Re: kern/43199: read(2) returns bad size in multithreaded programs

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR kern/43199; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/43199: read(2) returns bad size in multithreaded programs
Date: Fri, 23 Apr 2010 22:54:51 +0000

 On Fri, Apr 23, 2010 at 03:05:06PM +0000, Jason Thorpe wrote:
  >  That all depends on that sort of file descriptor it is and how you read
  >  from it.  If it is a regular file, the behavior is non-deterministic
  >  unless you use pread(2).  If it is a socket, correct behavior requires
  >  exclusive access to the socket for the duration of the read(2) call.
 
 C&V? As far as I understand, nothing about threads relaxes the basic
 requirement that read() calls be atomic. The Linux behavior of playing
 fast and loose violates POSIX; they've explicitly decided they don't
 care.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost