Attachment #24921 for bug #42512

View | Details | Raw Unified | Return to bug 42512
Collapse All | Expand All




<li><a href="#open">Open Discussion</a></li>
</ul>

<p>NOTE: As usual I missed some names, please add those I missed. During
  the discussion names have been abbreviated, for committers their
  FreeBSD.org username has been used, for non committers the initials
  are used.</p>

<h2>Attending:</h2>

<p>Committers In Person:</p>
<ul>
  <li>Robert Watson (rwatson)</li>
  <li>Julian Elischer (julian)</li>
  <li>John Baldwin (jhb)</li>
  <li>Matt Dillon (dillon)</li>
  <li>Warner Losh (warner)</li>
  <li>David O'Brien (obrien)</li>
  <li>Jeffery Hsu (hsu)</li>
  <li>Jennifer Yang (jennifer)</li>
  <li>Bosko Milekic (bmilekic)</li>
  <li>Alfred Perlstein (alfred)</li>
  <li>Doug Rabson (dfr)</li>
  <li>Paul Saab (ps)</li>
  <li>Brooks Davis (brooks)</li>
  <li>Murray Stokely (murray)</li>
  <li>Jonathan Mini (mini)</li>
  <li>Takanori Watanabe (takawata)</li>
  <li>Gordon Tetlow (gordon)</li>
  <li>Gregory Shapiro (gshapiro)</li>
  <li>Sam Leffler (sam)</li>
  <li>Bruce Mah (bmah)</li>
</ul>

<p>Also In Person:</p>
<ul>
  <li>Robert Watson (RW)</li>
  <li>Julian Elischer(JE)</li>
  <li>John Baldwin(JB)</li>
  <li>Matt Dillon (MD)</li>
  <li>Warner Losh (WL)</li>
  <li>David O'Brian (DO)</li>
  <li>Jeffery Xu (JX)</li>
  <li>Jennifer Ying (JY)</li>
  <li>Bosko Milekic (BM)</li>
  <li>Alfred Perlstein (AP)</li>
  <li>Doug Rabson (DR)</li>
  <li>Paul Saab (PS)</li>
  <li>Brooks Davis (BD)</li>
  <li>Murray Stokely (MS)</li>
  <li>Jonathan Mini (JM)</li>
  <li>Watanabe ???</li>
  <li>Gordon Tetlow (GT)</li>
  <li>Gregory Schapiro (GS)</li>
  <li>Sam Leffler (SL)</li>
  <li>Bruce Mah</li>
  <li>George Neville-Neil (gnn)</li>
  <!--<li>Unknown (??)</li>-->
</ul>

<p>On The Phone:</p>
<ul>
  <li>Alan Cox (alc)</li>
</ul>

<p>Via webcast:</p>

<p>Many people listened in via the stream and chatted on IRC to one
  another while the discussion took place.</p>

<p>The meeting followed a format where each section was led by an
  individual and then a discussion ensued.  Not all of the discussion


<div class="discussion">

<p><strong class="speaker">rwatson</strong> : What about userland?</p>

<p><strong class="speaker">julian</strong> : It can run different threads
in userland.  The primitives are all there it just needs a bit more
help.  I would like to put an idea out.  Is it a good idea to be able
to have non-threaded programs linking with threaded libraries?</p>

<p><strong class="speaker">rwatson</strong> : Putting async I/O into such a
thing would make sense.</p>

<p><strong class="speaker">julian</strong> : The library would not care
who was accessing it.</p>

<p><strong class="speaker">rwatson</strong> : For instance libc could be
threaded or not.</p>

<p><strong class="speaker">julian</strong> : That would be interesting.  I
don't know if the two interfaces are incompatible.</p>

<p><strong class="speaker">jhb</strong> : X does this.</p>

<p><strong class="speaker">dillon</strong> : It is very doable but you
have to make it non-preemptive.  If you're switching non-preemptively
you can use library routines which are non threaded.</p>

<p><strong class="speaker">julian</strong> : If I do what I'm thinking of
doing then each lib will have its own KSE group.</p>

<p><strong class="speaker">dillon</strong> : stdio does not have to be
thread aware if you don't schedule preemptively.  It all matters where
it blocks.</p>

<p><strong class="speaker">julian</strong> : Since you're a non-threaded
program you don't know that.</p>

<p><strong class="speaker">rwatson</strong> : If you're going to support
that, libc has to support threads.</p>

<p><strong class="speaker">rwatson</strong> : It sounds like some
complexity goes away.  Can we use 1 libc with has threading?</p>

