Created attachment 76841 [details] Kernel config file used. Kernel 3.6-rc1 Slackware 64 current (gcc 4.7.0) Core i7, 6GB Since 3.6-rc1, I get the following warning when booting with debug options (check memory use): WARNING: kmemcheck: Caught 32-bit read from uninitialized memory (ffff8801b6a6d950) ffffffff000000000000000000000000002c0400000000001c00000000000000 i i i i i i i i i i i i i i i i u u u u i i i i i i u u u u u u ^ Pid: 7839, comm: kdm Not tainted 3.6.0-rc1 #1 Acer Aspire 7750G/JE70_HR RIP: 0010:[<ffffffff811b3093>] [<ffffffff811b3093>] ext4_da_get_block_prep+0x243/0x2b0 RSP: 0018:ffff8801c29eb958 EFLAGS: 00010282 RAX: 0000000000000000 RBX: ffff8801b6a6d4d0 RCX: 0000000000000000 RDX: ffffffff811b3085 RSI: ffff8801b6a6d9e8 RDI: 0000000000000001 RBP: ffff8801c29eb9d8 R08: ffffffff820f4660 R09: ffffffff820c5ce0 R10: 0000000000000001 R11: 0000000000000000 R12: ffff8801a9606d00 R13: ffff8801c1109800 R14: 0000000000000000 R15: 0000000000000000 FS: 00007fd6e5908740(0000) GS:ffff8801c7e00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: ffff8801c6be0dd0 CR3: 00000001c29da000 CR4: 00000000000407f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff4ff0 DR7: 0000000000000400 [<ffffffff8116efc1>] __block_write_begin+0x1b1/0x540 [<ffffffff811b62f1>] ext4_da_write_begin+0xc1/0x1a0 [<ffffffff810f4c09>] generic_file_buffered_write+0x119/0x2a0 [<ffffffff810f6c3c>] __generic_file_aio_write+0x1bc/0x3c0 [<ffffffff810f6ec5>] generic_file_aio_write+0x85/0x110 [<ffffffff811afb24>] ext4_file_write+0xa4/0x4b0 [<ffffffff8113d2a7>] do_sync_write+0xa7/0xe0 [<ffffffff8113daf6>] vfs_write+0xa6/0x160 [<ffffffff8113ddf8>] sys_write+0x48/0xa0 [<ffffffff817e1aa2>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff In gdb, I get the following: (gdb) l *0xffffffff811b3093 0xffffffff811b3093 is in ext4_da_get_block_prep (fs/ext4/inode.c:1217). 1212 /* 1213 * ext4_calc_metadata_amount() has side effects, which we have 1214 * to be prepared undo if we fail to claim space. 1215 */ 1216 save_len = ei->i_da_metadata_calc_len; 1217 save_last_lblock = ei->i_da_metadata_calc_last_lblock; 1218 md_needed = EXT4_NUM_B2C(sbi, 1219 ext4_calc_metadata_amount(inode, lblock)); 1220 trace_ext4_da_reserve_space(inode, md_needed); 1221 This is new to 3.6-rc1, hence a regression.
It's a false positive in that this isn't going to actually cause an actual bug; we're just saving and restoring an uninitialized value if we have a error. The fact that we do this is no big deal, since if i_da_metadata_calc_len is zero, the value of i_data_metadata_calc_last_lblock is never consulted. So if we save an uninitialized value, and then later restore it, it really doesn't matter from a correctness point of view. That being said, triggering a kmemcheck failure is sad, even if it's not a real bug, so I agree we should fix it. The most expedient and most efficient fix is to simply initialize i_data_metadata_calc_last_lblock to zero in ext4_alloc_inode(); this is where we initialize i_data_metadata_calc_len to zero, and if we initialize the last_lblock to zero, it won't matter, and it saves us from having to adding an extra conditional in ext4_da_get_block_prep(). - Ted
A patch referencing this bug report has been merged in Linux v3.6-rc3: commit 7e731bc9a12339f344cddf82166b82633d99dd86 Author: Theodore Ts'o <tytso@mit.edu> Date: Sun Aug 5 23:28:16 2012 -0400 ext4: avoid kmemcheck complaint from reading uninitialized memory
Fixed with commit 7e731bc9a123