[PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Benjamin Herrenschmidt]

Some host drivers really do not like keyboards having an OUT endpoint.

For example, most UEFI forked from EDK2 before 2006 (or was it 2008 ?)
have a bug, they'll try to use the *last* interrupt EP in the
descriptor list and just assume it's an IN endpoint. Newer UEFIs
use the *first* interrupt endpoint instead. None of them checks the
direction :-(

This adds a "single_ep" option to f_hid which allows to specify that
only the IN path should be created. This should be used for keyboards
if they are ever to be used with such systems as host.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 drivers/usb/gadget/function/f_hid.c | 61 ++++++++++++++++++-----------
 drivers/usb/gadget/function/u_hid.h |  1 +
 2 files changed, 39 insertions(+), 23 deletions(-)

diff --git a/drivers/usb/gadget/function/f_hid.c b/drivers/usb/gadget/function/f_hid.c
index f3816a5c861e..ae393c4fb61b 100644
--- a/drivers/usb/gadget/function/f_hid.c
+++ b/drivers/usb/gadget/function/f_hid.c
@@ -44,6 +44,7 @@ struct f_hidg {
 	unsigned short			report_desc_length;
 	char				*report_desc;
 	unsigned short			report_length;
+	unsigned char			single_ep;
 
 	/* recv report */
 	struct list_head		completed_out_req;
@@ -249,7 +250,7 @@ static ssize_t f_hidg_read(struct file *file, char __user *buffer,
 	unsigned long flags;
 	int ret;
 
-	if (!count)
+	if (!count || hidg->single_ep)
 		return 0;
 
 	if (!access_ok(buffer, count))
@@ -612,15 +613,16 @@ static void hidg_disable(struct usb_function *f)
 	unsigned long flags;
 
 	usb_ep_disable(hidg->in_ep);
-	usb_ep_disable(hidg->out_ep);
-
-	spin_lock_irqsave(&hidg->read_spinlock, flags);
-	list_for_each_entry_safe(list, next, &hidg->completed_out_req, list) {
-		free_ep_req(hidg->out_ep, list->req);
-		list_del(&list->list);
-		kfree(list);
+	if (!hidg->single_ep) {
+		usb_ep_disable(hidg->out_ep);
+		spin_lock_irqsave(&hidg->read_spinlock, flags);
+		list_for_each_entry_safe(list, next, &hidg->completed_out_req, list) {
+			free_ep_req(hidg->out_ep, list->req);
+			list_del(&list->list);
+			kfree(list);
+		}
+		spin_unlock_irqrestore(&hidg->read_spinlock, flags);
 	}
-	spin_unlock_irqrestore(&hidg->read_spinlock, flags);
 
 	spin_lock_irqsave(&hidg->write_spinlock, flags);
 	if (!hidg->write_pending) {
@@ -770,10 +772,12 @@ static int hidg_bind(struct usb_configuration *c, struct usb_function *f)
 		goto fail;
 	hidg->in_ep = ep;
 
-	ep = usb_ep_autoconfig(c->cdev->gadget, &hidg_fs_out_ep_desc);
-	if (!ep)
-		goto fail;
-	hidg->out_ep = ep;
+	if (!hidg->single_ep) {
+		ep = usb_ep_autoconfig(c->cdev->gadget, &hidg_fs_out_ep_desc);
+		if (!ep)
+			goto fail;
+		hidg->out_ep = ep;
+	}
 
 	/* set descriptor dynamic values */
 	hidg_interface_desc.bInterfaceSubClass = hidg->bInterfaceSubClass;
@@ -784,11 +788,17 @@ static int hidg_bind(struct usb_configuration *c, struct usb_function *f)
 				cpu_to_le16(hidg->report_length);
 	hidg_hs_in_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
 	hidg_fs_in_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
-	hidg_ss_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
-	hidg_ss_out_comp_desc.wBytesPerInterval =
-				cpu_to_le16(hidg->report_length);
-	hidg_hs_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
-	hidg_fs_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
+	if (hidg->single_ep) {
+		hidg_interface_desc.bNumEndpoints = 1;
+	} else {
+		hidg_interface_desc.bNumEndpoints = 2;
+		hidg_ss_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
+		hidg_ss_out_comp_desc.wBytesPerInterval =
+			cpu_to_le16(hidg->report_length);
+		hidg_hs_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
+		hidg_fs_out_ep_desc.wMaxPacketSize = cpu_to_le16(hidg->report_length);
+	}
+
 	/*
 	 * We can use hidg_desc struct here but we should not relay
 	 * that its content won't change after returning from this function.
@@ -799,13 +809,15 @@ static int hidg_bind(struct usb_configuration *c, struct usb_function *f)
 
