Skip to contentSkip to conversation

mtd-utils: nanddump: Allow 64-bit lengths

PATCH In review Brian Norris 3 participants 28 messages Nov 3, 2010 – Nov 16, 2010 Memory technology devices (MTD)

lore
Share & tools
lore
#0 lore raw

PATCH 1/10 mtd-utils: nanddump: Allow 64-bit lengths

We should allow the dump length to be 64-bit, especially since the value was read in as a "long long" by strtoll().

---
 nanddump.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
nanddump.c +1 -1
index 86a71c9..fe29596 100644
@@ -83,7 +83,7 @@ static bool			noecc = false;		// don't error correct
 static bool			noskipbad = false;	// don't skip bad blocks
 static bool			omitoob = false;	// omit oob data
 static unsigned long long	start_addr;		// start address
-static unsigned long		length;			// dump length
+static unsigned long long	length;			// dump length
 static const char		*mtddev;		// mtd device name
 static const char		*dumpfile;		// dump file name
 static bool			omitbad = false;
Signature · 1 line
1.7.0.4
#1 lore raw

PATCH 2/10 mtd-utils: nandwrite: Comment, style fixups

Comment on "blockalign" default value is incorrect; it only defaults to a 1x multiplier of the actual block size. Perhaps this is a relic of early NAND where all block sizes were 16KB?

Reformatted a multi-line comment.

Changed separete "if" statements to a combined "if-else-if" since they were logically combinable. Should have no effect on results with minor effect on efficiency.

---
 nandwrite.c |   13 +++++++------
 1 files changed, 7 insertions(+), 6 deletions(-)
nandwrite.c +7 -6
index b0c4366..3fbfca9 100644
@@ -120,7 +120,7 @@ static bool		forcelegacy = false;
 static bool		noecc = false;
 static bool		noskipbad = false;
 static bool		pad = false;
-static int		blockalign = 1; /* default to using 16K block size */
+static int		blockalign = 1; /* default to using actual block size */
 
 static void process_options(int argc, char * const argv[])
 {
@@ -483,8 +483,7 @@ int main(int argc, char * const argv[])
 				if ((ret = ioctl(fd, MEMGETBADBLOCK, &offs)) < 0) {
 					perror("ioctl(MEMGETBADBLOCK)");
 					goto closeall;
-				}
-				if (ret == 1) {
+				} else if (ret == 1) {
 					baderaseblock = true;
 					if (!quiet)
 						fprintf(stderr, "Bad block at %x, %u block(s) "
@@ -520,9 +519,11 @@ int main(int argc, char * const argv[])
 
 			/* No padding needed - we are done */
 			if (tinycnt == 0) {
-				// For standard input, set the imglen to 0 to signal
-				// the end of the "file". For non standard input, leave
-				// it as-is to detect an early EOF
+				/*
+				 * For standard input, set imglen to 0 to signal
+				 * the end of the "file". For nonstandard input,
+				 * leave it as-is to detect an early EOF.
+				 */
 				if (ifd == STDIN_FILENO) {
 					imglen = 0;
 				}
Signature · 1 line
1.7.0.4
#2 lore raw

PATCH 3/10 mtd-utils: nandwrite: Clarify usage of aligned "erasesize"

In reply to Brian Norris

Due to the presence of the "--block-align" flag, nandwrite uses a blocksize throughout that, depeding on the execution parameters, may not be the actual erasesize of the NAND flash. In order to clarify this situation for the untrained viewer of nandwrite's code, we should not change the value of "meminfo.erasesize" itself; rather, we can utilize a separate, calculated "ebsize_aligned". Then, when a user actually wants to refer to the physical erasesize, it's straightforward.

---
 nandwrite.c |   30 +++++++++++++++++-------------
 1 files changed, 17 insertions(+), 13 deletions(-)
nandwrite.c +17 -13
index 3fbfca9..66a9ef7 100644
@@ -275,6 +275,7 @@ int main(int argc, char * const argv[])
 	// points to the OOB for the current page in filebuf
 	unsigned char *oobreadbuf = NULL;
 	unsigned char *oobbuf = NULL;
+	int ebsize_aligned;
 
 	process_options(argc, argv);
 
@@ -296,9 +297,12 @@ int main(int argc, char * const argv[])
 		exit(EXIT_FAILURE);
 	}
 
-	/* Set erasesize to specified number of blocks - to match jffs2
-	 * (virtual) block size */
-	meminfo.erasesize *= blockalign;
+	/*
+	 * Pretend erasesize is specified number of blocks - to match jffs2
+	 *   (virtual) block size
+	 * Use this value throughout unless otherwise necessary
+	 */
+	ebsize_aligned = meminfo.erasesize * blockalign;
 
 	if (mtdoffset & (meminfo.writesize - 1)) {
 		fprintf(stderr, "The start address is not page-aligned !\n"
@@ -435,7 +439,7 @@ int main(int argc, char * const argv[])
 	}
 
 	// Allocate a buffer big enough to contain all the data (OOB included) for one eraseblock
-	filebuf_max = pagelen * meminfo.erasesize / meminfo.writesize;
+	filebuf_max = pagelen * ebsize_aligned / meminfo.writesize;
 	filebuf = xmalloc(filebuf_max);
 	erase_buffer(filebuf, filebuf_max);
 
@@ -459,8 +463,8 @@ int main(int argc, char * const argv[])
 		 * skipped block(s) is also bad (number of blocks depending on
 		 * the blockalign).
 		 */
-		while (blockstart != (mtdoffset & (~meminfo.erasesize + 1))) {
-			blockstart = mtdoffset & (~meminfo.erasesize + 1);
+		while (blockstart != (mtdoffset & (~ebsize_aligned + 1))) {
+			blockstart = mtdoffset & (~ebsize_aligned + 1);
 			offs = blockstart;
 
 			// if writebuf == filebuf, we are rewinding so we must not
@@ -474,7 +478,7 @@ int main(int argc, char * const argv[])
 			baderaseblock = false;
 			if (!quiet)
 				fprintf(stdout, "Writing data to block %d at offset 0x%x\n",
-						 blockstart / meminfo.erasesize, blockstart);
+						 blockstart / ebsize_aligned, blockstart);
 
