Files
openipc-builder/devices/gk7205v200_otg_generic/general/package/all-patches/linux/0016-dwc3-device-mode-on-host-first-dual-role.patch
T
Dmitry IlyinandGitHub 96633d5577 gk7205v200: add an OTG profile that makes the camera a USB webcam (#113)
Adds gk7205v200_otg_generic, a firmware profile that puts the single DWC3
controller into dual-role so the board can either consume a USB webcam
(existing usbcam path, host role) or present itself to a PC as one (UVC
gadget, device role). The roles are mutually exclusive -- xHCI HCSPARAMS1
reports MaxPorts = 1 -- but usb-mode switches between them at runtime, so
one image covers both and no reboot is needed.

Kernel side, three patches:

0015 sets dr_mode = "otg" in the DTS so dwc3 registers both an xHCI host
and a UDC.

0016 makes device mode actually work on a part that has only ever run as
a host. Three fixes: re-run dwc3_event_buffers_setup() on gadget start
(an xHCI reset wipes GEVNTADR), soft-reset the core before taking the
lock so the PHY is re-initialised (without it high-speed chirp fails),
and swallow the run_stop timeout on pullup(0), which otherwise
permanently poisons cdev->deactivations.

0017 disables the vendor's UVC_SG_REQ path. It hands a whole frame to the
UDC as one scatter-gather request; dwc3 chains those sg entries into a
single isochronous service interval, which on a full-speed link is one
1023-byte packet, so every frame larger than that reached the host
truncated and H.264 decoders reported "bytestream -5" on roughly half the
frames. Goke never shipped this path enabled -- CONFIG_USB_CONFIGFS_F_UVC
is unset in every vendor defconfig, including gk7605v100's, and the SDK
carries no userspace UVC sample -- so it was never exercised. The
mainline path (32 plain one-packet requests) is used instead.

Userspace side: usb-mode switches roles with the unbind/flip/bind
sequencing dwc3_set_mode() does not do on its own, uvc-gadget-setup
composes the configfs gadget, and S72usbmode forces a concrete role at
boot -- not optional, since dr_mode = "otg" leaves GCTL.PRTCAPDIR in a
state this 4.9 dwc3 never moves off by itself.

The descriptors advertise exactly the one format, frame size and interval
majestic is configured to serve, read from /etc/majestic.yaml. majestic
answers every PROBE/COMMIT with its own bFormatIndex/bFrameIndex because
it has one stream and cannot follow the host, so anything extra is an
invitation to select something that will not arrive: asking for MJPEG got
raw H.264 (an SPS where a JPEG SOI belongs, garbage to every consumer),
and asking for 1080p got 720p.

Profile drops the air-unit packages (wifibroadcast, rtl8812au, mavfwd,
mavlink-router, vtund, linux-firmware, wireless-tools) and the whole
wireless kernel stack, which pays for usb_f_uvc.ko and for keeping the
full sensor list -- the fpv trim to IMX307/IMX335 is keyed on
OPENIPC_VARIANT="fpv", and a general-purpose USB webcam should not be
sensor-locked.

Measured on a gk7205v200 lab board over a full-speed link: H.264
1280x720 at 30 fps with zero decoder errors and no dropped buffers, and
MJPEG delivering valid JPEGs at the jpeg channel's own rate.

Requires a majestic built with ENABLE_UVC_GADGET to feed the gadget;
that is a separate change.
2026-07-28 23:21:12 +03:00