<p><strong class="speaker">julian</strong> : Do we want to go down this
path?</p>

<p><strong class="speaker">rwatson</strong> : Now or later?</p>

<p><strong class="speaker">julian</strong> : What do I design now to do
this?</p>

<p><strong class="speaker">jhb</strong> : For example libc_r does not
work with rfork.</p>

<p><strong class="speaker">julian</strong> : The answer is that yes we
should move forward.  Tricky issues, signals...</p>

<p><strong class="speaker">warner</strong> : Have people talked about
pthread programs and cancellation points?</p>

<p><strong class="speaker">julian</strong> : The pthreads library does not
assume that you're only going to change threads at yield() points.  We
are going to have cancellation points.  There is an unimplemented call
which will be able to send a thread targeted signal even into the
kernel.</p>

<p><strong class="speaker">julian</strong> : When a thread is scheduled
onto a KSE there is a mailbox that the userland thread scheduler
updates.</p>

<p><strong class="speaker">julian</strong> : Is there anyone else who has
some time or test it?  How many people should test this before I check
it in?  There is a patch that's continuously updated on my web site to
be able to patch it to -CURRENT.  There is a CVSUP target from cvsup

freefal there is a pointer there to a web page that explains how to
CVSUP from source.</p>

<p><strong class="speaker">rwatson</strong> : What about SMP locking for
this?</p>

<p><strong class="speaker">julian</strong> : Handled by the proc locking.
Has not been tried on SMP machines yet.</p>

<p><strong class="speaker">obrien</strong> : What about on Sparc?</p>

<p><strong class="speaker">julian</strong> : You may need to stub things
out.</p>

<p><strong class="speaker">jhb</strong> : Is the paper on the web site?</p>

<p><strong class="speaker">julian</strong> : The updated copy has disappeared.</p>  

<p><strong class="speaker">unknown</strong> : What's the different between
NetBSD and FreeBSD on this?</p>

<p><strong class="speaker">julian</strong> : Logically not a tremendous
difference but Net follows the paper closely and Free takes the idea
and makes it into a production system.  There were some tough battles
on -arch about this.  The tricky point is that the proc structure has

overwhelm the CPU when scheduling threads.  This is the major
difference.</p>

<p><strong class="speaker">julian</strong> : I greatly admire the NetBSD
way which is to take an idea and not dilute it.</p>

<p><strong class="speaker">julian</strong> : Net is also putting a Solaris
compatible threads package on top of their scheduler activations in
the Solaris ABI.</p>
</div>


<div class="discussion">

<p><strong class="speaker">jhb</strong> : Yesterday we talked about SMP
related things so I'll give a summary and then give a list of things
for 5.0.</p>

<p><strong class="speaker">jhb</strong> : The big thing for 5.0 is to
get the network stack out from under Giant.</p>

<p><strong class="speaker">jhb</strong> : Jefferey Xu and Jennifer Ying
were here to talk about this.  They have the PCBs checked in now.</p>

<p><strong class="speaker">jennifer</strong> : Interface Queues and SynCache
might be done.</p>
</div>



<div class="discussion">

<p><strong class="speaker">jhb</strong> : Aside from network the newbus
locking needs to be done (Warner Losh) and also CAM stuff.  No known
status on CAM.  Perhaps CAM is not needed for 5.0</p>

<p><strong class="speaker">jhb</strong> : Disk drive interrupts?  Would
help performance.  Going to talk to Poul Henning-Kamp</p>

<p><strong class="speaker">jhb</strong> : Alan Cox is working on the VM
system.  Working based on the old Mach stuff.  Objective for 5.0 is to
get zero fill and execute on write to work without Giant.  In future
he wants to look at locking down pmap() functions.</p>

<p><strong class="speaker">jhb</strong> : Still some stability issues.
UMA breaks some assumptions.  For instance sockets assume that once
memory is a socket its a socket forever, this is no longer true.</p>

<p><strong class="speaker">jhb</strong> : Talked to Mike Smith about
5.0 and have decided to stop adding features so that we can start
clean up 5.0 and make it a real release.  This might require hacks.</p>

<p><strong class="speaker">rwatson</strong> : For example in the UMA case
there could be a flag to just say "don't reclaim this zone" -- this
would help with issues such as the socket code assuming memory is type
stable.</p>

<p>Over to alc on the VM system.  Nothing to say.</p>

<p><strong class="speaker">bmilekic</strong> : As much as I might get hated
for this.  Will preemption stuff go in by 5.0?</p>

