feat: Add SD multiblock write support

Add support for CMD25 to write multiple sequential blocks in one go, in
an attempt to speed up uploads. It still needs to be benchmarked if this
actually results in a meaningful speedup.

Also fix logging from SD driver to be visible even when running under
micropython.

Signed-off-by: Matthias Blankertz <matthias@blankertz.org>

software/modules/rp2_sd/module.c

@@ -1,11 +1,14 @@
 // SPDX-License-Identifier: MIT
 // Copyright (c) 2025 Matthias Blankertz <matthias@blankertz.org>
 
+#include <stdarg.h>
+
 #include "py/obj.h"
 #include "sd.h"
 
 // Include MicroPython API.
 #include "py/mperrno.h"
+#include "py/mpprint.h"
 #include "py/runtime.h"
 
 // This module is RP2 specific
@@ -13,6 +16,14 @@
 
 #include <string.h>
 
+void sd_printf(const char *fmt, ...)
+{
+    va_list ap;
+    va_start(ap, fmt);
+    mp_vprintf(&mp_sys_stdout_print, fmt, ap);
+    va_end(ap);
+}
+
 const mp_obj_type_t sdcard_type;
 struct sdcard_obj {
     mp_obj_base_t base;
@@ -89,11 +100,14 @@ static mp_obj_t sdcard_writeblocks(mp_obj_t self_obj, mp_obj_t block_obj, mp_obj
     if (bufinfo.len % SD_SECTOR_SIZE != 0)
         mp_raise_ValueError(MP_ERROR_TEXT("Buffer length is invalid"));
     const int nblocks = bufinfo.len / SD_SECTOR_SIZE;
-    for (int block = 0; block < nblocks; block++) {
-        // TODO: Implement CMD25 write multiple blocks
-        if (!sd_writeblock(&self->sd_context, start_block + block, bufinfo.buf + block * SD_SECTOR_SIZE))
-            mp_raise_OSError(MP_EIO);
+    bool ret;
+    if (nblocks > 1) {
+        ret = sd_writeblocks(&self->sd_context, start_block, nblocks, bufinfo.buf);
+    } else {
+        ret = sd_writeblock(&self->sd_context, start_block, bufinfo.buf);
     }
+    if (!ret)
+        mp_raise_OSError(MP_EIO);
     return mp_const_none;
 }
 static MP_DEFINE_CONST_FUN_OBJ_3(sdcard_writeblocks_obj, sdcard_writeblocks);

software/modules/rp2_sd/sd.c

@@ -6,7 +6,7 @@
 #include <stdio.h>
 #include <string.h>
 
-// #define SD_DEBUG
+extern void sd_printf(const char *fmt, ...);
 
 #define SD_R1_ILLEGAL_COMMAND (1 << 2)
 
@@ -26,14 +26,14 @@ static bool sd_early_init(void)
     for (int i = 0; i < 500; ++i) {
         if (sd_cmd(0, 0, 1, &buf)) {
 #ifdef SD_DEBUG
-            printf("CMD0 resp %02hhx\n", buf);
+            sd_printf("CMD0 resp %02x\n", buf);
 #endif
             if (buf == 0x01) {
                 return true;
             }
         }
 #ifdef SD_DEBUG
-        printf("CMD0 timeout, try again...\n");
+        sd_printf("CMD0 timeout, try again...\n");
 #endif
     }
     return false;
@@ -44,14 +44,14 @@ static bool sd_check_interface_condition(void)
     uint8_t buf[5];
     if (sd_cmd(8, 0x000001AA, 5, buf)) {
         if ((buf[3] & 0xf) != 0x1 || buf[4] != 0xAA) {
-            printf("sd_init: check interface condition failed\n");
+            sd_printf("sd_init: check interface condition failed\n");
             return false;
         }
     } else {
         if (buf[0] & SD_R1_ILLEGAL_COMMAND) {
-            printf("sd_init: check interface condition returned illegal command - old card?\n");
+            sd_printf("sd_init: check interface condition returned illegal command - old card?\n");
         } else {
-            printf("sd_init: check interface condition failed\n");
+            sd_printf("sd_init: check interface condition failed\n");
             return false;
         }
     }
@@ -72,12 +72,12 @@ static bool sd_send_op_cond(void)
         if (!result) {
             if (use_acmd && buf & SD_R1_ILLEGAL_COMMAND) {
 #ifdef SD_DEBUG
-                printf("sd_init: card does not understand ACMD41, try CMD1...\n");
+                sd_printf("sd_init: card does not understand ACMD41, try CMD1...\n");
 #endif
                 use_acmd = false;
                 continue;
             } else if (buf != 0x01) {
-                printf("sd_init: send_op_cond failed\n");
+                sd_printf("sd_init: send_op_cond failed\n");
                 return false;
             } else {
                 continue;
@@ -87,7 +87,7 @@ static bool sd_send_op_cond(void)
             return true;
         }
     }
-    printf("sd_init: send_op_cond: timeout waiting for !idle\n");
+    sd_printf("sd_init: send_op_cond: timeout waiting for !idle\n");
     return false;
 }
 