 			/* Check all the blocks in an erase block for bad blocks */
 			if (noskipbad)
@@ -492,10 +496,10 @@ int main(int argc, char * const argv[])
 				}
 
 				if (baderaseblock) {
-					mtdoffset = blockstart + meminfo.erasesize;
+					mtdoffset = blockstart + ebsize_aligned;
 				}
-				offs +=  meminfo.erasesize / blockalign;
-			} while (offs < blockstart + meminfo.erasesize);
+				offs +=  ebsize_aligned / blockalign;
+			} while (offs < blockstart + ebsize_aligned);
 
 		}
 
@@ -642,7 +646,7 @@ int main(int argc, char * const argv[])
 			writebuf = filebuf;
 
 			erase.start = blockstart;
-			erase.length = meminfo.erasesize;
+			erase.length = ebsize_aligned;
 			fprintf(stderr, "Erasing failed write from %08lx-%08lx\n",
 				(long)erase.start, (long)erase.start+erase.length-1);
 			if (ioctl(fd, MEMERASE, &erase) != 0) {
@@ -654,14 +658,14 @@ int main(int argc, char * const argv[])
 			}
 
 			if (markbad) {
-				loff_t bad_addr = mtdoffset & (~(meminfo.erasesize / blockalign) + 1);
+				loff_t bad_addr = mtdoffset & (~meminfo.erasesize + 1);
 				fprintf(stderr, "Marking block at %08lx bad\n", (long)bad_addr);
 				if (ioctl(fd, MEMSETBADBLOCK, &bad_addr)) {
 					perror("MEMSETBADBLOCK");
 					goto closeall;
 				}
 			}
-			mtdoffset = blockstart + meminfo.erasesize;
+			mtdoffset = blockstart + ebsize_aligned;
 
 			continue;
 		}
Signature · 1 line
1.7.0.4
#3 lore raw

PATCH 4/10 mtd-utils: nandwrite: switch "oobsize" for "writesize"

In reply to Brian Norris

The text of a printf() states that we're printing OOB area, but the corresponding argument passes writesize. That probably wasn't the intent.

---
 nandwrite.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
nandwrite.c +1 -1
index 66a9ef7..a3ac968 100644
@@ -433,7 +433,7 @@ int main(int argc, char * const argv[])
 	// Check, if length fits into device
 	if (((imglen / pagelen) * meminfo.writesize) > (meminfo.size - mtdoffset)) {
 		fprintf(stderr, "Image %d bytes, NAND page %d bytes, OOB area %u bytes, device size %u bytes\n",
-				imglen, pagelen, meminfo.writesize, meminfo.size);
+				imglen, pagelen, meminfo.oobsize, meminfo.size);
 		perror("Input file does not fit into device");
 		goto closeall;
 	}
Signature · 1 line
1.7.0.4
#4 lore raw

PATCH 5/10 mtd-utils: nandwrite: Use libmtd to get correct mtd parameters

In reply to Brian Norris

Begin utilizing libmtd for MTD operations: use mtd_get_dev_info() to return a more detailed set of information about our MTD. Most importantly, libmtd will yield a 64-bit "size" parameter. This is necessary to properly detect devices larer than 4GB.

printf() arguments needed reformatted for the new mtd_dev_info data types. In addition, the printf() was restructured to keep lines shorter.

---
 nandwrite.c |   60 +++++++++++++++++++++++++++++++---------------------------
 1 files changed, 32 insertions(+), 28 deletions(-)
nandwrite.c +32 -28
index a3ac968..d05d257 100644
@@ -42,6 +42,7 @@
 #include <asm/types.h>
 #include "mtd/mtd-user.h"
 #include "common.h"
+#include <libmtd.h>
 
 // oob layouts to pass into the kernel as default
 static struct nand_oobinfo none_oobinfo = {
@@ -259,7 +260,7 @@ int main(int argc, char * const argv[])
 	int imglen = 0, pagelen;
 	bool baderaseblock = false;
 	int blockstart = -1;
-	struct mtd_info_user meminfo;
+	struct mtd_dev_info mtd;
 	struct mtd_oob_buf oob;
 	loff_t offs;
 	int ret;
@@ -275,6 +276,7 @@ int main(int argc, char * const argv[])
 	// points to the OOB for the current page in filebuf
 	unsigned char *oobreadbuf = NULL;
 	unsigned char *oobbuf = NULL;
+	libmtd_t mtd_desc;
 	int ebsize_aligned;
 
 	process_options(argc, argv);
@@ -290,24 +292,24 @@ int main(int argc, char * const argv[])
 		exit(EXIT_FAILURE);
 	}
 
+	mtd_desc = libmtd_open();
+	if (!mtd_desc)
+		return errmsg("can't initialize libmtd");
 	/* Fill in MTD device capability structure */
-	if (ioctl(fd, MEMGETINFO, &meminfo) != 0) {
-		perror("MEMGETINFO");
-		close(fd);
-		exit(EXIT_FAILURE);
-	}
+	if (mtd_get_dev_info(mtd_desc, mtd_device, &mtd) < 0)
+		return errmsg("mtd_get_dev_info failed");
 
 	/*
 	 * Pretend erasesize is specified number of blocks - to match jffs2
 	 *   (virtual) block size
 	 * Use this value throughout unless otherwise necessary
 	 */
-	ebsize_aligned = meminfo.erasesize * blockalign;
+	ebsize_aligned = mtd.eb_size * blockalign;
 
-	if (mtdoffset & (meminfo.writesize - 1)) {
+	if (mtdoffset & (mtd.min_io_size - 1)) {
 		fprintf(stderr, "The start address is not page-aligned !\n"
 				"The pagesize of this NAND Flash is 0x%x.\n",
-				meminfo.writesize);
+				mtd.min_io_size);
 		close(fd);
 		exit(EXIT_FAILURE);
 	}
@@ -373,7 +375,7 @@ int main(int argc, char * const argv[])
 			fprintf(stderr, "Use -f option to enforce legacy placement on autoplacement enabled mtd device\n");
 			goto restoreoob;
 		}
-		if (meminfo.oobsize == 8) {
+		if (mtd.oob_size == 8) {
 			if (forceyaffs) {
 				fprintf(stderr, "YAFSS cannot operate on 256 Byte page size");
 				goto restoreoob;
@@ -388,7 +390,7 @@ int main(int argc, char * const argv[])
 		}
 	}
 