<p><strong class="speaker">jhb</strong> :No, that's a 6.0 thing.  There
are things to do first.</p>

<p><strong class="speaker">unknown</strong> : Could this come in in
the life time of 5.? 5.1?</p>

<p><strong class="speaker">rwatson</strong> : This is a release issue really.</p>

<p><strong class="speaker">jhb</strong> : Yes, the kernel is pre-emptive.</p>

<p><strong class="speaker">rwatson</strong> : Perhaps we should talk about
is performance goals?  What are the comparisons to make?  Perhaps head
of 4 with head of 5.  We'll see a mix.</p>

<p><strong class="speaker">jhb</strong> : I need to run benchmarks.</p>

<p><strong class="speaker">rwatson</strong> : In terms of SMP features when
will VM be ready to be measured?  I can't put a date on it.</p>

<p><strong class="speaker">alc</strong> : I think I told John was in
time for release.  I'm already doing performance testing so we've
already started.</p>

<p><strong class="speaker">rwatson</strong> : We'll pick a date to start
doing measurements.  Perhaps 2 or 3 weeks from now.</p>

<p><strong class="speaker">alc</strong> : My guess is the the locking
pmap is going to take some time to shake out.  On the other hand the
next major module we should be working on is machine dependent level.
Last we should try approaching the vmobject level.  I'll start
worrying about performance in the near term.</p>

<p><strong class="speaker">rwatson</strong> : Will threading improve
latency or throughput for networking?</p>

<p><strong class="speaker">bmilekic</strong> : I would like if we could
actually start before.</p>

<p><strong class="speaker">rwatson</strong> : Do you have a timeline for
the interrupt threading stuff?</p>

<p><strong class="speaker">bmilekic</strong> : I finished some things last
night but there are still issues.  In a couple of weeks it should be
ready for first commit.</p>

<p><strong class="speaker">rwatson</strong> : Informally beginning to
measure performance now.  What are the right sets of tests?  Need to
discuss on -arch.</p>

<p><strong class="speaker">alc</strong> : It would be nice to have that
committed to the tools directory.</p>

<p><strong class="speaker">jhb</strong> : The statistics analysis
package are we using.</p>

<p><strong class="speaker">bmilekic</strong> : I have some good success with
netpipe for overall measurement.</p>

<p><strong class="speaker">rwatson</strong> : Need to be using consistent
compilers because of the compiler change.  Also all our debugging
stuff will slow down the benchmarking.</p>
</div>


<div class="discussion">

<p><strong class="speaker">dillon</strong> : Debug stuff on 5.0.  I think
it might be reasonable then to take the space hit and always have the
debugging in but turn it on and off with sysctl.</p>

<p><strong class="speaker">rwatson</strong> : We should commit an optimized
kernel configuration and benchmarking guidlines to the tree as
well.</p>
</div>


<div class="discussion">

<p><strong class="speaker">rwatson</strong> : I think we should continue
the performance discussion.  We want to accomplish a couple of things.
One is stability measurement.  What are the things we need to be
measuring?  What is our definition of useful?</p>

<p><strong class="speaker">hsu</strong> : End to end measurement
with gigabit cards.  For latency test connections per second.  Can use
ttcp or netbench in ports.</p>

<p><strong class="speaker">gnn</strong> : need to make sure we run
against all of 4.6</p>

<p><strong class="speaker">julian</strong> : Need to really have 3 tests.
4.6 (forever) 4.x (following updates) and -CURRENT.</p>

<p><strong class="speaker">rwatson</strong> : There are other dimensions.
Degree of parallelism for instance.  We might see degradation in uni
but get good stuff in multi case.</p>

<p><strong class="speaker">julian</strong> : Test for impact of KSE
complications as well.</p>

<p><strong class="speaker">alfred</strong> : I think as the results come
through people should not be too worried about it.  Perhaps we should
benchmark database type stuff as well.  Need to do something
comprehensive.</p>

<p><strong class="speaker">obrien</strong> : What does the test matrix
look like?  Different architectures and different numbers of
processors.</p>

<p><strong class="speaker">rwatson</strong> : Can we make a multi-processor
run uni-procesor.</p>

<p><strong class="speaker">alfred</strong> : Run queue and scheduler stuff?</p>

<p><strong class="speaker">julian</strong> : Will talk to Alfred.</p>

<p><strong class="speaker">rwatson</strong> : Is scalability testing important?</p>

