Ext4 reported as clean by fsck after hard reset: Is that normal

crashe2fsckext4fsck

My root-partition is formatted as ext4-filesystem.

I notice that, whenever my machine crashes and I have to hard-reset it, when booting up again and the root filesystem is checked this step takes a bit (like one to two seconds) longer than when booting from a cleanly shut down system, but it is reported as "clean" (and nothing like /dev/<rootpartition> was not cleanly unmounted, check forced). The filesystem is 92% full (352 GiB).

My question: I wonder if this is the normal and a safe behaviour of ext4 or some bug in the startup-scripts. I know that ext4 has much faster fsck than ext3, but I am worried about that it is reported as "clean" after a system crash.

When I run e2fsck -f manually on that partition the check lasts comparable to an ext2/ext3 filesystem. So I am worried and since beeing so i tuned my filesystem to be checked at every boot (tune2fs -c 1), which results in a full check taking as long as e2fsck -f every boot.

Edit, just to clarify:
After a non-clean reset, usually, on /var, which is reiserfs, fsck replays journal entries; on /boot, which is ext2, fsck runs, displays progress bar, and reports "clean" after running. Only on the root filesystem no "check forced" and no fsck-progress appears, which do appear for the other file systems even if they turn out to be clean. That is the worrying difference!

Best Answer

ext4 is a journaling filesystem, the and one of the main goals of journaling is to survive unclean shutdowns w/o damage, and thus not require a long fsck.

In short, a journaling filesystem like ext3/4 writes metadata (at least) changes twice. First, it writes them to the "journal". Then, once that's on disk, it writes to the actual filesystem metadata. (Writing to the journal is much faster, as the journal is sequential and doesn't require a bunch of seeks. At least on magnetic disks—seek penalty is greatly reduced on SSDs.)

The several extra seconds is likely the journal replay: basically, if the filesystem isn't unmounted cleanly, the next mount or fsck will read the journal and apply any changes that aren't yet in the main filesystem.

So, in short, sounds like its behaving as expected.

Related Solutions

How to avoid “Run fsck manually” messages while allowing experimenting with system time changes

I was going to suggest hacking e2fsck to disable the specific checks for a last mount time or last write times in the future. These are defined in problem.c / problem.h, and used in super.c. But in looking, I discovered that E2fsprogs 1.41.10 adds a new option to /etc/e2fsck.conf called broken_system_clock. This seems to be exactly what you need, and since you're using Red Hat Enterprise Linux 6, you should have 1.41.12, which includes this option. From the man page:

   broken_system_clock
          The e2fsck(8) program has some hueristics that assume  that  the
          system clock is correct.  In addition, many system programs make
          similar assumptions.  For example, the UUID library  depends  on
          time  not going backwards in order for it to be able to make its
          guarantees about issuing universally unique ID’s.  Systems  with
          broken  system clocks, are well, broken.  However, broken system
          clocks, particularly in embedded systems, do exist.  E2fsck will
          attempt  to  use  hueristics to determine if the time can no tbe
          trusted; and to skip time-based checks if this is true.  If this
          boolean  is set to true, then e2fsck will always assume that the
          system clock can not be trusted.

Yes, the man page can't spell "heuristics". Oops. But presumably the code works anyway. :)

Linux – What Does fsck -p (Preen) Do on ext4?

The answer to your question lies in the e2fsck/problems.c file of the e2fsprogs source code. Looking for the PR_PREEN_OK flag should get you started.

As the complete error handling is a bit more involved, due to the multitude of different error conditions that may occur, you are advised to have a closer look at the code if you are concerned about a specific case. However, the lists below were extracted from the comments to the error conditions and should give you a rough overview about the effects of the preen-mode.

