[MTD] [INFTL] Fix infinite loop in INFTL_foldchain
authorDaniel Rosenthal <danielrosenthal@acm.org>
Sun, 5 Oct 2008 21:43:10 +0000 (17:43 -0400)
committerDavid Woodhouse <David.Woodhouse@intel.com>
Wed, 8 Oct 2008 10:26:38 +0000 (11:26 +0100)
When iterating over a chain in reverse (oldest block first), this
patch correctly marks the PUtable[] entry of the second to last erase
block of a chain as BLOCK_NIL, regardless of whether or not it can
format the last block successfully. Before, the second to last block
was only marked as pointing to BLOCK_NIL if INFTL_formatblock()
succeeded on the last block of the chain, which could potentially
result in an infinite loop if the block was worn out and refused to
format.

Signed-off-by: Daniel Rosenthal <danielrosenthal@acm.org>
Acked-by: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
drivers/mtd/inftlcore.c

index c4f9d3378b24c658f246d8777d47735050cfdffe..50ce13887f63b6b81d2525a5070fbfc511643b1c 100644 (file)
@@ -388,6 +388,10 @@ static u16 INFTL_foldchain(struct INFTLrecord *inftl, unsigned thisVUC, unsigned
                if (thisEUN == targetEUN)
                        break;
 
+               /* Unlink the last block from the chain. */
+               inftl->PUtable[prevEUN] = BLOCK_NIL;
+
+               /* Now try to erase it. */
                if (INFTL_formatblock(inftl, thisEUN) < 0) {
                        /*
                         * Could not erase : mark block as reserved.
@@ -396,7 +400,6 @@ static u16 INFTL_foldchain(struct INFTLrecord *inftl, unsigned thisVUC, unsigned
                } else {
                        /* Correctly erased : mark it as free */
                        inftl->PUtable[thisEUN] = BLOCK_FREE;
-                       inftl->PUtable[prevEUN] = BLOCK_NIL;
                        inftl->numfreeEUNs++;
                }
        }