<p><strong class="speaker">obrienM</strong> : NFS testing.</p>

<p><strong class="speaker">rwatson</strong> : What about UI testing?</p>

<p><strong class="speaker">hsu</strong> : x11perf is the way to do that.</p>

<p><strong class="speaker">dillon</strong> : Currently we have a directory
for regression tests, should we do one for performance tests?</p>

<p><strong class="speaker">gnn</strong> : talk to sleepycat for DB
tests, see if they have some</p>

<p><strong class="speaker">alfred</strong> : Really nice to tests DB
applications that are heavily thread dependent.</p>

<p><strong class="speaker">hsu</strong> :Apache 2 has threads.</p>

<p><strong class="speaker">rwatson</strong> : What about commercial folks?
What do you do.</p>

<p><strong class="speaker">ps</strong> : Normally what we end
up doing is using the snapshot on some machines and see if the bugs
are out.  There is no performance testing really.</p>

<p><strong class="speaker">rwatson</strong> : Again, what about performance?</p>

<p><strong class="speaker">ps</strong> : We've really never had
one.  It's more just bugs.  We've just never found the performance to
be a problem.</p>

<p><strong class="speaker">rwatson</strong> : We need to create a forum for
talking about performance.  We need reproducible test cases.</p>

<p><strong class="speaker">ps</strong> : There's also other
things.  We've been doing lots of looking at this.  FreeBSD gets
kicked down by attacks for instance.  We have a lot of tools to get to
the project though.</p>

<p><strong class="speaker">rwatson</strong> : I will set up the mailing list.</p>
</div>
</div>



<div class="discussion">

<p><strong class="speaker">jhb</strong> : Questions about alpha?</p>

<p><strong class="speaker">rwatson</strong> : KSE on alpha?</p>

<p><strong class="speaker">julian</strong> : We have patches so it
compiles and runs non-KSE programs.  You can have the patched version
of the alpha kernel up and running though.</p>

<p><strong class="speaker">rwatson</strong> : Is the task owned of making
this work on Alpha?</p>

</div>


<div class="discussion">

<p><strong class="speaker">dfr</strong> : It works as far as I get to
use it.  It's not used in production right now.</p>

<p><strong class="speaker">ps</strong> : Intel shipped me a quad
processor IA64 board.  (McKinley is the name of the board).</p>

<p><strong class="speaker">rwatson</strong> : What does it need for 5.0?</p>

<p><strong class="speaker">dfr</strong> : It works, it works for SMP.
Self hosts, build worlds.  sysinstall compiles but needs more kicking
to work.</p>

<p><strong class="speaker">ps</strong> : Intel wants us to ship
a CD.</p>

<p><strong class="speaker">dfr</strong> : There is no thread support
right now (threading library needs to move to get/setcontext rather
than longjmp).</p>

<p><strong class="speaker">dfr</strong> : Need to move every driver to
use BUS DMA for large memory machines to get bounce buffers.</p>

<p><strong class="speaker">jhb</strong> : PHK is working on using a new
libwhisk so that sysinstall et al work on all systems.</p>

</div>


<div class="discussion">

<p><strong class="speaker">jake</strong> : Take control of KSE stuff
on Sparc 64.</p>

<p><strong class="speaker">rwatson</strong> : Do we have a Sparc 64 in the
cluster?</p>

<p><strong class="speaker">jake</strong> : It's not in the cluster
yet.  It's a serial cluster issue.</p>

<p><strong class="speaker">rwatson</strong> : Package building on S64?</p>

<p><strong class="speaker">jake</strong> : Perhaps a bunch of Ultra
60s for a package build.</p>

<p><strong class="speaker">obrien</strong> : 1500 build right now?</p>

<p><strong class="speaker">jake</strong> : Yes, but a lot of the
same bug in packages are broken.</p>

<p><strong class="speaker">jhb</strong> : Timeline for X?</p>

<p><strong class="speaker">jake</strong> : Not really.</p>

<p><strong class="speaker">rwatson</strong> : In terms of 5.0 how
comfortable are you?</p>

<p><strong class="speaker">jake</strong> : sysinstall is the only problem.</p>
</div>

<h3>PowerPC</h3>

<div class="discussion">

<p><strong class="speaker">benno</strong> : I got it up to
execing a fake init in the simulator and printing "hello world".
Trying to work with real hardware.  I now have some semblance of
busdma and am working on other stuff.  GEM on iMac is in an embryonic
state.  Should get to NFS mount in a few weeks.</p>

<p><strong class="speaker">rwatson</strong> : How do you feel about your
timeline?</p>

