadget layer.
>
> 2. GadgetFS performs some sanity checks on the provided USB descriptors,
> while Raw Gadget allows you to provide arbitrary data as responses to
> USB requests.
>
> 3. Raw Gadget provides a way to select a UDC device/driver to bind to,
> while GadgetFS currently binds to the first available UDC.
>
> 4. Raw Gadget uses predictable endpoint names (handles) across different
> UDCs (as long as UDCs have enough endpoints of each required transfer
> type).
>
> 5. Raw Gadget has ioctl-based interface instead of a filesystem-based one.
I like the idea, and you've shown that you can do neat stuff with this,
so that's a good proof-of-concept.
But, at the least, use a "real" char device node for this, not debugfs.
Switch this to a misc device and I'll be glad to review it.
Some minor nits that jump out at me other than debugfs:
> +static void raw_event_queue_destroy(struct raw_event_queue *queue)
> +{
> + int i;
> +
> + for (i = 0; i < queue->size; i++) {
> + pr_debug("freeing event[%d] = %px\n", i, queue->events[i]);
You still have a lot of debugging messages in here. Most of these can
be removed, as ftrace is useful instead now that you "know" the code
works properly.
> + kfree(queue->events[i]);
> + }
> + queue->size = 0;
> +}
> +
> +/*----------------------------------------------------------------------*/
> +
> +struct raw_dev;
> +
> +#define USB_RAW_MAX_ENDPOINTS 32
> +
> +enum ep_state {
> + STATE_EP_DISABLED,
> + STATE_EP_ENABLED,
> +};
> +
> +struct raw_ep {
> + struct raw_dev *dev;
> + enum ep_state state;
> + struct usb_ep *ep;
> + struct usb_request *req;
> + bool urb_queued;
> + bool disabling;
> + ssize_t status;
> +};
> +
> +enum dev_state {
> + STATE_DEV_INVALID = 0,
> + STATE_DEV_OPENED,
> + STATE_DEV_INITIALIZED,
> + STATE_DEV_RUNNING,
> + STATE_DEV_CLOSED,
> + STATE_DEV_FAILED
> +};
> +
> +struct raw_dev {
> + refcount_t count;
Why not just use a "real" struct device?
Or a kref at the least, no need to roll your own refcount logic here,
right?
> + spinlock_t lock;
> +
> + const char *udc_name;
> + struct usb_gadget_driver driver;
> +
> + /* Protected by lock: */
> + enum dev_state state;
> + bool gadget_registered;
> + struct usb_gadget *gadget;
> + struct usb_request *req;
> + bool ep0_in_pending;
> + bool ep0_out_pending;
> + bool ep0_urb_queued;
> + ssize_t ep0_status;
> + struct raw_ep eps[USB_RAW_MAX_ENDPOINTS];
> +
> + struct completion ep0_done;
> + struct raw_event_queue queue;
> +};
> +
> +static struct raw_dev *dev_new(void)
> +{
> + struct raw_dev *dev;
> +
> + dev = kzalloc(sizeof(*dev), GFP_KERNEL);
> + if (!dev)
> + return NULL;
> + refcount_set(&dev->count, 1); /* Matches dev_put() in raw_release(). */
> + spin_lock_init(&dev->lock);
> + init_completion(&dev->ep0_done);
> + raw_event_queue_init(&dev->queue);
> + pr_debug("device created\n");
> + return dev;
> +}
> +
> +static inline void dev_get(struct raw_dev *dev)
> +{
> + refcount_inc(&dev->count);
> +}
> +
> +static void dev_put(struct raw_dev *dev)
> +{
> + int i;
> +
> + if (likely(!refcount_dec_and_test(&dev->count)))
> + return;
unless you can measure it, don't use likely/unlikely as the compiler and
cpu can almost always get it right instead.
