svn commit: r276884 - head/sys/kern

[Robert Watson]

Author: rwatson
Date: Fri Jan  9 12:08:51 2015
New Revision: 276884
URL: https://svnweb.freebsd.org/changeset/base/276884

Log:
  Remove a 'This is dumb' comment that has been incorrect for at least a
  decade: m_pulldown() is willing to consider ordinary mbufs writable.
  Retain another, related, and also outdated comment, but with a caveat
  that it is partially stale.  Do not, for now, address the problem that
  it raises (that only EXT_CLUSTER external storage is considered
  writable, regardless of the results of M_WRITABLE() on the mbuf).
  
  MFC after:	3 days
  Sponsored by:	EMC / Isilon Storage Division

Modified:
  head/sys/kern/uipc_mbuf2.c

Modified: head/sys/kern/uipc_mbuf2.c
==============================================================================
--- head/sys/kern/uipc_mbuf2.c	Fri Jan  9 10:59:49 2015	(r276883)
+++ head/sys/kern/uipc_mbuf2.c	Fri Jan  9 12:08:51 2015	(r276884)
@@ -131,6 +131,8 @@ m_pulldown(struct mbuf *m, int off, int 
 	}
 
 	/*
+	 * The following comment is dated but still partially applies:
+	 *
 	 * XXX: This code is flawed because it considers a "writable" mbuf
 	 *      data region to require all of the following:
 	 *	  (i) mbuf _has_ to have M_EXT set; if it is just a regular
@@ -148,10 +150,6 @@ m_pulldown(struct mbuf *m, int off, int 
 	 * M_WRITABLE(). For now, we only evaluate once at the beginning and
 	 * live with this.
 	 */
-	/*
-	 * XXX: This is dumb. If we're just a regular mbuf with no M_EXT,
-	 *      then we're not "writable," according to this code.
-	 */
 	writable = 0;
 	if ((n->m_flags & M_EXT) == 0 ||
 	    (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n)))