ugh

CMakeLists.txt

@@ -1,4 +1,5 @@
 option(USE_USBCDC_FOR_STDIO "Export an extra USB-CDC interface for stdio, instead of echoing it to a UART port (and requiring UART loopback for receiving stdio output on a host computer)." OFF)
+option(PERSISTENT_STORAGE "Use persistent storage (usually on-chip/on-board flash) to save the current mode, and settings of modes" ON)
 set(FAMILY "rp2040" CACHE STRING "Board/MCU family, decides which drivers to use. Set to RP2040 by default.")
 set(BOARD "raspberry_pi_pico" CACHE STRING "Board used, determines the pinout. Defaults to the Raspberry Pi Pico.")
 
@@ -45,11 +46,12 @@ if(FAMILY STREQUAL "rp2040")
   # need uart stdio, usb is busy doing other stuff
   if(USE_USBCDC_FOR_STDIO)
     # we're going to manually implement this case
-	#pico_enable_stdio_uart(${PROJECT} 0)
+    #pico_enable_stdio_uart(${PROJECT} 0)
-    target_compile_definitions(${PROJECT} PUBLIC USE_USBCDC_FOR_STDIO=1 PICO_STDIO_USB=1)
+    target_compile_definitions(${PROJECT} PUBLIC USE_USBCDC_FOR_STDIO=1)
   else()
-	  #pico_enable_stdio_uart(${PROJECT} 1)
+    #pico_enable_stdio_uart(${PROJECT} 1)
   endif()
+  # TODO: separate flag for disabling this one?
   pico_enable_stdio_uart(${PROJECT} 1)
   pico_enable_stdio_usb(${PROJECT} 0)
 else()
@@ -69,6 +71,8 @@ target_sources(${PROJECT} PUBLIC
   ${CMAKE_CURRENT_SOURCE_DIR}/src/main.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/alloc.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/modeset.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/src/storage.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/src/storage_save.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/thread.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/tusb_plt.S
   ${CMAKE_CURRENT_SOURCE_DIR}/src/usb_descriptors.c
@@ -81,6 +85,8 @@ target_sources(${PROJECT} PUBLIC
   ${CMAKE_CURRENT_SOURCE_DIR}/src/m_jscan/jscan.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/m_sump/_sump.c
   ${CMAKE_CURRENT_SOURCE_DIR}/src/m_sump/cdc_sump.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/src/m_ftdi/_ftdi.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/src/m_ftdi/ftdi.c
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/cdc_uart.c
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/dap_jtag.c
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/dap_swd.c
@@ -91,12 +97,25 @@ target_sources(${PROJECT} PUBLIC
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/tempsensor.c
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_jscan/jscan_hw.c
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_sump/sump_hw.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/ftdi_hw.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/ftdi_proto.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/uart.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/mpsse.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/asyncbb.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/syncbb.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/mcuhost.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/fifo.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/cpufifo.c
 )
 if(USE_USBCDC_FOR_STDIO)
+  target_compile_definitions(${PROJECT} PUBLIC USE_USBCDC_FOR_STDIO=1)
   target_sources(${PROJECT} PUBLIC
     ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/cdc_stdio.c
   )
 endif()
+if (PERSISTENT_STORAGE)
+  target_compile_definitions(${PROJECT} PUBLIC PERSISTENT_STORAGE=1)
+endif()
 target_include_directories(${PROJECT} PUBLIC
   ${CMAKE_CURRENT_SOURCE_DIR}/src/
   ${CMAKE_CURRENT_SOURCE_DIR}/libco/
@@ -108,8 +127,9 @@ target_include_directories(${PROJECT} PUBLIC
   ${CMAKE_CURRENT_SOURCE_DIR}/bsp/default/
 )
 
-set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wextra -Werror=implicit-function-declaration -Werror=return-type")
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wextra -Werror=implicit-function-declaration -Werror=return-type -Werror=maybe-uninitialized")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra")
+set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -Wl,--cref")
 
 add_custom_target(fix_db ALL WORKING_DIRECTORY ${OUTPUT_DIR}
   COMMAND "${CMAKE_CURRENT_SOURCE_DIR}/scripts/fix_clang_db.py")
@@ -141,6 +161,9 @@ if(FAMILY STREQUAL "rp2040")
   pico_generate_pio_header(${PROJECT} ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/swo_uart_rx.pio)
   pico_generate_pio_header(${PROJECT} ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_default/swo_manchester_encoding.pio)
 
+  pico_generate_pio_header(${PROJECT} ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/ftdi_uart_rx.pio)
+  pico_generate_pio_header(${PROJECT} ${CMAKE_CURRENT_SOURCE_DIR}/bsp/${FAMILY}/m_ftdi/ftdi_uart_tx.pio)
+
   pico_add_extra_outputs(${PROJECT})
 
 else()

bsp/rp2040/bsp-info.h

@@ -20,6 +20,11 @@
 #else
 #define CFG_TUD_CDC 2
 #endif
-#define CFG_TUD_VENDOR 1
+#define CFG_TUD_VENDOR 3
+
+/* don't access storage for RAM-only builds */
+#if !PICO_NO_FLASH
+#define DBOARD_HAS_STORAGE
+#endif
 
 #endif

bsp/rp2040/bsp-storage.h

@@ -0,0 +1,66 @@
+// vim: set et:
+
+#ifndef BSP_STORAGE_H_
+#define BSP_STORAGE_H_
+
+#include <stddef.h>
+#include <stdint.h>
+#include <stdbool.h>
+
+#if !PICO_NO_FLASH
+
+#include <string.h>
+
+#include <hardware/regs/addressmap.h>
+#include <hardware/regs/xip.h>
+#include <hardware/structs/xip_ctrl.h>
+#include <hardware/flash.h>
+#include <hardware/sync.h>
+#include <pico/bootrom.h>
+
+#ifndef PICO_FLASH_SIZE_BYTES
+#error "PICO_FLASH_SIZE_BYTES not defined"
+#endif
+
+static inline size_t storage_get_program_size(void) {
+    extern uint8_t __flash_binary_start, __flash_binary_end;
+
+    return (size_t)&__flash_binary_end - (size_t)&__flash_binary_start;
+}
+
+static inline size_t storage_get_program_offset(void) {
+    extern uint8_t __flash_binary_start;
+
+    return (size_t)&__flash_binary_start - XIP_BASE;
+}
+
+#define STORAGE_SIZE             PICO_FLASH_SIZE_BYTES
+#define STORAGE_ERASEWRITE_ALIGN FLASH_SECTOR_SIZE
+// reads don't require any alignment
+
+// ---
+
+static inline void storage_read(void* dest, size_t offset, size_t size) {
+    // TODO: XIP/SSI DMA?
+    // * XIP DMA: used for loading stuff in the background while running code
+    // * SSI DMA: blocking & fast, code needs to run from RAM. a bit unwieldy
+    memcpy(dest, (uint8_t*)(XIP_BASE+offset), size);
+}
+static bool storage_erasewrite(size_t offset, const void* src, size_t size) {
+    // bad alignment => give an error in advance
+    if (offset & (FLASH_SECTOR_SIZE - 1)) return false;
+    if (size   & (FLASH_SECTOR_SIZE - 1)) return false;
+
+    // memcmp pre: if no changes, don't flash
+    if (!memcmp(src, (uint8_t*)(XIP_BASE+offset), size)) return true;
+
+    flash_range_erase(offset, size);
+    flash_range_program(offset, src, size);
+
+    return !memcmp(src, (uint8_t*)(XIP_BASE+offset), size);
+}
+
+#endif
+
+#endif
+

bsp/rp2040/m_default/dap_swo.c

@@ -61,6 +61,7 @@ uint32_t SWO_Mode_UART(uint32_t enable) {
         mode_enabled = false;
 
         if (swo_dmach >= 0) {
+            dma_channel_abort(swo_dmach);
             dma_channel_unclaim(swo_dmach); // ugh why is it "dma_channel_xyz" and "dma_xyz_channel"
             swo_dmach = -1;
         }
@@ -189,6 +190,7 @@ uint32_t SWO_Mode_Manchester(uint32_t enable) {
         mode_enabled = false;
 
         if (swo_dmach >= 0) {
+            dma_channel_abort(swo_dmach);
             dma_channel_unclaim(swo_dmach); // ugh why is it "dma_channel_xyz" and "dma_xyz_channel"
             swo_dmach = -1;
         }

bsp/rp2040/m_default/pinout.h

@@ -1,7 +1,7 @@
 // vim: set et ts=8:
 
-#ifndef PINOUT_H_
+#ifndef BSP_PINOUT_M_DEFAULT_H_
-#define PINOUT_H_
+#define BSP_PINOUT_M_DEFAULT_H_
 
 // UART config
 #define PINOUT_UART_TX         8

bsp/rp2040/m_ftdi/asyncbb.c

@@ -0,0 +1,28 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_asyncbb_init(struct ftdi_interface* itf) {
+    (void)itf;
+
+}
+void ftdi_if_asyncbb_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+
+}
+void ftdi_if_asyncbb_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+
+}
+
+void ftdi_if_asyncbb_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize) {
+    (void)itf; (void)data; (void)datasize;
+
+}
+size_t ftdi_if_asyncbb_read(struct ftdi_interface* itf, uint8_t* data, size_t maxsize) {
+    (void)itf; (void)data; (void)maxsize;
+
+    return 0;
+}
+
+

bsp/rp2040/m_ftdi/bsp-feature.h

@@ -0,0 +1,34 @@
+
+#ifndef BSP_FEATURE_M_FTDI_H_
+#define BSP_FEATURE_M_FTDI_H_
+
+#define DBOARD_HAS_FTDI
+
+/* TODO: more fine-grained FTDI support/not-support stuff? */
+
+#include "bsp-info.h"
+
+// not all that much here
+
+enum {
+    HID_N__NITF = 0
+};
+enum {
+#ifdef USE_USBCDC_FOR_STDIO
+    CDC_N_STDIO = 0,
+#endif
+
+    CDC_N__NITF
+};
+enum {
+    /*VND_N_FTDI_IFA = 0,
+    VND_N_FTDI_IFB,*/
+    VND_N_CFG = 0,
+
+    VND_N__NITF
+};
+
+#define VND_N_FTDI_IFA 42
+#define VND_N_FTDI_IFB 69
+
+#endif

bsp/rp2040/m_ftdi/cpufifo.c

@@ -0,0 +1,23 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_cpufifo_init(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_cpufifo_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_cpufifo_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+}
+
+void ftdi_if_cpufifo_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize) {
+    (void)itf; (void)data; (void)datasize;
+}
+size_t ftdi_if_cpufifo_read(struct ftdi_interface* itf, uint8_t* data, size_t maxsize) {
+    (void)itf; (void)data; (void)maxsize;
+
+    return 0;
+}
+

bsp/rp2040/m_ftdi/fifo.c

@@ -0,0 +1,22 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_fifo_init(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_fifo_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_fifo_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+}
+
+void ftdi_if_fifo_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize) {
+    (void)itf; (void)data; (void)datasize;
+}
+size_t ftdi_if_fifo_read(struct ftdi_interface* itf, uint8_t* data, size_t maxsize) {
+    (void)itf; (void)data; (void)maxsize;
+    return 0;
+}
+

bsp/rp2040/m_ftdi/ftdi_hw.c

@@ -0,0 +1,315 @@
+// vim: set et:
+
+/* include order matters here */
+#include <hardware/dma.h>
+#include <hardware/gpio.h>
+#include <hardware/irq.h>
+#include <hardware/pio.h>
+#include <hardware/structs/dma.h>
+
+#include "m_ftdi/ftdi_hw.h"
+
+struct ftdi_hw itf_bsp_data[2];
+
+static int piosm_to_dreq(int pio, int sm) {
+    return DREQ_PIO0_TX0 + pio * 8 + sm;
+}
+
+static void ftdihw_dma_isr();
+
+void ftdihw_init(struct ftdi_hw* fr, struct ftdi_interface* itf) {
+    memset(fr, 0, sizeof *fr);
+
+    fr->itf = itf;
+    fr->pio = ftdihw_itf_to_pio(itf);
+    fr->pinbase = ftdihw_itf_to_base(itf);
+
+    fr->rx.prg_off = 0xff;
+    fr->rx.piosm = 0xff;
+    fr->rx.dmach = 0xff;
+    fr->tx.prg_off = 0xff;
+    fr->tx.piosm = 0xff;
+    fr->tx.dmach = 0xff;
+
+    // we start with nothing to write out, but we can start filling a buffer already
+    fr->rx.dmabuf_dend = sizeof(fr->dma_in_buf) - 1;
+
+    irq_set_enabled(DMA_IRQ_0 + (fr->itf->index & 1), false);
+    irq_set_exclusive_handler(DMA_IRQ_0 + (fr->itf->index & 1), ftdihw_dma_isr);
+    irq_set_enabled(DMA_IRQ_0 + (fr->itf->index & 1), true);
+}
+void ftdihw_deinit(struct ftdi_hw* hw) {
+    irq_set_enabled(DMA_IRQ_0 + (hw->itf->index & 1), false);
+    irq_remove_handler(DMA_IRQ_0 + (hw->itf->index & 1), ftdihw_dma_isr);
+}
+
+bool ftdihw_dma_ch_init(struct ftdi_hw_ch* ch, struct ftdi_hw* hw, const void* prg) {
+    int off, sm, dmach;
+    sm = pio_claim_unused_sm(hw->pio, false);
+    if (sm == -1) return false;
+
+    dmach = dma_claim_unused_channel(false);
+    if (dmach == -1) {
+        pio_sm_unclaim(hw->pio, sm);
+        return false;
+    }
+
+    if (!pio_can_add_program(hw->pio, prg)) {
+        dma_channel_unclaim(dmach);
+        pio_sm_unclaim(hw->pio, sm);
+        return false;
+    }
+
+    off = pio_add_program(hw->pio, prg);
+
+    ch->prg = prg;
+    ch->prg_off = off;
+    ch->piosm = sm;
+    ch->dmach = dmach;
+
+    return true;
+}
+void ftdihw_dma_ch_deinit(struct ftdi_hw_ch* ch, struct ftdi_hw* hw) {
+    dma_channel_unclaim(ch->dmach);
+    pio_sm_set_enabled(hw->pio, ch->piosm, false);
+    pio_sm_unclaim(hw->pio, ch->piosm);
+    pio_remove_program(hw->pio, ch->prg, ch->prg_off);
+
+    ch->dmach = 0xff;
+    ch->piosm = 0xff;
+    ch->prg_off = 0xff;
+    ch->prg = NULL;
+}
+
+void ftdihw_dma_rx_setup(struct ftdi_hw* hw, bool start) {
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->rx.dmach, false);
+
+    if (hw->rx.dmabuf_dend == hw->rx.dmabuf_dstart)
+        --hw->rx.dmabuf_dend; // mod 256 automatically
+
+    dma_channel_config dcfg = dma_channel_get_default_config(hw->rx.dmach);
+    channel_config_set_read_increment(&dcfg, false);
+    channel_config_set_write_increment(&dcfg, true);
+    channel_config_set_dreq(&dcfg, piosm_to_dreq(hw->itf->index, hw->rx.piosm));
+    channel_config_set_transfer_data_size(&dcfg, DMA_SIZE_8);
+    channel_config_set_ring(&dcfg, true, 8); // 1<<8 -sized ring buffer on write end
+    dma_channel_configure(hw->rx.dmach, &dcfg,
+            &hw->dma_in_buf[hw->rx.dmabuf_dstart], &hw->pio->rxf[hw->rx.piosm],
+            (hw->rx.dmabuf_dend - hw->rx.dmabuf_dstart) % sizeof(hw->dma_in_buf), start);
+
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->rx.dmach, true);
+}
+void ftdihw_dma_rx_stop(struct ftdi_hw* hw) {
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->rx.dmach, false);
+    dma_channel_abort(hw->rx.dmach);
+}
+
+void ftdihw_dma_tx_setup(struct ftdi_hw* hw, bool start) {
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->tx.dmach, false);
+
+    if (hw->tx.dmabuf_dend == hw->tx.dmabuf_dstart)
+        --hw->tx.dmabuf_dend; // mod 256 automatically
+
+    dma_channel_config dcfg = dma_channel_get_default_config(hw->tx.dmach);
+    channel_config_set_read_increment(&dcfg, true);
+    channel_config_set_write_increment(&dcfg, false);
+    channel_config_set_dreq(&dcfg, piosm_to_dreq(hw->itf->index, hw->tx.piosm));
+    channel_config_set_transfer_data_size(&dcfg, DMA_SIZE_8);
+    channel_config_set_ring(&dcfg, false, 8); // 1<<8 -sized ring buffer on read end
+    dma_channel_configure(hw->tx.dmach, &dcfg,
+            &hw->pio->txf[hw->tx.piosm], &hw->dma_out_buf[hw->tx.dmabuf_dstart],
+            (hw->tx.dmabuf_dend - hw->tx.dmabuf_dstart) % sizeof(hw->dma_out_buf), start);
+
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->tx.dmach, true);
+}
+void ftdihw_dma_tx_stop(struct ftdi_hw* hw) {
+    dma_irqn_set_channel_enabled(hw->itf->index & 1, hw->tx.dmach, false);
+    dma_channel_abort(hw->tx.dmach);
+}
+
+size_t ftdihw_dma_read(struct ftdi_hw* hw, uint8_t* dest, size_t maxsize) {
+    // DMA is sending data from PIO to between dstart and dend, and is planned
+    // to continue up to dataend. when it finishes an xfer (between dstart and
+    // dend), it looks at dataend to see if it should continue sending more. if
+    // not, it pauses itself as tehre's no more work to be done, and sets a
+    // FIFO overrun flag in the FTDI error things
+    //
+    // here we read from dataend to dstart, careful not to touch the area the
+    // DMA is currently writing to (while holding the channel paused), and then
+    // moving the dataend marker. the channel is always reenabled, as either it
+    // was busy and now it has more space to put data in, or needs to be
+    // restarted as there is space now
+    //
+    // pausing the DMA channel is ok as it'll pause the PIO FIFO a bit, but
+    // nothing drastic
+
+    if (maxsize == 0) return 0;
+
+    size_t rv = 0;
+
+    // TODO: make time between pause & resume shorter by moving stuff around?
+    bool wasbusy = dma_hw->ch[hw->rx.dmach].ctrl_trig & DMA_CH0_CTRL_TRIG_BUSY_BITS;
+    hw_clear_bits(&dma_hw->ch[hw->rx.dmach].ctrl_trig, DMA_CH0_CTRL_TRIG_EN_BITS); // pause (also prevents IRQ)
+
+    // dstart can get modified by the IRQ; IRQ reads dataend
+    uint8_t dstart = hw->rx.dmabuf_dstart, dataend = hw->rx.dmabuf_dataend;
+    if (dataend > dstart) dstart += sizeof(hw->dma_in_buf);
+    // nothing ready yet - bail out
+    if (dstart == (dataend + 1) % sizeof(hw->dma_in_buf)) {
+        goto END;
+    }
+
+    __compiler_memory_barrier();
+
+    // copy from data in ringbuffer that was read in by PIO
+    rv = (size_t)dstart - (size_t)dataend - 1;
+    if (rv > maxsize) rv = maxsize;
+    for (size_t i = 0; i < rv; ++i) {
+        dest[i] = hw->dma_in_buf[(dataend + i) % sizeof(hw->dma_in_buf)];
+    }
+    uint8_t dataend_new = (dataend + rv) % sizeof(hw->dma_in_buf);
+
+    hw->rx.dmabuf_dataend = dataend_new;
+
+    hw->itf->modemstat &= ~sio_modem_fifoerr;
+
+END:
+    if (!wasbusy) {
+        hw->rx.dmabuf_dstart = hw->rx.dmabuf_dend; // not required to set in DMA hw, but need to keep track of it in code
+        hw->rx.dmabuf_dend = hw->rx.dmabuf_dataend;
+        ftdihw_dma_rx_setup(hw, true);
+    } else // if already busy easlier, simply reenable
+        hw_set_bits(&dma_hw->ch[hw->rx.dmach].ctrl_trig, DMA_CH0_CTRL_TRIG_EN_BITS); // resume (also reenables IRQ)
+    return rv;
+    return rv;
+}
+bool ftdihw_dma_write(struct ftdi_hw* hw, const uint8_t* src, size_t size) {
+    // DMA is sending data between dstart and dend to PIO, and is planned to
+    // continue up to dataend. when it finishes an xfer (between dstart and
+    // dend), it looks at dataned to see if it should continue sending more. if
+    // not, it pauses itself as there's no more work to be done
+    //
+    // here we insert some data in the ring buffer, careful to not overwrite
+    // what the DMA is currently transferring (while holding the channel
+    // paused), and then moving the dataend marker. the channel is always
+    // reenabled, as either it was busy and now has more work to do, or needs
+    // to be restarted as there's data now
+    //
+    // pausing the DMA channel is ok as it'll pause the PIO FIFO a bit, but
+    // nothing drastic
+
+    if (size >= 255) return false; // don't even bother
+
+    bool rv = true;
+
+    // TODO: make time between pause & resume shorter by moving stuff around?
+    bool wasbusy = dma_hw->ch[hw->tx.dmach].ctrl_trig & DMA_CH0_CTRL_TRIG_BUSY_BITS;
+    hw_clear_bits(&dma_hw->ch[hw->tx.dmach].ctrl_trig, DMA_CH0_CTRL_TRIG_EN_BITS); // pause (also prevents IRQ)
+
+    // dstart can get modified by the IRQ; IRQ reads dataend
+    uint8_t dstart = hw->tx.dmabuf_dstart, dataend = hw->tx.dmabuf_dataend;
+    if (dataend > dstart) dstart += sizeof(hw->dma_out_buf);
+    // no space to put it in - overrun error
+    if ((size_t)dstart - (size_t)dataend < size + 1) {
+        hw->itf->modemstat |= sio_modem_fifoerr;
+        rv = false;
+        goto END;
+    }
+
+    __compiler_memory_barrier();
+
+    // copy data to buffer, to be copied next
+    for (size_t i = 0; i < size; ++i) {
+        hw->dma_out_buf[(dataend + i) % sizeof(hw->dma_out_buf)] = src[i];
+    }
+    uint8_t dataend_new = (dataend + size) % sizeof(hw->dma_out_buf);
+
+    hw->tx.dmabuf_dataend = dataend_new;
+
+    hw->itf->modemstat &= ~sio_modem_temt;
+
+END:
+    if (!wasbusy) {
+        hw->tx.dmabuf_dstart = hw->tx.dmabuf_dend; // not required to set in DMA hw, but need to keep track of it in code
+        hw->tx.dmabuf_dend = hw->tx.dmabuf_dataend;
+        ftdihw_dma_tx_setup(hw, true);
+    } else // if already busy easlier, simply reenable
+        hw_set_bits(&dma_hw->ch[hw->tx.dmach].ctrl_trig, DMA_CH0_CTRL_TRIG_EN_BITS); // resume (also reenables IRQ)
+    return rv;
+}
+
+void ftdihw_dma_rx_flush(struct ftdi_hw* hw) {
+    ftdihw_dma_rx_stop(hw);
+    hw->rx.dmabuf_dstart = 0;
+    hw->rx.dmabuf_dend = 0;
+    hw->rx.dmabuf_dataend = 0;
+}
+void ftdihw_dma_tx_flush(struct ftdi_hw* hw) {
+    ftdihw_dma_tx_stop(hw);
+    hw->tx.dmabuf_dstart = 0;
+    hw->tx.dmabuf_dend = 0;
+    hw->tx.dmabuf_dataend = 0;
+}
+
+static void ftdihw_dma_rx_irq(struct ftdi_hw* hw) {
+    // an rx DMA transfer has finished. if in the meantime (between DMA xfer
+    // start and now) more room has become available, restart the DMA channel
+    // to write to the new space instead. otherwise, if there's no more space
+    // left, we're in a data overrun condition, so don't restart, and set the
+    // relevant FTDI error flags
+
+    uint8_t dend = hw->rx.dmabuf_dend, dataend = hw->rx.dmabuf_dataend;
+
+    if (dend == dataend) {
+        // data overrun, stop stuff (until read() restarts the DMA)
+        hw->itf->modemstat |= sio_modem_fifoerr;
+    } else {
+        hw->rx.dmabuf_dstart = dend;
+        hw->rx.dmabuf_dend = dataend;
+        ftdihw_dma_rx_setup(hw, true);
+    }
+}
+static void ftdihw_dma_tx_irq(struct ftdi_hw* hw) {
+    // a tx DMA transfer has finished. if in the meantile (between DMA xfer
+    // start and now) more data has become available, restart the DMA channel
+    // to read from the new data instead. otherwise, if there's no more data to
+    // be read, we have nothing to do and we can leave the DMA channel in its
+    // idle state.
+
+    uint8_t dend = hw->tx.dmabuf_dend, dataend = hw->tx.dmabuf_dataend;
+
+    if (dend == dataend) {
+        // nothing to do
+        hw->itf->modemstat |= sio_modem_temt;
+    } else {
+        hw->tx.dmabuf_dstart = dend;
+        hw->tx.dmabuf_dend = dataend;
+        ftdihw_dma_tx_setup(hw, true);
+    }
+}
+
+static void ftdihw_dma_isr() {
+    // interrupt service routine: dispatch to functions above depending on INTS0
+    uint32_t flags = dma_hw->ints0;
+    uint32_t allflg = 1u << itf_bsp_data[0].rx.dmach;
+    allflg |= 1u << itf_bsp_data[0].tx.dmach;
+    allflg |= 1u << itf_bsp_data[1].rx.dmach;
+    allflg |= 1u << itf_bsp_data[1].tx.dmach;
+
+    for (size_t i = 0; (i < 2) && (flags & allflg); ++i) {
+        uint32_t flg = 1u << itf_bsp_data[i].rx.dmach;
+        if (flags & flg) {
+            flags &= ~flg;
+            dma_hw->ints0 = flg; // ack int
+            ftdihw_dma_rx_irq(&itf_bsp_data[i]);
+        }
+
+        flg = 1u << itf_bsp_data[i].tx.dmach;
+        if (flags & flg) {
+            flags &= ~flg;
+            dma_hw->ints0 = flg; // ack int
+            ftdihw_dma_tx_irq(&itf_bsp_data[i]);
+        }
+    }
+}
+