The following errors/warnings are currently handled automatically when the -p flag is specified:

Relocate hint
Journal inode is invalid
Journal superblock is corrupt
Superblock has_journal flag is clear but has a journal
Superblock needs_recovery flag is set but not journal is present
Filesystem revision is 0, but feature flags are set
Superblock hint for external superblock
group descriptor N marked uninitialized without feature set.
group N block bitmap uninitialized but inode bitmap in use.
Group descriptor N has invalid unused inodes count.
Last group block bitmap uninitialized.
The test_fs flag is set (and ext4 is available)
Last mount time is in the future (fudged)
Last write time is in the future (fudged)
Block group checksum (latch question) is invalid.
Root directory has dtime set
Reserved inode has bad mode
Deleted inode has zero dtime
Inode in use, but dtime set
Zero-length directory
Inode has incorrect i_size
Inode has incorrect i_blocks
Bad superblock in group
Bad block group descriptors in group
Block claimed for no reason
Error allocating blocks for relocating metadata
Error allocating block buffer during relocation process
Relocating metadata group information from X to Y
Relocating metatdata group information to X
Block read error during relocation process
Block write error during relocation process
Immutable flag set on a device or socket inode
Non-zero size for device, fifo or socket inode
Filesystem revision is 0, but feature flags are set
Journal inode is not in use, but contains data
Journal has bad mode
INDEX_FL flag set on a non-HTREE filesystem
INDEX_FL flag set on a non-directory
Invalid root node in HTREE directory
Unsupported hash version in HTREE directory
Incompatible flag in HTREE root node
HTREE too deep
invalid inode->i_extra_isize
invalid ea entry->e_name_len
invalid ea entry->e_value_offs
invalid ea entry->e_value_block
invalid ea entry->e_value_size
invalid ea entry->e_hash
inode missing EXTENTS_FL, but is an extent inode
Inode should not have EOFBLOCKS_FL set
Directory entry has deleted or unused inode
Directory filetype not set
Directory filetype set on filesystem
Invalid HTREE root node
Invalid HTREE limit
Invalid HTREE count
HTREE interior node has out-of-order hashes in table
Inode found in group where _INODE_UNINIT is set
Inode found in group unused inodes area
i_blocks_hi should be zero
/lost+found not found
Unattached zero-length inode
Inode ref count wrong
Padding at end of inode bitmap is not set.
Padding at end of block bitmap is not set.
Block bitmap differences header
Block not used, but marked in bitmap
Block used, but not marked used in bitmap
Block bitmap differences end
Inode bitmap differences header
Inode not used, but marked in bitmap
Inode used, but not marked used in bitmap
Inode bitmap differences end
Free inodes count for group wrong
Directories count for group wrong
Free inodes count wrong
Free blocks count for group wrong
Free blocks count wrong
Block range not used, but marked in bitmap
Block range used, but not marked used in bitmap
Inode range not used, but marked in bitmap
Inode range used, but not marked used in bitmap
Group N block(s) in use but group is marked BLOCK_UNINIT
Group N inode(s) in use but group is marked INODE_UNINIT
Recreate journal if E2F_FLAG_JOURNAL_INODE flag is set

The following error conditions cause the non-interactive fsck process to abort, even if the -p flag is set:

