[v5,2/3] usb: gadget: udc: core: Invoke usb_gadget_connect only when started
Commit Message
usb_udc_connect_control() does not check to see if the udc has already
been started. This causes gadget->ops->pullup to be called through
usb_gadget_connect() when invoked from usb_udc_vbus_handler() even
before usb_gadget_udc_start() is called. Guard this by checking for
udc->started in usb_udc_connect_control() before invoking
usb_gadget_connect().
Guarding udc->vbus, udc->started, gadget->connect, gadget->deactivate
related functions with connect_lock. usb_gadget_connect_locked(),
usb_gadget_disconnect_locked(), usb_udc_connect_control_locked(),
usb_gadget_udc_start_locked(), usb_gadget_udc_stop_locked() are called
with this lock held as they can be simulataneously invoked from
different code paths.
Adding an additional check to make sure udc is started(udc->started)
before pullup callback is invoked.
This commit was reverted due to the crash reported in
https://lore.kernel.org/all/ZF4BvgsOyoKxdPFF@francesco-nb.int.toradex.com/.
commit 16737e78d190 ("usb: gadget: udc: core: Offload usb_udc_vbus_handler processing")
addresses the crash reported.
Cc: stable@vger.kernel.org
Fixes: 628ef0d273a6 ("usb: udc: add usb_udc_vbus_handler")
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
---
v5 is the first version in this series.
---
drivers/usb/gadget/udc/core.c | 148 ++++++++++++++++++++++++----------
1 file changed, 104 insertions(+), 44 deletions(-)
Comments
On Wed, May 31, 2023 at 04:02:02AM +0000, Badhri Jagan Sridharan wrote:
> usb_udc_connect_control() does not check to see if the udc has already
> been started. This causes gadget->ops->pullup to be called through
> usb_gadget_connect() when invoked from usb_udc_vbus_handler() even
> before usb_gadget_udc_start() is called. Guard this by checking for
> udc->started in usb_udc_connect_control() before invoking
> usb_gadget_connect().
After a merged version of patches 1/3 and 3/3 have been applied, it
seems like most of this will not be needed any more. Maybe not any of
it.
usb_udc_connect_control() gets called from only two places. One of them
is in gadget_bind_driver(), where we know that the UDC has been started
and connecting is allowed. The other place is the vbus work routine
queued by usb_udc_vbus_handler(). If that place checks the new
allow_connect flag before calling usb_gadget_connect(), nothing more
will be needed. You just have to make sure that the allow_connect flag
is set in gadget_bind_driver between the start and connect_control
calls, and it is cleared in gadget_unbind_driver before the
cancel_work_sync call.
It's possible that a new mutex will be needed to synchronize accesses to
the allow_connect flag. That's something you will have to study and
decide on. But if you can avoid adding one, that would be best.
> Guarding udc->vbus, udc->started, gadget->connect, gadget->deactivate
> related functions with connect_lock. usb_gadget_connect_locked(),
> usb_gadget_disconnect_locked(), usb_udc_connect_control_locked(),
> usb_gadget_udc_start_locked(), usb_gadget_udc_stop_locked() are called
> with this lock held as they can be simulataneously invoked from
> different code paths.
It's a general principle of kernel programming that locks protect data,
not code. So if this patch were to be accepted, you would have to
change this description to say that connect_lock guards various flags,
not various function calls.
Alan Stern
On Wed, May 31, 2023 at 10:55 AM Alan Stern <stern@rowland.harvard.edu> wrote:
>
> On Wed, May 31, 2023 at 04:02:02AM +0000, Badhri Jagan Sridharan wrote:
> > usb_udc_connect_control() does not check to see if the udc has already
> > been started. This causes gadget->ops->pullup to be called through
> > usb_gadget_connect() when invoked from usb_udc_vbus_handler() even
> > before usb_gadget_udc_start() is called. Guard this by checking for
> > udc->started in usb_udc_connect_control() before invoking
> > usb_gadget_connect().
>
> After a merged version of patches 1/3 and 3/3 have been applied, it
> seems like most of this will not be needed any more. Maybe not any of
> it.
Without the connect_lock introduced in this patch, wouldn't the
usb_gadget_connect()/
usb_gadget_disconnect() through soft_connect_store() race against
usb_gadget_connect()/ usb_gadget_disconnect() through
usb_udc_connect_control() ?
