Unnamed repository; edit this file 'description' to name the repository. http://cgit.235523.xyz/tinyusb.git/atom/?h=hil-add-ea4088qs 2026-09-03T11:19:17Z 2026-09-03T11:19:17Z hathach thach@tinyusb.org 2026-09-03T11:19:17Z urn:sha1:ea6739db33c7b6c9a984a8c115cb0384b479523f The board's probe is an LPC-Link2 with no VCOM, so its console is RTT. Two host tests could not run that way: test_host_cdc_msc_hid and test_host_msc_file_explorer opened the flasher's VCOM directly instead of going through open_board_console(), and the config gate rejecting is_cdc/is_msc fixtures on "logger": "rtt" boards existed only to keep them from dying mid-run on a port that does not exist. Route both through the console abstraction and drop the gate. The reset ordering differs per console -- RTT owns the probe so the board must be reset before the console opens, a VCOM survives the reset so it opens first -- and that rule now lives in open_console_reset() instead of being spelled out at each call site. The roster entry runs device, host and dual: USB1's Type-A drives a hub with a CH340 (TX-RX shorted, for the cdc echo) and a thumbdrive, while USB2 keeps serving the device tests. BUFFER_SIZE_DOWN=128 because that echo writes up to 64 bytes at once and SEGGER's ring keeps one byte free, which the 16-byte default cannot hold; 64 measured unreliable, 128 reliable. Measured on the rig: 19 passed, 0 failed, 0 skipped. usbtest 30/30, host MSC reads 682 KB/s, device CDC/MSC ~510 kB/s (full speed behind the rig's hub). 2026-09-03T11:19:17Z hathach thach@tinyusb.org 2026-09-03T11:19:17Z urn:sha1:74b846541c8a2b218753526c74aacbf19f68e288 console_write() forwards what the attached CDC device sent back to the board's console, and reached for board_uart_write() directly to stay non-blocking. On a board whose console is the debug probe there is no UART: board_uart_write() is a stub returning -1, so everything the device echoed was dropped, and the do/while loop retried a write that could never make progress and spun until more data arrived. Send it through SEGGER_RTT_Write() on LOGGER_RTT builds -- already non-blocking in the default NO_BLOCK_SKIP mode, which is what the UART path wanted board_uart_write() for. The retry is also bounded now. board_uart_write() is only documented to return the number of bytes sent, and the stubs disagree on how they say "no console here" -- lpc40 returns -1, apm32f0xx returns 0 -- so the loop cannot decide from the return value alone when to stop. It gives up after 10 ms without progress instead, which no draining console ever hits, and keeps the negative return as a fast path so a stub UART does not cost that wait on every read. The RTT header include lives in board_api.h so any example can reach it; the include path only exists when LOGGER=rtt defines LOGGER_RTT. host/cdc_msc_hid now passes on ea4088_quickstart, whose probe (LPC-Link2) has no VCOM: the CDC echo round-trips through a CH340 with TX-RX shorted. 2026-09-03T11:19:17Z hathach thach@tinyusb.org 2026-09-03T11:19:17Z urn:sha1:4e7be0b8d07c4b8b7278ac96ba290f58a2ed51c3 hcd_init() clears every pending interrupt flag before enabling its own set: OHCI_REG->interrupt_status = OHCI_REG->interrupt_status; If a device is already connected at that point, that write drops the RHSC which went with the connect, while the root hub keeps ConnectStatusChange set. The device stays put, so no new edge follows and the attach is never reported - the port sits at CCS=1, CSC=1, PES=0 forever. Host-only builds got away with it because the connect is usually detected after that clear. In a dual build the device stack initialises first and enables the shared USB interrupt, which moves the detection into the window that gets cleared, and the host then never sees the device. Scan the root-hub ports once at the end of hcd_init and raise the attach ourselves, clearing the stale change bit so the next RHSC does not report the same device again. Verified on ea4088_quickstart: dual/host_info_to_device_cdc now enumerates on a plain reset (HIL test passes, "Device 1: ID 1a86:8010 SN 7FD88F0604B5"), host/device_info still enumerates, and device/cdc_msc is unaffected. host/device_info also builds for lpcxpresso55s69, lpcxpresso1769 and mcb1800. 