@@ -107,7 +107,7 @@ static void sd_dump_cid [[maybe_unused]] (void)
         const uint8_t crc = sd_crc7(15, buf);
         const uint8_t card_crc = buf[15] >> 1;
         if (card_crc != crc) {
-            printf("CRC mismatch: Got %02hhx, expected %02hhx\n", card_crc, crc);
+            sd_printf("CRC mismatch: Got %02x, expected %02x\n", card_crc, crc);
             // Some cheap SD cards always report CRC=0, don't fail in that case
             if (card_crc != 0) {
                 return;
@@ -122,8 +122,8 @@ static void sd_dump_cid [[maybe_unused]] (void)
         uint32_t psn = buf[9] << 24 | buf[10] << 16 | buf[11] << 8 | buf[12];
         int mdt_year = 2000 + ((buf[13] & 0xf) << 4 | (buf[14] & 0xf0) >> 4);
         int mdt_month = buf[14] & 0x0f;
-        printf("CID: mid: %02hhx, oid: %.2s, pnm: %.5s, prv: %02hhx, psn: %08" PRIx32 ", mdt_year: %d, mdt_month: %d\n",
-               mid, oid, pnm, prv, psn, mdt_year, mdt_month);
+        sd_printf("CID: mid: %02x, oid: %.2s, pnm: %.5s, prv: %02x, psn: %08" PRIx32 ", mdt_year: %d, mdt_month: %d\n",
+                  mid, oid, pnm, prv, psn, mdt_year, mdt_month);
     }
 }
 
@@ -131,13 +131,13 @@ static bool sd_read_csd(struct sd_context *sd_context)
 {
     uint8_t buf[16];
     if (!sd_cmd_read(9, 0, 16, buf)) {
-        printf("Failed to read CSD\n");
+        sd_printf("Failed to read CSD\n");
         return false;
     }
     const uint8_t crc = sd_crc7(15, buf);
     const uint8_t card_crc = buf[15] >> 1;
     if (card_crc != crc) {
-        printf("CRC mismatch: Got %02hhx, expected %02hhx\n", card_crc, crc);
+        sd_printf("CRC mismatch: Got %02x, expected %02x\n", card_crc, crc);
         // Some cheap SD cards always report CRC=0, don't fail in that case
         if (card_crc != 0) {
             return false;
@@ -151,12 +151,12 @@ static bool sd_read_csd(struct sd_context *sd_context)
     switch (csd_ver) {
     case 0: {
         if (sd_context->sdhc_sdxc) {
-            printf("sd_init: Got CSD v1.0 but card is SDHC/SDXC?\n");
+            sd_printf("sd_init: Got CSD v1.0 but card is SDHC/SDXC?\n");
             return false;
         }
         const unsigned read_bl_len = buf[5] & 0xf;
         if (read_bl_len < 9 || read_bl_len > 11) {
-            printf("Invalid read_bl_len in CSD 1.0\n");
+            sd_printf("Invalid read_bl_len in CSD 1.0\n");
             return false;
         }
         blocksize = 1 << (buf[5] & 0xf);
@@ -176,15 +176,15 @@ static bool sd_read_csd(struct sd_context *sd_context)
         break;
     }
     case 2: {
-        printf("sd_init: Got CSD v3.0, but SDUC does not support SPI.\n");
+        sd_printf("sd_init: Got CSD v3.0, but SDUC does not support SPI.\n");
         return false;
     }
     }
     sd_context->blocks = blocks;
     sd_context->blocksize = blocksize;
 #ifdef SD_DEBUG
-    printf("CSD version %u.0, blocksize %u, blocks %u, capacity %llu MiB, max speed %u\n", version, blocksize, blocks,
-           ((uint64_t)blocksize * blocks) / (1024 * 1024), max_speed);
+    sd_printf("CSD version %u.0, blocksize %u, blocks %u, capacity %u MiB, max speed %u\n", version, blocksize, blocks,
+              (uint32_t)(((uint64_t)blocksize * blocks) / (1024 * 1024)), max_speed);
 #endif
     return true;
 }
@@ -205,11 +205,11 @@ bool sd_init(struct sd_context *sd_context, int mosi, int miso, int sck, int ss,
 
     uint32_t ocr;
     if (!sd_read_ocr(&ocr)) {
-        printf("sd_init: read OCR failed\n");
+        sd_printf("sd_init: read OCR failed\n");
         goto out_spi;
     }
     if ((ocr & 0x00380000) != 0x00380000) {
-        printf("sd_init: unsupported card voltage range\n");
+        sd_printf("sd_init: unsupported card voltage range\n");
         goto out_spi;
     }
 