On a side note, I am working on merging patches 1/3 and 3/3.
Thanks,
Badhri
>
> usb_udc_connect_control() gets called from only two places. One of them
> is in gadget_bind_driver(), where we know that the UDC has been started
> and connecting is allowed. The other place is the vbus work routine
> queued by usb_udc_vbus_handler(). If that place checks the new
> allow_connect flag before calling usb_gadget_connect(), nothing more
> will be needed. You just have to make sure that the allow_connect flag
> is set in gadget_bind_driver between the start and connect_control
> calls, and it is cleared in gadget_unbind_driver before the
> cancel_work_sync call.
>
> It's possible that a new mutex will be needed to synchronize accesses to
> the allow_connect flag. That's something you will have to study and
> decide on. But if you can avoid adding one, that would be best.
>
> > Guarding udc->vbus, udc->started, gadget->connect, gadget->deactivate
> > related functions with connect_lock. usb_gadget_connect_locked(),
> > usb_gadget_disconnect_locked(), usb_udc_connect_control_locked(),
> > usb_gadget_udc_start_locked(), usb_gadget_udc_stop_locked() are called
> > with this lock held as they can be simulataneously invoked from
> > different code paths.
>
> It's a general principle of kernel programming that locks protect data,
> not code. So if this patch were to be accepted, you would have to
> change this description to say that connect_lock guards various flags,
> not various function calls.
>
> Alan Stern
On Wed, May 31, 2023 at 03:40:26PM -0700, Badhri Jagan Sridharan wrote:
> On Wed, May 31, 2023 at 10:55 AM Alan Stern <stern@rowland.harvard.edu> wrote:
> >
> > On Wed, May 31, 2023 at 04:02:02AM +0000, Badhri Jagan Sridharan wrote:
> > > usb_udc_connect_control() does not check to see if the udc has already
> > > been started. This causes gadget->ops->pullup to be called through
> > > usb_gadget_connect() when invoked from usb_udc_vbus_handler() even
> > > before usb_gadget_udc_start() is called. Guard this by checking for
> > > udc->started in usb_udc_connect_control() before invoking
> > > usb_gadget_connect().
> >
> > After a merged version of patches 1/3 and 3/3 have been applied, it
> > seems like most of this will not be needed any more. Maybe not any of
> > it.
>
> Without the connect_lock introduced in this patch, wouldn't the
> usb_gadget_connect()/
> usb_gadget_disconnect() through soft_connect_store() race against
> usb_gadget_connect()/ usb_gadget_disconnect() through
> usb_udc_connect_control() ?
Okay, yes, that's a good point. It needs to be mentioned in the patch
description so that people will understand it is the real reason for
this change.
Alan Stern
On Wed, May 31, 2023 at 3:55 PM Alan Stern <stern@rowland.harvard.edu> wrote:
>
> On Wed, May 31, 2023 at 03:40:26PM -0700, Badhri Jagan Sridharan wrote:
> > On Wed, May 31, 2023 at 10:55 AM Alan Stern <stern@rowland.harvard.edu> wrote:
> > >
> > > On Wed, May 31, 2023 at 04:02:02AM +0000, Badhri Jagan Sridharan wrote:
> > > > usb_udc_connect_control() does not check to see if the udc has already
> > > > been started. This causes gadget->ops->pullup to be called through
> > > > usb_gadget_connect() when invoked from usb_udc_vbus_handler() even
> > > > before usb_gadget_udc_start() is called. Guard this by checking for
> > > > udc->started in usb_udc_connect_control() before invoking
> > > > usb_gadget_connect().
> > >
> > > After a merged version of patches 1/3 and 3/3 have been applied, it
> > > seems like most of this will not be needed any more. Maybe not any of
> > > it.
> >
> > Without the connect_lock introduced in this patch, wouldn't the
> > usb_gadget_connect()/
> > usb_gadget_disconnect() through soft_connect_store() race against
> > usb_gadget_connect()/ usb_gadget_disconnect() through
> > usb_udc_connect_control() ?
>
> Okay, yes, that's a good point. It needs to be mentioned in the patch
> description so that people will understand it is the real reason for
> this change.