211 lines
8.1 KiB
Diff

From: OpenIPC
Subject: [PATCH] usb: dwc3: make device mode work on a host-first dual-role part
On gk7205v200 the DTS boots in host mode and the role is switched at
runtime, so the gadget always starts from a controller that xHCI has
already owned and reset. Three things were missing for that to work; all
three are needed, and all three are things upstream's later
__dwc3_set_mode() rework does for the DWC3_GCTL_PRTCAP_DEVICE case.
1. The event buffer was never re-armed. xHCI's controller reset clears
GEVNTADRLO/HI and nothing programmed them again, so the core DMA'd its
event descriptors to physical address 0 -- not RAM on this SoC, DRAM
starts at 0x40000000. No event ever reached the driver: the link
chirped and counted SOFs in hardware and the host saw a device, but the
gadget never observed the USB reset.
2. The device side was never reset before being started, so the USB2 PHY
was left in whatever state xHCI put it in. The high-speed chirp then
failed, DSTS.CONNECTSPD came back full-speed, CONNECT_DONE was never
raised and the host timed out with -71.
3. Re-attach never worked, for two separate reasons -- see the comments in
dwc3_gadget_pullup().
With all three, enumeration is reliable across repeated attach/detach
cycles.
--- a/drivers/usb/dwc3/core.c
+++ b/drivers/usb/dwc3/core.c
@@ -144,7 +144,7 @@
* dwc3_core_soft_reset - Issues core soft reset and PHY reset
* @dwc: pointer to our context structure
*/
-static int dwc3_core_soft_reset(struct dwc3 *dwc)
+int dwc3_core_soft_reset(struct dwc3 *dwc)
{
u32 reg;
int retries = 1000;
@@ -316,7 +316,7 @@
*
* Returns 0 on success otherwise negative errno.
*/
-static int dwc3_event_buffers_setup(struct dwc3 *dwc)
+int dwc3_event_buffers_setup(struct dwc3 *dwc)
{
struct dwc3_event_buffer *evt;
--- a/drivers/usb/dwc3/core.h
+++ b/drivers/usb/dwc3/core.h
@@ -1155,6 +1155,8 @@
/* prototypes */
void dwc3_set_mode(struct dwc3 *dwc, u32 mode);
+int dwc3_event_buffers_setup(struct dwc3 *dwc);
+int dwc3_core_soft_reset(struct dwc3 *dwc);
u32 dwc3_core_fifo_space(struct dwc3_ep *dep, u8 type);
/* check whether we are on the DWC_usb31 core */
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1563,6 +1563,9 @@
return 0;
}
+static int __dwc3_gadget_start(struct dwc3 *dwc);
+static void __dwc3_gadget_stop(struct dwc3 *dwc);
+
static int dwc3_gadget_pullup(struct usb_gadget *g, int is_on)
{
struct dwc3 *dwc = gadget_to_dwc(g);
@@ -1571,8 +1574,80 @@
is_on = !!is_on;
+ if (!is_on) {
+ spin_lock_irqsave(&dwc->lock, flags);
+ ret = dwc3_gadget_run_stop(dwc, false, false);
+ spin_unlock_irqrestore(&dwc->lock, flags);
+ /*
+ * Never report a failure on the way down.
+ *
+ * dwc3_gadget_run_stop(0) polls DSTS.DEVCTRLHLT and returns
+ * -ETIMEDOUT if the core does not halt in time, which this one
+ * regularly does not. That error is not harmless: it propagates
+ * out through usb_gadget_disconnect() into
+ * usb_gadget_deactivate(), which then skips "gadget->deactivated
+ * = true", and into usb_function_deactivate(), which then skips
+ * "cdev->deactivations++". The refcount is left permanently
+ * unbalanced, so every later usb_function_activate() trips
+ * WARN_ON(cdev->deactivations == 0) and never calls
+ * usb_gadget_connect() -- the gadget can be disconnected once
+ * and then never comes back.
+ *
+ * RUN_STOP has been cleared regardless, so the pullup really is
+ * down; only the halt handshake was slow. Whatever state that
+ * leaves the core in does not matter, because the is_on path
+ * below soft-resets and re-initialises it from scratch.
+ */
+ if (ret) {
+ dev_dbg(dwc->dev,
+ "pullup(0): run_stop returned %d, ignoring so the deactivation refcount stays balanced\n",
+ ret);
+ ret = 0;
+ }
+ return ret;
+ }
+
+ /*
+ * Bringing the pullup back up after a disconnect needs a full device-side
+ * re-init, not just RUN_STOP.
+ *
+ * A gadget that has been disconnected once (f_uvc does this every time
+ * userspace closes /dev/videoN, via usb_function_deactivate()) leaves the