-	oob.length = meminfo.oobsize;
+	oob.length = mtd.oob_size;
 	oob.ptr = noecc ? oobreadbuf : oobbuf;
 
 	/* Determine if we are reading from standard input or from a file. */
@@ -403,7 +405,7 @@ int main(int argc, char * const argv[])
 		goto restoreoob;
 	}
 
-	pagelen = meminfo.writesize + ((writeoob) ? meminfo.oobsize : 0);
+	pagelen = mtd.min_io_size + ((writeoob) ? mtd.oob_size : 0);
 
 	/*
 	 * For the standard input case, the input size is merely an
@@ -431,20 +433,21 @@ int main(int argc, char * const argv[])
 	}
 
 	// Check, if length fits into device
-	if (((imglen / pagelen) * meminfo.writesize) > (meminfo.size - mtdoffset)) {
-		fprintf(stderr, "Image %d bytes, NAND page %d bytes, OOB area %u bytes, device size %u bytes\n",
-				imglen, pagelen, meminfo.oobsize, meminfo.size);
+	if (((imglen / pagelen) * mtd.min_io_size) > (mtd.size - mtdoffset)) {
+		fprintf(stderr, "Image %d bytes, NAND page %d bytes, OOB area %d"
+				" bytes, device size %lld bytes\n",
+				imglen, pagelen, mtd.oob_size, mtd.size);
 		perror("Input file does not fit into device");
 		goto closeall;
 	}
 
 	// Allocate a buffer big enough to contain all the data (OOB included) for one eraseblock
-	filebuf_max = pagelen * ebsize_aligned / meminfo.writesize;
+	filebuf_max = pagelen * ebsize_aligned / mtd.min_io_size;
 	filebuf = xmalloc(filebuf_max);
 	erase_buffer(filebuf, filebuf_max);
 
-	oobbuf = xmalloc(meminfo.oobsize);
-	erase_buffer(oobbuf, meminfo.oobsize);
+	oobbuf = xmalloc(mtd.oob_size);
+	erase_buffer(oobbuf, mtd.oob_size);
 
 	/*
 	 * Get data from input and write to the device while there is
@@ -454,7 +457,7 @@ int main(int argc, char * const argv[])
 	 * length or zero.
 	 */
 	while (((imglen > 0) || (writebuf < (filebuf + filebuf_len)))
-		&& (mtdoffset < meminfo.size)) {
+		&& (mtdoffset < mtd.size)) {
 		/*
 		 * New eraseblock, check for bad block(s)
 		 * Stay in the loop to be sure that, if mtdoffset changes because
@@ -504,8 +507,8 @@ int main(int argc, char * const argv[])
 		}
 
 		// Read more data from the input if there isn't enough in the buffer
-		if ((writebuf + meminfo.writesize) > (filebuf + filebuf_len)) {
-			int readlen = meminfo.writesize;
+		if ((writebuf + mtd.min_io_size) > (filebuf + filebuf_len)) {
+			int readlen = mtd.min_io_size;
 
 			int alreadyread = (filebuf + filebuf_len) - writebuf;
 			int tinycnt = alreadyread;
@@ -556,11 +559,11 @@ int main(int argc, char * const argv[])
 		}
 
 		if (writeoob) {
-			oobreadbuf = writebuf + meminfo.writesize;
+			oobreadbuf = writebuf + mtd.min_io_size;
 
 			// Read more data for the OOB from the input if there isn't enough in the buffer
-			if ((oobreadbuf + meminfo.oobsize) > (filebuf + filebuf_len)) {
-				int readlen = meminfo.oobsize;
+			if ((oobreadbuf + mtd.oob_size) > (filebuf + filebuf_len)) {
+				int readlen = mtd.oob_size;
 				int alreadyread = (filebuf + filebuf_len) - oobreadbuf;
 				int tinycnt = alreadyread;
 
@@ -619,7 +622,7 @@ int main(int argc, char * const argv[])
 				} else {
 					/* Set at least the ecc byte positions to 0xff */
 					start = old_oobinfo.eccbytes;
-					len = meminfo.oobsize - start;
+					len = mtd.oob_size - start;
 					memcpy(oobbuf + start,
 							oobreadbuf + start,
 							len);
@@ -634,7 +637,7 @@ int main(int argc, char * const argv[])
 		}
 
 		/* Write out the Page data */
-		if (pwrite(fd, writebuf, meminfo.writesize, mtdoffset) != meminfo.writesize) {
+		if (pwrite(fd, writebuf, mtd.min_io_size, mtdoffset) != mtd.min_io_size) {
 			erase_info_t erase;
 
 			if (errno != EIO) {
@@ -658,7 +661,7 @@ int main(int argc, char * const argv[])
 			}
 
 			if (markbad) {
-				loff_t bad_addr = mtdoffset & (~meminfo.erasesize + 1);
+				loff_t bad_addr = mtdoffset & (~mtd.eb_size + 1);
 				fprintf(stderr, "Marking block at %08lx bad\n", (long)bad_addr);
 				if (ioctl(fd, MEMSETBADBLOCK, &bad_addr)) {
 					perror("MEMSETBADBLOCK");
@@ -669,7 +672,7 @@ int main(int argc, char * const argv[])
 
 			continue;
 		}
-		mtdoffset += meminfo.writesize;
+		mtdoffset += mtd.min_io_size;
 		writebuf += pagelen;
 	}
 
@@ -679,6 +682,7 @@ closeall:
 	close(ifd);
 
 restoreoob:
+	libmtd_close(mtd_desc);
 	free(filebuf);
 	free(oobbuf);
 
Signature · 1 line
1.7.0.4
#5 lore raw

PATCH 6/10 mtd-utils: nandwrite: Use 64-bit offset

In reply to Brian Norris

To support large NAND devices, we need 64-bit data types for write offsets. This patch makes data type changes along with their corresponding printf() formats and the input conversion (i.e., use "strtoll()" instead of "strol()").

