Tuesday Feb 26, 2008

Got me a new laptop

Got me a new laptop two weeks ago - spiffy new Dell XPSM1530, dual core Intel T7500 cpu, 4gb ram, 320Gb sata disk, the ultrabright 1680x1050 screen, Intel 4965abg wireless, builtin webcam. Very nice.

Except that the builtin wired nic is a Marvell Yukon FE+. Not supported by skge, or yukonx from Marvell and while there's a patch for FreeBSD, it hasn't been ported or integrated into the myk driver that Masa Murayama wrote.

I logged 6660771 need GLDv3 driver support for Marvell Yukon FE+ in Solaris but it's not resolved yet.

Note for the unwary: when I tried the skge and yukonx drivers, I got system panics:

update_drv -v -a -i ' "pci11ab,22e" ' [skge|yukonx]

which results in a message like this:

ERROR: yukonx0: SkGeHwInnit: Currently not supported!

So being the Bright, Resourceful, Usually Correct and Exact person that I am, I emailed Masa directly asking for help.

A number of myk test iterations later and I've now got a working myk driver. Not totally sure when he's going to post the updated version to his website, but the version I've found success with is 2.6.0t9 - it's still missing a few things but it seems to be able to give me 11.mumble Mbyte/sec over my 100Mbit/sec switch to blinder (u40m2) - pretty good indeed.

I also needed to install the Opensound drivers but once PSARC/2008/043 is integrated I don't think that'll be necessary.

Now I can go off to the Sun TechDays conference next week with all the bits working together.

Thankyou Masa - you're a champ!

Sunday Oct 07, 2007

Opening up OpenSolaris just a little bit more

Get OpenSolaris

For the last few weeks I've been working with Jason King (jbk on #opensolaris) to integrate his clean-room re-implementation of libdisasm for SPARC.

Today, having received RTI approval, passed all the tests and checks and run many many nightly builds I was able to putback the changes to the ON gate. The heads up message is here.


The putback comments are

PSARC/2007/507 Unencumbered libdisasm for Sparc
6596739 need non-encumbered libdisasm for sparc
6396410 Update dis for preferred assembly language syntax
4751282 fp conversion ops decode registers incorrectly
4767086 fmovrq registers decoded wrong
4767091 pixel compare source registers decoded wrong
4767154 Registers for fmul8x16, fmul8sux16, fmul8ulx16 decoded wrong
4658958 dis misrepresents invalid opcodes
6193412 Support for new Olympus B/C instructions needed in disassemblers


I expect that there will be a few followup putbacks as people find edge cases, but the great thing about this putback is that \*you\* can make those changes if you want. You don't have to depend on Sun doing it for you :-)


Thankyou Jason - you've helped make OpenSolaris more open.

Technorati tags: OpenSolaris Jason King OpenSPARC UltraSPARC Solaris disassembler jbk libdisasm

Wednesday Jun 14, 2006

OpenSolaris... 1 year on

It's been a pretty wild year since Sun opened the lid on OpenSolaris. Well, I reckon it's been wild, at any rate :-)

As of a few minutes ago we've had the 100th community-member putback to the codebase completed. We've seen the opening of the vast majority of the ON gate (both kernel and userspace), the NWS gate (my favourite), the JDS consolidation .... lots!

There's a thriving irc channel (#opensolaris @ irc.freenode.net) which has been a fantastic support forum as well as being a meeting place, and we've got projects and sub-communities from A to Z.

I've noticed a real difference in the way that discussions (whether architectural or otherwise) now happen inside Sun, and it's a fantastic thing.... kinda like a caterpillar turning into a butterfly, if you will. There's been a visible crossover from discussions held in public, with the odd flamefest or two included.

Ok, now for the self-promotion part! I've used my blog to document a few procedures, provide some info on currently-supported Emulex FC hbas and QLogic FC hbas and to provide a link for my presentation Getting to know the SAN Stack.

And ok, sure, I blogged about other stuff too since I setup my blog early last year. But my point is that the opening of Solaris to the world has allowed Sun and non-Sun people to expand everybody's knowledge about what goes on under the hood, and it's coincided with a push from within to engage people outside Sun who might otherwise never have noticed. Even if you're not going to run Solaris or OpenSolaris, by being able to engage with Sun's staff on --- effectively --- a one-to-one basis, you have the opportunity to get new ideas. And so do we.

