dfu: add backend for MTD device
authorPatrick Delaunay <patrick.delaunay@st.com>
Mon, 14 Oct 2019 07:28:04 +0000 (09:28 +0200)
committerMarek Vasut <marek.vasut+renesas@gmail.com>
Thu, 31 Oct 2019 11:12:31 +0000 (12:12 +0100)
Add DFU backend for MTD device: allow to read
and write on all MTD device (NAND, SPI-NOR,
SPI-NAND,...)

For example :
> set dfu_alt_info "nand_raw raw 0x0 0x100000"
> dfu 0 mtd nand0

This MTD backend provides the same level than dfu nand
backend for NAND and dfu sf backend for SPI-NOR;
So it can replace booth of them but it also
add support of spi-nand.

> set dfu_alt_info "nand_raw raw 0x0 0x100000"
> dfu 0 mtd spi-nand0

The backend code is based on the "mtd" command
introduced by commit 5db66b3aee6f ("cmd: mtd:
add 'mtd' command")

Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
doc/README.dfu
drivers/dfu/Kconfig
drivers/dfu/Makefile
drivers/dfu/dfu.c
drivers/dfu/dfu_mtd.c [new file with mode: 0644]
include/dfu.h

index d1ef7e406455a98b86f6f330b33f71fafbf80507..96bced39d837d9c01182c164b78da21d98092b8a 100644 (file)
@@ -21,6 +21,7 @@ Overview:
   - NAND
   - RAM
   - SF (serial flash)
+  - MTD (all MTD device: NAND, SPI-NOR, SPI-NAND,...)
 
   These DFU backends are also used by
   - the dfutftp (see README.dfutftp)
@@ -30,6 +31,7 @@ Configuration Options:
   CONFIG_DFU
   CONFIG_DFU_OVER_USB
   CONFIG_DFU_MMC
+  CONFIG_DFU_MTD
   CONFIG_DFU_NAND
   CONFIG_DFU_RAM
   CONFIG_DFU_SF
@@ -104,6 +106,13 @@ Commands:
 
       with <partid> is the MTD partition index
 
+  "mtd" (all MTD device: NAND, SPI-NOR, SPI-NAND,...)
+    cmd: dfu 0 mtd <dev>
+      with <dev> the mtd identifier as defined in mtd command
+      (nand0, nor0, spi-nand0,...)
+    each element in "dfu_alt_info" =
+      <name> raw <offset> <size>  raw access to mtd device
+
   <interface> and <dev> are absent:
     the dfu command to use multiple devices
     cmd: dfu 0 list
@@ -114,6 +123,7 @@ Commands:
        nand <dev>=<alt1>;....;<altN>
        ram <dev>=<alt1>;....;<altN>
        sf <dev>=<alt1>;....;<altN>
+       mtd <dev>=<alt1>;....;<altN>
 
 
 Host tools:
index 1cbec81128535a6650563eab2f349d3217e88ecf..1e1dff33d25afec3c821fca4997dacdb541d4c03 100644 (file)
@@ -54,5 +54,11 @@ config DFU_SF_PART
          This option enables the support of "part" and "partubi" target in
          SPI flash DFU back end.
 
+config DFU_MTD
+       bool "MTD back end for DFU"
+       depends on MTD
+       help
+         This option enables using DFU to read and write to on any MTD device.
+
 endif
 endmenu
index 4164f342acc702fde160550893eb1d4ef91d9962..ebb119f3982683bf565b626763acd14038baabc3 100644 (file)
@@ -5,6 +5,7 @@
 
 obj-$(CONFIG_$(SPL_)DFU) += dfu.o
 obj-$(CONFIG_$(SPL_)DFU_MMC) += dfu_mmc.o
+obj-$(CONFIG_$(SPL_)DFU_MTD) += dfu_mtd.o
 obj-$(CONFIG_$(SPL_)DFU_NAND) += dfu_nand.o
 obj-$(CONFIG_$(SPL_)DFU_RAM) += dfu_ram.o
 obj-$(CONFIG_$(SPL_)DFU_SF) += dfu_sf.o
index 50919fcae9aa22f39a94f1b62ba008ac2067b988..4f4a07b790c75aaed1bb6bed22f196e689f9fc74 100644 (file)
@@ -462,6 +462,9 @@ static int dfu_fill_entity(struct dfu_entity *dfu, char *s, int alt,
        if (strcmp(interface, "mmc") == 0) {
                if (dfu_fill_entity_mmc(dfu, devstr, s))
                        return -1;
+       } else if (strcmp(interface, "mtd") == 0) {
+               if (dfu_fill_entity_mtd(dfu, devstr, s))
+                       return -1;
        } else if (strcmp(interface, "nand") == 0) {
                if (dfu_fill_entity_nand(dfu, devstr, s))
                        return -1;
@@ -566,7 +569,7 @@ int dfu_config_entities(char *env, char *interface, char *devstr)
 const char *dfu_get_dev_type(enum dfu_device_type t)
 {
        const char *const dev_t[] = {NULL, "eMMC", "OneNAND", "NAND", "RAM",
-                                    "SF"};
+                                    "SF", "MTD"};
        return dev_t[t];
 }
 