---
 nandwrite.c |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
nandwrite.c +6 -6
index d05d257..b362c29 100644
@@ -109,7 +109,7 @@ static void display_version(void)
 
 static const char	*standard_input = "-";
 static const char	*mtd_device, *img;
-static int		mtdoffset = 0;
+static long long	mtdoffset = 0;
 static bool		quiet = false;
 static bool		writeoob = false;
 static bool		rawoob = false;
@@ -201,7 +201,7 @@ static void process_options(int argc, char * const argv[])
 				writeoob = true;
 				break;
 			case 's':
-				mtdoffset = strtol(optarg, NULL, 0);
+				mtdoffset = strtoll(optarg, NULL, 0);
 				break;
 			case 'b':
 				blockalign = atoi(optarg);
@@ -213,7 +213,7 @@ static void process_options(int argc, char * const argv[])
 	}
 
 	if (mtdoffset < 0) {
-		fprintf(stderr, "Can't specify a negative device offset `%d'\n",
+		fprintf(stderr, "Can't specify a negative device offset `%lld'\n",
 				mtdoffset);
 		exit(EXIT_FAILURE);
 	}
@@ -259,7 +259,7 @@ int main(int argc, char * const argv[])
 	int ifd = -1;
 	int imglen = 0, pagelen;
 	bool baderaseblock = false;
-	int blockstart = -1;
+	long long blockstart = -1;
 	struct mtd_dev_info mtd;
 	struct mtd_oob_buf oob;
 	loff_t offs;
@@ -480,7 +480,7 @@ int main(int argc, char * const argv[])
 
 			baderaseblock = false;
 			if (!quiet)
-				fprintf(stdout, "Writing data to block %d at offset 0x%x\n",
+				fprintf(stdout, "Writing data to block %lld at offset 0x%llx\n",
 						 blockstart / ebsize_aligned, blockstart);
 
 			/* Check all the blocks in an erase block for bad blocks */
@@ -494,7 +494,7 @@ int main(int argc, char * const argv[])
 					baderaseblock = true;
 					if (!quiet)
 						fprintf(stderr, "Bad block at %x, %u block(s) "
-								"from %x will be skipped\n",
+								"from %llx will be skipped\n",
 								(int) offs, blockalign, blockstart);
 				}
 
Signature · 1 line
1.7.0.4
#6 lore raw

Re: [PATCH 06/10] mtd-utils: nandwrite: Use 64-bit offset

nandwrite.c static void process_options(int argc, char * const argv[]) · line 204

Quoted 2 lines · Brian Norris
nandwrite.c +1 -1
-				mtdoffset = strtol(optarg, NULL, 0);
+				mtdoffset = strtoll(optarg, NULL, 0);

Ideally this should check for strtoll failures, but this just a side note. Not sure if this is the best way, but I'd do it like this:

char *endp;

mtdoffset = strtoul(optarg, &endp, 0);
if (*endp != '\0' || endp == optarg)

	return errmsg("bad volume alignment: \"%s\"", optarg);

or something like this. The check for negative mtdoffset could be there as well.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#7 lore raw

Re: [PATCH 06/10] mtd-utils: nandwrite: Use 64-bit offset

Quoted 7 lines · Artem Bityutskiy
Quoted 2 lines · Brian Norris

-                             mtdoffset = strtol(optarg, NULL, 0);
+                             mtdoffset = strtoll(optarg, NULL, 0);

Ideally this should check for strtoll failures, but this just a side note. Not sure if this is the best way, but I'd do it like this:

Quoted 8 lines

char *endp;

mtdoffset = strtoul(optarg, &endp, 0);
if (*endp != '\0' || endp == optarg)
       return errmsg("bad volume alignment: \"%s\"", optarg);

or something like this. The check for negative mtdoffset could be there as well.

or use simple_strtoll() from common.h as that includes error handling ... -mike

#8 lore raw

Re: [PATCH 06/10] mtd-utils: nandwrite: Use 64-bit offset

Quoted 10 lines · Mike Frysinger
Quoted 9 lines · Artem Bityutskiy

nandwrite.c static void process_options(int argc, char * const argv[]) · line 204

Quoted 2 lines · Brian Norris
nandwrite.c +1 -1
-                             mtdoffset = strtol(optarg, NULL, 0);
+                             mtdoffset = strtoll(optarg, NULL, 0);

Ideally this should check for strtoll failures, but this just a side note. Not sure if this is the best way, but I'd do it like this:

char *endp;

Quoted 8 lines
Quoted 6 lines

mtdoffset = strtoul(optarg, &endp, 0);
if (*endp != '\0' || endp == optarg)

       return errmsg("bad volume alignment: \"%s\"", optarg);

or something like this. The check for negative mtdoffset could be there as well.

or use simple_strtoll() from common.h as that includes error handling ...

Oh, even better.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Битюцкий Артём)
#9 lore raw

PATCH 7/10 mtd-utils: nandwrite: avoid NULL buffer pointers

In reply to Brian Norris

Commit 07005d915d6a79dbdee14b0c4360df5058c3a98b made changes to the buffer allocation in nandwrite and did not handle all affected code areas properly. In particular, we were assigning:

	oob.ptr = noecc ? oobreadbuf : oobbuf;

However, since oobreadbuf and oobbuf are declared dynamically, they are NULL at this point. If they aren't properly assigned later, we unwittingly are passing a NULL pointer as oob buffer.

This assignment line is best moved after the buffer allocations and pointer assignment.