You win, I win, we all win.

I'm really glad to be a part of this community, both now and in the future.

Technorati tags: Solaris OpenSolaris Sun

Sunday May 28, 2006

On what "community" really means

Recently on the OpenSolaris discussion (osol-discuss) forum there has been a thread about Adobe Acrobat for Solaris x86.

One particular poster seems to find it difficult to do more than troll and be abusive, inflammatory and insulting about all the hard work that the OpenSolaris community (that's Sun people as well as non-Sun people) has done and continues to do in order to make (Open)Solaris better.

One Sun poster this afternoon asked:

Sure, and Mac OSX, Linux, Windows, \*BSD, and a host of other systems would be nice if they had DTrace, ZFS, Zones, SMF, et all...

Alas, Sun can't solve all the problems in the world, but they have been able to solve some pretty difficult ones, and they've released much of that technology to the open source world.

Have you made any contributions to Solaris yet, or tried?

A very valid question. After all, this is a community we're talking about here, and one of the benefits of being part of a community is that you get to voice your opinions. One of the concommitant responsibilities that comes with being a community member is the requirement to contribute in some way. That can be large (eg porting an app or driver, putting back code to $SRCTREE or evangelising the community) or small (logging bug reports, answering survey questions, answering questions on newsgroups or mailing lists, talking about your experiences using constructive criticism in both public and private fora) and anything in between.

Our non-Sun protagonist responded:

1) I'm not a programmer.
2) I shouldn't need to do something that is SUN's job.

Which I think totally misses the point.

As far as I can determine, our non-Sun protagonist deems everything about OpenSolaris to be Sun's responsibility. Well guess what, it's not wholly Sun's responsibility. Sure there are things which Sun can do that community members cannot (like work on the presently-closed parts of the source tree), but that doesn't mean that non-Sun people cannot contribute or should feel that they can't contribute. The OpenSolaris community is more than just Sun. Anybody who cares to can look at the discussions going on or even at the code which has been contributed by non-Sun people. There's even a page of metrics which right at the top shows that of the 13650 people registered on www.opensolaris.org, only 1400 are Sun employees.

I don't particularly care whether you're a programmer or not. What I care about is that you contribute constructively to the community, to make it better for all of us. If you want to give feedback, make it constructive. Offer to help the desktop, games, laptop, immigrants or any of the other groupings/projects/communities that make up this community as a whole.

That is what a community is --- a group of people who help each other.

Don't just whinge and irritate from the sidelines, get in there and be \*part\* of it.

Friday May 26, 2006

DTrace album cover

I just read Bryan Cantrill's blog entry on the FreeBSD DTrace port. He included a reference to the DTrace album cover artwork.

Hilarious.... just lose the red-eye (MWS deserves better) and it'll be a hit.

The one thing that worries me is that picture of Adam... when you look at this photo along with the photo in the DTrace caption contest there's room for concern. Adam, you still need to get a haircut :-D

Tuesday Apr 25, 2006

SOSUG #6 is on tomorrow night - you must come along!

Alan sent out the email earlier today: the 6th meeting of the Sydney OpenSolaris Users Group will take place tomorrow evening at the usual time and place. (See Alan's email to the group for more details).
We're incredibly privileged to have Brendan Gregg coming to talk about the DTraceToolkit and the part that he's played in helping create the second edition of Solaris Internals.
We are also incredibly privileged to have Bill Moore coming along for some q+a on ZFS and OpenSolaris development in general. Bill's actually on a well-deserved holiday so for him to take time out from that and come along is just fantastic.
So if you're in Sydney and can rock up to the iForce centre by 6:30pm tomorrow night (Wednesday 26th April), then please do.
And the other thing that I forgot to mention was that there will definitely be pizza, beer and wine provided.
If for no other reason than to show Bill that we know how to make a good red! :)