 	hidg_hs_in_ep_desc.bEndpointAddress =
 		hidg_fs_in_ep_desc.bEndpointAddress;
-	hidg_hs_out_ep_desc.bEndpointAddress =
-		hidg_fs_out_ep_desc.bEndpointAddress;
-
 	hidg_ss_in_ep_desc.bEndpointAddress =
 		hidg_fs_in_ep_desc.bEndpointAddress;
-	hidg_ss_out_ep_desc.bEndpointAddress =
-		hidg_fs_out_ep_desc.bEndpointAddress;
+
+	if (!hidg->single_ep) {
+		hidg_hs_out_ep_desc.bEndpointAddress =
+			hidg_fs_out_ep_desc.bEndpointAddress;
+		hidg_ss_out_ep_desc.bEndpointAddress =
+			hidg_fs_out_ep_desc.bEndpointAddress;
+	}
 
 	status = usb_assign_descriptors(f, hidg_fs_descriptors,
 			hidg_hs_descriptors, hidg_ss_descriptors, NULL);
@@ -924,6 +936,7 @@ CONFIGFS_ATTR(f_hid_opts_, name)
 F_HID_OPT(subclass, 8, 255);
 F_HID_OPT(protocol, 8, 255);
 F_HID_OPT(report_length, 16, 65535);
+F_HID_OPT(single_ep, 8, 1);
 
 static ssize_t f_hid_opts_report_desc_show(struct config_item *item, char *page)
 {
@@ -985,6 +998,7 @@ static struct configfs_attribute *hid_attrs[] = {
 	&f_hid_opts_attr_report_length,
 	&f_hid_opts_attr_report_desc,
 	&f_hid_opts_attr_dev,
+	&f_hid_opts_attr_single_ep,
 	NULL,
 };
 
@@ -1100,6 +1114,7 @@ static struct usb_function *hidg_alloc(struct usb_function_instance *fi)
 	hidg->minor = opts->minor;
 	hidg->bInterfaceSubClass = opts->subclass;
 	hidg->bInterfaceProtocol = opts->protocol;
+	hidg->single_ep = opts->single_ep;
 	hidg->report_length = opts->report_length;
 	hidg->report_desc_length = opts->report_desc_length;
 	if (opts->report_desc) {
diff --git a/drivers/usb/gadget/function/u_hid.h b/drivers/usb/gadget/function/u_hid.h
index 1594bfa312eb..d9303379d7fd 100644
--- a/drivers/usb/gadget/function/u_hid.h
+++ b/drivers/usb/gadget/function/u_hid.h
@@ -23,6 +23,7 @@ struct f_hid_opts {
 	unsigned short			report_length;
 	unsigned short			report_desc_length;
 	unsigned char			*report_desc;
+	unsigned char			single_ep;
 	bool				report_desc_alloc;
 
 	/*

Re: [PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Felipe Balbi]

Hi,

Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:

> Some host drivers really do not like keyboards having an OUT endpoint.
>
> For example, most UEFI forked from EDK2 before 2006 (or was it 2008 ?)
> have a bug, they'll try to use the *last* interrupt EP in the
> descriptor list and just assume it's an IN endpoint. Newer UEFIs
> use the *first* interrupt endpoint instead. None of them checks the
> direction :-(
>
> This adds a "single_ep" option to f_hid which allows to specify that
> only the IN path should be created. This should be used for keyboards
> if they are ever to be used with such systems as host.
>
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Could you come up with a slightly more descriptive name? single_ep
doesn't give me any hint of which endpoint will be left around.