@@ -219,11 +219,11 @@ bool sd_init(struct sd_context *sd_context, int mosi, int miso, int sck, int ss,
     sd_spi_set_bitrate(rate);
 
     if (!sd_read_ocr(&ocr)) {
-        printf("sd_init: read OCR failed\n");
+        sd_printf("sd_init: read OCR failed\n");
         goto out_spi;
     }
     if (!(ocr & (1 << 31))) {
-        printf("sd_init: card not powered up but !idle?\n");
+        sd_printf("sd_init: card not powered up but !idle?\n");
         goto out_spi;
     }
     sd_context->sdhc_sdxc = (ocr & (1 << 30));
@@ -234,20 +234,20 @@ bool sd_init(struct sd_context *sd_context, int mosi, int miso, int sck, int ss,
 
     if (sd_context->blocksize != SD_SECTOR_SIZE) {
         if (sd_context->blocksize != 1024 && sd_context->blocksize != 2048) {
-            printf("sd_init: Unsupported block size %u\n", sd_context->blocksize);
+            sd_printf("sd_init: Unsupported block size %u\n", sd_context->blocksize);
             goto out_spi;
         }
         // Attempt SET_BLOCKLEN command
         uint8_t resp[1];
         if (!sd_cmd(16, SD_SECTOR_SIZE, 1, resp)) {
-            printf("sd_init: SET_BLOCKLEN failed\n");
+            sd_printf("sd_init: SET_BLOCKLEN failed\n");
             goto out_spi;
         }
         // Successfully set blocksize to SD_SECTOR_SIZE, adjust context
         sd_context->blocks *= sd_context->blocksize / SD_SECTOR_SIZE;
 #ifdef SD_DEBUG
-        printf("Adjusted blocksize from %u to 512, card now has %u blocks\n", sd_context->blocksize,
-               sd_context->blocks);
+        sd_printf("Adjusted blocksize from %u to 512, card now has %u blocks\n", sd_context->blocksize,
+                  sd_context->blocks);
 #endif
         sd_context->blocksize = SD_SECTOR_SIZE;
     }
@@ -307,7 +307,7 @@ bool sd_readblock_complete(struct sd_context *sd_context)
 
 bool sd_readblock_is_complete(struct sd_context *sd_context) { return sd_cmd_read_is_complete(); }
 
-bool sd_writeblock(struct sd_context *sd_context, size_t sector_num, uint8_t buffer[const static SD_SECTOR_SIZE])
+bool sd_writeblock(struct sd_context *sd_context, const size_t sector_num, uint8_t buffer[const static SD_SECTOR_SIZE])
 {
     if (!sd_context->initialized || sector_num >= sd_context->blocks)
         return false;
@@ -319,3 +319,20 @@ bool sd_writeblock(struct sd_context *sd_context, size_t sector_num, uint8_t buf
     }
     return sd_cmd_write(24, addr, SD_SECTOR_SIZE, buffer);
 }
+
+bool sd_writeblocks(struct sd_context *sd_context, const size_t sector_num, const size_t sectors, uint8_t *const buffer)
+{
+    if (!sd_context->initialized || sector_num + sectors >= sd_context->blocks)
+        return false;
+
+    if (!sd_context->sdhc_sdxc) {
+        // Don't use multi-block writes for SDSC for now
+        // Need to configure WRITE_BL_LEN correctly
+        for (size_t sector = 0; sector < sectors; ++sector) {
+            if (!sd_writeblock(sd_context, sector_num + sector, buffer + sector * SD_SECTOR_SIZE))
+                return false;
+        }
+        return true;
+    }
+    return sd_cmd_write_multiple(25, sector_num, sectors, SD_SECTOR_SIZE, buffer);
+}

software/modules/rp2_sd/sd.h

@@ -24,3 +24,4 @@ bool sd_readblock_complete(struct sd_context *context);
 bool sd_readblock_is_complete(struct sd_context *context);
 