<p><strong class="speaker">benno</strong> : I'm not sure we'll have
something fully workable for 5.0.</p>

<p><strong class="speaker">rwatson</strong> : You're not at the point yet
on working on KSE are you?</p>

<p><strong class="speaker">benno</strong> : No, need a useful system
first.</p>

</div>

<h3>Adillon64</h3>

<div class="discussion">

<p><strong class="speaker">rwatson</strong> : I know that we're having
simulator problems.</p>

<p><strong class="speaker">obrien</strong> : The issues are about legal
and NDA.  Adillon decided on <a href="http://www.freebsdmall.com">FreeBSD
Mall</a> as the NDA person.  I have not had a working simulator since
September.</p>

<p><strong class="speaker">ps</strong> : I can make that happen, as
well as real hardware.</p>

<p><strong class="speaker">obrien</strong> :I've got a cross tool chain in
the tree.  I have some untested pmap stuff.  Hardware has been
available for a month or so.  We could boot FreeBSD 4.6 today if only
we had hardware.</p>

<p><strong class="speaker">rwatson</strong> : What do you think about 5.0?
Should we discuss that at another date?</p>

</div>

<h3>MIps</h3>

<div class="discussion">

<p><strong class="speaker">unknown</strong> :Juniper offered.</p>

<p><strong class="speaker">obrien</strong> : But we have no hardware.</p>

<p><strong class="speaker">unknown</strong> :Juniper thinks it's OK but
doesn't want to have it rot in the tree.</p>

<p><strong class="speaker">brooks</strong> : I have a line on a company
that does compact PCI with R6Ks.</p>

<p><strong class="speaker">rwatson</strong> : We're waiting for someone to
turn up.</p>

</div>

<a name="trust"></a>
<h2>Trusted BSD</h2>

<p><strong class="speaker">rwatson</strong> : Malc framework is what is of
interest today.</p>

<em>See Slides from Robert</em>

<div class="discussion">

<p><strong class="speaker">julian</strong> : Are the labels the same on
all structures?</p>

<p><strong class="speaker">rwatson</strong> : You can modify this but there
are issues with memory: is the space needed for a label too large to
add to an mbuf header, for example?  The label is small, but there
area lot of them?</p>

<p><strong class="speaker">bmilekic</strong> : When you're freeing the mbuf
do you write the label data?</p>

<p><strong class="speaker">rwatson</strong> : We blank it when we free it.</p>

<p><strong class="speaker">bmilekic</strong> : I do not think the 36 bytes
in the mbuf header is a problem.</p>

<p><strong class="speaker">julian</strong> : I'm more interested in the
"why" than the how.</p>

<p><strong class="speaker">rwatson</strong> : A lot of people are
interested in this.  Some of the things that do interest a lot of
people are things like doing on the fly security for a web server.</p>

<p><strong class="speaker">julian</strong> : Is there a black hatted TLA
interested?</p>

<p><strong class="speaker">rwatson</strong> : Yes and several gov'ts.  As
well as plenty of financial folks.</p>

<p><strong class="speaker">rwatson</strong> : There's a lot of userland
stuff that's not done yet.</p>
</div>
</div>

<a name="releng"></a>
<h2>Release Engineering</h2>

<p><strong class="speaker">murray</strong> : Shows a slide of releases.
4.6 is ready to go but having issues with ISO images.  DP1, a lot of
goals were met.  1000 packages were building on -CURRENT to get DP1
out.  Polished 4.2.  We need to start making decisions on 5.0.
November is still the date we're shooting for.  We're going to do a
4.7 and a 4.8.  DP3?</p>

<p>***GET samIDE FROM MURRAY***</p>

<div class="discussion">

<p><strong class="speaker">murray</strong> : Release engineering area of
the web site www.freebsd.org/releng.  For DP2 question about p4 or
CVS?  Will probably use p4 for DP2 as well.  USB subsystem?  Perl
removal?  KSE?</p>

<p><strong class="speaker">julian</strong> : KSE should be able to run
simple tests.</p>

<p><strong class="speaker">obrien</strong> : Is whatever you have
committed by DP2 be the same as the release.</p>

<p><strong class="speaker">julian</strong> : It will be a subset.</p>

<p><strong class="speaker">murray</strong> : What will the status be of
KSE in userland for 5.0?</p>

<p><strong class="speaker">julian</strong> : Can't answer that right
now. We're not removing the old libraries.  The userland work will
happen between DP2 and release.  The next step is MP as well as
UP.</p>

