001: Out of backing store
If the system runs out of backing store, the kernel panics.
002: Out of order syncs
Under certain circumstances the filesystem doesn't sync data to disk in the same order as ffs. If the system crashes while much data remains to be synced, the filesystem can be left in a state that will confuse fsck to the point that running fsck -y may conceivably trash the filesystem.
003: Recursive #!
echo #!/tmp/foo>/tmp/foo;chmod 755 /tmp/foo;/tmp/foo
causes the system to reboot without warning.
004: truncate()
truncate doesn't work correctly. In particular, truncating a file to a non-zero length will usually result in truncation to zero length and an error return.
005: umount -f
umount -f doesn't work correctly. In particular, it can crash the system.
006: /dev/fd
Certain uses of /dev/fd can lead to deadlock.
007: select()/console write
A process running on the console that selects on stdout for write will block forever. This may be an indication of some more general problem in select.
008: meaningless lseeks
Some devices for which lseek is meaningless will nevertheless allow you to lseek them.
009: setuid and setgid
The implementation of the setuid() library routine sets the real, effective, and saved user ids if the real user id of the caller is 0, and otherwise sets only the effective user id. It should set the real, effective, and saved user ids if the effective user id of the caller is 0, and otherwise set only the effective user id. A parallel problem exists in the implemntation of setgid(). [patch]
(As we receive additional bug reports, or additional information on the bugs listed above, it will be posted on this page.)
Last updated Wednesday, December 9, 1998 by Robert Haas (rhaas@eecs.harvard.edu).