Effects of this problem can be seen when writing oob data with the "-o" flag and without the "-n" flag:

	$ ./nandwrite -o /dev/mtd0 img.bin
	Writing data to block 0 at offset 0x0
	ioctl(MEMWRITEOOB): Bad address
	Data was only partially written due to error
	: Bad address
---
 nandwrite.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
nandwrite.c +2 -4
index b362c29..8ec5afe 100644
@@ -391,7 +391,6 @@ int main(int argc, char * const argv[])
 	}
 
 	oob.length = mtd.oob_size;
-	oob.ptr = noecc ? oobreadbuf : oobbuf;
 
 	/* Determine if we are reading from standard input or from a file. */
 	if (strcmp(img, standard_input) == 0) {
@@ -594,9 +593,7 @@ int main(int argc, char * const argv[])
 				}
 			}
 
-			if (noecc) {
-				oob.ptr = oobreadbuf;
-			} else {
+			if (!noecc) {
 				int i, start, len;
 				int tags_pos = 0;
 				/*
@@ -630,6 +627,7 @@ int main(int argc, char * const argv[])
 			}
 			/* Write OOB data first, as ecc will be placed in there */
 			oob.start = mtdoffset;
+			oob.ptr = noecc ? oobreadbuf : oobbuf;
 			if (ioctl(fd, MEMWRITEOOB, &oob) != 0) {
 				perror("ioctl(MEMWRITEOOB)");
 				goto closeall;
Signature · 1 line
1.7.0.4
#10 lore raw

PATCH 8/10 mtd-utils: nandwrite: prevent 32-bit overflow

In reply to Brian Norris

For large block- and page-sizes, the multiplication of ebsize_aligned and pagelen can overflow a 32-bit integer. This overflow can be prevented by a simple change in order of operations (i.e., do division first).

Since ebsize_aligned is always a multiple of mtd.min_io_size, this produces no change in results.

---
 nandwrite.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
nandwrite.c +1 -1
index 8ec5afe..364acdf 100644
@@ -441,7 +441,7 @@ int main(int argc, char * const argv[])
 	}
 
 	// Allocate a buffer big enough to contain all the data (OOB included) for one eraseblock
-	filebuf_max = pagelen * ebsize_aligned / mtd.min_io_size;
+	filebuf_max = ebsize_aligned / mtd.min_io_size * pagelen;
 	filebuf = xmalloc(filebuf_max);
 	erase_buffer(filebuf, filebuf_max);
 
Signature · 1 line
1.7.0.4
#11 lore raw

Re: [PATCH 08/10] mtd-utils: nandwrite: prevent 32-bit overflow

Quoted 7 lines · Brian Norris

For large block- and page-sizes, the multiplication of ebsize_aligned and pagelen can overflow a 32-bit integer.  This overflow can be prevented by a simple change in order of operations (i.e., do division first).

-       filebuf_max = pagelen * ebsize_aligned / mtd.min_io_size;
+       filebuf_max = ebsize_aligned / mtd.min_io_size * pagelen;

i'd insert a comment noting that order of operations matters to avoid 32bit overflow so someone doesnt switch it back by accident in the future
-mike

#12 lore raw

PATCH v2 8/10 mtd-utils: nandwrite: prevent 32-bit overflow

For large block- and page-sizes, the multiplication of ebsize_aligned and pagelen can overflow a 32-bit integer. This overflow can be prevented by a simple change in order of operations (i.e., do division first).

Since ebsize_aligned is always a multiple of mtd.min_io_size, this produces no change in results.

---
 nandwrite.c |    9 +++++++--
 1 files changed, 7 insertions(+), 2 deletions(-)
nandwrite.c +7 -2
index 8ec5afe..aea7572 100644
@@ -440,8 +440,13 @@ int main(int argc, char * const argv[])
 		goto closeall;
 	}
 
-	// Allocate a buffer big enough to contain all the data (OOB included) for one eraseblock
-	filebuf_max = pagelen * ebsize_aligned / mtd.min_io_size;
+	/*
+	 * Allocate a buffer big enough to contain all the data (OOB included)
+	 * for one eraseblock. The order of operations here matters; if ebsize
+	 * and pagelen are large enough, then "ebsize_aligned * pagelen" could
+	 * overflow a 32-bit data type.
+	 */
+	filebuf_max = ebsize_aligned / mtd.min_io_size * pagelen;
 	filebuf = xmalloc(filebuf_max);
 	erase_buffer(filebuf, filebuf_max);
 
Signature · 1 line
1.7.0.4
#13 lore raw

Re: [PATCH 08/10] mtd-utils: nandwrite: prevent 32-bit overflow

In reply to Brian Norris
Quoted 9 lines · Brian Norris

For large block- and page-sizes, the multiplication of ebsize_aligned and pagelen can overflow a 32-bit integer. This overflow can be prevented by a simple change in order of operations (i.e., do division first).

Since ebsize_aligned is always a multiple of mtd.min_io_size, this produces no change in results.

Brian, increasingly have less and less time for mtd, but I remember about your patches - I'll pick them as soon as I have some time. You can meanwhile address Mikes comments.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#14 lore raw

PATCH 9/10 mtd-utils: nanddump: type consistency

In reply to Brian Norris

Change type off "offs" for type consistency of 64-bit data types. Also change an accompanying printf().

---
 nandwrite.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
nandwrite.c +3 -3
index 364acdf..a28f7fb 100644
@@ -262,7 +262,7 @@ int main(int argc, char * const argv[])
 	long long blockstart = -1;
 	struct mtd_dev_info mtd;
 	struct mtd_oob_buf oob;
-	loff_t offs;
+	long long offs;
 	int ret;
 	int oobinfochanged = 0;
 	struct nand_oobinfo old_oobinfo;
@@ -492,9 +492,9 @@ int main(int argc, char * const argv[])
 				} else if (ret == 1) {
 					baderaseblock = true;
 					if (!quiet)
-						fprintf(stderr, "Bad block at %x, %u block(s) "
+						fprintf(stderr, "Bad block at %llx, %u block(s) "
 								"from %llx will be skipped\n",
-								(int) offs, blockalign, blockstart);
+								offs, blockalign, blockstart);
 				}
 
 				if (baderaseblock) {
Signature · 1 line
1.7.0.4
#15 lore raw

Re: [PATCH 09/10] mtd-utils: nanddump: type consistency

Quoted 1 line · Brian Norris

 nandwrite.c |    6 +++---

the subject says "nanddump" ...

nandwrite.c int main(int argc, char * const argv[]) · line 268

Quoted 11 lines
nandwrite.c +1 -1
@@ -262,7 +262,7 @@ int main(int argc, char * const argv[])
        long long blockstart = -1;
        struct mtd_dev_info mtd;
        struct mtd_oob_buf oob;
-       loff_t offs;
+       long long offs;
        int ret;
        int oobinfochanged = 0;
        struct nand_oobinfo old_oobinfo;

i dont think this is quite right. "offs" is given to an ioctl which takes __kernel_loff_t ... we shouldnt assume "long long" and "__kernel_loff_t" are always the same.
-mike

#16 lore raw

Re: [PATCH 09/10] mtd-utils: nanddump: type consistency

Quoted 4 lines · Mike Frysinger
Quoted 1 line · Brian Norris
 nandwrite.c |    6 +++---

the subject says "nanddump" ...