Thanks Alan !
I had posted the v6 version of the series with 1/3 and 3/3 of v5
squashed together. I have made changes to address your concerns in v5.
Instead of adding a new lock, I used the connect_lock in this patch to
protect the allow_connect flag.
Eager to know your thoughts !
Regards,
Badhri
>
> Alan Stern
@@ -37,6 +37,10 @@ static const struct bus_type gadget_bus_type;
* @vbus: for udcs who care about vbus status, this value is real vbus status;
* for udcs who do not care about vbus status, this value is always true
* @started: the UDC's started state. True if the UDC had started.
+ * @connect_lock: protects udc->vbus, udc->started, gadget->connect, gadget->deactivate related
+ * functions. usb_gadget_connect_locked, usb_gadget_disconnect_locked,
+ * usb_udc_connect_control_locked, usb_gadget_udc_start_locked, usb_gadget_udc_stop_locked are
+ * called with this lock held.
*
* This represents the internal data structure which is used by the UDC-class
* to hold information about udc driver and gadget together.
@@ -49,6 +53,7 @@ struct usb_udc {
bool vbus;
bool started;
struct work_struct vbus_work;
+ struct mutex connect_lock;
};
static struct class *udc_class;
@@ -688,17 +693,9 @@ int usb_gadget_vbus_disconnect(struct usb_gadget *gadget)
}
EXPORT_SYMBOL_GPL(usb_gadget_vbus_disconnect);
-/**
- * usb_gadget_connect - software-controlled connect to USB host
- * @gadget:the peripheral being connected
- *
- * Enables the D+ (or potentially D-) pullup. The host will start
- * enumerating this gadget when the pullup is active and a VBUS session
- * is active (the link is powered).
- *
- * Returns zero on success, else negative errno.
- */
-int usb_gadget_connect(struct usb_gadget *gadget)
+/* Internal version of usb_gadget_connect needs to be called with connect_lock held. */
+static int usb_gadget_connect_locked(struct usb_gadget *gadget)
+ __must_hold(&gadget->udc->connect_lock)
{
int ret = 0;
@@ -707,10 +704,12 @@ int usb_gadget_connect(struct usb_gadget *gadget)
goto out;
}
- if (gadget->deactivated) {
+ if (gadget->deactivated || !gadget->udc->started) {
/*
* If gadget is deactivated we only save new state.
* Gadget will be connected automatically after activation.
+ *
+ * udc first needs to be started before gadget can be pulled up.
*/
gadget->connected = true;
goto out;
@@ -725,22 +724,32 @@ int usb_gadget_connect(struct usb_gadget *gadget)
return ret;
}
-EXPORT_SYMBOL_GPL(usb_gadget_connect);
/**
- * usb_gadget_disconnect - software-controlled disconnect from USB host
- * @gadget:the peripheral being disconnected
- *
- * Disables the D+ (or potentially D-) pullup, which the host may see
- * as a disconnect (when a VBUS session is active). Not all systems
- * support software pullup controls.
+ * usb_gadget_connect - software-controlled connect to USB host
+ * @gadget:the peripheral being connected
*
- * Following a successful disconnect, invoke the ->disconnect() callback
- * for the current gadget driver so that UDC drivers don't need to.
+ * Enables the D+ (or potentially D-) pullup. The host will start
+ * enumerating this gadget when the pullup is active and a VBUS session
+ * is active (the link is powered).
*
* Returns zero on success, else negative errno.
*/
-int usb_gadget_disconnect(struct usb_gadget *gadget)
+int usb_gadget_connect(struct usb_gadget *gadget)
+{
+ int ret;
+
+ mutex_lock(&gadget->udc->connect_lock);
+ ret = usb_gadget_connect_locked(gadget);
+ mutex_unlock(&gadget->udc->connect_lock);
+
+ return ret;
+}
+EXPORT_SYMBOL_GPL(usb_gadget_connect);
+
+/* Internal version of usb_gadget_disconnect needs to be called with connect_lock held. */
+static int usb_gadget_disconnect_locked(struct usb_gadget *gadget)
+ __must_hold(&gadget->udc->connect_lock)
{
int ret = 0;
@@ -752,10 +761,12 @@ int usb_gadget_disconnect(struct usb_gadget *gadget)
if (!gadget->connected)
goto out;
- if (gadget->deactivated) {
+ if (gadget->deactivated || !gadget->udc->started) {
/*
* If gadget is deactivated we only save new state.
