On Jul 8,  5:24pm, chuq@chuq.com (Chuck Silvers) wrote:
-- Subject: Re: lib/30664: realpath and magic symlinks

|  > Perhaps readlink(2) should translate the magic symlink?  But that  
|  > would break ls(1).  Perhaps we could add a readlink2(2) that takes a  
|  > flags argument, allowing us to specify "raw" or "translated".   
|  > readlink(2) would default to "translated", and ls(1) would be changed  
|  > to use "raw".
|  
|  I think that the real problem is having the path interpretation be different
|  for symlinks vs. other path lookups done in the kernel.  it would make
|  a lot more sense to have the tranlation done for all pathname lookups
|  or none, and not this awkward halfway thing that "magic symlinks" provides.
|  
|  then readlink() would be fine as it is.

I totally agree with that. For now the best course of action is:

1. Leave magiclinks in, default off; mark it experimental and don't pull it
   up to 3.1
2. Create sysctl to turn on/off readlink() return behavior. This sysctl
   "proc.pid.magiclinks" (or something) defaults to 0, and is inherited
   across fork and exec. If proc.pid.magiclinks==0, readlink returns the
   translated path; if ==1 then it returns the raw path.
3. In the long term, move the whole mess to namei(), deprecate magiclinks.
   and rename the magiclinks mount option magicnames.

christos