Oops, I'll change that if I send a "v2".

Quoted 7 lines
Quoted 7 lines
nandwrite.c +0 -1
@@ -262,7 +262,7 @@ int main(int argc, char * const argv[])
       long long blockstart = -1;
       struct mtd_dev_info mtd;
       struct mtd_oob_buf oob;
-       loff_t offs;
Quoted 8 lines

nandwrite.c int main(int argc, char * const argv[]) · line 268

Quoted 4 lines
nandwrite.c +1 -0
+       long long offs;
       int ret;
       int oobinfochanged = 0;
       struct nand_oobinfo old_oobinfo;

i dont think this is quite right. "offs" is given to an ioctl which takes __kernel_loff_t ... we shouldnt assume "long long" and "__kernel_loff_t" are always the same.

This problem is sort of a "which comes first" question for me: I put this patch before patch 10, where in patch 10, the ioctl in question is replaced with an entirely different call to "mtd_is_bad".

In the new usage (patch 10), "mtd_is_bad" needs the eraseblock number as an int, so there's some arithmetic going on ("offs" divided by ebsize) that passes an int to "mtd_is_bad".

I was just trying to unify the usage of different 64-bit data types, and after patch 10, there's no need to be using the "loff_t" type. However, if it makes more sense, I can rearrange patch 9 and 10 such that the type change comes after the ioctl is no longer called directly.

Thanks,
Brian

#17 lore raw

Re: [PATCH 09/10] mtd-utils: nanddump: type consistency

Quoted 22 lines · Brian Norris
Quoted 16 lines · Mike Frysinger

nandwrite.c int main(int argc, char * const argv[]) · line 268

Quoted 11 lines · Brian Norris
nandwrite.c +1 -1
@@ -262,7 +262,7 @@ int main(int argc, char * const argv[])
        long long blockstart = -1;
        struct mtd_dev_info mtd;
        struct mtd_oob_buf oob;
-       loff_t offs;
+       long long offs;
        int ret;
        int oobinfochanged = 0;
        struct nand_oobinfo old_oobinfo;

i dont think this is quite right.  "offs" is given to an ioctl which takes __kernel_loff_t ... we shouldnt assume "long long" and "__kernel_loff_t" are always the same.

This problem is sort of a "which comes first" question for me: I put this patch before patch 10, where in patch 10, the ioctl in question is replaced with an entirely different call to "mtd_is_bad".

Quoted 8 lines

In the new usage (patch 10), "mtd_is_bad" needs the eraseblock number as an int, so there's some arithmetic going on ("offs" divided by ebsize) that passes an int to "mtd_is_bad".

I was just trying to unify the usage of different 64-bit data types, and after patch 10, there's no need to be using the "loff_t" type. However, if it makes more sense, I can rearrange patch 9 and 10 such that the type change comes after the ioctl is no longer called directly.

i'd leave the type unchanged when using the ioctl() and change it only when moving to a different API
-mike

#18 lore raw

PATCH v2 9/10 mtd-utils: nandwrite: full 64-bit support w/ libmtd

Several ioctls are replaced with libmtd calls which should give us 64-bit support for large devices. libmtd mostly provides drop-in replacements for the functionality we need. However, when we require erasure of a badly-written block, mtd_erase() only erases a single block, whereas MEMERASE could erase a larger region. In nandwrite, we may have a "virtual blocksize" of more than one (when blockalign > 1). Thus, I added a loop for this case.

The mtd_oob_buf struct is no longer needed, nor is "erase_info_t".

Error messages for the new libmtd calls reflect the style found in flash_erase.

Tested with nandsim and with NAND chips up to 4GB in size (I don't have a device that truly requires 64-bit addressing yet).

---
 nandwrite.c |   49 +++++++++++++++++++++++--------------------------
 1 files changed, 23 insertions(+), 26 deletions(-)
nandwrite.c +23 -26
index aea7572..00e7c28 100644
@@ -261,7 +261,6 @@ int main(int argc, char * const argv[])
 	bool baderaseblock = false;
 	long long blockstart = -1;
 	struct mtd_dev_info mtd;
-	struct mtd_oob_buf oob;
 	loff_t offs;
 	int ret;
 	int oobinfochanged = 0;
@@ -390,8 +389,6 @@ int main(int argc, char * const argv[])
 		}
 	}
 
-	oob.length = mtd.oob_size;
-
 	/* Determine if we are reading from standard input or from a file. */
 	if (strcmp(img, standard_input) == 0) {
 		ifd = STDIN_FILENO;
@@ -491,8 +488,8 @@ int main(int argc, char * const argv[])
 			if (noskipbad)
 				continue;
 			do {
-				if ((ret = ioctl(fd, MEMGETBADBLOCK, &offs)) < 0) {
-					perror("ioctl(MEMGETBADBLOCK)");
+				if ((ret = mtd_is_bad(&mtd, fd, offs / ebsize_aligned)) < 0) {
+					sys_errmsg("%s: MTD get bad block failed", mtd_device);
 					goto closeall;
 				} else if (ret == 1) {
 					baderaseblock = true;
@@ -631,43 +628,43 @@ int main(int argc, char * const argv[])
 				}
 			}
 			/* Write OOB data first, as ecc will be placed in there */
-			oob.start = mtdoffset;
-			oob.ptr = noecc ? oobreadbuf : oobbuf;
-			if (ioctl(fd, MEMWRITEOOB, &oob) != 0) {
-				perror("ioctl(MEMWRITEOOB)");
+			if (mtd_write_oob(mtd_desc, &mtd, fd, mtdoffset,
+						mtd.oob_size,
+						noecc ? oobreadbuf : oobbuf)) {
+				sys_errmsg("%s: MTD writeoob failure", mtd_device);
 				goto closeall;
 			}
 		}
 