bsp/rp2040/m_ftdi/ftdi_hw.h

@@ -0,0 +1,73 @@
+// vim: set et:
+
+#ifndef FTDI_BASE_H_
+#define FTDI_BASE_H_
+
+#include <stdint.h>
+#include <stdbool.h>
+
+#include "m_ftdi/pinout.h"
+#include "m_ftdi/ftdi.h"
+
+struct ftdi_hw_ch {
+    const void* prg;
+
+    uint8_t prg_off, piosm, dmach;
+
+    volatile uint8_t dmabuf_dstart, dmabuf_dend, dmabuf_dataend;
+};
+
+struct ftdi_hw {
+    struct ftdi_interface* itf;
+    PIO pio;
+
+    struct ftdi_hw_ch rx, tx;
+    uint8_t pinbase;
+
+    volatile uint8_t dma_in_buf[256/*CFG_TUD_VENDOR_TX_BUFSIZE*/];
+    volatile uint8_t dma_out_buf[256/*CFG_TUD_VENDOR_RX_BUFSIZE*/];
+};
+
+extern struct ftdi_hw itf_bsp_data[2];
+
+void ftdihw_init(struct ftdi_hw* fr, struct ftdi_interface* itf);
+void ftdihw_deinit(struct ftdi_hw* fr);
+
+bool ftdihw_dma_ch_init(struct ftdi_hw_ch* ch, struct ftdi_hw* fr, const void* prg);
+void ftdihw_dma_ch_deinit(struct ftdi_hw_ch* fr, struct ftdi_hw* hw);
+
+void ftdihw_dma_rx_setup(struct ftdi_hw* fr, bool start);
+void ftdihw_dma_rx_stop(struct ftdi_hw* fr);
+
+void ftdihw_dma_tx_setup(struct ftdi_hw* fr, bool start);
+void ftdihw_dma_tx_stop(struct ftdi_hw* fr);
+
+size_t ftdihw_dma_read(struct ftdi_hw* fr, uint8_t* dest, size_t maxsize);
+bool ftdihw_dma_write(struct ftdi_hw* fr, const uint8_t* src, size_t size);
+
+void ftdihw_dma_rx_flush(struct ftdi_hw* fr);
+void ftdihw_dma_tx_flush(struct ftdi_hw* fr);
+
+static inline int ftdihw_idx_to_base(int itf_idx) {
+    return itf_idx ? PINOUT_ITF_B_BASE : PINOUT_ITF_A_BASE;
+}
+static inline int ftdihw_itf_to_base(struct ftdi_interface* itf) {
+    return ftdihw_idx_to_base(itf->index);
+}
+
+static inline PIO ftdihw_idx_to_pio(int itf_idx) {
+    return itf_idx ? PINOUT_ITF_A_PIO : PINOUT_ITF_B_PIO;
+}
+static inline PIO ftdihw_itf_to_pio(struct ftdi_interface* itf) {
+    return ftdihw_idx_to_pio(itf->index);
+}
+
+static inline struct ftdi_hw* ftdihw_idx_to_hw(int itf_idx) {
+    return &itf_bsp_data[itf_idx & 1];
+}
+static inline struct ftdi_hw* ftdihw_itf_to_hw(struct ftdi_interface* itf) {
+    return ftdihw_idx_to_hw(itf->index);
+}
+
+#endif
+

bsp/rp2040/m_ftdi/ftdi_proto.c

@@ -0,0 +1,100 @@
+// vim: set et:
+
+#include <hardware/pio.h>
+
+#include "m_ftdi/ftdi.h"
+#include "m_ftdi/ftdi_hw.h"
+
+static void init_mode(struct ftdi_interface* itf, enum ftdi_mode mode) {
+    switch (mode) {
+    case ftmode_uart: ftdi_if_uart_init(itf); ftdi_if_uart_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_mpsse: ftdi_if_mpsse_init(itf); ftdi_if_mpsse_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_asyncbb: ftdi_if_asyncbb_init(itf); ftdi_if_asyncbb_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_syncbb: ftdi_if_syncbb_init(itf); ftdi_if_syncbb_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_mcuhost: ftdi_if_mcuhost_init(itf); ftdi_if_mcuhost_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_fifo: ftdi_if_fifo_init(itf); ftdi_if_fifo_set_baudrate(itf, itf->baudrate); break;
+    case ftmode_cpufifo: ftdi_if_cpufifo_init(itf); ftdi_if_cpufifo_set_baudrate(itf, itf->baudrate); break;
+    default: break;
+    }
+}
+static void deinit_mode(struct ftdi_interface* itf, enum ftdi_mode mode) {
+    switch (mode) {
+    case ftmode_uart: ftdi_if_uart_deinit(itf); break;
+    case ftmode_mpsse: ftdi_if_mpsse_deinit(itf); break;
+    case ftmode_asyncbb: ftdi_if_asyncbb_deinit(itf); break;
+    case ftmode_syncbb: ftdi_if_syncbb_deinit(itf); break;
+    case ftmode_mcuhost: ftdi_if_mcuhost_deinit(itf); break;
+    case ftmode_fifo: ftdi_if_fifo_deinit(itf); break;
+    case ftmode_cpufifo: ftdi_if_cpufifo_deinit(itf); break;
+    default: break;
+    }
+}
+
+void ftdi_if_init(struct ftdi_interface* itf) {
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    ftdihw_init(hw, itf);
+
+    init_mode(itf, ftdi_if_get_mode(itf));
+}
+void ftdi_if_deinit(struct ftdi_interface* itf) {
+    deinit_mode(itf, ftdi_if_get_mode(itf));
+
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    ftdihw_deinit(hw);
+}
+
+void ftdi_if_set_modemctrl(struct ftdi_interface* itf, uint8_t mask, uint8_t data) {
+    (void)itf; (void)mask; (void)data;
+    // TODO: what's this?
+}
+void ftdi_if_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    switch (ftdi_if_get_mode(itf)) {
+    case ftmode_uart: ftdi_if_uart_set_baudrate(itf, baudrate); break;
+    case ftmode_mpsse: ftdi_if_mpsse_set_baudrate(itf, baudrate); break;
+    case ftmode_asyncbb: ftdi_if_asyncbb_set_baudrate(itf, baudrate); break;
+    case ftmode_syncbb: ftdi_if_syncbb_set_baudrate(itf, baudrate); break;
+    case ftmode_mcuhost: ftdi_if_mcuhost_set_baudrate(itf, baudrate); break;
+    case ftmode_fifo: ftdi_if_fifo_set_baudrate(itf, baudrate); break;
+    case ftmode_cpufifo: ftdi_if_cpufifo_set_baudrate(itf, baudrate); break;
+    default: break;
+    }
+}
+enum ftdi_sio_modemstat ftdi_if_poll_modemstat(struct ftdi_interface* itf) {
+    (void)itf;
+
+    return sio_modem_cts | sio_modem_dts; // TODO: use this to read part of UART flow ctrl?
+}
+void ftdi_if_set_eventchar(struct ftdi_interface* itf, bool enable, uint8_t evchar) {
+    (void)itf; (void)enable; (void)evchar;
+    // TODO: when is this used? bitmode0-only? also ftd2xx headers make this look like its not just an "event on char" thing
+}
+void ftdi_if_set_errorchar(struct ftdi_interface* itf, bool enable, uint8_t erchar) {
+    (void)itf; (void)enable; (void)erchar;
+    // TODO: when is this used? bitmode0-only? also ftd2xx headers make this look like its not just an "error on char" thing
+}
+uint8_t ftdi_if_read_pins(struct ftdi_interface* itf) {
+    (void)itf;
+
+    return 0; // TODO: which pins does this return?
+}
+
+void ftdi_if_set_bitbang(struct ftdi_interface* itf, uint8_t dirmask,
+        enum ftdi_sio_bitmode bitmode, uint8_t olddir, enum ftdi_sio_bitmode oldmode) {
+    if (bitmode == oldmode && dirmask == olddir) return; // nothing to do
+
+    deinit_mode(itf, ftdi_get_mode_of(oldmode, itf->index ? FTDI_EEP_IFB_MODE : FTDI_EEP_IFA_MODE));
+    init_mode(itf, ftdi_if_get_mode(itf));
+}
+
+void ftdi_if_sio_reset(struct ftdi_interface* itf) { (void)itf; /* TODO: ? */ }
+void ftdi_if_sio_tciflush(struct ftdi_interface* itf) {
+    (void)itf; /* TODO: ? */
+}
+void ftdi_if_sio_tcoflush(struct ftdi_interface* itf) {
+    (void)itf; /* TODO: ? */
+}
+void ftdi_if_set_latency(struct ftdi_interface* itf, uint8_t latency) { (void)itf; (void)latency; /* TODO: ? */ }
+uint8_t ftdi_if_get_latency(struct ftdi_interface* itf) { return itf->latency; /* TODO: ? */ }
+

bsp/rp2040/m_ftdi/ftdi_uart_rx.pio

@@ -0,0 +1,94 @@
+;
+; Copyright (c) 2020 Raspberry Pi (Trading) Ltd.
+;
+; SPDX-License-Identifier: BSD-3-Clause
+;
+
+.program ftdi_uart_rx_mini
+
+; Minimum viable 8n1 UART receiver. Wait for the start bit, then sample 8 bits
+; with the correct timing.
+; IN pin 0 is mapped to the GPIO used as UART RX.
+; Autopush must be enabled, with a threshold of 8.
+
+    wait 0 pin 0        ; Wait for start bit
+    set x, 7 [10]       ; Preload bit counter, delay until eye of first data bit
+bitloop:                ; Loop 8 times
+    in pins, 1          ; Sample data
+    jmp x-- bitloop [6] ; Each iteration is 8 cycles
+
+% c-sdk {
+#include "hardware/clocks.h"
+#include "hardware/gpio.h"
+
+static inline void ftdi_uart_rx_mini_program_init(PIO pio, uint sm, uint offset, uint pin, uint baud) {
+    pio_sm_set_consecutive_pindirs(pio, sm, pin, 1, false);
+    pio_gpio_init(pio, pin);
+    gpio_pull_up(pin);
+
+    pio_sm_config c = ftdi_uart_rx_mini_program_get_default_config(offset);
+    sm_config_set_in_pins(&c, pin); // for WAIT, IN
+    // Shift to right, autopush enabled
+    sm_config_set_in_shift(&c, true, true, 8);
+    sm_config_set_fifo_join(&c, PIO_FIFO_JOIN_RX);
+    // SM transmits 1 bit per 8 execution cycles.
+    float div = (float)clock_get_hz(clk_sys) / (8 * baud);
+    sm_config_set_clkdiv(&c, div);
+
+    pio_sm_init(pio, sm, offset, &c);
+    pio_sm_set_enabled(pio, sm, true);
+}
+%}
+
+.program ftdi_uart_rx
+
+; Slightly more fleshed-out 8n1 UART receiver which handles framing errors and
+; break conditions more gracefully.
+; IN pin 0 and JMP pin are both mapped to the GPIO used as UART RX.
+
+start:
+    wait 0 pin 0        ; Stall until start bit is asserted
+    set x, 7    [10]    ; Preload bit counter, then delay until halfway through
+bitloop:                ; the first data bit (12 cycles incl wait, set).
+    in pins, 1          ; Shift data bit into ISR
+    jmp x-- bitloop [6] ; Loop 8 times, each loop iteration is 8 cycles
+    jmp pin good_stop   ; Check stop bit (should be high)
+
+    irq 4 rel           ; Either a framing error or a break. Set a sticky flag,
+    wait 1 pin 0        ; and wait for line to return to idle state.
+    jmp start           ; Don't push data if we didn't see good framing.
+
+good_stop:              ; No delay before returning to start; a little slack is
+    push                ; important in case the TX clock is slightly too fast.
+
+
+% c-sdk {
+static inline void ftdi_uart_rx_program_init(PIO pio, uint sm, uint offset, uint pin, uint baud) {
+    pio_sm_set_consecutive_pindirs(pio, sm, pin, 1, false);
+    pio_gpio_init(pio, pin);
+    gpio_pull_up(pin);
+
+    pio_sm_config c = ftdi_uart_rx_program_get_default_config(offset);
+    sm_config_set_in_pins(&c, pin); // for WAIT, IN
+    sm_config_set_jmp_pin(&c, pin); // for JMP
+    // Shift to right, autopull disabled
+    sm_config_set_in_shift(&c, true, false, 32);
+    // Deeper FIFO as we're not doing any TX
+    sm_config_set_fifo_join(&c, PIO_FIFO_JOIN_RX);
+    // SM transmits 1 bit per 8 execution cycles.
+    float div = (float)clock_get_hz(clk_sys) / (8 * baud);
+    sm_config_set_clkdiv(&c, div);
+
+    pio_sm_init(pio, sm, offset, &c);
+    pio_sm_set_enabled(pio, sm, true);
+}
+
+static inline char ftdi_uart_rx_program_getc(PIO pio, uint sm) {
+    // 8-bit read from the uppermost byte of the FIFO, as data is left-justified
+    io_rw_8 *rxfifo_shift = (io_rw_8*)&pio->rxf[sm] + 3;
+    while (pio_sm_is_rx_fifo_empty(pio, sm))
+        tight_loop_contents();
+    return (char)*rxfifo_shift;
+}
+
+%}