diff --git a/drivers/dfu/dfu_mtd.c b/drivers/dfu/dfu_mtd.c
new file mode 100644 (file)
index 0000000..13fc837
--- /dev/null
@@ -0,0 +1,232 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * dfu_mtd.c -- DFU for MTD device.
+ *
+ * Copyright (C) 2019,STMicroelectronics - All Rights Reserved
+ *
+ * Based on dfu_nand.c
+ */
+
+#include <common.h>
+#include <dfu.h>
+#include <mtd.h>
+
+static bool mtd_is_aligned_with_block_size(struct mtd_info *mtd, u64 size)
+{
+       return !do_div(size, mtd->erasesize);
+}
+
+static int mtd_block_op(enum dfu_op op, struct dfu_entity *dfu,
+                       u64 offset, void *buf, long *len)
+{
+       u64 off, lim, remaining;
+       struct mtd_info *mtd = dfu->data.mtd.info;
+       struct mtd_oob_ops io_op = {};
+       int ret = 0;
+       bool has_pages = mtd->type == MTD_NANDFLASH ||
+                        mtd->type == MTD_MLCNANDFLASH;
+
+       /* if buf == NULL return total size of the area */
+       if (!buf) {
+               *len = dfu->data.mtd.size;
+               return 0;
+       }
+
+       off = dfu->data.mtd.start + offset + dfu->bad_skip;
+       lim = dfu->data.mtd.start + dfu->data.mtd.size;
+
+       if (off >= lim) {
+               printf("Limit reached 0x%llx\n", lim);
+               *len = 0;
+               return op == DFU_OP_READ ? 0 : -EIO;
+       }
+       /* limit request with the available size */
+       if (off + *len >= lim)
+               *len = lim - off;
+
+       if (!mtd_is_aligned_with_block_size(mtd, off)) {
+               printf("Offset not aligned with a block (0x%x)\n",
+                      mtd->erasesize);
+               return 0;
+       }
+
+       /* first erase */
+       if (op == DFU_OP_WRITE) {
+               struct erase_info erase_op = {};
+
+               remaining = round_up(*len, mtd->erasesize);
+               erase_op.mtd = mtd;
+               erase_op.addr = off;
+               erase_op.len = mtd->erasesize;
+               erase_op.scrub = 0;
+
+               while (remaining) {
+                       if (erase_op.addr + remaining > lim) {
+                               printf("Limit reached 0x%llx while erasing at offset 0x%llx\n",
+                                      lim, off);
+                               return -EIO;
+                       }
+
+                       ret = mtd_erase(mtd, &erase_op);
+
+                       if (ret) {
+                               /* Abort if its not a bad block error */
+                               if (ret != -EIO) {
+                                       printf("Failure while erasing at offset 0x%llx\n",
+                                              erase_op.fail_addr);
+                                       return 0;
+                               }
+                               printf("Skipping bad block at 0x%08llx\n",
+                                      erase_op.addr);
+                       } else {
+                               remaining -= mtd->erasesize;
+                       }
+
+                       /* Continue erase behind bad block */
+                       erase_op.addr += mtd->erasesize;
+               }
+       }
+
+       io_op.mode = MTD_OPS_AUTO_OOB;
+       io_op.len = *len;
+       if (has_pages && io_op.len > mtd->writesize)
+               io_op.len = mtd->writesize;
+       io_op.ooblen = 0;
+       io_op.datbuf = buf;
+       io_op.oobbuf = NULL;
+
+       /* Loop over to do the actual read/write */
+       remaining = *len;
+       while (remaining) {
+               if (off + remaining > lim) {
+                       printf("Limit reached 0x%llx while %s at offset 0x%llx\n",
+                              lim, op == DFU_OP_READ ? "reading" : "writing",
+                              off);
+                       if (op == DFU_OP_READ) {
+                               *len -= remaining;
+                               return 0;
+                       } else {
+                               return -EIO;
+                       }
+               }
+
+               /* Skip the block if it is bad */
+               if (mtd_is_aligned_with_block_size(mtd, off) &&
+                   mtd_block_isbad(mtd, off)) {
+                       off += mtd->erasesize;
+                       dfu->bad_skip += mtd->erasesize;
+                       continue;
+               }
+
+               if (op == DFU_OP_READ)
+                       ret = mtd_read_oob(mtd, off, &io_op);
