[cgl_discussion] PSH HLD rev 1.0 comments

Randy.Dunlap rddunlap at osdl.org
Thu Oct 3 17:09:18 PDT 2002


Comments on "Platform Signal Handler - Event Log Interface"
Revision 1.0 of 9/13/02.


0.  Please delete the italics about what an HLD spec should contain.

1.  The PSH spec should contain more than an Event Log interface.
Actually the Event Log interface is questionable at the moment,
and even if it weren't, that would be the least interesting part
of a PSH spec, even if a necessary part of it.  IOW, it should be
a spec about PSH infrastructure/framework (is there any?), init
requirements of a PSH (if any), etc., as well as the classes of
events that a PSH must handle (which is pretty good in the current
document).

2.  (old news) Parts of this spec that relate to Driver Hardening v0.5
need to be addressed and modified.

Section 1.2, paragraph 1:  change PSH to PS (2 times)

Section 2, para. 1:  change PSH to PS

Section 3.1 questions (answers can go elsewhere if more appropriate):
a.  This section describes what happens after a PS event.  What does
    a PSH need to be before a PS event happens?
b.  How does a PSH hook into the hardware event mechanism?
c.  Can there be multiple PSH's?
d.  Can there be multiple PS event mechanisms?  If so, do they require
    a PSH each or can just one PSH handle all PS events?

Section 3.2.1, para. 1:  change PSH to PS

Section 3.2.1.2, para. 2:  "and cause a system reset."
	System reset is a policy decision.  This probably should be
	"and cause a system panic." and then panic() can implement
	the policy that is defined in /proc/sys/kernel/panic .


Thanks,
-- 
~Randy




More information about the cgl_discussion mailing list