re: kern/50475: sys_issetugid is missing locking

[matthew green <mrg%eterna.com.au@localhost>]

The following reply was made to PR kern/50475; it has been noted by GNATS.

From: matthew green <mrg%eterna.com.au@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
    netbsd-bugs%netbsd.org@localhost
Subject: re: kern/50475: sys_issetugid is missing locking
Date: Sun, 29 Nov 2015 10:13:54 +1100

 >    sys_issetugid() reads curproc->p_flag without taking any locks, but
 >    sys/proc.h says p_flag is supposed to be protected by p_lock.
 > 
 >    Granted reading ints is atomic on normal platforms, but it's wrong,
 >    and also other code is entitled to assume that it can temporarily
 >    leave invalid values in p_flag while it holds p_lock.
 > 
 >    If the behavior of sys_issetugid() is intended, it should be
 >    documented in proc.h.
 
 i think this is entirely intended.  these values are readable
 in process context without a lock.  it's not just PK_SUGID.
 this happens all over the tree, and i think it's entirely
 reasonable.
 
 proc.h comments need updating only here, i think.
 
 
 .mrg.