Skip to contentSkip to conversation

f2fs: off by one in garbage collection functions

bug report New Dan Carpenter 2 participants 3 messages Jan 18, 2013 – Jan 21, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hello Jaegeuk Kim,

The patch 7bc0900347e0: "f2fs: add garbage collection functions" from Nov 2, 2012, has an off-by-one bug.

   429  block_t start_bidx_of_node(unsigned int node_ofs)
   430  {
   431          unsigned int indirect_blks = 2 * NIDS_PER_BLOCK + 4;
   432          unsigned int bidx;
   433  
   434          if (node_ofs == 0)
   435                  return 0;
   436  
   437          if (node_ofs <= 2) {
   438                  bidx = node_ofs - 1;
   439          } else if (node_ofs <= indirect_blks) {
   440                  int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1);

If node_ofs == 3 here then (node_ofs - 4) is a very high positive number. We divide by 1019 and we get another still very high number but not so high that it is negative when cast as an int.

   441                  bidx = node_ofs - 2 - dec;

It means that bidx is much higher than intended here (4290752413).

I thought maybe there is an off by one somewhere. Perhaps the 4 should be a 3.

   442          } else {
   443                  int dec = (node_ofs - indirect_blks - 3) / (NIDS_PER_BLOCK + 1);
   444                  bidx = node_ofs - 5 - dec;
   445          }
   446          return bidx * ADDRS_PER_BLOCK + ADDRS_PER_INODE;
   447  }

regards,
dan carpenter

#1 lore raw

Hi,

2013-01-18 (금), 16:29 +0300, Dan Carpenter:

Quoted 18 lines

Hello Jaegeuk Kim,

The patch 7bc0900347e0: "f2fs: add garbage collection functions" from Nov 2, 2012, has an off-by-one bug.

   429  block_t start_bidx_of_node(unsigned int node_ofs)
   430  {
   431          unsigned int indirect_blks = 2 * NIDS_PER_BLOCK + 4;
   432          unsigned int bidx;
   433  
   434          if (node_ofs == 0)
   435                  return 0;
   436  
   437          if (node_ofs <= 2) {
   438                  bidx = node_ofs - 1;
   439          } else if (node_ofs <= indirect_blks) {
   440                  int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1);
Quoted 3 lines

If node_ofs == 3 here then (node_ofs - 4) is a very high positive number. We divide by 1019 and we get another still very high number but not so high that it is negative when cast as an int.

If the node_ofs is equal to 3, then it's a run-time bug on its caller, gc_data_segment(), not here itself.

In gc_data_segment(),
1. check_dnode(&nofs);
2. start_bidx_of_node(nofs);

Here, it means that we don't care about indirect node pages whose node offsets are 3, 3+1018+1, (3+1018+1)+1018+1, and so on. (Ref. *Index Structure* in Document/filesystems/f2fs.txt) Instead, we just check direct node pages only.

But, anyway, I'd better write comments to make it clear. Thank you for reviewing.

Signature · 2 lines
Jaegeuk Kim
Samsung
#2 lore raw

PATCH 1/1 f2fs: add comments of start_bidx_of_node

The caller of start_bidx_of_node() should give a proper node offsets which point only direct node blocks. Otherwise, it is a caller's bug. This patch adds comments to make it clear.

---
 fs/f2fs/gc.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
fs/f2fs/gc.c +5 -1
index 809cfec..c386910 100644
@@ -424,7 +424,11 @@ next_step:
 }
 
 /*
- * Calculate start block index that this node page contains
+ * Calculate start block index indicating the given node offset.
+ * Be careful, caller should give this node offset only indicating direct node
+ * blocks. If any node offsets, which point the other types of node blocks such
+ * as indirect or double indirect node blocks, are given, it must be a caller's
+ * bug.
  */
 block_t start_bidx_of_node(unsigned int node_ofs)
 {
Signature · 1 line
1.8.0.1.250.gb7973fb