bsp/rp2040/m_ftdi/ftdi_uart_tx.pio

@@ -0,0 +1,61 @@
+;
+; Copyright (c) 2020 Raspberry Pi (Trading) Ltd.
+;
+; SPDX-License-Identifier: BSD-3-Clause
+;
+
+.program ftdi_uart_tx
+.side_set 1 opt
+
+; An 8n1 UART transmit program.
+; OUT pin 0 and side-set pin 0 are both mapped to UART TX pin.
+
+    pull       side 1 [7]  ; Assert stop bit, or stall with line in idle state
+    set x, 7   side 0 [7]  ; Preload bit counter, assert start bit for 8 clocks
+bitloop:                   ; This loop will run 8 times (8n1 UART)
+    out pins, 1            ; Shift 1 bit from OSR to the first OUT pin
+    jmp x-- bitloop   [6]  ; Each loop iteration is 8 cycles.
+
+
+% c-sdk {
+#include "hardware/clocks.h"
+
+static inline void ftdi_uart_tx_program_init(PIO pio, uint sm, uint offset, uint pin_tx, uint baud) {
+    // Tell PIO to initially drive output-high on the selected pin, then map PIO
+    // onto that pin with the IO muxes.
+    pio_sm_set_pins_with_mask(pio, sm, 1u << pin_tx, 1u << pin_tx);
+    pio_sm_set_pindirs_with_mask(pio, sm, 1u << pin_tx, 1u << pin_tx);
+    pio_gpio_init(pio, pin_tx);
+
+    pio_sm_config c = ftdi_uart_tx_program_get_default_config(offset);
+
+    // OUT shifts to right, no autopull
+    sm_config_set_out_shift(&c, true, false, 32);
+
+    // We are mapping both OUT and side-set to the same pin, because sometimes
+    // we need to assert user data onto the pin (with OUT) and sometimes
+    // assert constant values (start/stop bit)
+    sm_config_set_out_pins(&c, pin_tx, 1);
+    sm_config_set_sideset_pins(&c, pin_tx);
+
+    // We only need TX, so get an 8-deep FIFO!
+    sm_config_set_fifo_join(&c, PIO_FIFO_JOIN_TX);
+
+    // SM transmits 1 bit per 8 execution cycles.
+    float div = (float)clock_get_hz(clk_sys) / (8 * baud);
+    sm_config_set_clkdiv(&c, div);
+
+    pio_sm_init(pio, sm, offset, &c);
+    pio_sm_set_enabled(pio, sm, true);
+}
+
+static inline void ftdi_uart_tx_program_putc(PIO pio, uint sm, char c) {
+    pio_sm_put_blocking(pio, sm, (uint32_t)c);
+}
+
+static inline void ftdi_uart_tx_program_puts(PIO pio, uint sm, const char *s) {
+    while (*s)
+        ftdi_uart_tx_program_putc(pio, sm, *s++);
+}
+
+%}

bsp/rp2040/m_ftdi/mcuhost.c

@@ -0,0 +1,38 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_mcuhost_init(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mcuhost_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mcuhost_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+}
+
+void ftdi_if_mcuhost_flush(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mcuhost_wait_io(struct ftdi_interface* itf, bool level) {
+    (void)itf; (void)level;
+}
+
+uint8_t ftdi_if_mcuhost_read8(struct ftdi_interface* itf, uint8_t addr) {
+    (void)itf; (void)addr;
+
+    return 0;
+}
+uint8_t ftdi_if_mcuhost_read16(struct ftdi_interface* itf, uint16_t addr) {
+    (void)itf; (void)addr;
+
+    return 0;
+}
+void ftdi_if_mcuhost_write8(struct ftdi_interface* itf, uint8_t addr, uint8_t value) {
+    (void)itf; (void)addr; (void)value;
+}
+void ftdi_if_mcuhost_write16(struct ftdi_interface* itf, uint16_t addr, uint8_t value) {
+    (void)itf; (void)addr; (void)value;
+}
+

bsp/rp2040/m_ftdi/mpsse.c

@@ -0,0 +1,82 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_mpsse_init(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mpsse_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mpsse_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+}
+
+
+void ftdi_if_mpsse_flush(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_mpsse_wait_io(struct ftdi_interface* itf, bool level) {
+    (void)itf; (void)level;
+}
+
+void ftdi_if_mpsse_set_dirval_lo(struct ftdi_interface* itf, uint8_t dir, uint8_t val) {
+    (void)itf; (void)dir; (void)val;
+}
+void ftdi_if_mpsse_set_dirval_hi(struct ftdi_interface* itf, uint8_t dir, uint8_t val) {
+    (void)itf; (void)dir; (void)val;
+}
+uint8_t ftdi_if_mpsse_read_lo(struct ftdi_interface* itf) {
+    (void)itf;
+
+    return 0;
+}
+uint8_t ftdi_if_mpsse_read_hi(struct ftdi_interface* itf) {
+    (void)itf;
+
+    return 0;
+}
+void ftdi_if_mpsse_loopback(struct ftdi_interface* itf, bool enable) {
+    (void)itf; (void)enable;
+}
+void ftdi_if_mpsse_set_clkdiv(struct ftdi_interface* itf, uint16_t div) {
+    (void)itf; (void)div;
+}
+
+uint8_t ftdi_if_mpsse_xfer_bits(struct ftdi_interface* itf, int flags, size_t nbits, uint8_t value) {
+    (void)itf; (void)flags; (void)nbits; (void)value;
+
+    return 0;
+}
+void ftdi_if_mpsse_xfer_bytes(struct ftdi_interface* itf, int flags, size_t nbytes, uint8_t* dst, const uint8_t* src) {
+    (void)itf; (void)flags; (void)nbytes; (void)dst; (void)src;
+}
+uint8_t ftdi_if_mpsse_tms_xfer(struct ftdi_interface* itf, int flags, size_t nbits, uint8_t value) {
+    (void)itf; (void)flags; (void)nbits; (void)value;
+
+    return 0;
+}
+
+void ftdi_if_mpsse_div5(struct ftdi_interface* itf, bool enable) {
+    (void)itf; (void)enable;
+}
+void ftdi_if_mpsse_data_3ph(struct ftdi_interface* itf, bool enable) {
+    (void)itf; (void)enable;
+}
+void ftdi_if_mpsse_adaptive(struct ftdi_interface* itf, bool enable) {
+    (void)itf; (void)enable;
+}
+
+void ftdi_if_mpsse_clockonly(struct ftdi_interface* itf, uint32_t cycles) {
+    (void)itf; (void)cycles;
+}
+void ftdi_if_mpsse_clock_wait_io(struct ftdi_interface* itf, bool level) {
+    (void)itf; (void)level;
+}
+void ftdi_if_mpsse_clockonly_wait_io(struct ftdi_interface* itf, bool level, uint32_t cycles) {
+    (void)itf; (void)level; (void)cycles;
+}
+void ftdi_if_mpsse_hi_is_tristate(struct ftdi_interface* itf, uint16_t pinmask) {
+    (void)itf; (void)pinmask;
+}
+

bsp/rp2040/m_ftdi/pinout.h

@@ -0,0 +1,119 @@
+
+#ifndef BSP_PINOUT_M_FTDI_H_
+#define BSP_PINOUT_M_FTDI_H_
+
+/* NOTE: no C code here! needs to be include-able from PIO asm! */
+
+#define PINOUT_ITF_A_BASE  2
+#define PINOUT_ITF_B_BASE 14
+
+// TODO: ?  or just different SMs on the same PIO? would complicate things tho
+#define PINOUT_ITF_A_PIO pio0
+#define PINOUT_ITF_B_PIO pio1
+
+#define PINOUT_DBUS0_OFF  0
+#define PINOUT_DBUS1_OFF  1
+#define PINOUT_DBUS2_OFF  2
+#define PINOUT_DBUS3_OFF  3
+#define PINOUT_DBUS4_OFF  4
+#define PINOUT_DBUS5_OFF  5
+#define PINOUT_DBUS6_OFF  6
+#define PINOUT_DBUS7_OFF  7
+#define PINOUT_CBUS0_OFF  8
+#define PINOUT_CBUS1_OFF  9
+#define PINOUT_CBUS2_OFF 10
+#define PINOUT_CBUS3_OFF 11
+
+#define PINOUT_UART_TXD_OFF     0
+#define PINOUT_UART_RXD_OFF     1
+#define PINOUT_UART_nRTS_OFF    2
+#define PINOUT_UART_nCTS_OFF    3
+#define PINOUT_UART_nDTR_OFF    4
+#define PINOUT_UART_nDSR_OFF    5
+#define PINOUT_UART_nDCD_OFF    6
+#define PINOUT_UART_nRI_OFF     7
+#define PINOUT_UART_TXDEN_OFF   8
+#define PINOUT_UART_nSLEEP_OFF  9
+#define PINOUT_UART_nRXLED_OFF 10
+#define PINOUT_UART_nTXLED_OFF 11
+
+#define PINOUT_FIFO_D0_OFF    0
+#define PINOUT_FIFO_D1_OFF    1
+#define PINOUT_FIFO_D2_OFF    2
+#define PINOUT_FIFO_D3_OFF    3
+#define PINOUT_FIFO_D4_OFF    4
+#define PINOUT_FIFO_D5_OFF    5
+#define PINOUT_FIFO_D6_OFF    6
+#define PINOUT_FIFO_D7_OFF    7
+#define PINOUT_FIFO_nRXF_OFF  8
+#define PINOUT_FIFO_nTXE_OFF  9
+#define PINOUT_FIFO_nRD_OFF  10
+#define PINOUT_FIFO_WR_OFF   11
+
+#define PINOUT_BBANG_D0_OFF    0
+#define PINOUT_BBANG_D1_OFF    1
+#define PINOUT_BBANG_D2_OFF    2
+#define PINOUT_BBANG_D3_OFF    3
+#define PINOUT_BBANG_D4_OFF    4
+#define PINOUT_BBANG_D5_OFF    5
+#define PINOUT_BBANG_D6_OFF    6
+#define PINOUT_BBANG_D7_OFF    7
+#define PINOUT_BBANG_nWR0_OFF  8
+#define PINOUT_BBANG_nRD0_OFF  9
+#define PINOUT_BBANG_nWR1_OFF 10
+#define PINOUT_BBANG_nRD1_OFF 11
+
+#define PINOUT_MPSSE_TCK_CK_OFF  0
+#define PINOUT_MPSSE_TDI_DO_OFF  1
+#define PINOUT_MPSSE_TDO_DI_OFF  2
+#define PINOUT_MPSSE_TMS_CS_OFF  3
+#define PINOUT_MPSSE_GPIOL0_OFF  4
+#define PINOUT_MPSSE_GPIOL1_OFF  5
+#define PINOUT_MPSSE_GPIOL2_OFF  6
+#define PINOUT_MPSSE_GPIOL3_OFF  7
+#define PINOUT_MPSSE_GPIOH0_OFF  8
+#define PINOUT_MPSSE_GPIOH1_OFF  9
+#define PINOUT_MPSSE_GPIOH2_OFF 10
+#define PINOUT_MPSSE_GPIOH3_OFF 11
+
+#define PINOUT_MCUHOST_AD0_OFF    0
+#define PINOUT_MCUHOST_AD1_OFF    1
+#define PINOUT_MCUHOST_AD2_OFF    2
+#define PINOUT_MCUHOST_AD3_OFF    3
+#define PINOUT_MCUHOST_AD4_OFF    4
+#define PINOUT_MCUHOST_AD5_OFF    5
+#define PINOUT_MCUHOST_AD6_OFF    6
+#define PINOUT_MCUHOST_AD7_OFF    7
+#define PINOUT_MCUHOST_IO0_OFF    8
+#define PINOUT_MCUHOST_IO1_OFF    9
+#define PINOUT_MCUHOST_IORDY_OFF 10
+#define PINOUT_MCUHOST_OSC_OFF   11
+// ---
+#define PINOUT_MCUHOST_A8_OFF     0
+#define PINOUT_MCUHOST_A9_OFF     1
+#define PINOUT_MCUHOST_AA_OFF     2
+#define PINOUT_MCUHOST_AB_OFF     3
+#define PINOUT_MCUHOST_AC_OFF     4
+#define PINOUT_MCUHOST_AD_OFF     5
+#define PINOUT_MCUHOST_AE_OFF     6
+#define PINOUT_MCUHOST_AF_OFF     7
+#define PINOUT_MCUHOST_nCS_OFF    8
+#define PINOUT_MCUHOST_ALE_OFF    9
+#define PINOUT_MCUHOST_nRD_OFF   10
+#define PINOUT_MCUHOST_nWR_OFF   11
+
+#define PINOUT_CPUFIFO_D0_OFF   0
+#define PINOUT_CPUFIFO_D1_OFF   1
+#define PINOUT_CPUFIFO_D2_OFF   2
+#define PINOUT_CPUFIFO_D3_OFF   3
+#define PINOUT_CPUFIFO_D4_OFF   4
+#define PINOUT_CPUFIFO_D5_OFF   5
+#define PINOUT_CPUFIFO_D6_OFF   6
+#define PINOUT_CPUFIFO_D7_OFF   7
+#define PINOUT_CPUFIFO_nCS_OFF  8
+#define PINOUT_CPUFIFO_A0_OFF   9
+#define PINOUT_CPUFIFO_nRD_OFF 10
+#define PINOUT_CPUFIFO_nWR_OFF 11
+
+#endif
+

bsp/rp2040/m_ftdi/syncbb.c

@@ -0,0 +1,22 @@
+// vim: set et:
+
+#include "m_ftdi/ftdi.h"
+
+void ftdi_if_syncbb_init(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_syncbb_deinit(struct ftdi_interface* itf) {
+    (void)itf;
+}
+void ftdi_if_syncbb_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    (void)itf; (void)baudrate;
+}
+
+
+void ftdi_if_syncbb_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize) {
+
+}
+size_t ftdi_if_syncbb_read(struct ftdi_interface* itf, uint8_t* data, size_t maxsize) {
+    return 0;
+}
+

bsp/rp2040/m_ftdi/uart.c

@@ -0,0 +1,143 @@
+// vim: set et:
+
+#include <hardware/dma.h>
+#include <hardware/gpio.h>
+#include <hardware/pio.h>
+#include <hardware/structs/dma.h>
+
+#include "m_ftdi/pinout.h"
+#include "m_ftdi/ftdi_hw.h"
+
+#include "ftdi_uart_rx.pio.h"
+#include "ftdi_uart_tx.pio.h"
+
+struct uart_state {
+    uint32_t baudrate;
+    bool enabled;
+};
+
+static struct uart_state state[2] = {
+    (struct uart_state){ .baudrate = 115200, .enabled = false },
+    (struct uart_state){ .baudrate = 115200, .enabled = false },
+};
+
+#define STATEOF(itf) (state[(itf)->index & 1])
+
+// set up PIO->dma_in_buf DMA:
+// * src=pio dst=buf 8bit 256words pacing=pio
+// * IRQ: set overrun bit
+// * start it
+// * SETUP AT MODE ENTER, STOP AT MODE EXIT
+//
+// set up dma_out_buf->PIO DMA:
+// * src=buf dst=pio 8bit <num>words pacing=pio
+// * ~~no IRQ I think?~~ IRQ: next ringbuffer part (see below)
+// * DO NOT start it on mode enter!
+// * STOP AT MODE EXIT
+//
+// read routine:
+// * abort DMA
+// * copy data from dma_in_buf (read xfer len? == remaining of 256 bytes)
+// * resetup & start DMA:
+//
+// write routine:
+// * if DMA running: set overrun bit, bail out?
+//   * should use ringbuffer-like structure
+//   * pointers: dma start, dma end, data end (after dma, contiguous)
+//     * dma end can be calculated from DMA MMIO, but, race conditions so no
+//   * use DMA IRQ for next block (and wraparound: dma cannot do wraparound manually)
+//     * do not start next block if data end == dma start
+//   * can we set DMA xfer len while in-flight? datasheet p92 2.5.1: nope. sad
+//   * only bail out when data end == dma start - 1
+// * copy data to dma_out_buf
+// * set up & start DMA
+//
+// * what with buffers larger than 256 bytes?
+//   * just drop for now ig
+// * what is the actual FTDI buffer size??
+//   * device-dependent so aaaa
+// * which bits get set on errors?
+// * do TCIFLUSH/TCOFLUSH influence these buffers, or only the USB proto
+//   handling buffers?
+
+void ftdi_if_uart_init(struct ftdi_interface* itf) {
+    if (STATEOF(itf).enabled) return; // shrug
+
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    if (!ftdihw_dma_ch_init(&hw->rx, hw, &ftdi_uart_rx_program)) return;
+    if (!ftdihw_dma_ch_init(&hw->tx, hw, &ftdi_uart_tx_program)) {
+        ftdihw_dma_ch_deinit(&hw->rx, hw);
+        return;
+    }
+
+    int pin_rx = hw->pinbase + PINOUT_UART_RXD_OFF,
+        pin_tx = hw->pinbase + PINOUT_UART_TXD_OFF;
+
+    ftdi_uart_rx_program_init(hw->pio, hw->rx.piosm, hw->rx.prg_off,
+            pin_rx, STATEOF(itf).baudrate);
+    ftdi_uart_tx_program_init(hw->pio, hw->tx.piosm, hw->tx.prg_off,
+            pin_tx, STATEOF(itf).baudrate);
+
+    gpio_set_function(pin_rx, GPIO_FUNC_PIO0 + itf->index);
+    gpio_set_function(pin_tx, GPIO_FUNC_PIO0 + itf->index);
+
+    ftdihw_dma_rx_setup(hw, true);
+    ftdihw_dma_tx_setup(hw, false);
+
+    STATEOF(itf).enabled = true;
+}
+void ftdi_if_uart_deinit(struct ftdi_interface* itf) {
+    if (!STATEOF(itf).enabled) return; // shrug
+
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    int pin_rx = hw->pinbase + PINOUT_UART_RXD_OFF,
+        pin_tx = hw->pinbase + PINOUT_UART_TXD_OFF;
+
+    ftdihw_dma_rx_flush(hw);
+    ftdihw_dma_tx_flush(hw);
+
+    ftdihw_dma_ch_deinit(&hw->rx, hw);
+    ftdihw_dma_ch_deinit(&hw->tx, hw);
+
+    gpio_set_function(pin_rx, GPIO_FUNC_NULL);
+    gpio_set_function(pin_tx, GPIO_FUNC_NULL);
+    gpio_set_pulls(pin_rx, false, false);
+    gpio_set_pulls(pin_tx, false, false);
+
+    STATEOF(itf).enabled = false;
+}
+void ftdi_if_uart_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate) {
+    if (!STATEOF(itf).enabled) return;
+
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    int pin_rx = ftdihw_itf_to_base(itf) + PINOUT_UART_RXD_OFF,
+        pin_tx = ftdihw_itf_to_base(itf) + PINOUT_UART_TXD_OFF;
+
+    ftdi_uart_rx_program_init(hw->pio, hw->rx.piosm, hw->rx.prg_off, pin_rx, baudrate);
+    ftdi_uart_tx_program_init(hw->pio, hw->tx.piosm, hw->tx.prg_off, pin_tx, baudrate);
+
+    STATEOF(itf).baudrate = baudrate;
+}
+
+
+void ftdi_if_set_flowctrl(struct ftdi_interface* itf, enum ftdi_flowctrl flow) {
+    (void)itf;  (void)flow; // TODO: bluh
+}
+void ftdi_if_set_lineprop(struct ftdi_interface* itf, enum ftdi_sio_lineprop lineprop) {
+    (void)itf; (void)lineprop; // TODO: break, stop, parity, #bits
+}
+
+void ftdi_if_uart_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize) {
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    ftdihw_dma_write(hw, data, datasize);
+}
+size_t ftdi_if_uart_read(struct ftdi_interface* itf, uint8_t* data, size_t maxsize) {
+    struct ftdi_hw* hw = ftdihw_itf_to_hw(itf);
+
+    return ftdihw_dma_read(hw, data, maxsize);
+}
+

bsp/rp2040/pico/assert.h

@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2020 Raspberry Pi (Trading) Ltd.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#ifndef _PICO_ASSERT_H
+#define _PICO_ASSERT_H
+
+#include <stdbool.h>
+
+#ifdef __cplusplus
+
+#include <cassert>
+
+extern "C" {
+#else
+#include <assert.h>
+#endif
+
+// PICO_CONFIG: PARAM_ASSERTIONS_ENABLE_ALL, Global assert enable, type=bool, default=0, group=pico_base
+// PICO_CONFIG: PARAM_ASSERTIONS_DISABLE_ALL, Global assert disable, type=bool, default=0, group=pico_base
+
+#ifndef PARAM_ASSERTIONS_ENABLE_ALL
+#define PARAM_ASSERTIONS_ENABLE_ALL 0
+#endif
+
+#ifndef PARAM_ASSERTIONS_DISABLE_ALL
+#define PARAM_ASSERTIONS_DISABLE_ALL 0
+#endif
+
+#define PARAM_ASSERTIONS_ENABLED(x) ((PARAM_ASSERTIONS_ENABLED_ ## x || PARAM_ASSERTIONS_ENABLE_ALL) && !PARAM_ASSERTIONS_DISABLE_ALL)
+
+#define invalid_params_if(x, test) ({if (PARAM_ASSERTIONS_ENABLED(x)) assert(!(test));})
+#define valid_params_if(x, test) ({if (PARAM_ASSERTIONS_ENABLED(x)) assert(test);})
+#define hard_assert_if(x, test) ({if (PARAM_ASSERTIONS_ENABLED(x)) hard_assert(!(test), #x);})
+
+#ifdef NDEBUG
+/*extern void hard_assertion_failure(void);
+static inline void hard_assert(bool condition, ...) {
+    if (!condition)
+        hard_assertion_failure();
+}*/
+#define hard_assert(cond, ...) do { if (!(cond)) panic("Hard assert: " __FILE__); } while (0)
+#else
+#define hard_assert assert
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+#endif

host/dpctl/__init__.py

@@ -52,6 +52,12 @@ def dpctl_do(args: Any) -> int:
             return 1
         return devcmds.sump_overclock_set(conn, oven)
 