 bool sd_writeblock(struct sd_context *context, size_t sector_num, uint8_t buffer[const static SD_SECTOR_SIZE]);
+bool sd_writeblocks(struct sd_context *sd_context, const size_t sector_num, const size_t sectors, uint8_t *buffer);

software/modules/rp2_sd/sd_spi.c

@@ -12,6 +12,8 @@
 #include <pico/time.h>
 #include <string.h>
 
+extern void sd_printf(const char *fmt, ...);
+
 typedef enum { DMA_READ_TOKEN, DMA_READ, DMA_IDLE } sd_dma_state;
 
 struct sd_dma_context {
@@ -222,7 +224,7 @@ bool sd_cmd_read_complete(void)
     sd_spi_read_blocking(0xff, &buf, 1);
     if (sd_spi_context.sd_dma_context.read_token_buf != 0xfe) {
 #ifdef SD_DEBUG
-        printf("read failed: invalid read token %02hhx\n", sd_spi_context.sd_dma_context.read_token_buf);
+        sd_printf("read failed: invalid read token %02x\n", sd_spi_context.sd_dma_context.read_token_buf);
 #endif
         return false;
     }
@@ -231,7 +233,7 @@ bool sd_cmd_read_complete(void)
     const uint16_t act_crc = sd_spi_context.sd_dma_context.crc_buf[0] << 8 | sd_spi_context.sd_dma_context.crc_buf[1];
     if (act_crc != expect_crc) {
 #ifdef SD_DEBUG
-        printf("read CRC fail: got %04hx, expected %04hx\n", act_crc, expect_crc);
+        sd_printf("read CRC fail: got %04x, expected %04x\n", act_crc, expect_crc);
 #endif
         return false;
     }
@@ -239,10 +241,9 @@ bool sd_cmd_read_complete(void)
     return true;
 }
 
-bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[const static datalen])
+static bool sd_cmd_write_begin(uint8_t cmd, uint32_t arg)
 {
-    uint8_t buf[2];
-    const uint16_t crc = sd_crc16(datalen, data);
+    uint8_t buf[1];
     sd_spi_cmd_send(cmd, arg);
     // Read up to 8 garbage bytes (0xff), followed by R1 (MSB is zero)
     bool got_r1 = false;
@@ -253,9 +254,20 @@ bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[cons
             break;
         }
     }
-    if (!got_r1 || buf[0] != 0x00)
-        goto abort;
-    buf[0] = 0xfe;
+    if (!got_r1 || buf[0] != 0x00) {
+#ifdef SD_DEBUG
+        sd_printf("write cmd fail: %02x\n", buf[0]);
+#endif
+        return false;
+    }
+    return true;
+}
+
+static bool sd_cmd_write_block(uint8_t token, unsigned datalen, uint8_t data[const static datalen])
+{
+    uint8_t buf[2];
+    const uint16_t crc = sd_crc16(datalen, data);
+    buf[0] = token;
     sd_spi_write_blocking(buf, 1);
     sd_spi_write_blocking(data, datalen);
     buf[0] = crc >> 8;
@@ -264,11 +276,16 @@ bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[cons
     sd_spi_read_blocking(0xff, buf, 1);
     if ((buf[0] & 0x1f) != 0x5) {
 #ifdef SD_DEBUG
-        printf("Write fail: %2hhx\n", buf[0]);
+        sd_printf("Write fail: %2x\n", buf[0]);
 #endif
-        goto abort;
+        return false;
     }
+    return true;
+}
 
+static bool sd_cmd_write_wait_nbusy(void)
+{
+    uint8_t buf[1];
     int timeout = 0;
     bool got_done = false;
     for (timeout = 0; timeout < 131072; ++timeout) {
@@ -279,18 +296,73 @@ bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[cons
         }
     }
 #ifdef SD_DEBUG
-    printf("dbg write end: %d, %2hhx\n", timeout, buf[0]);
+    sd_printf("dbg write end: %d, %2x\n", timeout, buf[0]);
 #endif
-    if (!got_done)
+    return got_done;
+}
+
+bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[const static datalen])
+{
+#ifdef SD_DEBUG
+    sd_printf("write 1 block at %u\n", arg);
+#endif
+    uint8_t buf[2];
+    if (!sd_cmd_write_begin(cmd, arg))
+        goto abort;
+    if (!sd_cmd_write_block(0xfe, datalen, data))
+        goto abort;
+    if (!sd_cmd_write_wait_nbusy())
         goto abort;
 