Wednesday Apr 19, 2006

Proof that Novell's COO doesn't understand Open Source

I saw a pointer to a story in The Australian's IT section today. The article is an interview with Novell's Robert Hovsepian in which he makes some statements which indicate to me that he just doesn't understand what OpenSource software is. Statements like this one:
"The whole spirit of open source is to have one base of code," he said. "Open-sourcing Solaris --- while it's appreciated that you can see the low-level pieces of code --- doesn't move the overall effort of the open-source community further down the track. It creates a fork, which none of us likes."
Huh? How is OpenSolaris a fork? Of what? Solaris? I don't think so. OpenSolaris is certainly not a fork of linux either. If you want to get historical and really pedantic, then it's possible (but disingenuous) to describe Solaris as a fork of AT&T's Unix known as SVR4. Wikipedia even has this you-beaut picture which shows the lineage. And the concept of having "one base of code" --- that's soooooo closed! So take that! OpenBSD, FreeBSD and NetBSD! Seriously, Mr Hovsepian should go back to WikiPedia and check out the description of Open Source and then perhaps he should have a wander over to The OpenSource Initiative and check a few facts about licensing as well as looking at the variety of OSI-certified OpenSource licenses and what's using them. I do appreciate his recognition that OpenSolaris' sourcecode is available for one-n-all to view. However, more than anything, I'm disappointed that the President and COO of a heavyweight in the industry that's built on Open Source software should display such ignorance. Mr Hovsepian, I know you know better than that.

Thursday Mar 30, 2006

Another entry in the "bash Sun over Java" battles

I see that one Peter Yared has penned an open letter to Jonathan Schwartz on "open sourcing" the Java JVM.
I think what Peter is getting at is that the license under which you can get the source for the JVM is not the GPL.
I'm so sick and tired of hearing people bash Sun because this company decided that the GPL was not the correct license for Java. Or for OpenSolaris for that matter.
The GPl is not the only Open Source license out there. Even RMS knows that. RMS is also enough of a realist to agree. If only the GPL zealots could pull their heads out and get a clue we could all get on with creating better applications and products instead of wasting time responding to straw man and ideological arguments.

As we were packing up the Sun stand at LinuxWorld Conference and Expo on Tuesday, a person (who I was told was Martin Fink) yelled out "When are you going to GPL Solaris?"
If only I'd been thinking quickly enough to respond: "How about you GPL your CPU first, then let's talk about the OS!"

Monday Jan 16, 2006

GPL v3 draft appears compatible in spirit with CDDL

I saw this email on the OpenSolaris Discuss mailing list saying that a draft of GPL v3 was available. In short order I found SunMink's report on the plenary session of the GPL v3 conference. What I found most interesting was the comment (and text of the draft) relating to patents. The provisions that are in the draft are really, really similar to those regarding patents in the CDDL: if you want to play in the sandpit, then you may not beat up on us with your patents. (See section 6 of CDDL). Overall, it appears to me that there is considerable agreement between this draft of the GPLv3 and the CDDL. This can only be a good thing for people who want to create, use and distribute Free and also OpenSource software.

Wednesday Dec 21, 2005

Getting started with your own CTF data

There was a <a href="http://www.opensolaris.org/jive/thread.jspa?threadID=4614&tstart=0">question</a> posed on the <a href="http://www.opensolaris.org/jive/forum.jspa?forumID=4">mdb-discuss</a> forum today, wondering how to add CTF data to kernel modules which you develop.

I gave a kinda-useless answer (basically <a href="http://cvs.opensolaris.org/source/search?q=ctfconvert">UTSL</a>;), but recalled that this question had been asked on #opensolaris a few weeks ago and I'd promised to write up a procedure.

So I found my copy of <a href="http://homepage2.nifty.com/mrym3/taiyodo/eng/index.htm">Murayama-san's</a> <tt>rh</tt> driver sourcecode, did a quick build of the <tt>fp</tt> module from the NWS consolidation, and then figured out J. Random Developer can add the CTF data in.

Firstly, grab yourself a copy of the <a href="http://www.genunix.org/mirror/SUNWonbld-20051116.i386.tar.bz2">ON build (onbld)</a> tools from <a href="http://www.genunix.org">genunix.org</a>. This package contains the utilities <tt>ctfconvert</tt> and <tt>ctfmerge</tt>. (Unfortunately they don't appear to have manpages yet).

Then with your driver, for each .o file that gets linked to make your driver, run

$ ctfconvert ---g ---l [label] objectfile.o

Then when each has been <tt>ctfconvert</tt>ed, run

$ ctfmerge ---l [label] ---o [output driver name] [list of .o files]

It's depressingly simple!