+    def sto_info(conn, args):
+        return devcmds.storage_info(conn)
+    def sto_flush(conn, args):
+        return devcmds.storage_flush(conn)
+    def sto_get(conn, args):
+        return devcmds.storage_get(conn, args.mode)
 
     #print(repr(args))
     cmds = {
@@ -59,6 +65,9 @@ def dpctl_do(args: Any) -> int:
         'get-mode-info': get_mode_info,
         'set-mode': set_mode,
         'bootloader': bootloader,
+        'storage-info': sto_info,
+        'storage-flush': sto_flush,
+        'storage-get': sto_get,
 
         'uart-cts-rts': uart_hw_flowctl,
         'tempsensor': tempsensor,
@@ -142,6 +151,15 @@ def main() -> int:
 
     bootloader = subcmds.add_parser("bootloader", help="Set the device in bootloader mode")
 
+    # persistent storage commands
+    storagehdr = subcmds.add_parser("storage-info", help="Get persistent storage info")
+
+    storageflush = subcmds.add_parser("storage-flush", help="Flush persistent storage data to storage medium")
+
+    storageget = subcmds.add_parser("storage-get", help="Get data of a particular mode")
+    storageget.add_argument('mode', default=None, nargs='?',
+                             help="Mode to get data of. Defaults to the current mode, 'all' means all modes.")
+
     # mode 1 commands
     usbhwfctl = subcmds.add_parser("uart-cts-rts", help="Get, enable/disable"+\
                                    " UART hardware flow control")

host/dpctl/commands.py

@@ -85,6 +85,56 @@ def set_mode(dev: DPDevice, mode: int) -> int:
 # ---
 
 
+def storage_info(dev: DPDevice) -> int:
+    try:
+        res = dev.storage_info()
+        print("magic: %s, version=%04x, current mode=%d, #modes=%d, DJB2(table)=%d" \
+              % (('OK' if res.magic == b'\xf0\x9f\x8f\xb3\xef\xb8\x8f\xe2\x80\x8d\xe2\x9a\xa7\xef\xb8\x8f' else 'BAD!'),
+                 res.version, res.curmode, res.nmodes, res.table_djb2))
+        for md in res.mode_data:
+            print("\tmode %d version %04x: 0x%x..+0x%x, DJB2=%d" % \
+                (md.mode, md.version, md.offset, md.datasize, md.data_djb2))
+        if len(res.mode_data) == 0:
+            print("No mode data");
+        return 0
+    except Exception as e:
+        print("Could not get storage info: %s" % str(e))
+        traceback.print_exc()
+        return 1
+
+
+def storage_flush(dev: DPDevice) -> int:
+    try:
+        res = dev.storage_flush()
+        print("storage saved" if res else "no write needed")
+        return 0
+    except Exception as e:
+        print("Could not flush persistent storage: %s" % str(e))
+        return 1
+
+
+def storage_get(dev: DPDevice, mode: str) -> int:
+    try:
+        if mode == 'all':
+            for m in dev.mode_info.keys():
+                res = dev.storage_get(mode)
+                print(repr(res)) # TODO
+            return 0
+        elif mode is None:
+            mode = dev.current_mode
+        else: mode = int(mode,0)
+
+        res = dev.storage_get(mode)
+        print(repr(res)) # TODO
+        return 0
+    except Exception as e:
+        print("Could not get storage data of mode %d: %s" % (mode, str(e)))
+        return 1
+
+
+# ---
+
+
 def uart_hw_flowctl_get(dev: DPDevice) -> int:
     try:
         res = dev.m1_usb_hw_flowctl_get()

host/dpctl/connection.py

@@ -32,10 +32,12 @@ class UsbConn(DevConn):
         import usb, usb.core
 
         cfg = dev.get_active_configuration()
+        #print("get active", cfg)
 
         if cfg is None:  # should be configured already, but eh
             dev.set_configuration()
             cfg = dev.get_active_configuration()
+            #print("set active", cfg)
 
             if cfg is None:
                 return "Couldn't get or set device configuration, aaaaa"
@@ -45,6 +47,7 @@ class UsbConn(DevConn):
                if i.bInterfaceClass == usb.CLASS_VENDOR_SPEC and
                   i.bInterfaceSubClass == UsbConn._SUBCLASS and
                   i.bInterfaceProtocol == UsbConn._PROTOCOL]
+        #print("vnd itf", itf)
 
         if len(itf) == 0:
             return "No vendor control interface found for device"
@@ -58,12 +61,18 @@ class UsbConn(DevConn):
         epin  = usb.util.find_descriptor(itf, custom_match =
             lambda e: usb.util.endpoint_direction(e.bEndpointAddress) == usb.util.ENDPOINT_IN)
 
+        #print("epout", epout, "epin", epin)
+
         try:
             # try to read the version number. if it throws, it means the usbdev
             # is in use by something else
+            #print("write")
             epout.write(b'\x00')
+            #print("wrote")
             resp = epin.read(4)
-        except usb.core.USBError:
+            #print("resp", resp)
+        except usb.core.USBError as e:
+            print("eep", e)
             return "Device is busy, already used by something else? (If you use "+\
                 "the kernel module, use a character device from /dev instead.)"
 
@@ -77,6 +86,7 @@ class UsbConn(DevConn):
         if verno > DevConn._VER_MAX:
             return "Version of device (%04x) too new, must be max. %04x" \
                 % (hex(verno, DevConn._VER_MAX))
+        #print("verno", verno)
 
         return UsbConn(dev, cfg, itf, epin, epout)
 
@@ -89,6 +99,8 @@ class UsbConn(DevConn):
         if dev is None or len(dev) != 1:
             return None
 
+        #print("devs", dev)
+
         rv = UsbConn._open_dev(dev[0])
         return None if isinstance(rv, str) else rv
 
@@ -127,6 +139,8 @@ class UsbConn(DevConn):
         if conntup is None:
             return "Could not open USB device '%s': not recognised" % conn
 
+        #print("conntup", conntup)
+
         dev = None
         if conn_busdev:
             if len(conntup) == 2:
@@ -148,9 +162,12 @@ class UsbConn(DevConn):
         return UsbConn._open_dev(dev[0])
 
     def read_raw(self, arr) -> int:
-        return self._epin.read(arr)
+        rv = self._epin.read(arr)
+        print("read", arr[:rv])
+        return rv
 
     def write_raw(self, b: bytes) -> int:
+        print("write", b)
         return self._epout.write(b)
 
     def __init__(self, dev, cfg, itf, epin, epout):
@@ -168,6 +185,7 @@ class UsbConn(DevConn):
 
         usb.util.release_interface(self._dev, self._itf)
         usb.util.dispose_resources(self._dev)
+        #print("released & disposed")
         self._epout = None
         self._epin  = None
         self._itf = None
@@ -205,6 +223,7 @@ class ChardevConn(DevConn):
                 "a Linux kernel module" % sys.platform
 
         try:
+            #print("try char", conn)
             fd = os.open(conn, os.O_RDWR)
             if fd < 0:
                 raise OSError("Negative file descriptor returned")

host/dpctl/protocol.py

@@ -18,6 +18,56 @@ STAT_BADARG     = 0x04
 STAT_ILLSTATE   = 0x05
 
 
+class StorageInfoMode(NamedTuple):
+    version: int
+    datasize: int
+    offset: int
+    mode: int
+    data_djb2: int
+
+    def from_bytes(b: bytes) -> StorageInfoMode:
+        assert len(b) == 12
+        v, ds, oam, d = struct.unpack('<HHII', b)
+        return StorageInfoMode(v, ds, oam & ((1<<28)-1), (oam >> 28) & 15, d)
+
+    def list_from_bytes(b: bytes) -> List[StorageInfoMode]:
+        nelem = len(b) // 12
+        assert nelem * 12 == len(b)
+
+        r = [None]*nelem
+        for i in range(nelem): r[i] = StorageInfoMode.from_bytes(b[(i*12):((i+1)*12)])
+        return [re for re in r if re.version != 0xffff and re.datasize != 0xffff]
+
+
+class StorageInfo(NamedTuple):
+    magic: bytes
+    version: int
+    curmode: int
+    nmodes: int
+    reserved: bytes
+    table_djb2: int
+    mode_data: List[StorageInfoMode]
+
+    def from_bytes(b: bytes) -> StorageInfo:
+        assert len(b) == 256
+
+        mag = b[:16]
+        ver, cm, nm = struct.unpack('<HBB', b[16:20])
+        res = b[20:28] + b[245-32:]
+        d2tab = struct.unpack('<I', b[28:32])[0]
+
+        mdat = StorageInfoMode.list_from_bytes(b[32:256-32])
+
+        assert len(mdat) == nm
+
+        return StorageInfo(mag, ver, cm, nm, res, d2tab, mdat)
+
+    def __str__(self):
+        return "StorageInfo(magic=%s, version=%d, curmode=%d, nmodes=%d, table_djb2=%d, mode_data=%s)" \
+            % (' '.join(hex(b) for b in self.magic), self.version,
+               self.curmode, self.nmodes, self.table_djb2, repr(self.mode_data))
+
+
 class JtagMatch(NamedTuple):
     tck: int
     tms: int
@@ -144,11 +194,11 @@ class DPDevice:
 
         if (plen & 0x80) != 0:
             plen &= 0x7f
-            plen |= self.read(1) << 7
+            plen |= self.read(1)[0] << 7
 
             if (plen & 0x4000) != 0:
                 plen &= 0x3fff
-                plen |= self.read(1) << 14
+                plen |= self.read(1)[0] << 14
 
         bs = self.read(plen)
         #print("==> got resp %d res %s" % (resp, repr(bs)))
@@ -236,6 +286,32 @@ class DPDevice:
 
         return { i for i in range(0, 8) if (pl[0] & (1<<i)) != 0 }
 
+    # persistent storage
+
+    def storage_info(self) -> StorageInfo:
+        cmd = bytearray(b'\x0c')
+        self.write(b'\x0c')
+        stat, pl = self.read_resp()
+        check_statpl(stat, pl, "get storage info", 256, 256)
+
+        return StorageInfo.from_bytes(pl)
+
+    def storage_flush(self) -> bool:
+        self.write(b'\x0e')
+        stat, pl = self.read_resp()
+        check_statpl(stat, pl, "flush storage", 1, 1)
+        return pl[0]
+
+    def storage_get(self, mode: int) -> bytes:
+        cmd = bytearray(b'\x0d\x00')
+        cmd[1] = mode
+        self.write(cmd)
+        stat, pl = self.read_resp()
+        check_statpl(stat, pl, "get storage data", None, None)
+
+        return pl # TODO: parse
+
+
     # mode 1 commands
 
     def m1_usb_hw_flowctl_get(self) -> bool:

src/.gitignore

@@ -0,0 +1 @@
+_storage_test/

src/m_default/_default.c

@@ -3,6 +3,7 @@
 #include <tusb.h>
 
 #include "mode.h"
+#include "storage.h"
 #include "thread.h"
 #include "usbstdio.h"
 #include "vnd_cfg.h"
@@ -35,6 +36,8 @@ enum m_default_feature {
     mdef_feat_tempsense = 1<<4,
 };
 
+static bool data_dirty = false;
+
 #ifdef DBOARD_HAS_UART
 static cothread_t uartthread;
 static uint8_t    uartstack[THREAD_STACK_SIZE];
@@ -81,6 +84,17 @@ static void enter_cb(void) {
     serprogthread = co_derive(serprogstack, sizeof serprogstack, serprog_thread_fn);
     thread_enter(serprogthread);  // will call cdc_serprog_init() on correct thread
 #endif
+
+    if (!data_dirty) { // only read when not read yet
+        struct mode_info mi = storage_mode_get_info(1);
+        if (mi.size != 0 && mi.version == 0x0010 /* TODO: version migration? */) {
+            uint8_t dst[2];
+            storage_mode_read(1, dst, 0, 2);
+
+            cdc_uart_set_hwflow(dst[0]);
+            tempsense_set_addr(dst[1]);
+        }
+    }
 }
 static void leave_cb(void) {
     // TODO: CMSISDAP?
@@ -144,7 +158,12 @@ static void handle_cmd_cb(uint8_t cmd) {
         break;
     case mdef_cmd_tempsense:
 #ifdef DBOARD_HAS_TEMPSENSOR
-        tempsense_bulk_cmd();
+        {
+            uint8_t addra = tempsense_get_addr();
+            tempsense_bulk_cmd();
+            uint8_t addrb = tempsense_get_addr();
+            data_dirty |= addra != addrb;
+        }
 #else
         vnd_cfg_write_str(cfg_resp_illcmd, "temperature sensor not implemented on this device");
 #endif
@@ -156,9 +175,10 @@ static void handle_cmd_cb(uint8_t cmd) {
             resp = cdc_uart_get_hwflow() ? 1 : 0;
             vnd_cfg_write_resp(cfg_resp_ok, 1, &resp);
         } else {
-            if (cdc_uart_set_hwflow(resp != 0))
+            if (cdc_uart_set_hwflow(resp != 0)) {
                 vnd_cfg_write_resp(cfg_resp_ok, 0, NULL);
-            else
+                data_dirty = true;
+            } else
                 vnd_cfg_write_str(cfg_resp_illcmd, "UART flow control setting not supported on this device");
         }
 #else
@@ -387,6 +407,19 @@ static bool my_vendor_control_xfer_cb(uint8_t rhport, uint8_t ep_addr,
 }
 #endif
 
+static uint16_t my_get_size(void) { return 2; }
+static void my_get_data(void* dst, size_t offset, size_t maxsize) {
+    (void)offset; (void)maxsize;
+
+    uint8_t* d = dst;
+
+    d[0] = cdc_uart_get_hwflow() ? 1 : 0;
+    d[1] = tempsense_get_addr();
+
+    data_dirty = false;
+}
+static bool my_is_dirty(void) { return data_dirty; }
+
 extern struct mode m_01_default;
 // clang-format off
 struct mode m_01_default = {
@@ -402,6 +435,13 @@ struct mode m_01_default = {
     .task  = task_cb,
     .handle_cmd = handle_cmd_cb,
 
+    .storage = {
+        .stclass = mode_storage_32b,
+        .get_size = my_get_size,
+        .get_data = my_get_data,
+        .is_dirty = my_is_dirty
+    },
+
 #if defined(DBOARD_HAS_CMSISDAP) && CFG_TUD_HID > 0
 #if 0
     .tud_hid_get_report_cb = my_hid_get_report_cb,

