Skip to contentSkip to conversation

nftl_format: prevent buffer overflow in BadUnitTable access

PATCH In review Anton Moryakov 2 participants 2 messages Jan 27 – Apr 13 Memory technology devices (MTD)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nftl_format: prevent buffer overflow in BadUnitTable access

In the INFTL formatting path, the code iterates over erase blocks from pezstart to maxzones (total blocks on device) and accesses BadUnitTable[ezone]. However, BadUnitTable is a fixed-size array of MAX_ERASE_ZONES elements (typically 1024 or 4096).

If the MTD device has more erase blocks than MAX_ERASE_ZONES, ezone can exceed the array bounds, causing a buffer overflow.

Fix by limiting the loop upper bound to MIN(maxzones, MAX_ERASE_ZONES).

---
 nand-utils/nftl_format.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
nand-utils/nftl_format.c +2 -1
index c8b8b50..0d38ddd 100644
@@ -372,7 +372,8 @@ int main(int argc, char **argv)
 		pezstart = startofs / meminfo.erasesize + 1;
 		pezend = startofs / meminfo.erasesize + ezones - 1;
 		numvunits = (ezones - 2) * PERCENTUSED / 100;
-		for (ezone = pezstart; ezone < maxzones; ezone++) {
+		unsigned long max_iter = (maxzones < MAX_ERASE_ZONES) ? maxzones : MAX_ERASE_ZONES;
+		for (ezone = pezstart; ezone < max_iter; ezone++) {
 			if (BadUnitTable[ezone] != ZONE_GOOD) {
 				if (numvunits > 1)
 					numvunits--;
Signature · 6 lines
2.39.2


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
#1 lore raw

Hi,

Quoted 8 lines · Anton Moryakov

However, BadUnitTable is a fixed-size array of MAX_ERASE_ZONES elements (typically 1024 or 4096).

If the MTD device has more erase blocks than MAX_ERASE_ZONES, ezone can exceed the array bounds, causing a buffer overflow.

Fix by limiting the loop upper bound to MIN(maxzones, MAX_ERASE_ZONES).

I get that this would silence the static analyzer, but wouldn't that introduce a different (logic) bug where the tool now silently fails to handle erase zones past a magic limit?

I think the correct thing to do here is to dynamically allocate the array based on the number of erase blocks.

Greetings,

David

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/