2026-09-03T11:19:17Z hathach thach@tinyusb.org 2026-09-03T11:19:17Z urn:sha1:b799fa0240416b1fd5215549852fde4a73a0be27 The board has both USB ports wired: U1 to the Type-A host connector and U2 to the micro-B device connector, and board_init() already puts the port select in that mode, so it can run the dual example. - Allow the example on LPC40XX. - Point the host stack at rhport 0: ohci maps the TinyUSB rhport onto its root-hub port index, and the host controller here is index 0, while the dual examples default to rhport 1. - Move the string-descriptor buffers out of the stack. They are written by the host controller, so on parts whose USB master cannot reach the CPU stack (this family) the strings came back as garbage. host/device_info already keeps its transfer buffer in CFG_TUH_MEM_SECTION; do the same here. Verified on ea4088_quickstart with a WCH-Link on the Type-A port and the micro-B plugged into a PC: mounted device 1 Device 1: ID 1a86:8010 SN 7FD88F0604B5 iManufacturer 1 wch.cn iProduct 2 WCH-Link 2026-09-03T11:19:17Z hathach thach@tinyusb.org 2026-09-03T11:19:17Z urn:sha1:4a57d2a5e22414621900eb2b87aebc5a45a6b3a5 Two things kept the OHCI host from ever seeing a device on the USB1 Type-A connector: - The family passed only CFG_TUSB_MEM_SECTION, but every example's tusb_config.h pre-defines CFG_TUD_MEM_SECTION/CFG_TUH_MEM_SECTION as empty, so tusb_option.h's fallback never applied and ohci_data landed in main SRAM at 0x1000_0000. The USB AHB master cannot reach it: the controller latched HcInterruptStatus.UE and never wrote the HCCA. Pass all three defines so the DMA structures land in peripheral SRAM. - VBUS for the host port is switched by an STMPS2171 whose enable is active high, but USB_PPWR1 asserts low, so muxing P1.19 to FUNC2 held the switch off and the connector never had power. NXP's own lpcusblib HAL muxes the same pin the same way, which is why the vendor host example fails here too. Keep P1.19 a GPIO and drive it: high when the host stack is enabled, low otherwise, so a device-only build does not leave the switch enable floating. host/device_info now enumerates: Device 1: ID 1a86:8010 SN 7FD88F0604B5 wch.cn / WCH-Link device/cdc_msc still passes on HIL. Note the module has no VBUS source of its own: its +5V rail must be fed (HDK USB, EXT_VIN or TRACE_5V pin 11) or the switch has no input to pass. 2026-08-28T19:41:13Z Ha Thach thach@tinyusb.org 2026-08-28T19:41:13Z urn:sha1:84dcec43c709643f8a1b1ebbab6d08fbbc605a6d rp2350: ETM trace board pico2_etm_trace 2026-08-28T18:08:40Z hathach thach@tinyusb.org 2026-08-28T09:40:25Z urn:sha1:f7eb812679598aecc1d7702264ba48e6c178d5da 2026-08-28T18:08:40Z hathach thach@tinyusb.org 2026-08-28T09:40:07Z urn:sha1:839c324458f6900a58182d9aa23d846e8ec55c94 Closes the sys-clock question. rp2350: the dynamic 156 MHz switch is removed - 150 MHz soak-tested clean (and a runtime switch truncates ETM capture on the trace carrier). rp2040 keeps its existing 120 MHz: soak sweeps show 120 = 8/8, stock 125 = 0/3 (bulk-OUT collapses, device NAKs ~600:1 with zero CRC errors on the wire), and 132 = 2/10 flaky despite an exact 12n/375k divider - no divider criterion predicts rp2040 PIO-USB health (the RX state machine samples at raw sysclk), so only soak-validated clocks ship. 2026-08-28T18:08:31Z hathach thach@tinyusb.org 2026-08-28T09:39:58Z urn:sha1:6e8e2caf7fc2dc5575f5e0c71e69c1b19b3f5699 Board files for the trace carrier (console GP12/13, LED GP10, I2C GP8/9, PIO-USB host on GP20, all retargeted in board.cmake so the SDK defaults cannot mux a trace pin), compile-time trace pin-conflict checks, the measured DBGPAUSE rationale, Ozone project, and skill/docs updates. Trace validated at the stock 150 MHz (75 MHz TRACECLK, +1 ns sampling): zero overflow through a 15 s throughput soak; V2 probe ceiling 120 MHz. 2026-08-28T09:09:22Z Ha Thach thach@tinyusb.org 2026-08-28T09:09:22Z urn:sha1:64d952c027df194ea8b107b3d5d0b6c4cb9a994b hil: drop the sysfs blindness subsystem and derive the recovery reserve