Status

Project Status Page

In lieu of a blog we will go a simple route here just posting notes with their associated dates. Jump to the bottom of the page to see the latest boot log output from the target.

We have completed process of pushing out code to the ppc-dev svn repository and the internal Sunlabs repository is now deprecated even though it still functional. While we could drop tarballs it doesn't seem to provide much value but can if requested.

Current tasks in process

1. Establish outward facing PPC development server vespa2 - done
2. Establish svn access with both pp-dev and labs polaris-trunk from vespa2 - done
3. Build vespa2:$HOME/usr/src/uts - done
4. Build vespa2:/$HOME/usr/src/lib - done - TODO: add rules for automated genassym and assym.h building.
5. Verify ODW boot and statically linked userland shell - the last development point when the Labs exited the project. - done
6. Rebuild all tools onbld, gcc, cmd/sgs and verify functionality based on all ppc-dev bits - in process
7. Update website and associated documentation - ongoing basis
8. Move druidia and 2nd ODW target to external site - pending
9. Fix lost MAC address bug to allow other 2 ODW targets to boot on VOF - pending
10. Sync the ppc-dev source to the latest OpenSolaris repository - pending
11. Move the current repo control from subversion to Mercurial - pending

Vespa2 as a development system is fully functional

We have succeeded in recreating and then moving one of the SunLabs development systems to be publicly available. A serious developer can now log into vespa2.sunlabs.com, pull the ppc-dev source, build it, download their kernel, boot it and debug it on a target Pegasos ODW system there. The statically linked init is a demonstration of a userland program that forks and execs, executes in from user space and returns to the kernel.

Vespa2 is now externally accessible

It is the first of two standalone development servers have been turned outward facing and is externally accessible by the community. These systems were the original project hardware used by the Pulsar Project and have been generously donated by SunLabs. They will be maintained at the Sun Menlo Park Bldg 16 facility but are not part of the Sun network or any project. This has not been a simple task and some details remain. Pictures are always good so here is the stuff in action. vespa2 shown on the bottom with an ODW board above. There is a network controlled powerstrip that can be used to recycle power and provide reset to the ODW target system.

jpg](http://www.opensolaris.org/os/project/ppc-dev/status/labdev4_small.jpg "The externally available ppc-dev system")

Now Have 2 Development Systems and 3 Targets -- 02/04/2009

We now have 2 available solaris build and boot servers for our odw targets. They are vespa and vespa2. Additionally we have drudia, a linux server which functions as a boot server for the third odw target. Anyone wanting to use druidia would build their bins on either vespa or vespa2 then scp over to druduia and execute on the odw served from drudia. We will be placing them on the external network shortly and they will be generally available for anyone who would like to do development.

Development Systems Up -- 01/25/2009

We have emerged from a self imposed hibernation. It seems that the folks really interested in getting a PPC port have returned and are now putting in the effort to make further progress.

Sunlabs has agreed to host the previous development systems and offer remote access. There will be two individual servers each with a ODW target and power / reset control. One setup will have a BDI debugger.

The svn repo has been updated with the final work done by Guy Shaw, this wraps up any direct work by SunLabs. There is a single user prompt in a statically linked init.

Updated SVN repository -- 02/18/2008

Now at rev 111. Major items to note

Bug fixes for missing BRL instruction in a number of critical assembly files
Implement lwp_pcb_exit()

Updated SVN repository -- 11/02/2007

Now at rev 97. Major items to note

We have sync'd with onnv bld 58 --hmmm aren't we a bit behind

Well this is the bulk of the commit, we have not done a merge since earlier this year. Why? It's a messy thing being on a parallel path here, but this will take us from 57 to 58. We will now be spending time on playing catchup to get back in sync with the releases. Guy does this by himself. There is alot more work to do now with each one of these since we are successfully building much more of the consolidation. The lack of accounting for a 3rd architecture can be painfully seen with each one of these but we grind away.

Updates to sgs. We can now build rtld (ld.so.1)

Updated SVN repository -- 10/29/2007

It is now at rev 95. Many updates to lib/ and a number of shared objects are building. We have yet to actually test any of this on the target since we need a functional rtld, which should be here soon.

You will need to build the libs with make -k install. We have a problem with building lint related items and that will get you temporarily past the issue.

To build you will also need the updated toolchain which has a reconfigured gcc which now handles c++.

Once you have the right shared objects you should also be able to build cmd/init and cmd/sh.

Sync Up with Sun Labs Source Development Complete -- 10/11/2007

Ok, we're now at rev 93 and the repo is in sync with the Sun Labs development. All of the latest code is out. It should build cleanly, just follow the instructions. If you have any questions, feedback, suggestions or want to jump in post to the discussion list.

We're in the process of updating the task pages and overall documentation on the project itself. Please take the time to review and provide feedback or information updating.