<p><strong class="speaker">obrien</strong> : Are we heading for a release?</p>

<p><strong class="speaker">murray</strong> : yes.</p>
 
<p><strong class="speaker">obrien</strong> : Then we have to stop having
major commits.</p>

<p><strong class="speaker">murray</strong> : Yes, the discussion today is
what are the major must have features.</p>

<p><strong class="speaker">rwatson</strong> : We need to decide if there
are major upcoming problems and reduce risk on things like KSE.</p>

<p><strong class="speaker">julian</strong> : That's why I want to get murray 3
in now.</p>

<p><strong class="speaker">rwatson</strong> : Do you think that KSE related
changes from later milestones are going to be isolated to KSE or
pervasive?</p>

<p><strong class="speaker">julian</strong> : Hard to say.  My guess is
that murray 4 stuff should be less pervasive.</p>

<p><strong class="speaker">rwatson</strong> : What happens if KSE just
doesn't work?</p>

<p><strong class="speaker">julian</strong> : Well it does work, the
patches work, it's a question of risk.  We need to check on new
things, like locking two threads in the same process.</p>

<p><strong class="speaker">dillon</strong> : KSEs only become fragile when
pthread uses them.  That's the turning point.</p>

<p><strong class="speaker">obrien</strong> : I'd like the rules for the
rest of the summer, I hope we'll talk about that.</p>

<p><strong class="speaker">murray</strong> : Earlier is better.</p>

<p><strong class="speaker">mini</strong> : I think the cutoff point for
KSE might be murray 3.</p>

<p><strong class="speaker">rwatson</strong> : It's the kind of thing where
if we need to back out we can.</p>

<p><strong class="speaker">julian</strong> : If you're not going to run
KSEs then you're OK.</p>

<p><strong class="speaker">rwatson</strong> : I think it's low risk.  Let's
avoid the risk is the message.</p>

<p><strong class="speaker">julian</strong> : The next DP2 (where we'd like
murray4).</p>

<p><strong class="speaker">alfred</strong> : We really need KSE so all
this concern about stuff that no one really uses is not a big deal.
People just need to play catch up.  We have performance problems and
we need to solve those.</p>

<p><strong class="speaker">obrien</strong> : We quickly need to figure out
our policy on multiple archs.</p>

<p><strong class="speaker">rwatson</strong> : I briefly want to respond to
Alfred.  We have asserted that KSE will be experimental.  It will be
in and 5.0 will go out but there might be issues.</p>

<p><strong class="speaker">jhb</strong> : Realistically for the network
stack is that IPv4 sockets will not be giant.  But this is only in the
network stack world.  Several people are working on it.</p>

<p><strong class="speaker">rwatson</strong> : The GEOM stuff will be
enabled by default in 5.0.  Sparc depends on it.  I do not know what
the impediments are to that though.</p>

<p><strong class="speaker">julian</strong> : The kernel stuff is there but
the user space is not.  It can't become the default until everything
is there.</p>

<p><strong class="speaker">warner</strong> : What level of control are you
going to exercise over the tree in the coming months?</p>

<p><strong class="speaker">murray</strong> : You're going to see more
level of control but we expect the requests to be reasonable.  It's a
very open process.</p>

<p><strong class="speaker">jhb</strong> : How are we going to address the 5/6 split?

<p><strong class="speaker">murray</strong> : Carefully is all I can
say.</p>

<p><strong class="speaker">rwatson</strong> : For 5. 0 we need to have a
more informed decision.  The release engineers will be trying to
reduce the number of large code changes more as time goes by.  We
don't have to wait for 5.x to be perfectly stable before we branch.</p>

<p><strong class="speaker">murray</strong> : Let's move it to more general
discussion of DP2?  Specific technologies.</p>

<p><strong class="speaker">bmilekic</strong> : Is there a strategy to lock
other protocols that are not locked down onw?</p>

<p><strong class="speaker">obrien</strong> : How much more do we need to
do before 5.0?</p>

<p><strong class="speaker">jhb</strong> : Bug fixing is what we're doing.</p>

<p><strong class="speaker">rwatson</strong> : The answer on the network
stack.  We need to choose a strategy on how to handle the other
protocols.</p>

<p><strong class="speaker">obrien</strong> : The crux is that socket
locking must be in 5.0.</p>

<p><strong class="speaker">rwatson</strong> : There are 2 or 3 problems.
Routing code is a problem.  See earlier discussions.</p>

<p><strong class="speaker">dfr</strong> : RCng is essentially done.
What it needs is testers.</p>