> + kfree(dev->udc_name);
> + kfree(dev->driver.udc_name);
> + if (dev->req) {
> + if (dev->ep0_urb_queued)
> + usb_ep_dequeue(dev->gadget->ep0, dev->req);
> + usb_ep_free_request(dev->gadget->ep0, dev->req);
> + }
> + raw_event_queue_destroy(&dev->queue);
> + for (i = 0; i < USB_RAW_MAX_ENDPOINTS; i++) {
> + if (dev->eps[i].state != STATE_EP_ENABLED)
> + continue;
> + usb_ep_disable(dev->eps[i].ep);
> + usb_ep_free_request(dev->eps[i].ep, dev->eps[i].req);
> + kfree(dev->eps[i].ep->desc);
> + dev->eps[i].state = STATE_EP_DISABLED;
> + }
> + kfree(dev);
> + pr_debug("device freed\n");
ftrace :)
> +static void gadget_unbind(struct usb_gadget *gadget)
> +{
> + struct raw_dev *dev = get_gadget_data(gadget);
> + unsigned long flags;
> +
> + if (WARN_ON(!dev))
> + return;
Why warn? How can this happen?
> + spin_lock_irqsave(&dev->lock, flags);
> + set_gadget_data(gadget, NULL);
> + spin_unlock_irqrestore(&dev->lock, flags);
> + dev_put(dev); /* Matches dev_get() in gadget_bind(). */
> + pr_debug("unbound\n");
> +}
> +
> +static int gadget_setup(struct usb_gadget *gadget,
> + const struct usb_ctrlrequest *ctrl)
> +{
> + int ret = 0;
> + struct raw_dev *dev = get_gadget_data(gadget);
> + unsigned long flags;
> +
> + pr_debug("bRequestType: 0x%x (%s), bRequest: 0x%x,\n"
> + " wValue: 0x%x, wIndex: 0x%x, wLength: %d\n",
> + ctrl->bRequestType,
> + (ctrl->bRequestType & USB_DIR_IN) ? "IN" : "OUT",
> + ctrl->bRequest, ctrl->wValue, ctrl->wIndex, ctrl->wLength);
> +
> + if (WARN_ON(!dev))
> + return -ENODEV;
Same here.
> + spin_lock_irqsave(&dev->lock, flags);
> + if (dev->state != STATE_DEV_RUNNING) {
> + pr_err("ignoring, device is not running\n");
> + ret = -ENODEV;
> + goto out_unlock;
> + }
> + pr_debug("in_pending: %d, out_pending: %d\n",
> + dev->ep0_in_pending, dev->ep0_out_pending);
> + if (dev->ep0_in_pending || dev->ep0_out_pending) {
> + pr_debug("stalling, already have pending request\n");
> + ret = -EBUSY;
> + goto out_unlock;
> + }
> + if ((ctrl->bRequestType & USB_DIR_IN) && ctrl->wLength) {
> + pr_debug("ep0_in_pending <= true\n");
> + dev->ep0_in_pending = true;
> + } else {
> + pr_debug("ep0_out_pending <= true\n");
> + dev->ep0_out_pending = true;
> + }
> + spin_unlock_irqrestore(&dev->lock, flags);
> +
> + ret = raw_queue_event(dev, USB_RAW_EVENT_CONTROL, sizeof(*ctrl), ctrl);
> + if (ret < 0)
> + pr_err("failed to queue event\n");
> + pr_debug("event queued\n");
> + goto out;
> +
> +out_unlock:
> + spin_unlock_irqrestore(&dev->lock, flags);
> +out:
> + return ret;
> +}
> +
> +static void gadget_disconnect(struct usb_gadget *gadget)
> +{
> + pr_debug("ignoring\n");
> +}
> +
> +static void gadget_suspend(struct usb_gadget *gadget)
> +{
> + pr_debug("ignoring\n");
> +}
> +
> +static void gadget_resume(struct usb_gadget *gadget)
> +{
> + pr_debug("ignoring\n");
> +}
> +
> +static void gadget_reset(struct usb_gadget *gadget)
> +{
> + pr_debug("ignoring\n");
> +}
I think you don't even need functions for these, right? If not, just
remove. If you do, we should fix the gadget core to not require them :)
> +
> +/*----------------------------------------------------------------------*/
> +
> +static int raw_open(struct inode *inode, struct file *fd)
> +{
> + struct raw_dev *dev;
> +
> + dev = dev_new();
> + if (!dev) {
> + pr_err("failed to created device");
So many error messages printed on failures, you only needed the original
one if memory was gone that the core sent out.