Successful Bulding of SVN repository -- 10/09/2007

The ppc-dev repo is now at rev 89 this is around the mid Aug 07 development timeline so we're now only a few months behind. Lots of ddi and interrupt development was going on. The additional good news here is that the repo is building on the outside by community members.

Updated SVN repository -- 09/28/2007

It is now at rev 82 this is now around the June 07 development timeline so the ppc-dev code base is catching up. This putback has updates to the VM, HAT layer and PMDB. A few more drivers completes the PPC driver build dirs and their associated Makefiles.

Updated SVN repository -- 09/27/2007

It is now at rev 80. This putback generally completes adding the PPC driver build dirs and their associated Makefiles. It also has the start of the CBE work along with update pci nexus and genral interrupt work. There are updates to the tool set that Guy has created, see the work in onbld. Added a bit of support for the EFIKA platform, but ODW continues to be the platform we develop on.

You may notice in the log alot of comments to code regarding double mappings. This was a problem we fought with for a while before finding an oversight related to the ODW firmware and active BAT regs that we thought were disabled.

Updated SVN repository -- 09/25/2007

It is now at rev 79. This putback includes a large number of the PPC driver build dirs and Makefiles. It also has the start of the rootnexus and ddi interface.

Updated SVN repository -- 09/20/2007

It is now at rev 78. At this point the interrupt framework and the cbe was just being added. The PMDB (Poor Mans Debugger) was updated and also some VM related work was added. The RHINE ethernet chip driver was added here.

Updated SVN repository -- 09/11/2007

It is now at rev 72. This is about half way through the merge of the ppc-dev tree with bld 57. Best to wait till the whole merge is pushed out before attempting to build and test. Note you will need to load ands run VOF on top of the ODW's firmware. A binary has been provided.

Updated SVN repository -- 08/28/2007

It is now at rev 62. This includes the beginning merge of the ppc-dev tree with bld 56 which is around the Jan 07 timeframe. Notable commits are related to mdb and other kernel boot funcs.

Updated SVN repository -- 07/23/2007

It is now at rev 38. This is up to the merge of the tree with bld 54. You will have to use the updated toolchain GNU/GCC/AS/solaris ld from now on to build. Get it here

Updated SVN repository -- 06/22/2007

It is now at rev 24. These updates were still related to general work that continued in the kernel bringup but you may also notice the cmd/sgs ld work that's was going on.

More SVN Updates -- 06/20/2007

Gotta love svn. The merge command works well on files that are already in place. but won't create new files. We've been using svn cat to create the new files, do an svn add, then go for the merge.

Created SVN repository -- 06/10/2007

Populated it with the source from the initial code release. As usual doing something for the 1st time always takes a bit longer to get the hang of it. The SCM management support wasn't available back at the time of the 1st source drop so the tarball was essentially it. You can access according to the std instructions. For example you would have svn+ssh://user@svn.opensolaris.org/svn/ppc-dev/ppc-dev. You also need to be affiliated with the project and have a public key available on your profile.

Updated GNU Assembler -- 11/27/2006

The GNU ld was covering up the incorrect symbol relocations, fancy that. The new one is here /opt/polaris/powerpc-solaris2/bin/as. The difference is that the new as should generate the proper relocations for unaligned 32 and 16 bit address relocations. With this new as we can go back to compiling with -g -gdwarf-2.

Merger with build 54 complete -- 11/22/2006

onnv-gate build 54 (2006-Nov-22) and the ppc-dev source base is complete. This is not a simple job, many files must be hand merged after the bulk is integrated.

Active topics -- 11/20/2006

1. Figured out problems with inetboot. Restricted reads to 1K as the ethernet driver in the OF drops back to back packets. Also randomized the initial port number so that back to back reboots won't have to wait for 4 minutes for the previous connections to time out.
2. Cross built ld. Required libconv, libelf, libld, liblddbg as well. Most of usr/src/cmd/sgs builds, but not all. Skipped over TLS (thread local storage) stuff for now. Also did nothing for 64-bit. Main issues surrounded no cross building was envisioned or testing. Byte flipping issues abound.
3. kernel and boot both can be properly built by the Solaris ld. We've switched over to the Solaris ld totally now and using the Solaris kernel mapping layout. Still cannot run at 0xe0000000 as OF has problems dealing with non real mode addresses.
4. Got krtld working. Apparently fully functional. Issues w/byte flipping in doreloc() and addend values were both in the addend field of the relocation and left at the memory location referred to by the relocation.
5. Worked on and analyzed issues with usr/src/tools/ctf
6. Started working on mdb/kmdb. Makefile tree in place.
7. Created a work around solution for genassym/assym.h until the ctf tools are functional. Compile the offsets (data table only) with the GNU gcc cross compiler into an assembler file. Use that assembler file as input into Sun x86 c compiler. Relies on the fact that the x86 and PowerPC assemblers are very similar.