<p><strong class="speaker">alfred</strong> : What about libh (I think libh
is wrong but this is what I heard)?</p>

<p><strong class="speaker">jhb</strong> : It's very far along but not a
5.0 thing.</p>

<p><strong class="speaker">warner</strong> : Problems with interrupt
routing in alcPCI?</p>

<p><strong class="speaker">takawata</strong> : Cannot handle PCI PCI
interrupt routing.  Many 802.11x have this problem.</p>

<p><strong class="speaker">julian</strong> : Is it a problem from Intel?</p>

<p><strong class="speaker">takawata</strong> : This is not an Intel
problem but a problem on our side.  PCI PCI routing code should be
added.  New code is necessary.</p>



UFS2		rcNG		KSE M3			CAM SMPng

GEOM		TrustedBSD Malc	BusDMA			Newbus SMPng

C++		Cardbus		libwhisk/sysinstall	KOBJ? (no!)
				sparc64

Perl Removal	alcPI		Alpha SMP Stability	Pkgs for
							sparc64, IA64

devd		PCI intr route	document hints		release docs

							platform
</pre>

<p><strong class="speaker">unknown</strong> : Firewire?</p>

<p><strong class="speaker">rwatson</strong> : What hardware shipping on
IA64?</p>

<p><strong class="speaker">dfr</strong> : Intel stuff</p>

<p><strong class="speaker">rwatson</strong> : What about on Sparc64?</p>

<p><strong class="speaker">obrien</strong> : Very limited (hme...)</p>

<p><strong class="speaker">rwatson</strong> : KOBJ extensions discussed at
BSDCon?</p>

<p><strong class="speaker">warner</strong> : Not sure, probably not for
5.0.  Pervasive, so no.</p>

<p><strong class="speaker">rwatson</strong> : How broken is C++?</p>

<p><strong class="speaker">obrien</strong> : Only on sparc64.  Don't
really know yet, but it's probably a library issue.  The compiler is a
pre-release snapshot.  The diffs are now getting large from May 5 to
now.  We should attempt to be as far along this gcc branch as possible


<div class="discussion">

<p><strong class="speaker">gordon</strong> : Talking about rc.d stuff.
Import from NetBSD.  Right now we have patches out there that are
translated from the current boot order.  It's in perforce.  After the
conference it will go into the mainline.  Single toggle for
booting.</p>

<p><strong class="speaker">rwatson</strong> : How in sync are the bits in
the new stuff with the old stuff.</p>

<p><strong class="speaker">gordon</strong> : Last patch is from June 3rd,
but it's tracking closely.</p>

<p><strong class="speaker">rwatson</strong> : What is the schedule for
committing to the main tree.</p>

<p><strong class="speaker">gordon</strong> : We have large patches so
we're going to re-import from NetBSD.</p>

<p><strong class="speaker">rwatson</strong> : How about you have it done by
July 1?</p>

<p><strong class="speaker">gordon</strong> : We could probably do that.
Definitely want to be in DP2.</p>

<p><strong class="speaker">gshapiro</strong> : How long will we keep the old
stuff for?</p>

<p><strong class="speaker">gordon</strong> : We'll keep them both in for a
while.  Not more than 1.5 months though.</p>

<p><strong class="speaker">julian</strong> : Have you had a look at all at
the Mac OS/X startup code?</p>

<p><strong class="speaker">gordon</strong> : No.</p>

<p><strong class="speaker">julian</strong> : Do you deal with dependencies?</p>

<p><strong class="speaker">gordon</strong> : There is meta data in each
script that says what needs what.  There is a program that orders
everything correctly.</p>

<p><strong class="speaker">unknown</strong> : How does this effect the rc
script for ports install?</p>

<p><strong class="speaker">gordon</strong> : We could make this available
to ports but won't on the first version.</p>

<p><strong class="speaker">alfred</strong> : Can I recommend that you
recommend this to ports?</p>

<p><strong class="speaker">gordon</strong> : Yes, the problem is that we
have so many ports.</p>

<p><strong class="speaker">rwatson</strong> : The reason for this is for
rebundlers of FreeBSD in their environments.  We don't have to have it
for DP2 but it should be an ultimate goal.  We might need to have a
policy statement on this.  That at date X all ports must use the new


<div class="discussion">

<p><strong class="speaker">sam</strong> : I've been working on hardware
crypto.  I'm looking for consensus on getting hardware crypto in the
kernel.  This will not happen in 5.0.</p>