Now here's what I did for the <a href="http://homepage2.nifty.com/mrym3/taiyodo/eng/index.htm">rh</a> driver:

$ /usr/ccs/bin/make
$ cd i386
$ /ws/onnv-tools/onbld/bin/i386/ctfconvert ---g ---l RH gem.o
$ /ws/onnv-tools/onbld/bin/i386/ctfconvert ---g ---l RH rh_gem.o
$ /ws/onnv-tools/onbld/bin/i386/ctfmerge ---l RH ---o rh rh_gem.o gem.o
$ cd amd64
$ /ws/onnv-tools/onbld/bin/i386/ctfconvert ---g ---l RH gem.o
$ /ws/onnv-tools/onbld/bin/i386/ctfconvert ---g ---l RH rh_gem.o
$ /ws/onnv-tools/onbld/bin/i386/ctfmerge ---l RH ---o rh rh_gem.o gem.o

Then I did the usual <tt>make install</tt> and <tt>adddrv.sh</tt> and I could now do this in a session of <tt>mdb ---k</tt>:

> ::status
debugging live kernel (64-bit) on doppio
operating system: 5.11 onnv-gate:2005-12-11 (i86pc)
> ::modinfo ! grep rh
232 fffffffff4feed58     adf8   1 rh (via rhine nic driver v1.0.24)
> ::print ---t struct gem_stats
    uint32_t intr
    uint32_t crc
    uint32_t errrcv
    uint32_t overflow
    uint32_t frame
    uint32_t missed
    uint32_t runt
    uint32_t frame_too_long
    uint32_t norcvbuf
    uint32_t collisions
    uint32_t first_coll
    uint32_t multi_coll
    uint32_t excoll
    uint32_t nocarrier
    uint32_t defer
    uint32_t errxmt
    uint32_t underflow
    uint32_t xmtlatecoll

And even better, with a development version of Solaris CAT I can use the <tt>stype</tt> command:

SolarisCAT(live/11X)> stype gem_stats
struct gem_stats { (size: 0x48 bytes)
   typedef uint32_t = unsigned intr; (offset 0x0 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned crc; (offset 0x4 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned errrcv; (offset 0x8 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned overflow; (offset 0xc bytes, size 0x4 bytes)
   typedef uint32_t = unsigned frame; (offset 0x10 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned missed; (offset 0x14 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned runt; (offset 0x18 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned frame_too_long; (offset 0x1c bytes, size 0x4 bytes)
   typedef uint32_t = unsigned norcvbuf; (offset 0x20 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned collisions; (offset 0x24 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned first_coll; (offset 0x28 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned multi_coll; (offset 0x2c bytes, size 0x4 bytes)
   typedef uint32_t = unsigned excoll; (offset 0x30 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned nocarrier; (offset 0x34 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned defer; (offset 0x38 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned errxmt; (offset 0x3c bytes, size 0x4 bytes)
   typedef uint32_t = unsigned underflow; (offset 0x40 bytes, size 0x4 bytes)
   typedef uint32_t = unsigned xmtlatecoll; (offset 0x44 bytes, size 0x4 bytes)
} ;

Monday Nov 21, 2005

Schillix 0.3 is here!

I just saw Joerg's sort-of announcement on the #opensolaris IRC channel:
(05:55:13) schily: I did just release SchilliX-0.3 ... (05:56:15) schily: ftp://ftp.berlios.de/pub/schillix/ ... (05:57:05) schily: we are really close to a self hosting platform now!
Having a non-Sun-supplied distribution get to the oh-so-close-to-self-hosting stage is fantastic news. With Joerg's expertise and skill I'm sure it won't be too long now before SchilliX is at that point. Self-hosting (bootstrapping the bootstrap?) shows that it's quite possible to produce a distribution without Sun's help (except for the closed binaries) --- a confidence boost if ever I could think of one. Well done Joerg! The announcement (is it official?) is to be found here in the OpenSolaris Discussions area.

