Skip to contentSkip to conversation

nand: adjust erase/read/write partition/chip size for bad blocks

PATCH New Harvey Chapman 2 participants 3 messages Feb 27, 2013

lore
Share & tools
lore
#0 lore raw

cover nand: adjust erase/read/write partition/chip size for bad blocks

Accidentally let a test sneak in to the patch. Corrected.

Harvey Chapman (1):
nand: adjust erase/read/write partition/chip size for bad blocks

 common/cmd_nand.c |   35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)
Signature · 1 line
1.7.10.4
#1 lore raw

PATCH 1/1 nand: adjust erase/read/write partition/chip size for bad blocks

Adjust the sizes calculated for whole partition/chip operations by removing the size of bad blocks so we don't try to erase/read/write past a partition/chip boundary.

---
 common/cmd_nand.c |   35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)
common/cmd_nand.c +35 -0
index 495610c..460b370 100644
@@ -428,6 +428,31 @@ static int raw_access(nand_info_t *nand, ulong addr, loff_t off, ulong count,
 	return ret;
 }
 
+/* Adjust a chip/partition size down for bad blocks so we don't
+ * read/write/erase past the end of a chip/partition by accident.
+ */
+static void adjust_size_for_badblocks(loff_t *size, loff_t offset, int dev)
+{
+	/* We grab the nand info object here fresh because this is usually
+	 * called after arg_off_size() which can change the value of dev.
+	 */
+	nand_info_t *nand = &nand_info[dev];
+	loff_t maxoffset = offset + *size;
+	int badblocks = 0;
+
+	/* count badblocks in NAND from offset to offset + size */
+	for (; offset < maxoffset; offset += nand->erasesize) {
+		if (nand_block_isbad(nand, offset))
+			badblocks++;
+	}
+	/* adjust size if any bad blocks found */
+	if (badblocks) {
+		*size -= badblocks * nand->erasesize;
+		printf("size adjusted to 0x%llx (%d bad blocks)\n",
+		       (unsigned long long)*size, badblocks);
+	}
+}
+
 static int do_nand(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 {
 	int i, ret = 0;
@@ -524,6 +549,7 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 		int scrub = !strncmp(cmd, "scrub", 5);
 		int spread = 0;
 		int args = 2;
+		int adjust_size = 0;
 		const char *scrub_warn =
 			"Warning: "
 			"scrub option will erase all factory set bad blocks!\n"
@@ -540,8 +566,10 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 				spread = 1;
 			} else if (!strcmp(&cmd[5], ".part")) {
 				args = 1;
+				adjust_size = 1;
 			} else if (!strcmp(&cmd[5], ".chip")) {
 				args = 0;
+				adjust_size = 1;
 			} else {
 				goto usage;
 			}
@@ -560,6 +588,10 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 		if (arg_off_size(argc - o, argv + o, &dev, &off, &size) != 0)
 			return 1;
 
+		/* size is unspecified */
+		if (adjust_size && !scrub)
+			adjust_size_for_badblocks(&size, off, dev);
+
 		nand = &nand_info[dev];
 
 		memset(&opts, 0, sizeof(opts));
@@ -644,6 +676,9 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 						&off, &size) != 0)
 				return 1;
 
+			/* size is unspecified */
+			if (argc < 5)
+				adjust_size_for_badblocks(&size, off, dev);
 			rwsize = size;
 		}
 
Signature · 1 line
1.7.10.4
#2 lore raw

PATCH 1/1 nand: adjust erase/read/write partition/chip size for bad blocks

Quoted 8 lines · Harvey Chapman

Adjust the sizes calculated for whole partition/chip operations by removing the size of bad blocks so we don't try to erase/read/write past a partition/chip boundary.

---
 common/cmd_nand.c |   35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

I don't think we want to do this adjustment for erase -- the value you pass in there is already defined as being the range of blocks on the flash, rather than the amount of data you want to be able to store in the range (except for the ".spread" variant, but that can't be used with .part or .chip).

This makes me wonder if a better approach would be a flag that tells the nand_read|write_skip_bad() function whether "size" is meant as "good data size" or "actual size"? Then we could just set that flag for whole-chip/whole-partition accesses, and let the existing bad block skipping code handle it with some minor tweaks. This would be similar to opts->spread for erase.

-Scott