> + return -ENOMEM;
> + }
> + fd->private_data = dev;
> + dev->state = STATE_DEV_OPENED;
> + pr_debug("device opened");
> + return 0;
> +}
> +
> +static int raw_release(struct inode *inode, struct file *fd)
> +{
> + int ret = 0;
> + struct raw_dev *dev = fd->private_data;
> + unsigned long flags;
> + bool unregister = false;
> +
> + if (!dev)
> + return -EBUSY;
How can that happen?
> +
> + spin_lock_irqsave(&dev->lock, flags);
> + dev->state = STATE_DEV_CLOSED;
> + pr_debug("device is closed\n");
> + if (!dev->gadget) {
> + spin_unlock_irqrestore(&dev->lock, flags);
> + goto out_put;
> + }
> + if (dev->gadget_registered)
> + unregister = true;
> + dev->gadget_registered = false;
> + spin_unlock_irqrestore(&dev->lock, flags);
> +
> + if (unregister) {
> + ret = usb_gadget_unregister_driver(&dev->driver);
> + WARN_ON(ret != 0);
> + dev_put(dev); /* Matches dev_get() in raw_ioctl_run(). */
> + }
> +
> +out_put:
> + dev_put(dev); /* Matches dev_new() in raw_open(). */
> + pr_debug("device released");
> + return ret;
> +}
> +
> +/*----------------------------------------------------------------------*/
> +
> +#define UDC_NAME_LENGTH_MAX 128
> +
> +static int raw_ioctl_init(struct raw_dev *dev, unsigned long value)
> +{
> + int ret = 0;
> + struct usb_raw_init arg;
> + char *udc_driver_name;
> + char *udc_device_name;
> + unsigned long flags;
> +
> + ret = copy_from_user(&arg, (void __user *)value, sizeof(arg));
> + if (ret)
> + return ret;
> +
> + switch (arg.speed) {
> + case USB_SPEED_LOW:
> + case USB_SPEED_FULL:
> + case USB_SPEED_HIGH:
> + case USB_SPEED_SUPER:
> + break;
> + default:
> + arg.speed = USB_SPEED_HIGH;
> + }
> +
> + udc_driver_name = kmalloc(UDC_NAME_LENGTH_MAX, GFP_KERNEL);
> + if (!udc_driver_name)
> + return -ENOMEM;
> + ret = strncpy_from_user(udc_driver_name, arg.driver_name,
> + UDC_NAME_LENGTH_MAX);
> + if (ret < 0) {
> + kfree(udc_driver_name);
> + return ret;
> + }
> + ret = 0;
> + pr_debug("udc_driver_name: %s\n", udc_driver_name);
> +
> + udc_device_name = kmalloc(UDC_NAME_LENGTH_MAX, GFP_KERNEL);
> + if (!udc_device_name) {
> + kfree(udc_driver_name);
> + return -ENOMEM;
> + }
> + ret = strncpy_from_user(udc_device_name, arg.device_name,
> + UDC_NAME_LENGTH_MAX);
You are burying the max size of names in the .c code, put it in the .h
to give userspace a chance.
> +static bool check_ep_caps(struct usb_ep *ep,
> + struct usb_endpoint_descriptor *desc)
> +{
> + switch (desc->bmAttributes & USB_ENDPOINT_XFERTYPE_MASK) {
> + case USB_ENDPOINT_XFER_ISOC:
We have functions/macros for endpoint types, right? Can't you use them
here?
> + if (!ep->caps.type_iso)
> + return false;
> + break;
> + case USB_ENDPOINT_XFER_BULK:
> + if (!ep->caps.type_bulk)
> + return false;
> + break;
> + case USB_ENDPOINT_XFER_INT:
> + if (!ep->caps.type_int)
> + return false;
> + break;
> + default:
> + return false;
> + }
> +
> + switch (desc->bEndpointAddress & USB_ENDPOINT_DIR_MASK) {
> + case USB_DIR_IN:
Same here, we have macros for this.
> +static int __init raw_init(void)
> +{
> + raw_file = debugfs_create_file("raw-gadget", 0600,
> + usb_debug_root, NULL, &raw_ops);
> + if (!raw_file) {
this return value will never be NULL no matter what. So you just tested
for an impossiblity :)
In the future, there's never a new to check any return value for any
debugfs calls, it's not an issue.