Move to solaris ld -- 11/09/2006

It is available in /opt/onbld/usr/ccs/bin/ld. There are also the libraries which are in /opt/onbld/lib/. In order to run the new linker you must set in your environment LD=/opt/onbld/usr/ccs/bin/ld then export LD.

But now we run into some very strange issues with symbol unaligned relocations. Since we use the Metrowerks PowerTAP and Codewarrior for source level debuging we need to have DWARF symbols even though we can build successfully without them. The solaris ld complains, but the GNU ld never did.

(for sh and bsh at least).

KRTLD update -- 11/07/2006

krtld is running.
inetboot loads /platform/chrp/kernel/unix.
inetboot reads unix and observes that misc/krtld
is required. inetboot reads in /kernel/misc/krtld.
Control is transferred to krtld. krtld relocates
itself and then reads unix for its symbols and
well as reading that genunix is needed. krtld
then reads in genunix and relocates it and
binds everything together. Control is then
transferred to unix which starts execution.

Initial project contribution -- 10/02/2006

This was quite a bit of work just to get to this point but now there is a working development system available to all who would like to particpate in any area of this project. There is no closed code, all of the platform source is open. The build is sync with ON a bit past BLD_47. The Polaris task map was updated also.

In simple terms the kernel boots and outputs to the console to a point and then will panic. This is currently during the VM initialization. Development continues not only on the VM/HAT but in parallel on KRTLD, RTLD. At this time however we statically link unix/genunix to facilitate debugging. The task map is the best way to obtain details on development status of each area.

--

Important target notes for now 11/1/08

you must add these lines to your asy.conf file in your $ROOT/kernel/drv

You must enter in the following once the ODW boots and your at the smartfirmware prompt. This will route the ethernet irq to a non shared number.

The current status is that we are at the point of running our first code from userland. As you can see there is a bit more work to do but progress has been made. In this run we have replaced the "usual" init with a application that is statically linked that simply does an open of the console device, a write of
"Hello, world!" and then exits.

> elfexec

elfexec: level = 0
> elfexec -> getelfhead
< elfexec -> getelfhead
> elfexec -> getelfphdr
< elfexec -> getelfphdr
exec_args: size = 4096
hat_unload_callback: vaddr=dfffd000, len=2000
hat_unload_callback: cb = NULL
ppcmmu_alloc: > vmem_alloc
ppcmmu_alloc: < vmem_alloc
ppcmmu_setup: not kernel hat
ppcmmu_setup: > mmu_segload
ppcmmu_setup: < mmu_segload
> exec_args -> stk_copyout

\-> grow
\-> grow_internal: sp=dfffefc0, growszc=0
<- grow_internal: newsize=1000
as_fault: seg->s_ops->fault = e0204ff4 = unix:segvn_fault+0
<- grow
< exec_args -> stk_copyout
as_fault: seg->s_ops->fault = e0204ff4 = unix:segvn_fault+0
< elfexec
> exec_init -- rc=0

Hello, world! <--This is what we have been waiting for!!!

WARNING: init(1M) exited with status 0: restarting automatically

To browse the latest project source just go here Note that the cvs viewer appears to be unable to view the recent putbacks on the svn repo.

01/26/2009 target boot log

--
Quite a bit of progress had been made since the October release, see below

CONNECT THE SERIAL PORT TO THE TARGET

RESET THE ODW TARGET BY POWER CYCLING OR USING THE BDI-2000 "HALT-RESET-GO"

AFTER LOADING KRTLD AND THE KERNEL HERE IS THE FIRST OUTPUT

LIKE IT SAID IT IS IN MAIN

SO ABOVE IS THE MENU OF THE PMDB BUT WE JUST RUN THE MACRO "M1 GO"