Monday Oct 24, 2005

Zero hour approaches

I was doing my best to catch up with the mail traffic from the ZFS team, and I noticed an interesting term being used: "z day" I guess it's a lot like Zero Hour, but for ZFS. If you're still skeptical (sceptical?) about whether or not ZFS actually exists, I'd ask you to be patient for a little bit longer.... Not long now, I promise! The impending public release of ZFS is something which is dear to my heart, and something I really really really want to demo at a SOSUG meeting RSN. And then there's the NWS or SAN consolidation (see the roadmap) which is also imminent. Since that's the team that I'm a part of I think I'll reserve for myself the task of presenting something about it at a SOSUG meeting sometime soonish. There's nothing quite like talking about your field of endeavour :-D OpenSolaris: Innovation MattersOpenSolaris User Group

Sunday Oct 16, 2005

Brief report on last night's SOSUG meeting

Last night's SOSUG was a roaring success. We had a fairly full room --- about 30 people --- come along to hear Bryan Cantrill talk about some of the new stuff he's doing with DTrace. Oh, and a few people were interested in hearing me talk about ZFS too. :-) The slideware for the talk is here if you are interested. Jeff Bonwick reckoned that I wouldn't be able to give that talk in under an hour, and he was right. (At StorageNetworkingWorld last month I didn't take questions, skipped a few slides and didn't breathe in order to deliver it in 15 minutes. Urk!) It was great being able deliver to a group of people who are the ones going to see a direct benefit to their daily lives if they migrate to ZFS. I said to one attendee that if he migrated his environment to ZFS then he'd be able to spend more of his time on the interesting stuff that he wants to do, rather than the tedious storage admin that he has to do at the moment with SVM, VxVM, UFS and VxFS. And when we got to the admin commands example (slides 20-23) everybody was able to see that it was so. I even got some questions from Brendan Gregg which sounded like Dorothy Dixers but which really, really weren't --- and I was able to say "Wait a sec, that's all on the next/second next/... slide" and it was and he was happy! Enough about me... Bryan Cantrill was the real star of the evening. He's out here for a few days to present a tutorial and plenary address at AUUG's 2005 Conference. So of course he's being dragged to customer meetings all over the country while we've got him. He flew from LAX direct to Brisbane, did a few meetings there, hopped on a plane to come to Sydney, came and presented at SOSUG and then tomorrow afternoon is off to Melbourne for some more meetings. Then I think on Friday afternoon he gets to fly home. So anyway, back to his presentation. Since we had about 50% of the room having actually used DTrace, he didn't give us the fairly stock/standard presentation. He showed us some RFEs that he's implemented, not all of which have been integrated. Some of them are doozies: new options for sorting the output of aggregations, new keywords and predicates for getting to userland functions, the "-v" option to show stability levels (see attributes(5)) and a few other things too. I figure Bryan was somewhat jetlagged because he didn't talk at a million miles an hour.... but I reckon that every single attendee last night went away impressed with not only (cue deep off-screen voice) The Awesome Power Of DTrace but also with Bryan's skill as a presenter. We had two video cameras going, so once Boyd gets through with transferring the tapes (5 dats!) to disk and putting all the bits together, I'll post a followup with the link on mediacast so you can see Bryan in full flight. It's a joy to behold :-D

The NWS Consolidation is coming soon to OpenSolaris

The group I work in is known as SAN Engineering Product Development, which is part of the Data Management Group (the division formerly-known-as Network Storage or NWS). One of the first questions I asked when I started was "Are we going to place the NWS consolidation under CDDL and release it to the OpenSolaris community?" The answer was "yes, except for the bits which we've licensed from 3rd parties and can't release" which is exactly what I'd hoped would be the case. Yesterday I saw an email on our team mailing list saying that progress on this is going ahead at full speed, and we're getting very close to a release. If you wander over to the roadmap document, then we're on track with the delivery target. I don't know about you, but I'm excited --- we're going to have a community for the consolidation, and you can join it / subscribe to it just like any of the other communities that are under the OpenSolaris umbrella. Bring it on! OpenSolaris: Innovation MattersOpenSolaris User Group

Sunday Oct 09, 2005

Apparently, GNU/OpenSolaris is only a month away

According to this message on a OpenSolaris discussion forum, a GNU/OpenSolaris distribution is only about a month away. There's even a screenshot to prove it's not a pipedream. I figure there are more than a few people out there who'd love to use the (Open)Solaris kernel with a consistent GNU environment wrapped around it. While I probably wouldn't use it myself because my work is tied up pretty deeply with our SAN stack implementation and I want a consistent san(d)box to play in, I reckon it would be a really good thing to have. Bring it on!

I work at Oracle in the Solaris group. The opinions expressed here are entirely my own, and neither Oracle nor any other party necessarily agrees with them.


« April 2014