Block bitmap not in group
Inode bitmap not in group
Inode table not in group
Filesystem size is wrong
Inode count in superblock is incorrect
The Hurd does not support the filetype feature
Journal has an unknown superblock type
Ask if we should clear the journal
Journal superblock has an unknown read-only feature flag set
Journal superblock has an unknown incompatible feature flag set
Journal has unsupported version number
Ask if we should run the journal anyway
Reserved blocks w/o resize_inode
Resize_inode not enabled, but resize inode is non-zero
Resize inode invalid
Last mount time is in the future
Last write time is in the future
group descriptor N checksum is invalid.
Root directory is not an inode
Block bitmap conflicts with some other fs block
Inode bitmap conflicts with some other fs block
Inode table conflicts with some other fs block
Block bitmap is on a bad block
Inode bitmap is on a bad block
Illegal blocknumber in inode
Block number overlaps fs metadata
Inode has illegal blocks (latch question)
Too many bad blocks in inode
Illegal block number in bad block inode
Bad block inode has illegal blocks (latch question)
Bad block used as bad block indirect block
Inconsistency can't be fixed prompt
Bad primary block prompt
Suppress messages prompt
Imagic flag set on an inode when filesystem doesn't support it
Compression flag set on an inode when filesystem doesn't support it
Deal with inodes that were part of orphan linked list
Deal with inodes that were part of corrupted orphan linked list (latch question)
Error reading extended attribute block
Invalid extended attribute block
Extended attribute reference count incorrect
Multiple EA blocks not supported
Error EA allocation collision
Bad extended attribute name
Bad extended attribute value
Inode too big (latch question)
Directory too big
Regular file too big
Symlink too big
Bad block has indirect block that conflicts with filesystem block
Resize inode failed
inode appears to be a directory
Error while reading extent tree
Failure to iterate extents
Bad starting block in extent
Extent ends beyond filesystem
EXTENTS_FL flag set on a non-extents filesystem
inode has extents, superblock missing INCOMPAT_EXTENTS feature
Fast symlink has EXTENTS_FL set
Extents are out of order
Inode has an invalid extent node
Clone duplicate/bad blocks?
Bad inode number for '.'
Directory entry has bad inode number
Directry entry is link to '.'
Directory entry points to inode now located in a bad block
Directory entry contains a link to a directory
Directory entry contains a link to the root directry
Directory entry has illegal characters in its name
Missing '.' in directory inode
Missing '..' in directory inode
First entry in directory inode doesn't contain '.'
Second entry in directory inode doesn't contain '..'
i_faddr should be zero
i_file_acl should be zero
i_dir_acl should be zero
i_frag should be zero
i_fsize should be zero
inode has bad mode
directory corrupted
filename too long
Directory inode has a missing block (hole)
'.' is not NULL terminated
'..' is not NULL terminated
Illegal character device inode
Illegal block device inode
Duplicate '.' entry
Duplicate '..' entry
Final rec_len is wrong
Error reading directory block
Error writing directory block
Directory entry for '.' is big. Split?
Illegal FIFO inode
Illegal socket inode
Directory filetype incorrect
Directory filename is null
Invalid symlink
i_file_acl (extended attribute block) is bad
Filesystem contains large files, but has no such flag in sb
Clear invalid HTREE directory
Bad block in htree interior node
Duplicate directory entry found
Non-unique filename found
i_blocks_hi should be zero
Unexpected HTREE block
Root inode not allocated
No room in lost+found
Unconnected directory inode
.. entry is incorrect
Lost+found not a directory
Unattached inode
Superblock corrupt
Fragments not supported
Error determing physical device size of filesystem
The external journal has (unsupported) multiple filesystems
Can't find external journal
External journal has bad superblock
Superblock has a bad journal UUID
Error allocating inode bitmap
Error allocating block bitmap
Error allocating icount link information
Error allocating directory block array
Error while scanning inodes
Error while iterating over blocks
Error while storing inode count information
Error while storing directory block information
Error while reading inode (for clearing)
Error allocating refcount structure
Error reading Extended Attribute block while fixing refcount
Error writing Extended Attribute block while fixing refcount
Error allocating EA region allocation structure
Error while scanning inodes
Error allocating inode bitmap
Internal error: couldn't find dir_info
Error allocating icount structure
Error iterating over directory blocks
Error deallocating inode
Error adjusting EA refcount
Error allocating inode bitmap
Error creating root directory
Root inode is not directory; aborting
Cannot proceed without a root inode.
Internal error: couldn't find dir_info
Programming error: bitmap endpoints don't match
Internal error: fudging end of bitmap
Error copying in replacement inode bitmap
Error copying in replacement block bitmap

Best Answer

Related Solutions

How to avoid “Run fsck manually” messages while allowing experimenting with system time changes

Linux – What Does fsck -p (Preen) Do on ext4?

Related Question