+               else
+                       ret = mtd_write_oob(mtd, off, &io_op);
+
+               if (ret) {
+                       printf("Failure while %s at offset 0x%llx\n",
+                              op == DFU_OP_READ ? "reading" : "writing", off);
+                       return -EIO;
+               }
+
+               off += io_op.retlen;
+               remaining -= io_op.retlen;
+               io_op.datbuf += io_op.retlen;
+               io_op.len = remaining;
+               if (has_pages && io_op.len > mtd->writesize)
+                       io_op.len = mtd->writesize;
+       }
+
+       return ret;
+}
+
+static int dfu_get_medium_size_mtd(struct dfu_entity *dfu, u64 *size)
+{
+       *size = dfu->data.mtd.info->size;
+
+       return 0;
+}
+
+static int dfu_read_medium_mtd(struct dfu_entity *dfu, u64 offset, void *buf,
+                              long *len)
+{
+       int ret = -1;
+
+       switch (dfu->layout) {
+       case DFU_RAW_ADDR:
+               ret = mtd_block_op(DFU_OP_READ, dfu, offset, buf, len);
+               break;
+       default:
+               printf("%s: Layout (%s) not (yet) supported!\n", __func__,
+                      dfu_get_layout(dfu->layout));
+       }
+
+       return ret;
+}
+
+static int dfu_write_medium_mtd(struct dfu_entity *dfu,
+                               u64 offset, void *buf, long *len)
+{
+       int ret = -1;
+
+       switch (dfu->layout) {
+       case DFU_RAW_ADDR:
+               ret = mtd_block_op(DFU_OP_WRITE, dfu, offset, buf, len);
+               break;
+       default:
+               printf("%s: Layout (%s) not (yet) supported!\n", __func__,
+                      dfu_get_layout(dfu->layout));
+       }
+
+       return ret;
+}
+
+static int dfu_flush_medium_mtd(struct dfu_entity *dfu)
+{
+       return 0;
+}
+
+static unsigned int dfu_polltimeout_mtd(struct dfu_entity *dfu)
+{
+       return DFU_DEFAULT_POLL_TIMEOUT;
+}
+
+int dfu_fill_entity_mtd(struct dfu_entity *dfu, char *devstr, char *s)
+{
+       char *st;
+       struct mtd_info *mtd;
+       bool has_pages;
+
+       mtd = get_mtd_device_nm(devstr);
+       if (IS_ERR_OR_NULL(mtd))
+               return -ENODEV;
+       put_mtd_device(mtd);
+
+       dfu->dev_type = DFU_DEV_MTD;
+       dfu->data.mtd.info = mtd;
+
+       has_pages = mtd->type == MTD_NANDFLASH || mtd->type == MTD_MLCNANDFLASH;
+       dfu->max_buf_size = has_pages ? mtd->erasesize : 0;
+
+       st = strsep(&s, " ");
+       if (!strcmp(st, "raw")) {
+               dfu->layout = DFU_RAW_ADDR;
+               dfu->data.mtd.start = simple_strtoul(s, &s, 16);
+               s++;
+               dfu->data.mtd.size = simple_strtoul(s, &s, 16);
+       } else {
+               printf("%s: (%s) not supported!\n", __func__, st);
+               return -1;
+       }
+
+       dfu->get_medium_size = dfu_get_medium_size_mtd;
+       dfu->read_medium = dfu_read_medium_mtd;
+       dfu->write_medium = dfu_write_medium_mtd;
+       dfu->flush_medium = dfu_flush_medium_mtd;
+       dfu->poll_timeout = dfu_polltimeout_mtd;
+
+       /* initial state */
+       dfu->inited = 0;
+
+       return 0;
+}
index 7d60ffc2282156afd20a5a71a6837e51882c2cf9..924952f805c83a4c50bd0a29087383c6883ed569 100644 (file)
@@ -22,6 +22,7 @@ enum dfu_device_type {
        DFU_DEV_NAND,
        DFU_DEV_RAM,
        DFU_DEV_SF,
+       DFU_DEV_MTD,
 };
 
 enum dfu_layout {
@@ -55,6 +56,14 @@ struct mmc_internal_data {
        unsigned int part;
 };
 
+struct mtd_internal_data {
+       struct mtd_info *info;
+
+       /* RAW programming */
+       u64 start;
+       u64 size;
+};
+
 struct nand_internal_data {
        /* RAW programming */
        u64 start;
@@ -105,6 +114,7 @@ struct dfu_entity {
 
        union {
                struct mmc_internal_data mmc;
+               struct mtd_internal_data mtd;
                struct nand_internal_data nand;
                struct ram_internal_data ram;
                struct sf_internal_data sf;
@@ -249,6 +259,17 @@ static inline int dfu_fill_entity_sf(struct dfu_entity *dfu, char *devstr,
 }
 #endif
 
+#if CONFIG_IS_ENABLED(DFU_MTD)
+int dfu_fill_entity_mtd(struct dfu_entity *dfu, char *devstr, char *s);
+#else
+static inline int dfu_fill_entity_mtd(struct dfu_entity *dfu, char *devstr,
+                                     char *s)
+{
+       puts("MTD support not available!\n");
+       return -1;
+}
+#endif
+
 /**
  * dfu_tftp_write - Write TFTP data to DFU medium
  *