RE: [cgl_discussion] RE: about

Tariq,

I've had some of the same issues that LKML did with the original Driver
Hardening spec (too confining, new APIs that can't be imposed on all
drivers, not everyone wants the new event logging, etc.).  Which is why I
went my own way with the CGL raid/scsi/adaptec hardening.  

Also, Dave Howell is becoming involved with HDD for drivers used by
Langley/Pr systems, so you guys may want to collaborate.

Here's my take on a rework of the Driver Hardening efforts with a practical
bent.  
   Level1 - fix any bugs found:
		. during use case testing, 
		. from kernel janitor programs (source parsing), or 
            . from mailing list patches (backports)
		A driver would be considered Level-1 hardened if the janitor
programs passed and the 
		use-case testing passed.  The tests that were run should be
posted on the project when 
		they are done.
   Level2 - make sure that the driver exposes:
            . statistics/tallies   (/proc by default)
            . diagnostics/controls  (write to /proc, sysfs, or /dev ioctls,
etc.)
            . useful log messages for exceptions and versioning (info for
servicability)
            . unique device identifiers (for service/FRU, and hot-plug)
            and add any service/administration tools as needed
   Level3 - do some management of statistics and diagnostics (via RSM or now
PCP) 
            do some analysis of errors/statistics for predictive failure

My thinking is that if any form of these Level2 functions is supported,
conforming the functions to a common interface (with a PCP module, for
example) won't be that hard.

Use Case Testing categories:  
   feature tests, 
   induced-error tests (could include FI), 
   stress tests (overnight, etc.), 
   performance tests.

As I understand it, the scope of HDD has been reduced, for now, to just
encompass Level 1 stuff, is that right?

Anyway, this is my $0.02.  Comments/questions welcome.

Andy


-----Original Message-----
From: Shureih, Tariq [mailto:[email protected]] 
Sent: Friday, February 14, 2003 1:05 PM
To: Lynch, Rusty; 'Steven Brown'
Cc: 'Arun Prasad'; [email protected];
[email protected]
Subject: RE: [cgl_discussion] RE: [Hardeneddrivers-discuss] about


I am not on thehardeneddrivers mailing list seems like it.

I took over the HDD initiative after LKML basically slaughtered the original
proposal.

I am of the opinion that hardened drivers is an incomplete statement and
more of the opinion of a hardened linux.

Since then I have been working with the kernel Janitor project and actively
following the 2.5.x development to achieve two things.

Establish a trend in the janitor project for auditing device drivers (for a
start) based on the project's TODO list and clean device drivers of known
problems/issues.  Further, publish a Drivers HOWTO-Dos and DON'Ts for
writing a robust linux kernel device driver based on best coding practices
as well as best linux details.

Second, establish the idea of a hardened linux kernel rather than just a
hardened device driver spec.  This stems from my believe that a hardened
driver (and kernel over all) is nothing more than a well thought off design,
good error management, in-place fault recovery and the obvious, good coding
discipline.

I'm barely getting my feet wet with the driver's howto, but things should
start surfacing soon.

--
Tariq Shureih
*My opinions are my own and don't represent my employer.*

-----Original Message-----
From: Lynch, Rusty [mailto:[email protected]] 
Sent: Friday, February 14, 2003 9:53 AM
To: 'Steven Brown'; Lynch, Rusty
Cc: 'Arun Prasad'; [email protected];
[email protected]
Subject: [cgl_discussion] RE: [Hardeneddrivers-discuss] about

You are correct; the last couple of weeks have been slow.  This has been
mainly due to the fact that two of the active developers have been on
vacation (Chinese New Year), and also due to the fact that I have been
focusing on a couple of other activities.

As for as the device driver hardening specification... the idea suffered a
bloody death when it was finally pitched to LKML.  Although there were a
couple of aspects to driver hardening that the community (LKML) did find
interesting, one of those was the idea enabling fault injection testing and
the other creating a HOWTO document to aid driver developers in creating
robust device drivers.

The fault injection fall out spawned http://fault-injection.sf.net, and that
is what I focused on.

IMHO the original needs that triggered the whole device driver hardening
specification are still relevant, and might be addressed (in some form) in
the version 2 CGL requirements that are currently being debated.