* Gadget will stay disconnected after activation.
+ *
+ * udc should have been started before gadget being pulled down.
*/
gadget->connected = false;
goto out;
@@ -775,6 +786,30 @@ int usb_gadget_disconnect(struct usb_gadget *gadget)
return ret;
}
+
+/**
+ * usb_gadget_disconnect - software-controlled disconnect from USB host
+ * @gadget:the peripheral being disconnected
+ *
+ * Disables the D+ (or potentially D-) pullup, which the host may see
+ * as a disconnect (when a VBUS session is active). Not all systems
+ * support software pullup controls.
+ *
+ * Following a successful disconnect, invoke the ->disconnect() callback
+ * for the current gadget driver so that UDC drivers don't need to.
+ *
+ * Returns zero on success, else negative errno.
+ */
+int usb_gadget_disconnect(struct usb_gadget *gadget)
+{
+ int ret;
+
+ mutex_lock(&gadget->udc->connect_lock);
+ ret = usb_gadget_disconnect_locked(gadget);
+ mutex_unlock(&gadget->udc->connect_lock);
+
+ return ret;
+}
EXPORT_SYMBOL_GPL(usb_gadget_disconnect);
/**
@@ -795,10 +830,11 @@ int usb_gadget_deactivate(struct usb_gadget *gadget)
if (gadget->deactivated)
goto out;
+ mutex_lock(&gadget->udc->connect_lock);
if (gadget->connected) {
- ret = usb_gadget_disconnect(gadget);
+ ret = usb_gadget_disconnect_locked(gadget);
if (ret)
- goto out;
+ goto unlock;
/*
* If gadget was being connected before deactivation, we want
@@ -808,6 +844,8 @@ int usb_gadget_deactivate(struct usb_gadget *gadget)
}
gadget->deactivated = true;
+unlock:
+ mutex_unlock(&gadget->udc->connect_lock);
out:
trace_usb_gadget_deactivate(gadget, ret);
@@ -831,6 +869,7 @@ int usb_gadget_activate(struct usb_gadget *gadget)
if (!gadget->deactivated)
goto out;
+ mutex_lock(&gadget->udc->connect_lock);
gadget->deactivated = false;
/*
@@ -838,7 +877,8 @@ int usb_gadget_activate(struct usb_gadget *gadget)
* while it was being deactivated, we call usb_gadget_connect().
*/
if (gadget->connected)
- ret = usb_gadget_connect(gadget);
+ ret = usb_gadget_connect_locked(gadget);
+ mutex_unlock(&gadget->udc->connect_lock);
out:
trace_usb_gadget_activate(gadget, ret);
@@ -1079,19 +1119,22 @@ EXPORT_SYMBOL_GPL(usb_gadget_set_state);
/* ------------------------------------------------------------------------- */
-static void usb_udc_connect_control(struct usb_udc *udc)
+/* Acquire connect_lock before calling this function. */
+static void usb_udc_connect_control_locked(struct usb_udc *udc) __must_hold(&udc->connect_lock)
{
- if (udc->vbus)
- usb_gadget_connect(udc->gadget);
+ if (udc->vbus && udc->started)
+ usb_gadget_connect_locked(udc->gadget);
else
- usb_gadget_disconnect(udc->gadget);
+ usb_gadget_disconnect_locked(udc->gadget);
}
static void vbus_event_work(struct work_struct *work)
{
struct usb_udc *udc = container_of(work, struct usb_udc, vbus_work);
- usb_udc_connect_control(udc);
+ mutex_lock(&udc->connect_lock);
+ usb_udc_connect_control_locked(udc);
+ mutex_unlock(&udc->connect_lock);
}
/**
@@ -1139,7 +1182,7 @@ void usb_gadget_udc_reset(struct usb_gadget *gadget,
EXPORT_SYMBOL_GPL(usb_gadget_udc_reset);
/**
- * usb_gadget_udc_start - tells usb device controller to start up
+ * usb_gadget_udc_start_locked - tells usb device controller to start up
* @udc: The UDC to be started
*
* This call is issued by the UDC Class driver when it's about
@@ -1150,8 +1193,11 @@ EXPORT_SYMBOL_GPL(usb_gadget_udc_reset);
* necessary to have it powered on.