> + pr_err("failed to create raw-gadget in debugfs\n");
> + return -ENOMEM;
> + }
> + return 0;
> +}
> +
> +static void __exit raw_exit(void)
> +{
> + if (!raw_file)
> + return;
No need to check:
> + debugfs_remove(raw_file);
that call can take NULL
> + raw_file = NULL;
No need for this :)
> +/*
> + * struct usb_raw_init - argument for USB_RAW_IOCTL_INIT ioctl.
> + * @speed: The speed of the emulated USB device, takes the same values as
> + * the usb_device_speed enum: USB_SPEED_FULL, USB_SPEED_HIGH, etc.
> + * @driver_name: The name of the UDC driver.
> + * @device_name: The name of a UDC instance.
> + *
> + * The last two fields identify a UDC the gadget driver should bind to.
> + * For example, Dummy UDC has "dummy_udc" as its driver_name and "dummy_udc.N"
> + * as its device_name, where N in the index of the Dummy UDC instance.
> + * At the same time the dwc2 driver that is used on Raspberry Pi Zero, has
> + * "20980000.usb" as both driver_name and device_name.
> + */
> +struct usb_raw_init {
> + uint64_t speed;
> + const char *driver_name;
> + const char *device_name;
If you have structures crossing the user/kernel boundry, always use the
correct data types. Those start with "__". So this would be __u32 and
__u8 * instead.
> +};
> +
> +/* The type of event fetched with the USB_RAW_IOCTL_EVENT_FETCH ioctl. */
> +enum usb_raw_event_type {
> + USB_RAW_EVENT_INVALID,
> +
> + /* This event is queued when the driver has bound to a UDC. */
> + USB_RAW_EVENT_CONNECT,
> +
> + /* This event is queued when a new control request arrived to ep0. */
> + USB_RAW_EVENT_CONTROL,
> +
> + /* The list might grow in the future. */
> +};
> +
> +/*
> + * struct usb_raw_event - argument for USB_RAW_IOCTL_EVENT_FETCH ioctl.
> + * @type: The type of the fetched event.
> + * @length: Length of the data buffer. Updated by the driver and set to the
> + * actual length of the fetched event data.
> + * @data: A buffer to store the fetched event data.
> + *
> + * Currently the fetched data buffer is empty for USB_RAW_EVENT_CONNECT,
> + * and contains struct usb_ctrlrequest for USB_RAW_EVENT_CONTROL.
> + */
> +struct usb_raw_event {
> + uint32_t type;
> + uint32_t length;
> + char data[0];
__u32 please.
> +};
> +
> +#define USB_RAW_IO_FLAGS_ZERO 0x0001
> +#define USB_RAW_IO_FLAGS_MASK 0x0001
> +
> +static int usb_raw_io_flags_valid(uint16_t flags)
> +{
> + return (flags & ~USB_RAW_IO_FLAGS_MASK) == 0;
> +}
> +
> +static int usb_raw_io_flags_zero(uint16_t flags)
> +{
> + return (flags & USB_RAW_IO_FLAGS_ZERO);
> +}
> +
> +/*
> + * struct usb_raw_ep_io - argument for USB_RAW_IOCTL_EP0/EP_WRITE/READ ioctls.
> + * @ep: Endpoint handle as returned by USB_RAW_IOCTL_EP_ENABLE for
> + * USB_RAW_IOCTL_EP_WRITE/READ. Ignored for USB_RAW_IOCTL_EP0_WRITE/READ.
> + * @flags: When USB_RAW_IO_FLAGS_ZERO is specified, the zero flag is set on
> + * the submitted USB request, see include/linux/usb/gadget.h for details.
> + * @length: Length of data.
> + * @data: Data to send for USB_RAW_IOCTL_EP0/EP_WRITE. Buffer to store received
> + * data for USB_RAW_IOCTL_EP0/EP_READ.
> + */
> +struct usb_raw_ep_io {
> + uint16_t ep;
> + uint16_t flags;
__u16
> + uint32_t length;
> + char data[0];
> +};
> +
> +/*
> + * Initializes a Raw Gadget instance.