By original needs, I mean that everyone and their dog is releasing "hardened
device drivers" for Linux.  What exactly does that mean? An attempt to
specify how the driver is implemented surfaced a lot of fundamental problems
with the approach.  

Personally I think the answer might be to specify a general test plan
instead of specifying API's and coding styles that the driver uses.  I would
expect that test plan to include fault injection testing to prove the driver
responds _correctly_ to fault conditions.

    --rustyl

> -----Original Message-----
> From: Steven Brown [mailto:[email protected]]
> Sent: Friday, February 14, 2003 9:25 AM
> To: Lynch, Rusty
> Cc: 'Arun Prasad'; [email protected]
> Subject: Re: [Hardeneddrivers-discuss] about
> 
> Rusty,
> 
> Prior to your and Parun's posts I haven't seen any activity on this
> group since I subscribed a couple of weeks ago, so . . . .
> 
> Does your response imply that since there is the project for fault
> injection, that device driver hardening is being actively worked on for
> the 2.6 kernel ? If so, then do you happen to the the state of that
> activity and the current revision of the DDH specfication ?
> 
> Thanks.
> 
> swb
> 
> ------------------------
> Steven W. Brown
> 
> Lynch, Rusty wrote:
> 
> >Hello Arun,
> >
> >The fault injection project is actively working on a 2.5 kernel based
> >implementation at http://fault-injection.sf.net.  At that site you will
> see
> >a pointer to a bitkeeper tree that has our absolute latest code, and then
> we
> >also throw patches up on our sourceforge download page (but not very
> >frequently since every seems to be happy pulling from the bk tree.)
> >
> >The project has stopped focusing on the user space fault injection tools
> in
> >favor of focusing on developing a core fault injection infrastructure for
> >the 2.6 kernel.
> >
> >To start contributing, sign up on the project mailing list, clone the bk
> >tree (and if that is a problem the request a new diff on the mailing
> list),
> >and start submitting patches to the list.
> >
> >    --rustyl
> >
> >
> >
> >>-----Original Message-----
> >>From: Arun Prasad [mailto:[email protected]]
> >>Sent: Friday, February 14, 2003 12:26 AM
> >>To: [email protected]
> >>Subject: [Hardeneddrivers-discuss] about
> >>
> >>Hi all,
> >>    Iam interested in contribute (participate) in the Driver Harness
> >>tool for Linux (FITH).  I browsed through
> >>the FITH hld and DDH-05 spec...  I like to clarify the below doubts...
> >>
> >>* Does this project still going on actively?
> >>
> >>* Whats the current status of the project?
> >>
> >>* Can I start contributing to this project from this stage?
> >>
> >>
> >>Please clarify the doubts...
> >>
> >>Thanks
> >>-arun
> >>
> >>
> >>
> >>-------------------------------------------------------
> >>This SF.NET email is sponsored by: FREE  SSL Guide from Thawte
> >>are you planning your Web Server Security? Click here to get a FREE
> >>Thawte SSL guide and find the answers to all your  SSL security issues.
> >>http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0026en
> >>_______________________________________________
> >>Hardeneddrivers-discuss mailing list
> >>[email protected]
> >>https://lists.sourceforge.net/lists/listinfo/hardeneddrivers-discuss
> >>
> >>
> >
> >
> >-------------------------------------------------------
> >This SF.NET email is sponsored by: FREE  SSL Guide from Thawte
> >are you planning your Web Server Security? Click here to get a FREE
> >Thawte SSL guide and find the answers to all your  SSL security issues.
> >http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0026en
> >_______________________________________________
> >Hardeneddrivers-discuss mailing list
> >[email protected]
> >https://lists.sourceforge.net/lists/listinfo/hardeneddrivers-discuss
> >
> >
> 
> --
> swb
> 
> -------------------------------
> Steven W. Brown
> x86 Entry Systems (XES)
> Sun Mircosystems, Inc.
> 12 Network Circle, MS:MPK12-225
> access line: 650-352-5018/x40018
> 
> [email protected]
> --------------------------------

_______________________________________________
cgl_discussion mailing list
[email protected]
http://lists.osdl.org/mailman/listinfo/cgl_discussion
_______________________________________________
cgl_discussion mailing list
[email protected]
http://lists.osdl.org/mailman/listinfo/cgl_discussion


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf