Re: ls and stat segfault on loopback mounted image

From: Valdis.Kletnieksat_private
Date: Tue Jan 22 2002 - 10:37:23 PST

  • Next message: Brian Coyle: "Re: ls and stat segfault on loopback mounted image"

    On Mon, 21 Jan 2002 22:20:53 EST, Brian Coyle <brianat_private>  said:
    
    > Copies of the subject disk images were mounted (ro,loop,nodev,noexec) such 
    > that the original filesystem structure was maintained.
    > 
    > $CRACKED_BOX/home/ftp appears to be a chroot jail (can you confirm Lance?) 
    > containing bin, etc, lib and pub directories.  Navigating into bin, etc and 
    > pub followed by an `ls -l` each works fine.
    > 
    > However, in lib, `ls -l` and `stat *` both fail with a segfault.  Omitting
    > the -l option for ls does list the files.  debugfs (ls -l) has no problem 
    > with the directory.
    
    I'm wondering if you managed to get a bad copy of the disk image, and there's
    a busticated inode belonging to some file in lib/.  
    
    To test:
    
    1) cd lib/
    2) /bin/ls     (you say this works)
    3) (bash/ksh)  for i in `/bin/ls`; do echo $i; /bin/ls -l $i; done
    
    This will find if there's a specific file that's giving ls and company
    indigestion.  Then go poke that file's inode (which you can get with
    'ls -i broken.file') with debugfs and see if it's corrupted in some way.
    
    I'm wondering if the $CRACKED_BOX had a kernel module loaded that used
    a previously reserved bit in the inode as a "hide me please" flag, and
    a modified lsattr/chattr command to set the bit, and 'ls' and 'stat'
    both die when they see a non-zero in the reserved bit, but debugfs
    ignores it (since debugfs is *expecting to see broken inodes, and 'ls'
    is expecting to be handed correct ones).
    
    -- 
    				Valdis Kletnieks
    				Computer Systems Senior Engineer
    				Virginia Tech
    
    
    
    



    This archive was generated by hypermail 2b30 : Tue Jan 22 2002 - 11:10:13 PST