src/m_ftdi/_ftdi.c

@@ -0,0 +1,265 @@
+// vim: set et:
+
+#include "tusb_config.h"
+#include <tusb.h>
+
+#include "mode.h"
+#include "thread.h"
+#include "usbstdio.h"
+#include "vnd_cfg.h"
+
+#include "m_ftdi/bsp-feature.h"
+#include "m_ftdi/ftdi.h"
+
+static bool data_dirty = false;
+void ftdi_eeprom_dirty_set(bool v) { (void)v; data_dirty = true; }
+bool ftdi_eeprom_dirty_get(void) { return data_dirty; }
+
+#ifdef DBOARD_HAS_FTDI
+static cothread_t ftdithread_ifa, ftdithread_ifb;
+static uint8_t    ftdistack_ifa[THREAD_STACK_SIZE>>1], ftdistack_ifb[THREAD_STACK_SIZE>>1];
+
+static void ftdi_thread_fn_ifa(void) {
+
+    while (1) {
+        printf("fn ifa thread!\n");
+        //ftdi_task_ifa();
+        thread_yield();
+    }
+}
+static void ftdi_thread_fn_ifb(void) {
+    while (1) {
+        printf("fn ifb thread!\n");
+        //ftdi_task_ifb();
+        thread_yield();
+    }
+}
+#endif
+
+static void enter_cb(void) {
+    printf("mode5 enter begin VND_N_CONFIG=%d\n", VND_N_CFG);
+
+#ifdef USE_USBCDC_FOR_STDIO
+    stdio_usb_set_itf_num(CDC_N_STDIO);
+#endif
+    //vnd_cfg_set_itf_num(VND_N_CFG);
+
+#ifdef DBOARD_HAS_FTDI
+    /*ftdithread_ifa = co_derive(ftdistack_ifa, sizeof ftdistack_ifa, ftdi_thread_fn_ifa);
+    ftdithread_ifb = co_derive(ftdistack_ifb, sizeof ftdistack_ifb, ftdi_thread_fn_ifb);*/
+#endif
+
+    /*if (!data_dirty) {
+        struct mode_info mi = storage_mode_get_info(5);
+        if (mi.size != 0 && mi.version == 0x0010) {
+            storage_mode_read(5, ftdi_eeprom, 0, sizeof ftdi_eeprom);
+        }
+    }*/
+
+    //ftdi_init();
+    printf("mode5 enter end\n");
+}
+static void leave_cb(void) {
+    printf("mode5 leave\n");
+#ifdef DBOARD_HAS_FTDI
+    //ftdi_deinit();
+#endif
+}
+
+static void task_cb(void) {
+    printf("mode5 task\n");
+#ifdef DBOARD_HAS_FTDI
+    //tud_task();
+    //thread_enter(ftdithread_ifa);
+    //tud_task();
+    //thread_enter(ftdithread_ifb);
+#endif
+}
+
+static void handle_cmd_cb(uint8_t cmd) {
+    printf("mode5 handlecmd %02x\n", cmd);
+    uint8_t resp = 0;
+
+    switch (cmd) {
+    case mode_cmd_get_features:
+        resp = 0; // TODO: what do we put here?
+        vnd_cfg_write_resp(cfg_resp_ok, 1, &resp);
+        break;
+    default:
+        vnd_cfg_write_strf(cfg_resp_illcmd, "unknown mode5 command %02x", cmd);
+        break;
+    }
+}
+
+enum {
+    STRID_LANGID = 0,
+    STRID_MANUFACTURER,
+    STRID_PRODUCT,
+    STRID_SERIAL,
+
+    STRID_CONFIG,
+
+    STRID_IF_VND_CFG,
+    STRID_IF_VND_FTDI_IFA,
+    STRID_IF_VND_FTDI_IFB,
+    STRID_IF_CDC_STDIO,
+};
+enum {
+    /*ITF_NUM_VND_FTDI_IFA,
+    ITF_NUM_VND_FTDI_IFB,*/
+
+#if CFG_TUD_VENDOR > 0
+    ITF_NUM_VND_CFG,
+#endif
+#ifdef USE_USBCDC_FOR_STDIO
+    ITF_NUM_CDC_STDIO_COM,
+    ITF_NUM_CDC_STDIO_DATA,
+#endif
+
+    ITF_NUM__TOTAL
+};
+enum {
+    CONFIG_TOTAL_LEN
+        = TUD_CONFIG_DESC_LEN
+        /*+ TUD_VENDOR_DESC_LEN
+        + TUD_VENDOR_DESC_LEN*/
+#if CFG_TUD_VENDOR > 0
+        + TUD_VENDOR_DESC_LEN
+#endif
+#ifdef USE_USBCDC_FOR_STDIO
+        + TUD_CDC_DESC_LEN
+#endif
+};
+
+/*#define EPNUM_VND_FTDI_IFA_OUT  0x02
+#define EPNUM_VND_FTDI_IFA_IN   0x81
+#define EPNUM_VND_FTDI_IFB_OUT  0x04
+#define EPNUM_VND_FTDI_IFB_IN   0x83
+
+#define EPNUM_VND_CFG_OUT       0x05
+#define EPNUM_VND_CFG_IN        0x85
+#define EPNUM_CDC_STDIO_OUT     0x06
+#define EPNUM_CDC_STDIO_IN      0x86
+#define EPNUM_CDC_STDIO_NOTIF   0x87*/
+
+#define EPNUM_VND_CFG_OUT       0x01
+#define EPNUM_VND_CFG_IN        0x81
+#define EPNUM_CDC_STDIO_OUT     0x02
+#define EPNUM_CDC_STDIO_IN      0x82
+#define EPNUM_CDC_STDIO_NOTIF   0x83
+
+// clang-format off
+static const uint8_t desc_configuration[] = {
+    TUD_CONFIG_DESCRIPTOR(1, ITF_NUM__TOTAL, STRID_CONFIG, CONFIG_TOTAL_LEN,
+        TUSB_DESC_CONFIG_ATT_REMOTE_WAKEUP, 100),
+
+    /*TUD_VENDOR_DESCRIPTOR(ITF_NUM_VND_FTDI_IFA, STRID_IF_VND_FTDI_IFA,
+        EPNUM_VND_FTDI_IFA_OUT, EPNUM_VND_FTDI_IFA_IN, CFG_TUD_VENDOR_RX_BUFSIZE),
+    TUD_VENDOR_DESCRIPTOR(ITF_NUM_VND_FTDI_IFB, STRID_IF_VND_FTDI_IFB,
+        EPNUM_VND_FTDI_IFB_OUT, EPNUM_VND_FTDI_IFB_IN, CFG_TUD_VENDOR_RX_BUFSIZE),*/
+
+#if CFG_TUD_VENDOR > 0
+    TUD_VENDOR_DESCRIPTOR_EX(ITF_NUM_VND_CFG, STRID_IF_VND_CFG, EPNUM_VND_CFG_OUT,
+        EPNUM_VND_CFG_IN, CFG_TUD_VENDOR_RX_BUFSIZE, VND_CFG_SUBCLASS, VND_CFG_PROTOCOL),
+#endif
+
+#ifdef USE_USBCDC_FOR_STDIO
+    TUD_CDC_DESCRIPTOR(ITF_NUM_CDC_STDIO_COM, STRID_IF_CDC_STDIO, EPNUM_CDC_STDIO_NOTIF,
+        CFG_TUD_CDC_RX_BUFSIZE, EPNUM_CDC_STDIO_OUT, EPNUM_CDC_STDIO_IN, CFG_TUD_CDC_RX_BUFSIZE),
+#endif
+};
+static const char* string_desc_arr[] = {
+    NULL,
+
+    [STRID_CONFIG]            = "Configuration descriptor",
+    // max string length check:  |||||||||||||||||||||||||||||||
+    [STRID_IF_VND_CFG  ]      = "Device cfg/ctl interface",
+    [STRID_IF_VND_FTDI_IFA]   = "DragonProbe FT2232D interface A",
+    [STRID_IF_VND_FTDI_IFB]   = "DragonProbe FT2232D interface B",
+#ifdef USE_USBCDC_FOR_STDIO
+    [STRID_IF_CDC_STDIO]      = "stdio CDC interface (debug)",
+#endif
+};
+// clang-format on
+
+static tusb_desc_device_t desc_device = {
+    .bLength         = sizeof(tusb_desc_device_t),
+    .bDescriptorType = TUSB_DESC_DEVICE,
+    .bcdUSB          = 0x0110,  // TODO: 0x0200 ? is an eeprom option
+    .bDeviceClass    = 0x00,
+    .bDeviceSubClass = 0x00,
+    .bDeviceProtocol = 0x00,
+    .bMaxPacketSize0 = CFG_TUD_ENDPOINT0_SIZE,
+
+    .idVendor  = 0x0403, // ?
+    .idProduct = 0x6010, // ?
+    .bcdDevice = 0x0500, // required!
+
+    .iManufacturer = STRID_MANUFACTURER,
+    .iProduct      = STRID_PRODUCT,
+    .iSerialNumber = STRID_SERIAL,
+
+    .bNumConfigurations = 0x01
+};
+
+static const uint8_t* my_descriptor_device_cb(void) {
+    return (const uint8_t*)&desc_device;
+}
+
+#if CFG_TUD_CDC > 0
+static void my_cdc_line_coding_cb(uint8_t itf, cdc_line_coding_t const* line_coding) {
+    printf("mode5 linecoding %02x, %lu\n", itf, line_coding->bit_rate);
+    switch (itf) {
+#ifdef USE_USBCDC_FOR_STDIO
+        case CDC_N_STDIO:
+            stdio_usb_line_coding_cb(line_coding);
+            break;
+#endif
+    }
+}
+#endif
+
+static uint16_t my_get_size(void) { return sizeof ftdi_eeprom; }
+static void my_get_data(void* dst, size_t offset, size_t maxsize) {
+    memcpy(dst, (const uint8_t*)ftdi_eeprom + offset, maxsize);
+
+    data_dirty = false;
+}
+static bool my_is_dirty(void) { return data_dirty; }
+
+extern struct mode m_05_ftdi;
+// clang-format off
+struct mode m_05_ftdi = {
+    .name = "FTDI FT2232D emulation mode",
+    .version = 0x0010,
+    .n_string_desc = sizeof(string_desc_arr)/sizeof(string_desc_arr[0]),
+
+    .usb_desc = desc_configuration,
+    .string_desc = string_desc_arr,
+
+    .enter = enter_cb,
+    .leave = leave_cb,
+    .task  = task_cb,
+    .handle_cmd = handle_cmd_cb,
+
+    /*.storage = {
+        .stclass = mode_storage_512b,
+        .get_size = my_get_size,
+        .get_data = my_get_data,
+        .is_dirty = my_is_dirty
+    },*/
+
+#ifdef DBOARD_HAS_FTDI
+    //.tud_descriptor_device_cb = my_descriptor_device_cb,
+#endif
+
+#if CFG_TUD_CDC > 0
+    .tud_cdc_line_coding_cb = my_cdc_line_coding_cb,
+#endif
+
+#ifdef DBOARD_HAS_FTDI
+    //.tud_vendor_control_xfer_cb = ftdi_control_xfer_cb,
+#endif
+};
+// clang-format on
+

src/m_ftdi/ftdi.c

@@ -0,0 +1,486 @@
+// vim: set et:
+
+#include "tusb_config.h"
+#include <tusb.h>
+
+#include "thread.h"
+
+#include "m_ftdi/bsp-feature.h"
+#include "m_ftdi/ftdi.h"
+
+// index = interface number (A/B)
+//              out/in bRequest wValue wIndex data wLength
+// RESET:       out    0        0      index
+// TCIFLUSH:    out    0        2      index
+// TCOFLUSH:    out    0        1      index
+// SETMODEMCTRL:out    1        (mask:8<<8 | data:8) // bit0=dtr bit1=rts
+// SETFLOWCTRL: out    2        xon?1:0 (flowctrl | index) // flowctrl: 0=disable, 1=ctsrts, 2=dtrdsr 4=xonxoff FROM VALUE, not set/reset!
+// SETBAUDRATE: out    3        brate  brate24|index // 48 MHz clocks, /16? , baudrate is 24bit, highest byte in index MSB
+// SETLINEPROP: out    4        (break:1<<14 | stop:2<<11 | parity:3<<8 | bits:8) ; break: off/on, stop=1/15/2, parity=none/odd/even/mark/space; bits=7/8
+// POLLMODEMSTAT:in    5        0      index  &modemstat len=2 // first byte: bit0..3=0 bit4=cts bit5=dts bit6=ri bit7=rlsd ; second byte: bit0=dr bit1=oe bit2=pe bit3=fe bit4=bi bit5=thre bit6=temt bit7=fifoerr
+// SETEVENTCHAR:out    6        (endis:1<<8 | char)
+// SETERRORCHAR:out    7        (endir:1<<8 | char)
+//  <there is no bReqest 8>
+// SETLATENCY:  out    9        latency(1..255)
+// GETLATENCY:  in     0xa      0      index  &latency len=1
+// SETBITBANG:  out    0xb      dirmask:8<<8 | mode:8 // mode: MPSSE mode: 0=serial/fifo, 1=bitbang, 2=mpsse, 4=syncbb, 8=mcu, 16=opto
+// READPINS:    in     0xc      0      index  &pins len=1
+// READEEPROM:  in     0x90     0      eepaddr &val 2 // eepaddr in 16-bit units(?)
+// WRITEEEPROM: out    0x91     value  eepaddr
+// ERASEEEPROM: out    0x92     0      0
+
+// eeprom layout:
+// max size is 128 words (256 bytes), little-endian
+// 00: type, driver, .. stuff (chanA byte, chanB byte) -> TODO: what does this mean exactly?
+//     0=UART 1=FIFO 2=opto 4=CPUFIFO 8=FT1284
+// 01: VID
+// 02: PID
+// 03: bcdDevice
+// 04: byte08: flags: bit7=1 bit6=1=selfpowered bit5=1=remotewakeup ; byte09=max power, 2mA units
+// 05: bit0=epin isochr, bit1=epout isochr, bit2=suspend pulldn, bit3=serialno use, bit4=usbver change, bit5..7=0
+// 06: bcdUSB
+// 07: byte0E=manufstr.off-0x80, byte0F=manufstr.len (in bytes, but 16-bit ascii)
+// 08: ^ but product
+// 09: ^ but serial
+// 0a: chip type (66 etc -> we emulate '00', internal)
+// "strings start at 0x96 for ft2232c" (byte addr)
+// checksum:
+//     initial: 0xaaaa
+//     digest 1 byte:
+//         checksum = (checksum ^ word) <<< 1;
+// placed at last word of EEPROM
+
+// ftdi_write_data(), ftdi_read_data(): bulk xfer
+
+// input/tristate: 200K pullup
+// SI/WU: let's ignore this
+// UART mode: standard stuff
+// FIFO mode: RDF#=0: enable output. RD# rising when RXF#=0: fetch next byte
+//            TXF#=0: enable input. WR falling when TXF#=0: write byte
+// bitbang mode: ^ similar, no RDF#/TXF#, RD#/WR# (now WR# jenai WR) pos depends on UART mode std (UART vs FIFO)
+// sync bitbang: doesn't use RD#/WR#, clocked by baudrate control
+// MPSSE: lots of magic. TCK/SK, TDI/DO, TDO/DI, TMS/CS, GPIOL0..3, GPIOH0..3(7?)
+
+// bulk xfer formats: (interface index <-> bulk epno)
+// * UART: ok I guess
+// * FIFO: same
+// * bitbang, sync bitbang: ???? (just data or also clock stuff?? ?)
+// * MPSSE, MCU: see separate PDF
+// * CPUFIFIO: same as FIFO ig
+// * opto, FT1284: not supported
+
+#ifdef DBOARD_HAS_FTDI
+
+int __builtin_ctz(unsigned int v);
+
+uint16_t ftdi_eeprom[128];
+
+struct ftdi_interface ftdi_ifa, ftdi_ifb;
+
+void ftdi_init(void) {
+    // init eeprom defaults
+    memset(ftdi_eeprom, 0xff, sizeof ftdi_eeprom);
+    ftdi_eeprom[ 0] = 0x0000; // both default to UART
+    ftdi_eeprom[ 1] = 0x0403; // idVendor
+    ftdi_eeprom[ 2] = 0x6010; // idProduct
+    ftdi_eeprom[ 3] = 0x0500; // bcdDevice
+    ftdi_eeprom[ 4] = 0x5080; // 100 mA, no flags
+    ftdi_eeprom[ 5] = 0x0000; // more flags
+    ftdi_eeprom[ 6] = 0x0110; // bcdUSB
+    ftdi_eeprom[ 7] = 0x0000; // no manuf. str (TODO?)
+    ftdi_eeprom[ 8] = 0x0000; // no prod. str  (TODO?)
+    ftdi_eeprom[ 9] = 0x0000; // no serial str (TODO?)
+    ftdi_eeprom[10] = 0x0000; // internal chip
+    ftdi_eeprom[0x7f] = ftdi_eeprom_checksum_calc(ftdi_eeprom, 0x7f);
+
+    memset(&ftdi_ifa, 0, sizeof ftdi_ifa);
+    memset(&ftdi_ifb, 0, sizeof ftdi_ifb);
+    ftdi_ifa.lineprop = sio_bits_8 | sio_stop_1; // 8n1
+    ftdi_ifb.lineprop = sio_bits_8 | sio_stop_1; // 8n1
+    ftdi_ifa.index = 0;
+    ftdi_ifb.index = 1;
+    ftdi_ifa.modemstat = sio_modem_temt;
+    ftdi_ifb.modemstat = sio_modem_temt;
+    ftdi_if_init(&ftdi_ifa);
+    ftdi_if_init(&ftdi_ifb);
+}
+void ftdi_deinit(void) {
+    ftdi_if_deinit(&ftdi_ifa);
+    ftdi_if_deinit(&ftdi_ifb);
+}
+
+static uint8_t vnd_read_byte(struct ftdi_interface* itf, int itfnum) {
+    while (itf->rxavail <= 0) {
+        if (!tud_vendor_n_mounted(itfnum) || !tud_vendor_n_available(itfnum)) {
+            thread_yield();
+            continue;
+        }
+
+        itf->rxpos   = 0;
+        itf->rxavail = tud_vendor_n_read(itfnum, itf->bufbuf, sizeof itf->bufbuf);
+
+        if (itf->rxavail == 0) thread_yield();
+    }
+
+    uint8_t rv = itf->bufbuf[itf->rxpos];
+    ++itf->rxpos;
+    --itf->rxavail;
+
+    return rv;
+}
+
+typedef void (*ftfifo_write_fn)(struct ftdi_interface*, const uint8_t*, size_t);
+typedef size_t (*ftfifo_read_fn)(struct ftdi_interface*, uint8_t*, size_t);
+struct ftfifo_fns {
+    ftfifo_write_fn write;
+    ftfifo_read_fn  read ;
+};
+static const struct ftfifo_fns fifocbs[] = {
+    { ftdi_if_uart_write, ftdi_if_uart_read }, // technically mpsse
+    { ftdi_if_asyncbb_write, ftdi_if_asyncbb_read },
+    { ftdi_if_syncbb_write, ftdi_if_syncbb_read },
+    { NULL, NULL }, // mcuhost
+    { ftdi_if_fifo_write, ftdi_if_fifo_read },
+    { NULL, NULL }, // opto
+    { ftdi_if_cpufifo_write, ftdi_if_cpufifo_read },
+    { NULL, NULL }, // ft1284
+};
+
+// for handling USB bulk commands
+static void ftdi_task(int itfnum) {
+    if (!tud_vendor_n_mounted(itfnum)) return; // can't do much in this case
+
+    struct ftdi_interface* itf;
+    if (itfnum == VND_N_FTDI_IFA) itf = &ftdi_ifa;
+    else if (itfnum == VND_N_FTDI_IFB) itf = &ftdi_ifb;
+    else return;
+
+
+    // for UART, FIFO, asyncbb, syncbb, and CPUfifo modes, operation is
+    // relatively straightforward: the device acts like some sort of FIFO, so
+    // it's just shoving bytes to the output pins. MPSSE and MCU host emulation
+    // modes are more difficult, as the bulk data actually has some form of
+    // protocol.
+    enum ftdi_mode mode = ftdi_if_get_mode(itf);
+    struct ftfifo_fns fifocb = fifocbs[(mode == 0) ? 0 : __builtin_ctz(mode)];
+    uint32_t avail;
+    uint8_t cmdbyte = 0;
+    switch (mode) {
+    case ftmode_uart   : case ftmode_fifo  : case ftmode_cpufifo:
+    case ftmode_asyncbb: case ftmode_syncbb:
+        if (fifocb.read == NULL || fifocb.write == NULL) goto CASE_DEFAULT; // welp
+
+        avail = tud_vendor_n_available(itfnum);
+        if (avail) {
+            tud_vendor_n_read(itfnum, itf->writebuf, avail);
+            fifocb.write(itf, itf->writebuf, avail);
+        }
+
+        do {
+            avail = fifocb.read(itf, itf->readbuf, sizeof itf->readbuf);
+
+            if (avail) tud_vendor_n_write(itfnum, itf->readbuf, avail);
+        } while (avail == sizeof itf->readbuf);
+        break;
+    case ftmode_mpsse:
+        avail = 0;
+
+        switch ((cmdbyte = vnd_read_byte(itf, itfnum))) {
+        case ftmpsse_set_dirval_lo: // low byte of output gpio, not to low level
+            itf->writebuf[0] = vnd_read_byte(itf, itfnum); // val
+            itf->writebuf[1] = vnd_read_byte(itf, itfnum); // dir
+            ftdi_if_mpsse_set_dirval_lo(itf, itf->writebuf[1], itf->writebuf[0]);
+            break;
+        case ftmpsse_set_dirval_hi:
+            itf->writebuf[0] = vnd_read_byte(itf, itfnum); // val
+            itf->writebuf[1] = vnd_read_byte(itf, itfnum); // dir
+            ftdi_if_mpsse_set_dirval_hi(itf, itf->writebuf[1], itf->writebuf[0]);
+            break;
+        case ftmpsse_read_lo:
+            itf->readbuf[0] = ftdi_if_mpsse_read_lo(itf);
+            avail = 1;
+            break;
+        case ftmpsse_read_hi:
+            itf->readbuf[0] = ftdi_if_mpsse_read_hi(itf);
+            avail = 1;
+            break;
+        case ftmpsse_loopback_on : ftdi_if_mpsse_loopback(itf, true ); break;
+        case ftmpsse_loopback_off: ftdi_if_mpsse_loopback(itf, false); break;
+        case ftmpsse_set_clkdiv:
+            avail  = vnd_read_byte(itf, itfnum);
+            avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            ftdi_if_mpsse_set_clkdiv(itf, (uint16_t)avail);
+            avail = 0;
+            break;
+
+        case ftmpsse_flush: ftdi_if_mpsse_flush(itf); break;
+        case ftmpsse_wait_io_hi: ftdi_if_mpsse_wait_io(itf, true ); break;
+        case ftmpsse_wait_io_lo: ftdi_if_mpsse_wait_io(itf, false); break;
+
+        case ftmpsse_div5_disable: ftdi_if_mpsse_div5(itf, false); break;
+        case ftmpsse_div5_enable : ftdi_if_mpsse_div5(itf, true ); break;
+        case ftmpsse_data_3ph_en : ftdi_if_mpsse_data_3ph(itf, true ); break;
+        case ftmpsse_data_3ph_dis: ftdi_if_mpsse_data_3ph(itf, false); break;
+        case ftmpsse_clockonly_bits: ftdi_if_mpsse_clockonly(itf, vnd_read_byte(itf, itfnum)); break;
+        case ftmpsse_clockonly_bytes:
+            avail  = vnd_read_byte(itf, itfnum);
+            avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            ftdi_if_mpsse_clockonly(itf, avail);
+            avail = 0;
+            break;
+        case ftmpsse_clock_wait_io_hi: ftdi_if_mpsse_clock_wait_io(itf, true ); break;
+        case ftmpsse_clock_wait_io_lo: ftdi_if_mpsse_clock_wait_io(itf, false); break;
+        case ftmpsse_adapclk_enable : ftdi_if_mpsse_adaptive(itf, true ); break;
+        case ftmpsse_adapclk_disable: ftdi_if_mpsse_adaptive(itf, false); break;
+        case ftmpsse_clock_bits_wait_io_hi:
+            avail  = vnd_read_byte(itf, itfnum);
+            avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            ftdi_if_mpsse_clockonly_wait_io(itf, true, avail);
+            avail = 0;
+            break;
+        case ftmpsse_clock_bits_wait_io_lo:
+            avail  = vnd_read_byte(itf, itfnum);
+            avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            ftdi_if_mpsse_clockonly_wait_io(itf, false, avail);
+            avail = 0;
+            break;
+        case ftmpsse_hi_is_tristate:
+            avail  = vnd_read_byte(itf, itfnum);
+            avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            ftdi_if_mpsse_hi_is_tristate(itf, avail);
+            avail = 0;
+            break;
+
+        default: CASE_DEFAULT:
+            if (!(cmdbyte & ftmpsse_specialcmd)) {
+                if (cmdbyte & ftmpsse_tmswrite) {
+                    if (cmdbyte & ftmpsse_bitmode) {
+                        itf->writebuf[0] = vnd_read_byte(itf, itfnum); // number of bits
+                        itf->writebuf[1] = vnd_read_byte(itf, itfnum); // data bits to output
+                        itf->readbuf[0] = ftdi_if_mpsse_tms_xfer(itf, cmdbyte, itf->writebuf[0], itf->writebuf[1]);
+                        if (cmdbyte & ftmpsse_tdoread) avail = 1;
+                        break;
+                    }
+                    // else: fallthru to error code
+                } else {
+                    if (cmdbyte & ftmpsse_bitmode) {
+                        itf->writebuf[0] = vnd_read_byte(itf, itfnum); // number of bits
+                        if (cmdbyte & ftmpsse_tdiwrite)
+                            itf->writebuf[1] = vnd_read_byte(itf, itfnum); // data bits to output
+                        itf->readbuf[0] = ftdi_if_mpsse_xfer_bits(itf, cmdbyte, itf->writebuf[0], itf->writebuf[1]);
+                        if (cmdbyte & ftmpsse_tdoread) avail = 1;
+                        break;
+                    } else {
+                        avail  = vnd_read_byte(itf, itfnum);
+                        avail |= (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+
+                        for (size_t i = 0; i < avail; i += 64) {
+                            uint32_t thisbatch = avail - i;
+                            if (thisbatch > 64) thisbatch = 64;
+                            for (size_t j = 0; j < thisbatch; ++j)
+                                itf->writebuf[j] = vnd_read_byte(itf, itfnum);
+                            ftdi_if_mpsse_xfer_bytes(itf, cmdbyte, thisbatch, itf->readbuf, itf->writebuf);
+                            tud_vendor_n_write(itfnum, itf->readbuf, thisbatch);
+                        }
+
+                        avail = 0;
+                        break;
+                    }
+                }
+            }
+
+            itf->readbuf[0] = 0xfa;
+            itf->readbuf[1] = cmdbyte;
+            avail = 2;
+            break;
+        }
+
+        if (avail) tud_vendor_n_write(itfnum, itf->readbuf, avail);
+        break;
+    case ftmode_mcuhost:
+        avail = 0;
+        switch ((cmdbyte = vnd_read_byte(itf, itfnum))) {
+        case ftmcu_flush: ftdi_if_mcuhost_flush(itf); break;
+        case ftmcu_wait_io_hi: ftdi_if_mcuhost_wait_io(itf, true ); break;
+        case ftmcu_wait_io_lo: ftdi_if_mcuhost_wait_io(itf, false); break;
+
+        case ftmcu_read8:
+            itf->readbuf[0] = ftdi_if_mcuhost_read8(itf, vnd_read_byte(itf, itfnum));
+            avail = 1;
+            break;
+        case ftmcu_read16:
+            avail = (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            avail |= vnd_read_byte(itf, itfnum);
+            itf->readbuf[0] = ftdi_if_mcuhost_read16(itf, (uint16_t)avail);
+            avail = 1;
+            break;
+        case ftmcu_write8:
+            itf->writebuf[0] = vnd_read_byte(itf, itfnum);
+            itf->writebuf[1] = vnd_read_byte(itf, itfnum);
+            ftdi_if_mcuhost_write8(itf, itf->writebuf[0], itf->writebuf[1]);
+            break;
+        case ftmcu_write16:
+            avail  = (uint32_t)vnd_read_byte(itf, itfnum) << 8;
+            avail |= vnd_read_byte(itf, itfnum);
+            itf->writebuf[0] = vnd_read_byte(itf, itfnum);
+            ftdi_if_mcuhost_write8(itf, avail, itf->writebuf[0]);
+            avail = 0;
+            break;
+
+        default: // send error response when command doesn't exist
+            itf->readbuf[0] = 0xfa;
+            itf->readbuf[1] = cmdbyte;
+            avail = 2;
+            break;
+        }
+
+        if (avail) tud_vendor_n_write(itfnum, itf->readbuf, avail);
+        break;
+    default: // drop incoming data so that the pipes don't get clogged. can't do much else
+        avail = tud_vendor_n_available(itfnum);
+        if (avail) tud_vendor_n_read(itfnum, itf->writebuf, avail);
+        break;
+    }
+}
+void ftdi_task_ifa(void) { ftdi_task(VND_N_FTDI_IFA); }
+void ftdi_task_ifb(void) { ftdi_task(VND_N_FTDI_IFB); }
+
+#define FT2232D_CLOCK (48*1000*1000)
+
+uint32_t ftdi_if_decode_baudrate(uint32_t enc_brate) { // basically reversing libftdi ftdi_to_clkbits
+    static const uint8_t ftdi_brate_frac_lut[8] = { 0, 4, 2, 1, 3, 5, 6, 7 };
+
+    // special cases
+    if (enc_brate == 0) return FT2232D_CLOCK >> 4;
+    else if (enc_brate == 1) return FT2232D_CLOCK / 24;
+    else if (enc_brate == 2) return FT2232D_CLOCK >> 5;
+
+    uint32_t div = (enc_brate & 0x7fff) << 3; // integer part
+    div = div | ftdi_brate_frac_lut[(enc_brate >> 14) & 7];
+    uint32_t baud = FT2232D_CLOCK / div;
+
+    if (baud & 1) baud = (baud >> 1) + 1; // raunding
+    else baud >>= 1;
+    return baud;
+}
+
+static uint8_t control_buf[2];
+
+bool ftdi_control_xfer_cb(uint8_t rhport, uint8_t stage,
+        tusb_control_request_t const* req) {
+    // return true: don't stall
+    // return false: stall
+
+    // not a vendor request -> not meant for this code
+    if (req->bmRequestType_bit.type != TUSB_REQ_TYPE_VENDOR) return true;
+
+    // data stage not needed: data stages only for device->host xfers, never host->device
+    if (stage != CONTROL_STAGE_SETUP) return true;
+
+    // tud_control_status(rhport, req); : acknowledge
+    // tud_control_xfer(rhport, req, bufaddr, size); : ack + there's a data phase with stuff to do
+    //   write: send bufaddr value. read: data stage will have bufaddr filled out
+
+    // do EEPROM stuff first, as these don't use wIndex as channel select
+    uint16_t tmp16;
+    switch (req->bRequest) {
+    case sio_readeeprom:
+        tmp16 = ftdi_eeprom[req->wIndex & 0x7f];
+        control_buf[0] = tmp16 & 0xff;
+        control_buf[1] = tmp16 >> 8;
+        return tud_control_xfer(rhport, req, control_buf, 2);
+    case sio_writeeeprom:
+        ftdi_eeprom[req->wIndex & 0x7f] = req->wValue;
+        return tud_control_status(rhport, req);
+    case sio_eraseeeprom:
+        memset(ftdi_eeprom, 0xff, sizeof ftdi_eeprom);
+        return tud_control_status(rhport, req);
+    }
+
+    int itfnum = req->wIndex & 0xff;
+    struct ftdi_interface* itf = &ftdi_ifa; // default
+
+    if (itfnum > 2) return false; // bad interface number
+    else if (itfnum == 1) itf = &ftdi_ifa;
+    else if (itfnum == 2) itf = &ftdi_ifb;
+
+    switch (req->bRequest) {
+    case sio_cmd:
+        if (req->wValue == sio_reset) {
+            itf->modem_mask = 0;
+            itf->modem_data = 0;
+            itf->flow = ftflow_none;
+            itf->lineprop = sio_bits_8 | sio_stop_1; // 8n1
+            itf->charen = 0;
+            itf->bb_dir = 0;
+            itf->bb_mode = sio_mode_reset;
+            itf->mcu_addr_latch = 0;
+            itf->rxavail = 0; itf->rxpos = 0;
+            ftdi_if_sio_reset(itf);
+        } else if (req->wValue == sio_tciflush) {
+            itf->rxavail = 0; itf->rxpos = 0;
+            ftdi_if_sio_tciflush(itf);
+        } else if (req->wValue == sio_tcoflush) {
+            // nothing extra to clear here I think
+            ftdi_if_sio_tcoflush(itf);
+        } else return false; // unk
+        return tud_control_status(rhport, req);
+    case sio_setmodemctrl:
+        ftdi_if_set_modemctrl(itf,
+                itf->modem_mask = (req->wValue >> 8),
+                itf->modem_data = (req->wValue & 0xff));
+        return tud_control_status(rhport, req);
+    case sio_setflowctrl: {
+        enum ftdi_flowctrl flow = (req->wIndex >> 8);
+        if (!req->wValue)
+            flow = (flow & ~ftflow_xonxoff) | (itf->flow & ftflow_xonxoff);
+        ftdi_if_set_flowctrl(itf, itf->flow = flow);
+        } return tud_control_status(rhport, req);
+    case sio_setbaudrate: {
+        uint32_t enc_brate = (uint32_t)req->wValue | ((uint32_t)(req->wIndex & 0xff00) << 8);
+        uint32_t brate = ftdi_if_decode_baudrate(enc_brate);
+        ftdi_if_set_baudrate(itf, itf->baudrate = brate);
+        } return tud_control_status(rhport, req);
+    case sio_setlineprop:
+        ftdi_if_set_lineprop(itf, itf->lineprop = req->wValue);
+        return tud_control_status(rhport, req);
+    case sio_pollmodemstat:
+        tmp16 = ftdi_if_poll_modemstat(itf);
+        control_buf[0] = tmp16 & 0xff;
+        control_buf[1] = tmp16 >>   8;
+        return tud_control_xfer(rhport, req, control_buf, 2);
+    case sio_seteventchar:
+        if (req->wValue >> 8) itf->charen |= eventchar_enable;
+        else if (itf->charen & eventchar_enable) itf->charen ^= eventchar_enable;
+        ftdi_if_set_eventchar(itf, req->wValue >> 8,
+                itf->eventchar = (req->wValue & 0xff));
+        return tud_control_status(rhport, req);
+    case sio_seterrorchar:
+        if (req->wValue >> 8) itf->charen |= errorchar_enable;
+        else if (itf->charen & errorchar_enable) itf->charen ^= errorchar_enable;
+        ftdi_if_set_errorchar(itf, req->wValue >> 8,
+                itf->errorchar = (req->wValue & 0xff));
+        return tud_control_status(rhport, req);
+    case sio_setlatency:
+        ftdi_if_set_latency(itf, itf->latency = (req->wValue & 0xff));
+        return tud_control_status(rhport, req);
+    case sio_getlatency:
+        control_buf[0] = ftdi_if_get_latency(itf);
+        return tud_control_xfer(rhport, req, control_buf, 1);
+    case sio_setbitbang: {
+        uint8_t olddir = itf->bb_dir;
+        enum ftdi_sio_bitmode oldmode = itf->bb_mode;
+        ftdi_if_set_bitbang(itf,
+                itf->bb_dir  = (req->wValue >> 8),
+                itf->bb_mode = (req->wValue & 0xff),
+                olddir, oldmode);
+        } return tud_control_status(rhport, req);
+    case sio_readpins:
+        control_buf[0] = ftdi_if_read_pins(itf);
+        return tud_control_xfer(rhport, req, control_buf, 1);
+    default: return false; // stall if not recognised
+    }
+}
+
+#endif /* DBOARD_HAS_FTDI */
+