*
* Returns zero on success, else negative errno.
+ *
+ * Caller should acquire connect_lock before invoking this function.
*/
-static inline int usb_gadget_udc_start(struct usb_udc *udc)
+static inline int usb_gadget_udc_start_locked(struct usb_udc *udc)
+ __must_hold(&udc->connect_lock)
{
int ret;
@@ -1168,7 +1214,7 @@ static inline int usb_gadget_udc_start(struct usb_udc *udc)
}
/**
- * usb_gadget_udc_stop - tells usb device controller we don't need it anymore
+ * usb_gadget_udc_stop_locked - tells usb device controller we don't need it anymore
* @udc: The UDC to be stopped
*
* This call is issued by the UDC Class driver after calling
@@ -1177,8 +1223,11 @@ static inline int usb_gadget_udc_start(struct usb_udc *udc)
* The details are implementation specific, but it can go as
* far as powering off UDC completely and disable its data
* line pullups.
+ *
+ * Caller should acquire connect lock before invoking this function.
*/
-static inline void usb_gadget_udc_stop(struct usb_udc *udc)
+static inline void usb_gadget_udc_stop_locked(struct usb_udc *udc)
+ __must_hold(&udc->connect_lock)
{
if (!udc->started) {
dev_err(&udc->dev, "UDC had already stopped\n");
@@ -1337,6 +1386,7 @@ int usb_add_gadget(struct usb_gadget *gadget)
udc->gadget = gadget;
gadget->udc = udc;
+ mutex_init(&udc->connect_lock);
udc->started = false;
@@ -1539,11 +1589,15 @@ static int gadget_bind_driver(struct device *dev)
if (ret)
goto err_bind;
- ret = usb_gadget_udc_start(udc);
- if (ret)
+ mutex_lock(&udc->connect_lock);
+ ret = usb_gadget_udc_start_locked(udc);
+ if (ret) {
+ mutex_unlock(&udc->connect_lock);
goto err_start;
+ }
usb_gadget_enable_async_callbacks(udc);
- usb_udc_connect_control(udc);
+ usb_udc_connect_control_locked(udc);
+ mutex_unlock(&udc->connect_lock);
kobject_uevent(&udc->dev.kobj, KOBJ_CHANGE);
return 0;
@@ -1575,12 +1629,14 @@ static void gadget_unbind_driver(struct device *dev)
kobject_uevent(&udc->dev.kobj, KOBJ_CHANGE);
cancel_work_sync(&udc->vbus_work);
- usb_gadget_disconnect(gadget);
+ mutex_lock(&udc->connect_lock);
+ usb_gadget_disconnect_locked(gadget);
usb_gadget_disable_async_callbacks(udc);
if (gadget->irq)
synchronize_irq(gadget->irq);
udc->driver->unbind(gadget);
- usb_gadget_udc_stop(udc);
+ usb_gadget_udc_stop_locked(udc);
+ mutex_unlock(&udc->connect_lock);
mutex_lock(&udc_lock);
driver->is_bound = false;
@@ -1666,11 +1722,15 @@ static ssize_t soft_connect_store(struct device *dev,
}
if (sysfs_streq(buf, "connect")) {
- usb_gadget_udc_start(udc);
- usb_gadget_connect(udc->gadget);
+ mutex_lock(&udc->connect_lock);
+ usb_gadget_udc_start_locked(udc);
+ usb_gadget_connect_locked(udc->gadget);
+ mutex_unlock(&udc->connect_lock);
} else if (sysfs_streq(buf, "disconnect")) {
- usb_gadget_disconnect(udc->gadget);
- usb_gadget_udc_stop(udc);
+ mutex_lock(&udc->connect_lock);
+ usb_gadget_disconnect_locked(udc->gadget);
+ usb_gadget_udc_stop_locked(udc);
+ mutex_unlock(&udc->connect_lock);
} else {
dev_err(dev, "unsupported command '%s'\n", buf);
ret = -EINVAL;