 		/* Write out the Page data */
-		if (pwrite(fd, writebuf, mtd.min_io_size, mtdoffset) != mtd.min_io_size) {
-			erase_info_t erase;
-
+		if (mtd_write(&mtd, fd, mtdoffset / mtd.eb_size, mtdoffset % mtd.eb_size,
+					writebuf, mtd.min_io_size)) {
+			int i;
 			if (errno != EIO) {
-				perror("pwrite");
+				sys_errmsg("%s: MTD write failure", mtd_device);
 				goto closeall;
 			}
 
 			/* Must rewind to blockstart if we can */
 			writebuf = filebuf;
 
-			erase.start = blockstart;
-			erase.length = ebsize_aligned;
-			fprintf(stderr, "Erasing failed write from %08lx-%08lx\n",
-				(long)erase.start, (long)erase.start+erase.length-1);
-			if (ioctl(fd, MEMERASE, &erase) != 0) {
-				int errno_tmp = errno;
-				perror("MEMERASE");
-				if (errno_tmp != EIO) {
-					goto closeall;
+			fprintf(stderr, "Erasing failed write from %#08llx to %#08llx\n",
+				blockstart, blockstart + ebsize_aligned - 1);
+			for (i = blockstart; i < blockstart + ebsize_aligned; i += mtd.eb_size) {
+				if (mtd_erase(mtd_desc, &mtd, fd, mtd.eb_size)) {
+					int errno_tmp = errno;
+					sys_errmsg("%s: MTD Erase failure", mtd_device);
+					if (errno_tmp != EIO) {
+						goto closeall;
+					}
 				}
 			}
 
 			if (markbad) {
-				loff_t bad_addr = mtdoffset & (~mtd.eb_size + 1);
-				fprintf(stderr, "Marking block at %08lx bad\n", (long)bad_addr);
-				if (ioctl(fd, MEMSETBADBLOCK, &bad_addr)) {
-					perror("MEMSETBADBLOCK");
+				fprintf(stderr, "Marking block at %08llx bad\n",
+						mtdoffset & (~mtd.eb_size + 1));
+				if (mtd_mark_bad(&mtd, fd, mtdoffset / mtd.eb_size)) {
+					sys_errmsg("%s: MTD Mark bad block failure", mtd_device);
 					goto closeall;
 				}
 			}
Signature · 1 line
1.7.0.4
#19 lore raw

Re: [PATCH v2 09/10] mtd-utils: nandwrite: full 64-bit support w/ libmtd

Quoted 7 lines · Brian Norris

Several ioctls are replaced with libmtd calls which should give us 64-bit support for large devices. libmtd mostly provides drop-in replacements for the functionality we need. However, when we require erasure of a badly-written block, mtd_erase() only erases a single block, whereas MEMERASE could erase a larger region. In nandwrite, we may have a "virtual blocksize" of more than one (when blockalign > 1). Thus, I added a loop for this case.

You could as well just introduce another libmtd helper - would be cleaner, IMO.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#20 lore raw

Re: [PATCH v2 09/10] mtd-utils: nandwrite: full 64-bit support w/ libmtd

Quoted 2 lines · Artem Bityutskiy

You could as well just introduce another libmtd helper - would be cleaner, IMO.

OK, I'll look at this and the "simple_strtoll()" suggestion sometime. I should probably examine "common.h" more before patching mtd-utils next time!

And thanks for the push.

Brian

#21 lore raw

Re: [PATCH v2 09/10] mtd-utils: nandwrite: full 64-bit support w/ libmtd

Quoted 6 lines · Brian Norris
Quoted 2 lines · Artem Bityutskiy

You could as well just introduce another libmtd helper - would be cleaner, IMO.

OK, I'll look at this and the "simple_strtoll()" suggestion sometime. I should probably examine "common.h" more before patching mtd-utils next time!

it's an evolving helper, so if you come across generally useful funcs that are in an app or two, feel free to propose moving them to the header for all to benefit
-mike

#22 lore raw

PATCH v2 10/10 mtd-utils: nandwrite: type consistency

In reply to Mike Frysinger

Change type off "offs" for type consistency of 64-bit data types. The "loff_t" type is no longer needed for the MEMGETBADBLOCK ioctl since it isn't called dirently anymore - this is handled by mtd_is_bad().

Also change an accompanying printf().

---
 nandwrite.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
nandwrite.c +3 -3
index 00e7c28..bbe69b0 100644
@@ -261,7 +261,7 @@ int main(int argc, char * const argv[])
 	bool baderaseblock = false;
 	long long blockstart = -1;
 	struct mtd_dev_info mtd;
-	loff_t offs;
+	long long offs;
 	int ret;
 	int oobinfochanged = 0;
 	struct nand_oobinfo old_oobinfo;
@@ -494,9 +494,9 @@ int main(int argc, char * const argv[])
 				} else if (ret == 1) {
 					baderaseblock = true;
 					if (!quiet)
-						fprintf(stderr, "Bad block at %x, %u block(s) "
+						fprintf(stderr, "Bad block at %llx, %u block(s) "
 								"from %llx will be skipped\n",
-								(int) offs, blockalign, blockstart);
+								offs, blockalign, blockstart);
 				}
 
 				if (baderaseblock) {
Signature · 1 line
1.7.0.4
#23 lore raw

PATCH 10/10 mtd-utils: nandwrite: full 64-bit support w/ libmtd

In reply to Brian Norris

Several ioctls are replaced with libmtd calls which should give us 64-bit support for large devices. libmtd mostly provides drop-in replacements for the functionality we need. However, when we require erasure of a badly-written block, mtd_erase() only erases a single block, whereas MEMERASE could erase a larger region. In nandwrite, we may have a "virtual blocksize" of more than one (when blockalign > 1). Thus, I added a loop for this case.

The mtd_oob_buf struct is no longer needed, nor is "erase_info_t".

Error messages for the new libmtd calls reflect the style found in flash_erase.