src/m_ftdi/ftdi.h

@@ -0,0 +1,375 @@
+// vim: set et:
+
+#ifndef FTDI_H_
+#define FTDI_H_
+
+#include "tusb_config.h"
+#include <tusb.h>
+
+// USB command handling, and mode interfacing stuff
+
+void ftdi_init(void);
+void ftdi_deinit(void);
+
+// have separate tasks for the separate bulk endpoints so that a wait-for-data
+// in one ep won't cause the otehr to stall, too
+void ftdi_task_ifa(void);
+void ftdi_task_ifb(void);
+
+bool ftdi_control_xfer_cb(uint8_t rhport, uint8_t ep_addr,
+        tusb_control_request_t const* req);
+
+void ftdi_eeprom_dirty_set(bool v);
+bool ftdi_eeprom_dirty_get(void);
+
+extern uint16_t ftdi_eeprom[128];
+
+#define FTDI_EEP_IFA_MODE   (ftdi_eeprom[0] & 0xff)
+#define FTDI_EEP_IFB_MODE   (ftdi_eeprom[0] >>   8)
+#define FTDI_EEP_IDVENDOR   (ftdi_eeprom[1])
+#define FTDI_EEP_IDPRODUCT  (ftdi_eeprom[2])
+#define FTDI_EEP_BCDDEVICE  (ftdi_eeprom[3])
+#define FTDI_EEP_PWFLAGS    (ftdi_eeprom[4] & 0xff)
+#define FTDI_EEP_MAXAMP     (ftdi_eeprom[4] >>   8)
+#define FTDI_EEP_USBFLAGS   (ftdi_eeprom[5])
+#define FTDI_EEP_BCDUSB     (ftdi_eeprom[6])
+#define FTDI_EEP_MANUF_OFF  (ftdi_eeprom[7] & 0xff)
+#define FTDI_EEP_MANUF_LEN  (ftdi_eeprom[7] >>   8)
+#define FTDI_EEP_PROD_OFF   (ftdi_eeprom[8] & 0xff)
+#define FTDI_EEP_PROD_LEN   (ftdi_eeprom[8] >>   8)
+#define FTDI_EEP_SERIAL_OFF (ftdi_eeprom[9] & 0xff)
+#define FTDI_EEP_SERIAL_LEN (ftdi_eeprom[9] >>   8)
+#define FTDI_EEP_CHIPTYPE   (ftdi_eeprom[10])
+
+static inline uint16_t ftdi_eeprom_checksum_init(void) {
+    return 0xaaaa;
+}
+static inline uint16_t ftdi_eeprom_checksum_digest(uint16_t val, const uint16_t* data, size_t len) {
+    for (size_t i = 0; i < len; ++i) {
+        val = (val ^ data[i]);
+        val = (val << 1) | (val >> 15);
+    }
+
+    return val;
+}
+static inline uint16_t ftdi_eeprom_checksum_calc(const uint16_t* data, size_t len) {
+    return ftdi_eeprom_checksum_digest(ftdi_eeprom_checksum_init(), data, len);
+}
+
+// USB protocol stuff
+
+enum ftdi_request {
+    sio_cmd           =  0,
+    sio_setmodemctrl  =  1,
+    sio_setflowctrl   =  2,
+    sio_setbaudrate   =  3,
+    sio_setlineprop   =  4,
+    sio_pollmodemstat =  5,
+    sio_seteventchar  =  6,
+    sio_seterrorchar  =  7,
+
+    sio_setlatency    =  9,
+    sio_getlatency    = 10,
+    sio_setbitbang    = 11,
+    sio_readpins      = 12,
+
+    sio_readeeprom    = 0x90,
+    sio_writeeeprom   = 0x91,
+    sio_eraseeeprom   = 0x92
+};
+
+enum ftdi_sio_cmd {
+    sio_reset    = 0,
+    sio_tciflush = 2,
+    sio_tcoflush = 1
+};
+
+enum ftdi_sio_lineprop {
+    sio_break_on    = 1<<14,
+    sio_break_off   = 0<<14,
+    sio_break__mask = 1<<14,
+
+    sio_stop_1     = 0<<11,
+    sio_stop_15    = 1<<11, // deprecated?
+    sio_stop_2     = 2<<11,
+    sio_stop__mask = 3<<11,
+
+    sio_parity_none  = 0<<8,
+    sio_parity_odd   = 1<<8,
+    sio_parity_even  = 2<<8,
+    sio_parity_mark  = 3<<8,
+    sio_parity_space = 4<<8,
+    sio_parity__mask = 7<<8,
+
+    sio_bits_7     =    7<<0,
+    sio_bits_8     =    8<<0,
+    sio_bits__mask = 0xff<<0,
+};
+
+enum ftdi_sio_modemstat {
+    sio_modem_cts  = 1<< 4, // Clear to Send active
+    sio_modem_dts  = 1<< 5, // Data Set Ready active
+    sio_modem_ri   = 1<< 6, // Ring Indicator active
+    sio_modem_rlsd = 1<< 7, // Receive Line Signal Detect active
+
+    sio_modem_dr   = 1<< 8, // Data Ready
+    sio_modem_oe   = 1<< 9, // Overrun Error
+    sio_modem_pe   = 1<<10, // Parity Error
+    sio_modem_fe   = 1<<11, // Framing Error
+    sio_modem_bi   = 1<<12, // Break Interrupt
+    sio_modem_thre = 1<<13, // Transmitter Holding REgister
+    sio_modem_temt = 1<<14, // Transmitter Empty
+    sio_modem_fifoerr = 1<<15 // Error in receive FIFO
+};
+
+enum ftdi_sio_bitmode {
+    sio_mode_reset   =  0, // i.e. from EEPROM
+    sio_mode_bitbang =  1,
+    sio_mode_mpsse   =  2,
+    sio_mode_syncbb  =  4,
+    sio_mode_mcu     =  8,
+
+    // 0x10: opto
+    // 0x20: cbus bitbang (R-type only)
+    // 0x40: sync fifo (2232h) // like regular fifo mode, but with clock output
+    // 0x80: ft1284 (232h, not 2232d)
+};
+
+enum ftdi_eep_defmode {
+    fteep_mode_uart   = 0,
+    fteep_mode_fifo   = 1,
+    fteep_mode_opto   = 2, // not implementing this here
+    fteep_mode_cpu    = 4,
+    fteep_mode_ft1284 = 8, // not impl. on 2232d, 232h-only
+};
+
+// mpsse, mcuhost commands
+
+// if bit 7 of an MPSSE command byte 
+
+enum ftdi_mpsse_cflg {
+    ftmpsse_negedge_wr = 1<<0, // if 0, output bits on positive clock edige
+    ftmpsse_bitmode    = 1<<1, // if 0, byte mode
+    ftmpsse_negedge_rd = 1<<2, // if 0, input bits on positive clock edge
+    ftmpsse_lsbfirst   = 1<<3, // if 0, msb first
+    ftmpsse_tdiwrite   = 1<<4, // 1 == do perform output
+    ftmpsse_tdoread    = 1<<5, // 1 == do perform input
+    ftmpsse_tmswrite   = 1<<6, // 1 == do perform output?
+
+    ftmpsse_specialcmd = 1<<7  // see below enum if set
+};
+// bitmode: 1 length byte, max=7 (#bits = length+1) for separate bits
+// bytemode: 2 length bytes == number of bytes that follow
+// both tdiwrite and tdoread high: only one length value/equal number of bits in/out!
+// if both tdiwrite and tdoread are high, negedge_wr and negedge_rd must differ
+// tms read/writes: readback is from tdo, bit7 in databyte is tdi output, held constant
+// tdiwrite always 0, bitmode 1 in impls, can be ignored I guess
+// also always lsbfirst, but not too hard to support msbfirst too
+// idle levels (of eg. tms/cs) -> set_dirval?
+
+enum ftdi_mpssemcu_cmd {
+    ftmpsse_set_dirval_lo = 0x80, // sets initial clock level!
+    ftmpsse_set_dirval_hi = 0x82,
+    ftmpsse_read_lo       = 0x81,
+    ftmpsse_read_hi       = 0x83,
+    ftmpsse_loopback_on   = 0x84,
+    ftmpsse_loopback_off  = 0x85,
+    ftmpsse_set_clkdiv    = 0x86, // period = 12MHz / ((1 + value16) * 2)
+
+    ftmpsse_flush      = 0x87, // flush dev->host usb buffer
+    ftmpsse_wait_io_hi = 0x88, // wait for gpiol1/io1 to be high
+    ftmpsse_wait_io_lo = 0x89, // wait for gpiol1/io1 to be low
+
+    // technically ft2232h-only but we can support these, too
+    ftmpsse_div5_disable = 0x8a, // ft2232h internally has a 5x faster clock, but slows it down by default
+    ftmpsse_div5_enable  = 0x8b, // for backwards compat. these two commands enable/disable that slowdown
+    ftmpsse_data_3ph_en  = 0x8c, // enable 3-phase data
+    ftmpsse_data_3ph_dis = 0x8d, // disable 3-phase data
+    ftmpsse_clockonly_bits  = 0x8e, // enable clock for n bits, no data xfer
+    ftmpsse_clockonly_bytes = 0x8f, // enable clock for n bytes, no data xfer
+    ftmpsse_clock_wait_io_hi = 0x94, // wait_io_hi + clockonly
+    ftmpsse_clock_wait_io_lo = 0x95, // wait_io_lo + clockonly
+    ftmpsse_adapclk_enable  = 0x96, // enable ARM JTAG adaptive clocking (rtck gpiol3 input)
+    ftmpsse_adapclk_disable = 0x97, // disable ARM JTAG adaptive clocking (rtck gpiol3 input)
+    ftmpsse_clock_bits_wait_io_hi = 0x9c, // clock_wait_io_hi + clockonly_bits
+    ftmpsse_clock_bits_wait_io_lo = 0x9d, // clock_wait_io_lo + clockonly_bits
+    ftmpsse_hi_is_tristate = 0x9e, // turns 1 output to tristate for selected outputs
+
+    ftmcu_flush      = 0x87, // flush dev->host usb buffer
+    ftmcu_wait_io_hi = 0x88, // wait for gpiol1/io1 to be high
+    ftmcu_wait_io_lo = 0x89, // wait for gpiol1/io1 to be low
+
+    ftmcu_read8   = 0x90,
+    ftmcu_read16  = 0x91,
+    ftmcu_write8  = 0x92,
+    ftmcu_write16 = 0x93
+};
+
+// internal use only types
+
+enum ftdi_mode { // combines EEPROM setting + bitmode
+    ftmode_uart    = 0,
+
+    ftmode_mpsse   =  1,
+    ftmode_asyncbb =  2,
+    ftmode_syncbb  =  4,
+    ftmode_mcuhost =  8,
+
+    ftmode_fifo    = 0x10,
+    ftmode_opto    = 0x20, // not implementing this here
+    ftmode_cpufifo = 0x40,
+    ftmode_ft1284  = 0x80, // not impl. on 2232d
+};
+
+enum ftdi_flowctrl {
+    ftflow_none    = 0,
+    ftflow_ctsrts  = 1,
+    ftflow_dtrdts  = 2,
+    ftflow_xonxoff = 4,
+};
+
+struct ftdi_interface {
+    // TODO soft fields maybe, because it's a mess with lots of padding right now
+    int index;
+
+    uint8_t modem_mask;
+    uint8_t modem_data;
+
+    enum ftdi_flowctrl flow;
+
+    uint32_t baudrate;
+    enum ftdi_sio_lineprop lineprop;
+
+    enum ftdi_sio_modemstat modemstat;
+
+    uint8_t eventchar, errorchar;
+    enum { eventchar_enable = 1<<0, errorchar_enable = 1<<1 } charen;
+
+    uint8_t latency; // latency timer. TODO: implement this
+
+    uint8_t bb_dir; // high/1 bit = output, 0=input
+    enum ftdi_sio_bitmode bb_mode;
+
+    uint16_t mcu_addr_latch;
+
+    // these are for USB bulk cmds etc.
+    // "write" means write to hardware output pins
+    // "read" means read from hardware input pins
+    uint8_t writebuf[CFG_TUD_VENDOR_RX_BUFSIZE];
+    uint8_t readbuf [CFG_TUD_VENDOR_TX_BUFSIZE];
+    uint8_t bufbuf  [CFG_TUD_VENDOR_RX_BUFSIZE]; // for buffered IO
+    uint32_t rxavail, rxpos;
+};
+
+extern struct ftdi_interface ftdi_ifa, ftdi_ifb;
+
+// interface control stuff
+
+static inline enum ftdi_mode ftdi_get_mode_of(enum ftdi_sio_bitmode bb_mode, uint8_t eepmode) {
+    if (bb_mode == 0x10) return ftmode_opto;
+
+    if (bb_mode == sio_mode_reset) {
+        return (eepmode << 4) & 0xf0;
+    } else return bb_mode & 0xf;
+}
+static inline enum ftdi_mode ftdi_if_get_mode(struct ftdi_interface* itf) {
+    return ftdi_get_mode_of(itf->bb_mode, itf->index ? FTDI_EEP_IFB_MODE : FTDI_EEP_IFA_MODE);
+}
+uint32_t ftdi_if_decode_baudrate(uint32_t enc_brate);
+
+// control request stuff. implemented by bsp driver
+void ftdi_if_init(struct ftdi_interface* itf);
+void ftdi_if_deinit(struct ftdi_interface* itf);
+
+void ftdi_if_sio_reset(struct ftdi_interface* itf);
+void ftdi_if_sio_tciflush(struct ftdi_interface* itf);
+void ftdi_if_sio_tcoflush(struct ftdi_interface* itf);
+void ftdi_if_set_modemctrl(struct ftdi_interface* itf, uint8_t mask, uint8_t data);
+void ftdi_if_set_flowctrl(struct ftdi_interface* itf, enum ftdi_flowctrl flow);
+void ftdi_if_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_set_lineprop(struct ftdi_interface* itf, enum ftdi_sio_lineprop lineprop);
+enum ftdi_sio_modemstat ftdi_if_poll_modemstat(struct ftdi_interface* itf);
+void ftdi_if_set_eventchar(struct ftdi_interface* itf, bool enable, uint8_t evchar);
+void ftdi_if_set_errorchar(struct ftdi_interface* itf, bool enable, uint8_t erchar);
+void ftdi_if_set_latency(struct ftdi_interface* itf, uint8_t latency);
+uint8_t ftdi_if_get_latency(struct ftdi_interface* itf);
+void ftdi_if_set_bitbang(struct ftdi_interface* itf, uint8_t dirmask, enum ftdi_sio_bitmode,
+        uint8_t olddir, enum ftdi_sio_bitmode oldmode);
+uint8_t ftdi_if_read_pins(struct ftdi_interface* itf);
+
+// bulk commands (also implemented by bsp driver)
+
+// "write" means write to hardware output pins
+// "read" means read from hardware input pins
+void   ftdi_if_uart_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize);
+size_t ftdi_if_uart_read (struct ftdi_interface* itf,       uint8_t* data, size_t  maxsize);
+
+void   ftdi_if_fifo_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize);
+size_t ftdi_if_fifo_read (struct ftdi_interface* itf,       uint8_t* data, size_t  maxsize);
+void   ftdi_if_cpufifo_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize);
+size_t ftdi_if_cpufifo_read (struct ftdi_interface* itf,       uint8_t* data, size_t  maxsize);
+
+void   ftdi_if_asyncbb_write(struct ftdi_interface* itf, const uint8_t* data, size_t datasize);
+size_t ftdi_if_asyncbb_read (struct ftdi_interface* itf,       uint8_t* data, size_t  maxsize);
+void   ftdi_if_syncbb_write (struct ftdi_interface* itf, const uint8_t* data, size_t datasize);
+size_t ftdi_if_syncbb_read  (struct ftdi_interface* itf,       uint8_t* data, size_t  maxsize);
+
+
+void ftdi_if_mpsse_flush(struct ftdi_interface* itf);
+void ftdi_if_mpsse_wait_io(struct ftdi_interface* itf, bool level);
+
+void ftdi_if_mpsse_set_dirval_lo(struct ftdi_interface* itf, uint8_t dir, uint8_t val);
+void ftdi_if_mpsse_set_dirval_hi(struct ftdi_interface* itf, uint8_t dir, uint8_t val);
+uint8_t ftdi_if_mpsse_read_lo(struct ftdi_interface* itf);
+uint8_t ftdi_if_mpsse_read_hi(struct ftdi_interface* itf);
+void ftdi_if_mpsse_loopback(struct ftdi_interface* itf, bool enable);
+void ftdi_if_mpsse_set_clkdiv(struct ftdi_interface* itf, uint16_t div);
+
+uint8_t ftdi_if_mpsse_xfer_bits(struct ftdi_interface* itf, int flags, size_t nbits, uint8_t value);
+void ftdi_if_mpsse_xfer_bytes(struct ftdi_interface* itf, int flags, size_t nbytes, uint8_t* dst, const uint8_t* src);
+uint8_t ftdi_if_mpsse_tms_xfer(struct ftdi_interface* itf, int flags, size_t nbits, uint8_t value);
+
+void ftdi_if_mpsse_div5(struct ftdi_interface* itf, bool enable);
+void ftdi_if_mpsse_data_3ph(struct ftdi_interface* itf, bool enable);
+void ftdi_if_mpsse_adaptive(struct ftdi_interface* itf, bool enable);
+
+void ftdi_if_mpsse_clockonly(struct ftdi_interface* itf, uint32_t cycles);
+void ftdi_if_mpsse_clock_wait_io(struct ftdi_interface* itf, bool level);
+void ftdi_if_mpsse_clockonly_wait_io(struct ftdi_interface* itf, bool level, uint32_t cycles);
+void ftdi_if_mpsse_hi_is_tristate(struct ftdi_interface* itf, uint16_t pinmask);
+
+
+void ftdi_if_mcuhost_flush(struct ftdi_interface* itf);
+void ftdi_if_mcuhost_wait_io(struct ftdi_interface* itf, bool level);
+
+uint8_t ftdi_if_mcuhost_read8 (struct ftdi_interface* itf, uint8_t  addr);
+uint8_t ftdi_if_mcuhost_read16(struct ftdi_interface* itf, uint16_t addr);
+void ftdi_if_mcuhost_write8 (struct ftdi_interface* itf, uint8_t  addr, uint8_t value);
+void ftdi_if_mcuhost_write16(struct ftdi_interface* itf, uint16_t addr, uint8_t value);
+
+void ftdi_if_uart_init(struct ftdi_interface* itf);
+void ftdi_if_mpsse_init(struct ftdi_interface* itf);
+void ftdi_if_asyncbb_init(struct ftdi_interface* itf);
+void ftdi_if_syncbb_init(struct ftdi_interface* itf);
+void ftdi_if_mcuhost_init(struct ftdi_interface* itf);
+void ftdi_if_fifo_init(struct ftdi_interface* itf);
+void ftdi_if_cpufifo_init(struct ftdi_interface* itf);
+
+void ftdi_if_uart_deinit(struct ftdi_interface* itf);
+void ftdi_if_mpsse_deinit(struct ftdi_interface* itf);
+void ftdi_if_asyncbb_deinit(struct ftdi_interface* itf);
+void ftdi_if_syncbb_deinit(struct ftdi_interface* itf);
+void ftdi_if_mcuhost_deinit(struct ftdi_interface* itf);
+void ftdi_if_fifo_deinit(struct ftdi_interface* itf);
+void ftdi_if_cpufifo_deinit(struct ftdi_interface* itf);
+
+void ftdi_if_uart_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_mpsse_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_asyncbb_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_syncbb_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_mcuhost_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_fifo_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+void ftdi_if_cpufifo_set_baudrate(struct ftdi_interface* itf, uint32_t baudrate);
+
+#endif
+

