| Age | Commit message (Collapse) | Author |
|---|
|
Even if CFG_TUH_MIDI_RX_BUFSIZE=100*TUH_EPSIZE_BULK_MAX, calling tuh_midi_stream_read without a loop can return only one 4-byte packet and preventing subsequent transfer.
Signed-off-by: Zixun LI <[email protected]>
|
|
|
|
Follow-up to #3239. tuh_midi_stream_read terminates on cable-number
transitions, leaving residue in the FIFO. With the default RX FIFO
sized equal to one bulk packet, the next bulk IN transfer fails to
queue and the driver silently stops receiving. Raising the default
to 2x bulk gives single-call apps a full packet of headroom and
documents the drain-loop expectation.
Reproduced with Akai LPD8 mk1 (VID 09E8 PID 0075) on STM32H753 DWC2
host; fixed with this patch. See #3613 for full repro + captures.
|
|
- hcd_edpt_clear_stall: use ep_addr instead of hardcoded 0 (control endpoint)
- hcd: add TUP_USBIP_FSDEV_DRD define for MCUs with host support (C0, G0, H5,
U3, U5) and use it in hcd compile guard instead of enumerating MCUs
- dcd_edpt_close_all: use FSDEV_EP_COUNT instead of CFG_TUD_ENDPPOINT_MAX for
PMA btable offset to match handle_bus_reset
- midi host tuh_midi_stream_write: add missing cable_num to system messages,
SysEx, and real-time MIDI packets. Add 0xF mask for SysEx CIN checks.
Aligns with midi_device.c tud_midi_n_stream_write implementation.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
|
|
|
|
for clarity and flexibility
|
|
|
|
The existing tud_midi_n_stream_read() accepts a cable_num parameter but
ignores it — all cables share a single FIFO and stream parser state, so
data from different virtual cables is silently mixed together.
Add tud_midi_n_demux_stream_read() which returns the cable number of the
data that was actually read. It peeks at each USB-MIDI event packet
header before consuming it and stops when the next packet belongs to a
different cable, allowing callers to dispatch per-cable without losing
data.
Implementation details:
- Mirrors the host-side tuh_midi_stream_read() approach: tu_edpt_stream_peek
for cable inspection, CIN-based byte count (USB MIDI 1.0 Table 4-1),
leftover handling via existing midi_driver_stream_t
- *p_cable_num initialized to 0xff sentinel so callers can detect
"no data" even when return value is 0
- Cable-change check (total_read > 0 guard) covers both leftover-originated
reads and freshly consumed packets
- TU_VERIFY uses explicit != NULL comparisons, consistent with codebase style
- Note: shares stream->buffer with tud_midi_n_stream_read(); do not mix
calls on the same interface
- Adds single-interface convenience wrapper tud_midi_demux_stream_read()
Closes #1838
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
|
|
of bool. help to simplify parsing configuration
|
|
|
|
|
|
|
|
tu_edpt_stream_open() does not clear fifo, allow for persistent stream when disconnect/reconnect
|
|
also add tud_midi_n_packet_write/read_n()
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
fix compiling with nuc family
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
Signed-off-by: HiFiPhile <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
512 bytes for bulk in fullspeed mode.
|
|
|
|
change tuh_midi_rx/tx_cb() to have xferred_bytes
rename tuh_midi_get_num_rx/tx_cables() to tuh_midi_get_rx/tx_cable_count()
use default empty callback instead of weak null to be compatible with keil compiler
|
|
|
|
than 1 midi per device.
|
|
|
|
to parse/extract this information if needed
rename tuh_midi_configure() to mounted() for consistency
|
|
- add tuh_midi_packet_read_n() and tuh_midi_packet_write_n()
- add CFG_TUH_MIDI_STREAM_API to opt out stream API
|
|
- rename midi_stream_t to midi_driver_stream_t and move to midi.h (common for device and host)
|
|
xfer_cb as well as ep read()
- de-dup tuh_midi_get_num_rx/tx_cables
- add tuh_midi_read_available()
|
|
|
|
- comment out tuh_descriptor_device_cb/tuh_desc_configuration_cb since it is unrelated to this PR
|
|
|
|
|
|
|