     gpio_put(sd_spi_context.ss, true);
     sd_spi_read_blocking(0xff, buf, 1);
+#ifdef SD_DEBUG
+    sd_printf("write ok\n");
+#endif
     return true;
 
 abort:
     gpio_put(sd_spi_context.ss, true);
     sd_spi_read_blocking(0xff, buf, 1);
+#ifdef SD_DEBUG
+    sd_printf("write fail\n");
+#endif
+    return false;
+}
+
+bool sd_cmd_write_multiple(uint8_t cmd, uint32_t arg, unsigned blocks, unsigned datalen, uint8_t *const data)
+{
+#ifdef SD_DEBUG
+    sd_printf("write %u blocks at %u\n", blocks, arg);
+#endif
+    uint8_t buf[2];
+    if (!sd_cmd_write_begin(cmd, arg))
+        goto abort;
+    for (unsigned i = 0; i < blocks; ++i) {
+        if (!sd_cmd_write_block(0b11111100, datalen, data + datalen * i))
+            goto abort;
+        if (!sd_cmd_write_wait_nbusy())
+            goto abort;
+    }
+    buf[0] = 0b11111101;
+    buf[1] = 0xff;
+    sd_spi_write_blocking(buf, 2);
+    if (!sd_cmd_write_wait_nbusy())
+        goto abort;
+
+    gpio_put(sd_spi_context.ss, true);
+    sd_spi_read_blocking(0xff, buf, 1);
+#ifdef SD_DEBUG
+    sd_printf("write ok\n");
+#endif
+    return true;
+
+abort:
+    gpio_put(sd_spi_context.ss, true);
+    sd_spi_read_blocking(0xff, buf, 1);
+#ifdef SD_DEBUG
+    sd_printf("write fail\n");
+#endif
     return false;
 }
 

software/modules/rp2_sd/sd_spi.h

@@ -27,3 +27,4 @@ bool sd_cmd_read_complete(void);
 bool sd_cmd_read_is_complete(void);
 
 bool sd_cmd_write(uint8_t cmd, uint32_t arg, unsigned datalen, uint8_t data[const static datalen]);
+bool sd_cmd_write_multiple(uint8_t cmd, uint32_t arg, unsigned blocks, unsigned datalen, uint8_t *const data);

software/tools/standalone-mp3/CMakeLists.txt

@@ -15,6 +15,8 @@ option(ENABLE_PLAY_TEST "Enable mp3 playback test" OFF)
 option(ENABLE_SD_READ_CRC "Enable crc check when reading from sd card" OFF)
 option(ENABLE_SD_DEBUG "Enable debug output for sd card driver" OFF)
 
+set(PICO_USE_FASTEST_SUPPORTED_CLOCK 1)
+
 # initialize the Raspberry Pi Pico SDK
 pico_sdk_init()
 
@@ -62,7 +64,7 @@ add_subdirectory(../../lib/helix_mp3 helix_mp3)
 
 target_link_libraries(standalone_mp3 PRIVATE pico_stdlib hardware_dma hardware_spi hardware_sync hardware_pio helix_mp3)
 target_include_directories(standalone_mp3 PRIVATE ${SD_LIB_DIR})
-target_compile_options(standalone_mp3 PRIVATE -Og -DSD_DEBUG)
+target_compile_options(standalone_mp3 PRIVATE -Og)
 
 
 pico_add_extra_outputs(standalone_mp3)

software/tools/standalone-mp3/main.c

@@ -2,6 +2,7 @@
 #include "sd.h"
 
 #include <math.h>
+#include <stdarg.h>
 #include <stdio.h>
 #include <string.h>
 
@@ -20,6 +21,14 @@ extern void sd_spi_dbg_clk(const int div, const int frac);
 
 extern void sd_spi_dbg_loop(void);
 
+void sd_printf(const char *fmt, ...)
+{
+    va_list ap;
+    va_start(ap, fmt);
+    vprintf(fmt, ap);
+    va_end(ap);
+}
+
 #define MAX_VOLUME 0x8000u
 
 void __time_critical_func(volume_adjust)(int16_t *restrict buf, size_t samples, uint16_t scalef)
@@ -157,7 +166,7 @@ static void write_test(struct sd_context *sd_context)
             data_buffer[i] ^= 0xff;
         }
 
-        if (!sd_writeblock(sd_context, 0, data_buffer)) {
+        if (!sd_writeblocks(sd_context, 0, sizeof(data_buffer) / SD_SECTOR_SIZE, data_buffer)) {
             printf("sd_writeblock failed\n");
             return;
         }