+ * core half torn down: the USB2 PHY keeps whatever state the disconnect
+ * left it in and the event buffer registers are stale. Re-asserting
+ * RUN_STOP on its own gets the host as far as SET_ADDRESS -- the device
+ * reaches "addressed" -- and then control transfers fail, so enumeration
+ * never completes. Only the first connect after dwc3_gadget_start() ever
+ * worked, because that path does the reset for us.
+ *
+ * So do here what dwc3_gadget_start() does: quiesce, soft reset (which
+ * re-runs usb_phy_init()/phy_init()), then start the device side again.
+ * dwc3_event_buffers_setup() is called from __dwc3_gadget_start().
+ *
+ * Callable only from process context: dwc3_core_soft_reset() sleeps, so it
+ * must run outside dwc->lock. That holds for every caller we have --
+ * usb_gadget_connect() reaches us from usb_gadget_activate() (userspace
+ * opening the video node) and from udc_bind_to_driver(). A platform that
+ * called usb_udc_vbus_handler() from hard IRQ would need reworking here.
+ */
spin_lock_irqsave(&dwc->lock, flags);
- ret = dwc3_gadget_run_stop(dwc, is_on, false);
+ __dwc3_gadget_stop(dwc);
+ spin_unlock_irqrestore(&dwc->lock, flags);
+
+ ret = dwc3_core_soft_reset(dwc);
+ if (ret) {
+ dev_err(dwc->dev, "pullup: soft reset failed: %d\n", ret);
+ return ret;
+ }
+
+ spin_lock_irqsave(&dwc->lock, flags);
+ ret = __dwc3_gadget_start(dwc);
+ if (ret) {
+ spin_unlock_irqrestore(&dwc->lock, flags);
+ dev_err(dwc->dev, "pullup: gadget start failed: %d\n", ret);
+ return ret;
+ }
+ ret = dwc3_gadget_run_stop(dwc, true, false);
spin_unlock_irqrestore(&dwc->lock, flags);
return ret;
@@ -1652,6 +1727,27 @@
int ret = 0;
u32 reg;
+ /*
+ * Re-program the event buffer address before starting the device side.
+ *
+ * dwc3_core_init() sets GEVNTADRLO/HI up once at probe, but on a
+ * dual-role part the host is brought up straight afterwards and xHCI's
+ * controller reset clears them again. Nothing else re-arms them, so on
+ * any host->device transition (and on this SoC, that includes the very
+ * first one, because the DTS boots in host mode) the core is left with
+ * an event buffer address of 0. It then DMAs its event descriptors to
+ * physical address 0 -- which is not RAM here, DRAM starts at
+ * 0x40000000 -- so no event ever reaches the driver: the link chirps and
+ * counts SOFs in hardware, the host sees a high-speed device, but the
+ * gadget never observes the USB reset and enumeration dies with -71.
+ *
+ * Upstream fixed this by calling dwc3_event_buffers_setup() from
+ * __dwc3_set_mode() when switching to DWC3_GCTL_PRTCAP_DEVICE; this
+ * tree predates that rework, so re-arm here instead. This covers both
+ * the gadget-bind and the resume path, and is idempotent.
+ */
+ dwc3_event_buffers_setup(dwc);
+
reg = dwc3_readl(dwc->regs, DWC3_DCFG);
reg &= ~(DWC3_DCFG_SPEED_MASK);
@@ -1751,6 +1847,29 @@
int ret = 0;
int irq;
+ /*
+ * Re-initialise the controller and its PHYs before handing the port to
+ * the device side.
+ *
+ * On a dual-role part the host is started first, so by the time a gadget
+ * driver binds, the USB2 PHY is still in the state xHCI left it in and
+ * the device side has never been reset. Symptom: the link comes up but
+ * loses the high-speed chirp handshake, so DSTS.CONNECTSPD reads
+ * full-speed, DWC3_DEVICE_EVENT_CONNECT_DONE is never raised, and the
+ * host -- which is driving high-speed -- times out with -71.
+ *
+ * dwc3_core_soft_reset() re-runs usb_phy_init()/phy_init() and resets
+ * the device side, which is exactly what upstream's __dwc3_set_mode()
+ * does for the DWC3_GCTL_PRTCAP_DEVICE case. This tree predates that
+ * rework. Must be called outside dwc->lock: it sleeps (udelay loops and
+ * PHY init).
+ */
+ ret = dwc3_core_soft_reset(dwc);
+ if (ret) {
+ dev_err(dwc->dev, "soft reset before gadget start failed: %d\n", ret);
+ return ret;
+ }
+
irq = dwc->irq_gadget;
ret = request_threaded_irq(irq, dwc3_interrupt, dwc3_thread_interrupt,
IRQF_SHARED, "dwc3", dwc->ev_buf);