src/m_sump/_sump.c

@@ -3,6 +3,7 @@
 #include <tusb.h>
 
 #include "mode.h"
+#include "storage.h"
 #include "thread.h"
 #include "usbstdio.h"
 #include "vnd_cfg.h"
@@ -20,6 +21,8 @@ enum m_sump_feature {
     msump_feat_sump      = 1<<0,
 };
 
+static bool data_dirty = false;
+
 #ifdef DBOARD_HAS_SUMP
 static cothread_t sumpthread;
 static uint8_t    sumpstack[THREAD_STACK_SIZE];
@@ -44,6 +47,16 @@ static void enter_cb(void) {
     sumpthread = co_derive(sumpstack, sizeof sumpstack, sump_thread_fn);
     thread_enter(sumpthread);
 #endif
+
+    if (!data_dirty) { // only read when not read yet
+        struct mode_info mi = storage_mode_get_info(4);
+        if (mi.size != 0 && mi.version == 0x0010 /* TODO: version migration? */) {
+            uint8_t dst[1];
+            storage_mode_read(4, dst, 0, 1);
+
+            sump_hw_set_overclock(dst[0]);
+        }
+    }
 }
 static void leave_cb(void) {
 #ifdef DBOARD_HAS_SUMP
@@ -133,7 +146,6 @@ enum {
 #define EPNUM_CDC_STDIO_NOTIF   0x85
 
 // clang-format off
-// TODO: replace magic 64s by actual buffer size macros
 static const uint8_t desc_configuration[] = {
     TUD_CONFIG_DESCRIPTOR(1, ITF_NUM__TOTAL, STRID_CONFIG, CONFIG_TOTAL_LEN,
         TUSB_DESC_CONFIG_ATT_REMOTE_WAKEUP, 100),
@@ -178,6 +190,17 @@ static void my_cdc_line_coding_cb(uint8_t itf, cdc_line_coding_t const* line_cod
 }
 #endif
 
+static uint16_t my_get_size(void) { return 1; }
+static void my_get_data(void* dst, size_t offset, size_t maxsize) {
+    (void)offset; (void)maxsize;
+
+    uint8_t* d = dst;
+    d[0] = sump_hw_get_overclock();
+
+    data_dirty = false;
+}
+static bool my_is_dirty(void) { return data_dirty; }
+
 extern struct mode m_04_sump;
 // clang-format off
 struct mode m_04_sump = {
@@ -188,6 +211,13 @@ struct mode m_04_sump = {
     .usb_desc = desc_configuration,
     .string_desc = string_desc_arr,
 
+    .storage = {
+        .stclass = mode_storage_32b,
+        .get_size = my_get_size,
+        .get_data = my_get_data,
+        .is_dirty = my_is_dirty
+    },
+
     .enter = enter_cb,
     .leave = leave_cb,
     .task  = task_cb,

src/main.c

@@ -7,6 +7,7 @@
 #include "tusb.h"
 
 #include "mode.h"
+#include "storage.h"
 #include "thread.h"
 #include "usbstdio.h"
 #include "vnd_cfg.h"
@@ -24,6 +25,8 @@ static void vndcfg_thread_fn(void) {
 }
 
 int main() {
+    stdio_uart_init();
+
     thread_init();
 
     board_init();  // tinyusb hardware support function
@@ -31,15 +34,24 @@ int main() {
     vndcfg_thread = co_derive(vndcfg_stack, sizeof vndcfg_stack, vndcfg_thread_fn);
     thread_enter(vndcfg_thread);
 
-    modes_init();
+#if defined(PERSISTENT_STORAGE) && defined(DBOARD_HAS_STORAGE)
+    int startupmode = storage_init();
+#else
+    int startupmode = -1;
+#endif
+
+    // FIXME: put elsewhere?
+#ifdef USE_USBCDC_FOR_STDIO
+    stdio_usb_set_itf_num(0);
+#endif
+
+    modes_init(startupmode);
     if (mode_current) mode_current->enter();
 
     tusb_init();
 
     // FIXME: put elsewhere?
 #ifdef USE_USBCDC_FOR_STDIO
-    stdio_usb_set_itf_num(0);
-
     stdio_usb_init();
 #endif
 

src/mode.h

@@ -9,6 +9,8 @@
 #include "tusb_config.h"
 #include <tusb.h>
 
+#include "storage.h"
+
 // clang-format off
 
 struct mode {
@@ -19,6 +21,8 @@ struct mode {
     const uint8_t* usb_desc;
     const char** string_desc;
 
+    struct mode_storage storage;
+
     void (*enter)(void); // claim required hardware. no tusb calls here please
     void (*leave)(void); // release current in-use hardware. no tusb calls here please
     void (*task )(void);
@@ -35,7 +39,7 @@ struct mode {
     void (*tud_cdc_line_coding_cb)(uint8_t itf, cdc_line_coding_t const* line_coding);
 #endif
 //#if CFG_TUD_VENDOR > 0
-    bool (*tud_vendor_control_xfer_cb)(uint8_t rhport, uint8_t ep_addr,
+    bool (*tud_vendor_control_xfer_cb)(uint8_t rhport, uint8_t stage,
             tusb_control_request_t const* req);
 //#endif
 
@@ -46,8 +50,9 @@ struct mode {
 };
 
 // call this BEFORE tusb_init!
-void modes_init(void);
+void modes_init(int init_mode);
 
+// IMMEDIATELY switches mode. use "mode_next_id' to do a graceful switch
 void modes_switch(uint8_t newmode);
 
 extern int mode_current_id;

src/modeset.c

@@ -5,8 +5,9 @@
 #include "alloc.h"
 #include "board.h" /* bsp_reset_bootloader() */
 #include "mode.h"
+#include "storage.h"
 
-extern struct mode m_01_default, m_03_jscan, m_04_sump;
+extern struct mode m_01_default, m_03_jscan, m_04_sump, m_05_ftdi;
 
 // clang-format off
 const struct mode* const mode_list[16] = {
@@ -15,6 +16,7 @@ const struct mode* const mode_list[16] = {
     NULL, // mode 2 (hw chip programming stuff) not implemented yet
     &m_03_jscan,
     &m_04_sump,
+    &m_05_ftdi,
     NULL, // terminating entry
 };
 // clang-format on
@@ -37,9 +39,13 @@ enum tusbgot_index {
 
 #define ORDEF(a,b) ((a != NULL) ? a : b)
 
-void modes_init(void) {
+void modes_init(int newid) {
-    // switch to the default mode, but without doing a USB reboot thing
+    if (newid < 0 || newid >= 16 || mode_list[newid] == NULL) {
-    mode_current_id = &mode_default - mode_list;
+        // switch to the default mode, but without doing a USB reboot thing
+        mode_current_id = &mode_default - mode_list;
+    } else {
+        mode_current_id = newid;
+    }
     mode_next_id = -1;
 
     //if (!mode_default) return;
@@ -94,6 +100,8 @@ void modes_switch(uint8_t newmode) {
     // maybe wait a second or so for the host to notice this
     sleep_ms(500/2);
 
+    printf("disconnect\n");
+
     if (newmode == 0) bsp_reset_bootloader();
 
     // now apply the new tusb settings
@@ -159,11 +167,25 @@ void modes_switch(uint8_t newmode) {
         // clang-format on
     }
 
+    printf("reconnect\n");
+
     // and reconnect
     tud_connect();
     sleep_ms(500/2);
     //while (!tud_mounted()) sleep_ms(5);
 
+    printf("enter\n");
+
     if (mode_current) mode_current->enter();
 }
 
+#if defined(PERSISTENT_STORAGE) && defined(DBOARD_HAS_STORAGE)
+void tud_umount_cb(void) {
+    storage_flush_data();
+}
+void tud_suspend_cb(bool remote_wakeup_en) {
+    (void)remote_wakeup_en;
+    storage_flush_data();
+}
+#endif
+

