Re: fatfs locking

[Thomas Bushnell, BSG]

Roland McGrath <roland@frob.com> writes:

> It occurs to me that without this change, if a directory in a ufs or ext2fs
> filesystem contains a link to itself by a name other than "..", then a
> lookup of that name will deadlock the directory node.  (That is probably an
> invalid state that fsck would fix, but still.)  Am I right about that?

It is assumed that (with the exception of "..", and links to self) the
structure of links is a DAG.

The rule that you must hold a lock on the directory to acquire the
node in the directory is very old in Unix, and works around the
following situation:

lock dir
look up name, record inode number N
unlock dir

  other thread now unlinks inode N, which then gets allocated as some
  new node

lock inode N

whoops--wrong node!

(There is also a problem if the node is freed but not allocated; this
problem is more easily detectable and can be worked around.)

The ".." case works by:

lock subdir
look up .., record parent inode number N
unlock subdir -- holding a reference to prevent reallocation
lock parent inode N
lock subdir
look up ".." again, make sure it still points at N
unlock subdir

Thomas