Tested with nandsim and with NAND chips up to 4GB in size (I don't have a device that truly requires 64-bit addressing yet).

---
 nandwrite.c |   49 +++++++++++++++++++++++--------------------------
 1 files changed, 23 insertions(+), 26 deletions(-)
nandwrite.c +23 -26
index a28f7fb..37d2204 100644
@@ -261,7 +261,6 @@ int main(int argc, char * const argv[])
 	bool baderaseblock = false;
 	long long blockstart = -1;
 	struct mtd_dev_info mtd;
-	struct mtd_oob_buf oob;
 	long long offs;
 	int ret;
 	int oobinfochanged = 0;
@@ -390,8 +389,6 @@ int main(int argc, char * const argv[])
 		}
 	}
 
-	oob.length = mtd.oob_size;
-
 	/* Determine if we are reading from standard input or from a file. */
 	if (strcmp(img, standard_input) == 0) {
 		ifd = STDIN_FILENO;
@@ -486,8 +483,8 @@ int main(int argc, char * const argv[])
 			if (noskipbad)
 				continue;
 			do {
-				if ((ret = ioctl(fd, MEMGETBADBLOCK, &offs)) < 0) {
-					perror("ioctl(MEMGETBADBLOCK)");
+				if ((ret = mtd_is_bad(&mtd, fd, offs / ebsize_aligned)) < 0) {
+					sys_errmsg("%s: MTD get bad block failed", mtd_device);
 					goto closeall;
 				} else if (ret == 1) {
 					baderaseblock = true;
@@ -626,43 +623,43 @@ int main(int argc, char * const argv[])
 				}
 			}
 			/* Write OOB data first, as ecc will be placed in there */
-			oob.start = mtdoffset;
-			oob.ptr = noecc ? oobreadbuf : oobbuf;
-			if (ioctl(fd, MEMWRITEOOB, &oob) != 0) {
-				perror("ioctl(MEMWRITEOOB)");
+			if (mtd_write_oob(mtd_desc, &mtd, fd, mtdoffset,
+						mtd.oob_size,
+						noecc ? oobreadbuf : oobbuf)) {
+				sys_errmsg("%s: MTD writeoob failure", mtd_device);
 				goto closeall;
 			}
 		}
 
 		/* Write out the Page data */
-		if (pwrite(fd, writebuf, mtd.min_io_size, mtdoffset) != mtd.min_io_size) {
-			erase_info_t erase;
-
+		if (mtd_write(&mtd, fd, mtdoffset / mtd.eb_size, mtdoffset % mtd.eb_size,
+					writebuf, mtd.min_io_size)) {
+			int i;
 			if (errno != EIO) {
-				perror("pwrite");
+				sys_errmsg("%s: MTD write failure", mtd_device);
 				goto closeall;
 			}
 
 			/* Must rewind to blockstart if we can */
 			writebuf = filebuf;
 
-			erase.start = blockstart;
-			erase.length = ebsize_aligned;
-			fprintf(stderr, "Erasing failed write from %08lx-%08lx\n",
-				(long)erase.start, (long)erase.start+erase.length-1);
-			if (ioctl(fd, MEMERASE, &erase) != 0) {
-				int errno_tmp = errno;
-				perror("MEMERASE");
-				if (errno_tmp != EIO) {
-					goto closeall;
+			fprintf(stderr, "Erasing failed write from %#08llx to %#08llx\n",
+				blockstart, blockstart + ebsize_aligned - 1);
+			for (i = blockstart; i < blockstart + ebsize_aligned; i += mtd.eb_size) {
+				if (mtd_erase(mtd_desc, &mtd, fd, mtd.eb_size)) {
+					int errno_tmp = errno;
+					sys_errmsg("%s: MTD Erase failure", mtd_device);
+					if (errno_tmp != EIO) {
+						goto closeall;
+					}
 				}
 			}
 
 			if (markbad) {
-				loff_t bad_addr = mtdoffset & (~mtd.eb_size + 1);
-				fprintf(stderr, "Marking block at %08lx bad\n", (long)bad_addr);
-				if (ioctl(fd, MEMSETBADBLOCK, &bad_addr)) {
-					perror("MEMSETBADBLOCK");
+				fprintf(stderr, "Marking block at %08llx bad\n",
+						mtdoffset & (~mtd.eb_size + 1));
+				if (mtd_mark_bad(&mtd, fd, mtdoffset / mtd.eb_size)) {
+					sys_errmsg("%s: MTD Mark bad block failure", mtd_device);
 					goto closeall;
 				}
 			}
Signature · 1 line
1.7.0.4
Acked-by on v1 1/10
#25 lore raw
In reply to Brian Norris
Quoted 13 lines · Brian Norris

We should allow the dump length to be 64-bit, especially since the value was read in as a "long long" by strtoll().

---
 nanddump.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
nanddump.c +0 -0
index 86a71c9..fe29596 100644

nanddump.c static bool noecc = false; // don't error correct · line 89

Quoted 8 lines
nanddump.c +1 -1
@@ -83,7 +83,7 @@ static bool			noecc = false;		// don't error correct
 static bool			noskipbad = false;	// don't skip bad blocks
 static bool			omitoob = false;	// omit oob data
 static unsigned long long	start_addr;		// start address
-static unsigned long		length;			// dump length
+static unsigned long long	length;			// dump length
 static const char		*mtddev;		// mtd device name
 static const char		*dumpfile;		// dump file name
 static bool			omitbad = false;

This patch does not apply. Which version of MTD utils do you use? This change was done long time ago by the following commit:

commit b16c1b630491a461b3ebb55d714d7bb0cd122737

Author: Grant Erickson <[email protected]>
Date:   Sun Sep 7 20:45:21 2008 +0000

nanddump: Qualifier Clean-up

Static-qualified all globals except 'main' because they have no use beyond file scope.
Constant-qualified MTD device and input positional parameter globals. Constant-qualified argv array.

Forgot to git-pull ? :-))))

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#26 lore raw
Quoted 1 line · Artem Bityutskiy

Forgot to git-pull ? :-))))

Brian, ignore this e-mail please :-) I've just pushed this patch, thanks!

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)