<h3>Syscall vector change for 64bits</h3>


<p><strong class="speaker">dillon</strong> : Two ways to go.  Need to
create a new syscall vector.  The other is to do a 1 off replacement.
Prefer the former.</p>

<p><strong class="speaker">rwatson</strong> : Perhaps we need to create a
FreeBSD 5 syscall vector.  Could be a new ABI.</p>

<p><strong class="speaker">julian</strong> : Aren't there enough other numbers?</p>

<p><strong class="speaker">rwatson</strong> : That's one way to look at it
and other platforms have done that?  Is that too heavy weight?</p>

<p><strong class="speaker">julian</strong> : It sounds that way to me.
You end up having to replicate the old ones into the new one.</p>

<p><strong class="speaker">dillon</strong> : The issue is about pollution.</p>

<p><strong class="speaker">dfr</strong> : Seems like too much work for 5.x</p>

<p><strong class="speaker">julian</strong> : It's more work.  There are
now two places.  Why not talk to OpenBSD?</p>

<p><strong class="speaker">rwatson</strong> : Should there be a BSD alfredI?
Tough to do across projects.</p>

<p><strong class="speaker">obrien</strong> : Who here is going to see that
through?  We have not talked to NetBSD about even SMP.</p>

<p><strong class="speaker">alfred</strong> : Does changing the syscall
table allow us to do clean up?</p>

<p><strong class="speaker">rwatson</strong> : We could do that without
doing 64bit syscall table.</p>

<h3>5.x ABI stability</h3>


<p><strong class="speaker">rwatson</strong> : There are new functions in
5.x.  At what point do we stop changing?</p>

<p><strong class="speaker">dfr</strong> : When people start really using it.</p>

<p><strong class="speaker">rwatson</strong> : How do we tell?  How did Solaris do it?</p>

<p><strong class="speaker">Everyone</strong> : No one knows.</p>

<p><strong class="speaker">dfr</strong> : It's too hard to add a
syscall vector.  Library issues are a problem.</p>

<p><strong class="speaker">obrien</strong> : We can use ELF to handle that.</p>

<p><strong class="speaker">dfr</strong> : Let's just add 20 new
syscalls instead of adding new work that we don't really really need.</p>

<p><strong class="speaker">rwatson</strong> : Punt on lack of time to do
this.</p>

<p><strong class="speaker">dillon</strong> : I see obrien's point with the
libraries but I have done this with time_t at 64 bits.</p>

<h3>devd</h3>


<p><strong class="speaker">rwatson</strong> : The devd stuff was to
integrate cardbus, newbus, etc.</p>

<p><strong class="speaker">julian</strong> : To monitor requests to mount
or create new devices.</p>

<p><strong class="speaker">rwatson</strong> : Is this a 5.0 requirement?
Is there anyone to do this?</p>

<!-- Which Gordon was this ? -->
<p><strong class="speaker">gordon (from IRC)</strong> : PHK has patches
that make having devd unnecessary.</p>

<p><strong class="speaker">brooks</strong> : Need something that does what
pccardd did.  </p>

<p><strong class="speaker">julian</strong> : Need to be able to do this
through a file.  </p>

<p><strong class="speaker">warner</strong> : (from IRC): That's a 6.0
feature.</p>

<p><strong class="speaker">julian</strong> : It would not be a large step
to put something in the middle to handle this.</p>

<p><strong class="speaker">julian</strong> : Sometime in the 5 lifetime we
need this.</p>

<p><strong class="speaker">warner</strong> : There is no way to monitor
events in newbus but it would be easy to add.</p>

<p><strong class="speaker">julian</strong> : I'm not sure I understood you
correctly.</p>

<p><strong class="speaker">warner</strong> : What happens now in a PCI is
that it makes a call to pci_get_devid() and the driver would say "yes
I am " or "no I'm not" so you'd have to change each of the busses to
do this but that's not too tough because we have a small # of
busses.</p>

<p><strong class="speaker">jhb</strong> : Mike Smith gave us an
informal tour of OS/X.  OS/X uses XML to do this.  They have the DEVID
in XML.</p>

<p><strong class="speaker">brooks</strong> : I looked at some PCI drivers
and some work that way but some don't.</p>

<p><strong class="speaker">julian</strong> : It seems to me we need to not
have to modify every single driver.  If you've got a device that's not
supported you ask all drivers.  At the point when you run out you make
an outcall.  The outcall returns does a substitution.</p>

<p><strong class="speaker">rwatson</strong> : Time up, time to wrap up.</p>
</div>

  &footer;

Return to bug 42512