src/storage.c

@@ -0,0 +1,194 @@
+// vim: set et:
+
+#include "bsp-info.h"
+#include "bsp-storage.h"
+#include "mode.h"
+#include "storage.h"
+
+#if !defined(PERSISTENT_STORAGE) || !defined(DBOARD_HAS_STORAGE)
+int storage_init(void) { return -1; }
+bool storage_flush_data(void) { return false; }
+struct mode_info storage_mode_get_info(int _) {
+    (void)_; return (struct mode_info){ .size = 0, .version = 0 };
+}
+void storage_mode_read(int _, void* __, size_t ___, size_t ____) {
+    (void)_; (void)__; (void)___; (void)____;
+}
+#else
+
+#include "storage_internal.h"
+
+uint32_t str_hash_djb2_digest(uint32_t hash, const void* data, size_t len) {
+    const uint8_t* d = data;
+
+    for (size_t i = 0; i < len; ++i) hash += hash * 33 + d[i];
+
+    return hash;
+}
+
+bool header_valid = false;
+struct storage_header header_tmp;
+uint8_t data_tmp[256];
+uint16_t mode_bad = 0;
+
+static int in_init = 0;
+
+static void storage_init_defaults(void) {
+    memcpy(header_tmp.magic, STORAGE_MAGIC, STORAGE_MAGIC_LEN);
+    header_tmp.fwversion = STORAGE_VER;
+    header_tmp.curmode = mode_current_id;
+    header_tmp.nmodes = 0;
+    memset(header_tmp.reserved, 0xff, sizeof(header_tmp.reserved));
+    memset(header_tmp.mode_data_table, 0xff, sizeof(header_tmp.mode_data_table));
+    header_tmp.table_djb2 = str_hash_djb2(header_tmp.mode_data_table,
+            sizeof(struct mode_data)*MAX_MDT_ELEMENTS);
+
+    header_valid = true;
+}
+
+int storage_init(void) {
+    ++in_init;
+
+    mode_bad = 0;
+    storage_read(&header_tmp, STORAGE_SIZE - sizeof(struct storage_header),
+            sizeof(struct storage_header));
+
+    bool bad = false;
+    if (memcmp(header_tmp.magic, STORAGE_MAGIC, STORAGE_MAGIC_LEN)) {
+        storage_init_defaults();
+        if (in_init == 1) storage_flush_data();
+        --in_init;
+        return -1;
+    }
+
+    if (header_tmp.fwversion != STORAGE_VER) {
+        // TODO: migrate... if there were any older versions
+        header_valid = false;
+        --in_init;
+        return -1;
+    }
+
+    if (header_tmp.nmodes >= 16) bad = true;
+    else if (str_hash_djb2(header_tmp.mode_data_table,
+                sizeof(struct mode_data)*MAX_MDT_ELEMENTS) != header_tmp.table_djb2)
+        bad = true;
+    else if (header_tmp.curmode >= 16 || header_tmp.curmode == 0
+            || mode_list[header_tmp.curmode] == NULL)
+        bad = true;
+
+    if (bad) {
+        storage_init_defaults();
+        if (in_init == 1) storage_flush_data();
+        --in_init;
+        return -1;
+    }
+
+    header_valid = true;
+    --in_init;
+    return header_tmp.curmode;
+}
+
+struct mode_info storage_mode_get_info(int mode) {
+    #define DEF_RETVAL ({ \
+        if (mode < 16 && mode > 0 && header_valid && header_tmp.nmodes != 0) mode_bad |= 1<<mode; \
+        (struct mode_info){ .size = 0, .version = 0 }; \
+    }) \
+
+
+    if (mode >= 16 || !header_valid || mode <= 0 || header_tmp.nmodes == 0)
+        return DEF_RETVAL;
+
+    for (size_t i = 0; i < header_tmp.nmodes; ++i) {
+        struct mode_data md = header_tmp.mode_data_table[i];
+        int mdmode = (uint8_t)(md.offsetandmode >> 28);
+        uint16_t mdsize = md.datasize;
+        uint32_t mdoffset = md.offsetandmode & ((1<<28)-1);
+
+        if (mdmode != mode) continue;
+        if (mdsize == 0xffff || md.version == 0xffff || md.offsetandmode == 0xffffffffu)
+            continue; // empty (wut?)
+
+        // found it!
+
+        if (mdsize == 0) return DEF_RETVAL; // no data stored
+        if (mdoffset == 0 || mdoffset >= STORAGE_SIZE)
+            return DEF_RETVAL; // bad offset
+        // program code collision cases
+        if (mdoffset < storage_get_program_offset() && mdoffset+mdsize >=
+                storage_get_program_offset()) return DEF_RETVAL;
+        if (mdoffset < storage_get_program_offset()+storage_get_program_size()
+                && mdoffset+mdsize >= storage_get_program_offset()+storage_get_program_size())
+            return DEF_RETVAL;
+        if (mdoffset >= storage_get_program_offset()
+                && mdoffset+mdsize <= storage_get_program_offset()+storage_get_program_size())
+            return DEF_RETVAL;
+
+        // now check whether the data hash is corrupted
+        uint32_t hash = str_hash_djb2_init();
+
+        for (size_t i = 0; i < mdsize; i += sizeof(data_tmp)) {
+            size_t toread = sizeof(data_tmp);
+            if (mdsize - i < toread) toread = mdsize - i;
+
+            storage_read(data_tmp, mdoffset + i, toread);
+
+            hash = str_hash_djb2_digest(hash, data_tmp, toread);
+        }
+
+        if (hash != md.data_djb2) return DEF_RETVAL;
+
+        return (struct mode_info) {
+            .size = mdsize,
+            .version = md.version
+        };
+    }
+
+    return DEF_RETVAL;
+
+    #undef DEF_RETVAL
+}
+void storage_mode_read(int mode, void* dst, size_t offset, size_t maxlen) {
+<<<<<<< HEAD
+    if (mode >= 16 || !header_valid || mode <= 0 || header_tmp.nmodes == 0) return;
+=======
+    if (mode >= 16 || !header_valid || mode <= 0) return;
+>>>>>>> 62e3181 (PERSISTENT_STORAGE usage flag)
+
+    for (size_t i = 0; i < header_tmp.nmodes; ++i) {
+        struct mode_data md = header_tmp.mode_data_table[i];
+        int mdmode = (uint8_t)(md.offsetandmode >> 28);
+        uint16_t mdsize = md.datasize;
+        uint32_t mdoffset = md.offsetandmode & ((1<<28)-1);
+
+        if (mdmode != mode) continue;
+        if (mdsize == 0xffff || md.version == 0xffff || md.offsetandmode == 0xffffffffu)
+            continue; // empty (wut?)
+
+        // found it!
+
+        if (mdsize == 0) { mode_bad |= 1<<mode; return; /* no data stored */ }
+        if (mdoffset == 0 || mdoffset >= STORAGE_SIZE) {
+            mode_bad |= 1<<mode; return; /* bad offset */
+        }
+        // program code collision cases
+        if (mdoffset < storage_get_program_offset() && mdoffset+mdsize >=
+                storage_get_program_offset()) { mode_bad |= 1<<mode; return; }
+        if (mdoffset < storage_get_program_offset()+storage_get_program_size()
+                && mdoffset+mdsize >= storage_get_program_offset()+storage_get_program_size()) {
+            mode_bad |= 1<<mode; return;
+        }
+        if (mdoffset >= storage_get_program_offset()
+                && mdoffset+mdsize <= storage_get_program_offset()+storage_get_program_size()) {
+            mode_bad |= 1<<mode; return;
+        }
+
+        if (offset >= mdsize) return;
+
+        // skip hash check in this case
+        storage_read(dst, mdoffset + offset, mdsize < maxlen ? mdsize : maxlen);
+        return;
+    }
+}
+
+#endif /* BOARD_HAS_STORAGE */
+

src/storage.h

@@ -0,0 +1,50 @@
+// vim: set et:
+
+#ifndef STORAGE_H_
+#define STORAGE_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define STORAGE_VER 0x0010
+
+enum mode_storage_class {
+    mode_storage_none, // this mode has no storage
+    mode_storage_32b , // this mode typically won't use more than   32 bytes of data
+    mode_storage_128b, // this mode typically won't use more than  128 bytes of data
+    mode_storage_512b, // this mode typically won't use more than  512 bytes of data
+    mode_storage_big   // this mode uses a lot of data
+};
+
+// mode callbacks used by the storage subsystem
+struct mode_storage {
+    enum mode_storage_class stclass;
+    uint16_t (*get_size)(void);
+    // if stclass < 512b, offset & maxsize can be ignored
+    void     (*get_data)(void* dst, size_t offset, size_t maxsize);
+    bool     (*is_dirty)(void); // if data was changed since last mode_read/get_data call
+};
+
+struct mode_info {
+    uint32_t size;
+    uint16_t version;
+};
+
+// functions mode-specific code can use to retrieve the save data
+
+struct mode_info storage_mode_get_info(int mode); // returns size=0 for none found
+void storage_mode_read(int mode, void* dst, size_t offset, size_t maxlen);
+
+// global functions
+
+// reads all data, creates table if needed
+int storage_init(void);
+
+// flush edits if anything has been edited
+bool storage_flush_data(void);
+
+bool storage_priv_mode_has(int mode);
+void* storage_priv_get_header_ptr(void);
+
+#endif
+

src/storage_internal.h

@@ -0,0 +1,74 @@
+// vim: set et:
+
+#ifndef STORAGE_INTERNAL_H_
+#define STORAGE_INTERNAL_H_
+
+inline static uint32_t str_hash_djb2_init(void) {
+    return 5381;
+}
+uint32_t str_hash_djb2_digest(uint32_t hash, const void* data, size_t len);
+inline static uint32_t str_hash_djb2(const void* data, size_t len) {
+    return str_hash_djb2_digest(str_hash_djb2_init(), data, len);
+}
+
+/*
+ * storage header (in last 256b of flash) (endianness is whatever is native to the device):
+ *
+ *                3            7            b            f
+ *  --+----------------------------------------------------+
+ *  0 | f0 9f 8f b3  ef b8 8f e2  80 8d e2 9a  a7 ef b8 8f |  magic number
+ * 10 | fwver mm nm  <reserved (0xff)       >  <tbl djb2>  |  fwver: current version (bcd)
+ * 20 | <mode data table ...>                              |  mm: current mode, nm: number of modes saved (~size of "mode data table")
+ * 30 |                                                    |  tbl djb2: djb2-hash of "mode data table" (entire table, not nmodes)
+ * 40 | <reserved (0xff) ...>                              |  empty "mode data table" entries are 0xff-filled
+ *
+ * mode data table: array of:
+ *     struct mode_data {
+ *         uint16_t version;
+ *         uint16_t datasize;
+ *         uint28_t flashoffset; // from beginning of flash
+ *         uint4_t modeid;
+ *         uint32_t data_djb2;
+ *     };
+ *
+ * mode data blobs are typically allocated smallest to largest (according to
+ * mode_storage_class), from the last page of flash (cf. bsp-storage.h) down
+ */
+
+#define STORAGE_MAGIC "\xf0\x9f\x8f\xb3\xef\xb8\x8f\xe2\x80\x8d\xe2\x9a\xa7\xef\xb8\x8f"
+#define STORAGE_MAGIC_LEN 16
+
+__attribute__((__packed__)) struct mode_data {
+    uint16_t version;
+    uint16_t datasize;
+    uint32_t offsetandmode; // mode ID stored in MSNybble
+    uint32_t data_djb2;
+};
+
+#define MAX_MDT_ELEMENTS ((256 - 64) / sizeof(struct mode_data)) /* should be 16 */
+
+__attribute__((__packed__)) struct storage_header {
+    // +0
+    uint8_t magic[STORAGE_MAGIC_LEN];
+    // +16
+    uint16_t fwversion;
+    uint8_t curmode;
+    uint8_t nmodes; // *stored* modes, not modes it knows of
+    uint8_t reserved[8];
+    uint32_t table_djb2;
+    // +32
+    struct mode_data mode_data_table[MAX_MDT_ELEMENTS]; // 192 bytes in size
+    // +224
+    uint8_t reserved2[32];
+};
+
+// TODO: static assert sizeof(struct storage_header) == 256
+// TODO: static assert MAX_MDT_ELEMENTS >= 16
+
+extern bool header_valid;
+extern struct storage_header header_tmp;
+extern uint8_t data_tmp[256];
+extern uint16_t mode_bad;
+
+#endif
+

src/storage_save.c

@@ -0,0 +1,172 @@
+// vim: set et:
+
+#include "bsp-info.h"
+#include "bsp-storage.h"
+#include "mode.h"
+#include "storage.h"
+
+#if defined(PERSISTENT_STORAGE) && defined(DBOARD_HAS_STORAGE)
+
+#include "storage_internal.h"
+
+bool storage_priv_mode_has(int i) {
+    if (mode_list[i]->storage.stclass == mode_storage_none) return false;
+    if (mode_list[i]->storage.get_size == NULL) return false;
+    if (mode_list[i]->storage.get_data == NULL) return false;
+    if (mode_list[i]->storage.is_dirty == NULL) return false;
+
+    return true;
+}
+void* storage_priv_get_header_ptr(void) {
+    return &header_tmp;
+}
+
+static struct mode_storage msto[16];
+
+static size_t storage_allocate_new(void) {
+    static const size_t stclass_sz[] = { 0, 32, 128, 512, 0xffffffff };
+
+    memcpy(header_tmp.magic, STORAGE_MAGIC, STORAGE_MAGIC_LEN);
+    memset(header_tmp.reserved, 0xff, sizeof(header_tmp.reserved));
+    memset(header_tmp.reserved2, 0xff, sizeof(header_tmp.reserved2));
+    memset(header_tmp.mode_data_table, 0xff, sizeof(header_tmp.mode_data_table));
+
+    header_tmp.fwversion = STORAGE_VER;
+    header_tmp.curmode = mode_current_id;
+    header_tmp.nmodes = 0;
+
+    size_t current_page = STORAGE_SIZE - STORAGE_ERASEWRITE_ALIGN,
+           current_page_end = STORAGE_SIZE - sizeof(struct storage_header);
+    size_t current_wrhead = current_page;
+    size_t npages = 1;
+
+    for (enum mode_storage_class stcls = mode_storage_32b; stcls <= mode_storage_big; ++stcls) {
+        for (int mode = 1; mode < 16; ++mode) {
+            if (mode_list[mode] == NULL || !storage_priv_mode_has(mode)) continue;
+            if (mode_list[mode]->storage.stclass != stcls) continue;
+
+            // too big for the class? don't write the data, then
+            uint16_t dsize = mode_list[mode]->storage.get_size();
+            if (dsize > stclass_sz[stcls]) continue;
+
+            if (current_wrhead + dsize > current_page_end) { // FIXME: data that is >1 page size (do we want to support this?)
+                current_page_end = current_page;
+                current_page -= STORAGE_ERASEWRITE_ALIGN;
+                current_wrhead = current_page;
+                ++npages;
+
+                if (current_page < storage_get_program_offset() + storage_get_program_size())
+                    return 0; // welp, out of space
+            }
+
+            struct mode_data* md = &header_tmp.mode_data_table[header_tmp.nmodes];
+            md->version = mode_list[mode]->version;
+            md->datasize = dsize;
+            md->offsetandmode = current_wrhead | ((uint32_t)mode << 28);
+            msto[header_tmp.nmodes] = mode_list[mode]->storage; // copy to RAM because mode_list is in rodata!
+
+            current_wrhead += stclass_sz[stcls];
+
+            uint32_t hash = str_hash_djb2_init();
+
+            for (size_t i = 0; i < dsize; i += sizeof(data_tmp)) {
+                size_t tohash = sizeof(data_tmp);
+                if (dsize - i < tohash) tohash = dsize - i;
+
+                mode_list[mode]->storage.get_data(data_tmp, i, tohash);
+
+                hash = str_hash_djb2_digest(hash, data_tmp, tohash);
+            }
+
+            md->data_djb2 = hash;
+            ++header_tmp.nmodes;
+        }
+    }
+
+    header_tmp.table_djb2 = str_hash_djb2(header_tmp.mode_data_table,
+            sizeof(header_tmp.mode_data_table));
+
+    return npages;
+}
+
+static void st_safe_memcpy(void* dst, const void* src, size_t size) {
+    const uint64_t* s = src;
+    uint64_t* d = dst;
+    for (size_t i = 0; i < (size>>3); ++i) {
+        d[i] = s[i];
+        asm volatile("":::"memory");
+    }
+
+    if (size & 7) {
+        for (size_t i = (size>>3)<<3; i < size; ++i) {
+            ((uint8_t*)dst)[i] = ((const uint8_t*)src)[i];
+            asm volatile("":::"memory");
+        }
+    }
+}
+
+static void storage_serialize_xip(size_t page,
+        size_t pagestart, size_t pageend, void* dest) {
+    if (page == 0) {
+        st_safe_memcpy((uint8_t*)dest + pageend - pagestart,
+                &header_tmp, sizeof(struct storage_header));
+    }
+
+    for (size_t i = 0; i < header_tmp.nmodes; ++i) {
+        struct mode_data* md = &header_tmp.mode_data_table[i];
+        uint32_t mdoffset = md->offsetandmode & ((1<<28)-1);
+        uint32_t mdsize = md->datasize;
+
+        if (mdoffset < pagestart || mdoffset + mdsize >= pageend)
+            continue; // must be fully within the page
+
+        msto[i].get_data((uint8_t*)dest + mdoffset - pagestart, 0, mdsize);
+    }
+}
+
+static void storage_write_data(void) {
+    size_t npages = storage_allocate_new();
+    if (npages == 0) {
+        storage_init();
+        return; // TODO: error, somehow
+    }
+
+    static uint8_t base[STORAGE_ERASEWRITE_ALIGN]; // TODO FIXME: HUGE RAM HOG!
+
+    size_t current_page = STORAGE_SIZE - STORAGE_ERASEWRITE_ALIGN,
+           current_page_end = STORAGE_SIZE - sizeof(struct storage_header);
+    for (size_t page = 0; page < npages; ++page) {
+        storage_serialize_xip(page, current_page, current_page_end, base);
+
+        if (!storage_erasewrite(current_page, base, STORAGE_ERASEWRITE_ALIGN)) {
+            storage_init();
+            return; // TODO: error, somehow
+        }
+
+        current_page_end = current_page;
+        current_page -= STORAGE_ERASEWRITE_ALIGN;
+    }
+
+    // TODO:
+    // * save on a timer event?
+    // * try to save when unplugging???
+}
+
+bool storage_flush_data(void) {
+    if (mode_bad != 0 || mode_current_id != header_tmp.curmode) {
+        storage_write_data();
+        return true;
+    } else for (int i = 1; i < 16; ++i) {
+        if (mode_list[i] == NULL || !storage_priv_mode_has(i)) continue;
+
+        if (mode_list[i]->storage.is_dirty()) {
+            storage_write_data();
+            return true;
+        }
+    }
+
+    return false;
+}
+
+#endif
+

src/usbstdio.h

@@ -13,5 +13,7 @@ void stdio_usb_set_itf_num(int itf);
 void stdio_usb_line_coding_cb(cdc_line_coding_t const* line_coding);
 #endif
 
+void stdio_uart_init(void);
+
 #endif
 

src/vnd_cfg.c

@@ -18,6 +18,8 @@ static uint32_t rxavail, rxpos, txpos;
 
 static int VND_N_CFG = 0;
 
+extern uint8_t data_tmp[256];
+
 void vnd_cfg_init(void) {
     rxavail = 0;
     rxpos   = 0;
@@ -125,6 +127,8 @@ void vnd_cfg_task(void) {
     uint8_t cmd = vnd_cfg_read_byte();
     uint8_t verbuf[2];
 
+    //printf("vcfg %02x\n", cmd);
+
     if (cmd & 0xf0) {
         uint8_t mode = (uint8_t)(cmd & 0xf0) >> 4;
         uint8_t mcmd = cmd & 0x0f;
@@ -191,11 +195,54 @@ void vnd_cfg_task(void) {
         case cfg_cmd_get_infostr:
             vnd_cfg_write_str(cfg_resp_ok, INFO_PRODUCT(INFO_BOARDNAME));
             break;
+#if defined(PERSISTENT_STORAGE) && defined(DBOARD_HAS_STORAGE)
+        case cfg_cmd_storage_get_header:
+            vnd_cfg_write_resp(cfg_resp_ok, 256, storage_priv_get_header_ptr());
+            break;
+        case cfg_cmd_storage_get_modedata:
+            verbuf[0] = vnd_cfg_read_byte();
+            if (verbuf[0] == 0 || verbuf[0] >= 16 || mode_list[verbuf[0]] == NULL) {
+                vnd_cfg_write_resp(cfg_resp_nosuchmode, 0, NULL);
+            } else if (!storage_priv_mode_has(verbuf[0])) {
+                vnd_cfg_write_resp(cfg_resp_badarg, 0, NULL);
+            } else {
+                uint32_t len = storage_mode_get_info(verbuf[0]).size;
+                vnd_cfg_write_byte(cfg_resp_ok);
+                if (len < (1<<7)) {
+                    vnd_cfg_write_byte(len);
+                } else if (len < (1<<14)) {
+                    vnd_cfg_write_byte((len & 0x7f) | 0x80);
+                    vnd_cfg_write_byte((len >> 7) & 0x7f);
+                } else {
+                    vnd_cfg_write_byte((len & 0x7f) | 0x80);
+                    vnd_cfg_write_byte(((len >> 7) & 0x7f) | 0x80);
+                    vnd_cfg_write_byte(((len >> 14) & 0x7f));
+                }
+
+                for (size_t i = 0; i < len; i += sizeof data_tmp) {
+                    size_t tosend = sizeof data_tmp;
+                    if (tosend > len - i) tosend = len - i;
+                    storage_mode_read(verbuf[0], data_tmp, i, tosend);
+
+                    for (size_t ii = 0; ii < tosend; ++ii)
+                        vnd_cfg_write_byte(data_tmp[ii]);
+                }
+
+                vnd_cfg_write_flush();
+            }
+            break;
+        case cfg_cmd_storage_flush_data:
+            verbuf[0] = storage_flush_data() ? 1 : 0;
+            vnd_cfg_write_resp(cfg_resp_ok, 1, verbuf);
+            break;
+#endif
         default:
             vnd_cfg_write_resp(cfg_resp_illcmd, 0, NULL);
             break;
         }
     }
+
+    //printf("vnd cfg cmd=%02x done\n", cmd);
 }
 #else /* CFG_TUD_VENDOR == 0 */
 void vnd_cfg_init(void) { }

src/vnd_cfg.h

@@ -30,6 +30,12 @@ enum cfg_cmd {
     cfg_cmd_get_cur_mode = 0x02,
     cfg_cmd_set_cur_mode = 0x03,
     cfg_cmd_get_infostr  = 0x04,
+
+#if defined(PERSISTENT_STORAGE) && defined(DBOARD_HAS_STORAGE)
+    cfg_cmd_storage_get_header   = 0x0c,
+    cfg_cmd_storage_get_modedata = 0x0d,
+    cfg_cmd_storage_flush_data   = 0x0e,
+#endif
 };
 
 // common commands for every mode