PERFORCE change 79535 for review

[Robert Watson]

http://perforce.freebsd.org/chv.cgi?CH=79535

Change 79535 by rwatson at rwatson_paprika on 2005/07/04 11:15:16

	On amd64, check access to /dev/io against CAP_SYS_RAWIO.
	
	On amd64, use CAP_SYS_RAWIO to control access to setting user space
	debugging registers on kernel addresses.  However, it strikes me
	that this should likely be entirely disallowed.

Affected files ...

.. //depot/projects/trustedbsd/sebsd/sys/amd64/amd64/io.c#2 edit
.. //depot/projects/trustedbsd/sebsd/sys/amd64/amd64/machdep.c#11 edit

Differences ...

==== //depot/projects/trustedbsd/sebsd/sys/amd64/amd64/io.c#2 (text+ko) ====

@@ -28,6 +28,7 @@
 __FBSDID("$FreeBSD: src/sys/amd64/amd64/io.c,v 1.1 2004/08/01 11:40:50 markm Exp $");
 
 #include <sys/param.h>
+#include <sys/capability.h>
 #include <sys/conf.h>
 #include <sys/fcntl.h>
 #include <sys/lock.h>
@@ -54,7 +55,7 @@
 {
 	int error;
 
-	error = suser(td);
+	error = cap_check(td, CAP_SYS_RAWIO);
 	if (error != 0)
 		return (error);
 	error = securelevel_gt(td->td_ucred, 0);

==== //depot/projects/trustedbsd/sebsd/sys/amd64/amd64/machdep.c#11 (text+ko) ====

@@ -1726,9 +1726,12 @@
 		 * address space is written into from within the kernel
 		 * ... wouldn't that still cause a breakpoint to be generated
 		 * from within kernel mode?
+		 *
+		 * XXXRW: Should we allow privileged user processes to do
+		 * this at all?
 		 */
 
-		if (suser(td) != 0) {
+		if (cap_check(td, CAP_SYS_RAWIO) != 0) {
 			if (dbregs->dr[7] & 0x3) {
 				/* dr0 is enabled */
 				if (dbregs->dr[0] >= VM_MAXUSER_ADDRESS)
To Unsubscribe: send mail to majordomo at trustedbsd.org
with "unsubscribe trustedbsd-cvs" in the body of the message