> + * Accepts a pointer to the usb_raw_init struct as an argument.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_INIT _IOW('U', 0, struct usb_raw_init)
> +
> +/*
> + * Instructs Raw Gadget to bind to a UDC and start emulating a USB device.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_RUN _IO('U', 1)
> +
> +/*
> + * A blocking ioctl that waits for an event and returns fetched event data to
> + * the user.
> + * Accepts a pointer to the usb_raw_event struct.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_EVENT_FETCH _IOR('U', 2, struct usb_raw_event)
> +
> +/*
> + * Queues an IN (OUT for READ) urb as a response to the last control request
> + * received on endpoint 0, provided that was an IN (OUT for READ) request and
> + * waits until the urb is completed. Copies received data to user for READ.
> + * Accepts a pointer to the usb_raw_ep_io struct as an argument.
> + * Returns length of trasferred data on success or negative error code on
> + * failure.
> + */
> +#define USB_RAW_IOCTL_EP0_WRITE _IOW('U', 3, struct usb_raw_ep_io)
> +#define USB_RAW_IOCTL_EP0_READ _IOWR('U', 4, struct usb_raw_ep_io)
> +
> +/*
> + * Finds an endpoint that supports the transfer type specified in the
> + * descriptor and enables it.
> + * Accepts a pointer to the usb_endpoint_descriptor struct as an argument.
> + * Returns enabled endpoint handle on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_EP_ENABLE _IOW('U', 5, struct usb_endpoint_descriptor)
> +
> +/* Disables specified endpoint.
> + * Accepts endpoint handle as an argument.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_EP_DISABLE _IOW('U', 6, int)
__u32, right?
> +
> +/*
> + * Queues an IN (OUT for READ) urb as a response to the last control request
> + * received on endpoint usb_raw_ep_io.ep, provided that was an IN (OUT for READ)
> + * request and waits until the urb is completed. Copies received data to user
> + * for READ.
> + * Accepts a pointer to the usb_raw_ep_io struct as an argument.
> + * Returns length of trasferred data on success or negative error code on
> + * failure.
> + */
> +#define USB_RAW_IOCTL_EP_WRITE _IOW('U', 7, struct usb_raw_ep_io)
> +#define USB_RAW_IOCTL_EP_READ _IOWR('U', 8, struct usb_raw_ep_io)
> +
> +/*
> + * Switches the gadget into the configured state.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_CONFIGURE _IO('U', 9)
> +
> +/*
> + * Constrains UDC VBUS power usage.
> + * Accepts current limit in 2 mA units as an argument.
> + * Returns 0 on success or negative error code on failure.
> + */
> +#define USB_RAW_IOCTL_VBUS_DRAW _IOW('U', 10, uint32_t)
__u32
thanks,
greg k-h
next prev parent reply other threads:[~2019-11-08 21:17 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-11-08 18:26 [PATCH 0/1] usb: gadget: add raw-gadget interface Andrey Konovalov
2019-11-08 18:26 ` [PATCH 1/1] " Andrey Konovalov
2019-11-08 21:17 ` Greg Kroah-Hartman [this message]
2019-11-08 22:17 ` Andrey Konovalov
2019-11-14 15:08 ` Andrey Konovalov
2019-11-14 15:33 ` Andrey Konovalov
2019-11-15 3:20 ` Greg Kroah-Hartman
2019-11-15 3:19 ` Greg Kroah-Hartman
2019-11-08 21:45 ` [PATCH 0/1] " Alan Stern
2019-11-08 22:18 ` Andrey Konovalov
2019-11-14 15:10 ` Andrey Konovalov
2019-11-14 15:56 ` Alan Stern
2019-11-13 5:22 ` kbuild test robot
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191108211745.GA1282512@kroah.com \
--to=gregkh@linuxfoundation.org \
--cc=andreyknvl@google.com \
--cc=balbi@kernel.org \
--cc=corbet@lwn.net \
--cc=dvyukov@google.com \
--cc=elver@google.com \
--cc=glider@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-usb@vger.kernel.org \
--cc=stern@rowland.harvard.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.