Perhaps call it 'disable_output_report'?

-- 
balbi

Re: [PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Benjamin Herrenschmidt]

On Fri, 2019-08-09 at 08:31 +0300, Felipe Balbi wrote:
> Hi,
> 
> Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:
> 
> > Some host drivers really do not like keyboards having an OUT
> > endpoint.
> > 
> > For example, most UEFI forked from EDK2 before 2006 (or was it 2008
> > ?)
> > have a bug, they'll try to use the *last* interrupt EP in the
> > descriptor list and just assume it's an IN endpoint. Newer UEFIs
> > use the *first* interrupt endpoint instead. None of them checks the
> > direction :-(
> > 
> > This adds a "single_ep" option to f_hid which allows to specify
> > that
> > only the IN path should be created. This should be used for
> > keyboards
> > if they are ever to be used with such systems as host.
> > 
> > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> 
> Could you come up with a slightly more descriptive name? single_ep
> doesn't give me any hint of which endpoint will be left around.
> 
> Perhaps call it 'disable_output_report'?

Sure. Or more concice "input_only" maybe ?

Cheers,
Ben.

Re: [PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Felipe Balbi]

Hi,

Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:

> On Fri, 2019-08-09 at 08:31 +0300, Felipe Balbi wrote:
>> Hi,
>> 
>> Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:
>> 
>> > Some host drivers really do not like keyboards having an OUT
>> > endpoint.
>> > 
>> > For example, most UEFI forked from EDK2 before 2006 (or was it 2008
>> > ?)
>> > have a bug, they'll try to use the *last* interrupt EP in the
>> > descriptor list and just assume it's an IN endpoint. Newer UEFIs
>> > use the *first* interrupt endpoint instead. None of them checks the
>> > direction :-(
>> > 
>> > This adds a "single_ep" option to f_hid which allows to specify
>> > that
>> > only the IN path should be created. This should be used for
>> > keyboards
>> > if they are ever to be used with such systems as host.
>> > 
>> > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>> 
>> Could you come up with a slightly more descriptive name? single_ep
>> doesn't give me any hint of which endpoint will be left around.
>> 
>> Perhaps call it 'disable_output_report'?
>
> Sure. Or more concice "input_only" maybe ?

that works too. Another option would to introduce two options,
has_input_report and has_output_report and have them true by default.

Then user can even produce an output-only HID device, like these odd
USB-controlled relays.

-- 
balbi

Re: [PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Benjamin Herrenschmidt]

On Fri, 2019-08-09 at 11:08 +0300, Felipe Balbi wrote:
> 
> that works too. Another option would to introduce two options,
> has_input_report and has_output_report and have them true by default.
> 
> Then user can even produce an output-only HID device, like these odd
> USB-controlled relays.

Ideally we could just parse the descriptor :-) But that's a bit
trickier.

One thing I've been meaning to look at is a way to properly handle
keyboards (and other "on/off" buttons) by having the kernel latch the
state so it can properly respond to things like get report, or resend
all the "down" keys after a reset (so that things like "keep some key
pressed during boot for magic to happen") actually works...

But one thing at a time :-)

Cheers,
Ben.

Re: [PATCH] [RFC] usb: gadget: hid: Add "single_ep" option

[Felipe Balbi]

Hi,

Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:

> On Fri, 2019-08-09 at 11:08 +0300, Felipe Balbi wrote:
>> 
>> that works too. Another option would to introduce two options,
>> has_input_report and has_output_report and have them true by default.
>> 
>> Then user can even produce an output-only HID device, like these odd
>> USB-controlled relays.
>
> Ideally we could just parse the descriptor :-) But that's a bit
> trickier.
>
> One thing I've been meaning to look at is a way to properly handle
> keyboards (and other "on/off" buttons) by having the kernel latch the
> state so it can properly respond to things like get report, or resend
> all the "down" keys after a reset (so that things like "keep some key
> pressed during boot for magic to happen") actually works...

cool idea

> But one thing at a time :-)

yup

-- 
balbi