THE KERNEL BANNER: WE ARE ALIVE

NOW WE ARE STARTING INIT - BUT WE HAVE A STATICALLY LINKED USERLAND SHELL

INSTEAD AND HAVE RENAMED IT FOR INIT AND IT GIVES US A PROMPT

SILLY ME I TRIED TO EXECUTE "HELLO" BUT THAT IS NOT THE CORRECT PROGRAM NAME

ONCE AGAIN I MISTYPD THE PROGRAM NAME

HERE WE FORCE AN EXCEPTION AND IT DUMPS US BACK INTO PMDB

thread = e48587e0:
lwp = e4d5e3a0
lwp errno = 0
t_sysnum = 8e
t_astflag = 1
t_sig_check = 0
traceback: %fp = e4f95c90
traceback: fp=0xe4f95c90, nextfp=0xe4f95d30
e4f95c90 e007d8c8 unix:panic+88
traceback: fp=0xe4f95d30, nextfp=0xe4f95d50
e4f95d30 e0051800 unix:unimplemented_fn+38
traceback: fp=0xe4f95d50, nextfp=0xe4f95d60
e4f95d50 e0051d08 unix:lwp_stk_fini+20
traceback: fp=0xe4f95d60, nextfp=0xe4f95e20
e4f95d60 e0185e04 unix:lwp_create+49c
traceback: fp=0xe4f95e20, nextfp=0xe4f95e80
e4f95e20 e0189054 unix:forklwp+154
traceback: fp=0xe4f95e80, nextfp=0xe4f95f20
e4f95e80 e0153ab0 unix:cfork+708
traceback: fp=0xe4f95f20, nextfp=0xe4f95f50
e4f95f20 e01532fc unix:forksys+6c
traceback: fp=0xe4f95f50, nextfp=0x0
e4f95f50 e00023dc unix:sys_call+f0
end of traceback
MMU state:
--
IBAT0={0, 0}
IBAT1={0, 0}
IBAT2={0, 0}
IBAT3={0, 0}
DBAT0={0, 0}
DBAT1={0, 0}
DBAT2={0, 0}
DBAT3={0, 0}
Segment registers:

00: 20000040={T=0,Ks=0,Kp=1,N=0,VSID=40} *
01: 20000041={T=0,Ks=0,Kp=1,N=0,VSID=41} *
02: 20000042={T=0,Ks=0,Kp=1,N=0,VSID=42} *
03: 20000043={T=0,Ks=0,Kp=1,N=0,VSID=43} *
04: 20000044={T=0,Ks=0,Kp=1,N=0,VSID=44} *
05: 20000045={T=0,Ks=0,Kp=1,N=0,VSID=45} *
06: 20000046={T=0,Ks=0,Kp=1,N=0,VSID=46} *
07: 20000047={T=0,Ks=0,Kp=1,N=0,VSID=47} *
08: 20000048={T=0,Ks=0,Kp=1,N=0,VSID=48} *
09: 20000049={T=0,Ks=0,Kp=1,N=0,VSID=49} *
10: 2000004a={T=0,Ks=0,Kp=1,N=0,VSID=4a} *
11: 2000004b={T=0,Ks=0,Kp=1,N=0,VSID=4b} *
12: 2000004c={T=0,Ks=0,Kp=1,N=0,VSID=4c} *
13: 2000004d={T=0,Ks=0,Kp=1,N=0,VSID=4d} *
14: 2000001e={T=0,Ks=0,Kp=1,N=0,VSID=1e}
15: 2000001f={T=0,Ks=0,Kp=1,N=0,VSID=1f}
SDR1=fc0003f={HTABORG=fc00000=252M,HTABMASK=3f} ptsize=4M
HID0=0x8410c09c
=8410c09c HID1=0x2c80 =2c80

Statistics:
mutex_enter: enter=519192, owned=2227, lost_resv=0
mutex_tryenter: enter=35281, owned=0, lost_resv=0
atomic: interference=0, lost_resv=0, ov32=0, ovfld=0
PMDB configuration:
Startup:
Unit test (ut) : 0 Copy handlers (ch) : 1 Enable Interrupts (ei) : 1 Read-only ktext (ro) : 1 Run configure (config) : 1 Run ttyinit (ttyinit) : 1 Debug init (dbinit) : 1 Run demo (demo) : 0 Debug: Trap debug (td) : 0 CBE debug (cbe) : 0 Genassym (ga) : 0 MMU Flight Rec (mfr) : 0 Int Flight Rec (ifr) : 0 Panic Watch (pwatch) : 0 kmem~_flags (kmf) : d PMDB = Poor Man's DeBugger Commands are: mmufr Show MMU flight recorder mod Show modules vmem Show vmems: heap, module {text, data, ctf} dump Show virtual memory (takes address and 32-bit count) trans Show PROM virtual to physical memory translations ptrans Show translations from pagetable ptstat Show pagetable statistics ptegp Show entire PTEG pair for a given virtual address devs Show device tree intr Show information about interrupts reboot Reboot the target go Exit debugger and continue running Solaris pmdb panic:~.~.~. Unknown command, ~.~.~. Type '?' for help on PMDB commands. pmdb panic:~{{code}}

{{/code}}======> VOF is now loaded

> Use this simple VOF command that will load a kernel from your

> bootserver nfsroot proto/root

> but first let's get inetboot from tftpboot

> Got KRTLD, let's get rolling

> See where you have loaded the kernel

> Here are the first statements from the kernel

Here are the loaded modules

id text-va text-sz name file-name flags
------------
0 e0001000 3c8a30 unix /platform/chrp/kernel/unix
1 e03c9a30 2a2dc krtld misc/krtld
2 0 0 cl_bootstrap misc/cl_bootstrap
3 e03f5000 8c1c specfs fs/specfs
4 0 0 dtrace dtrace
5 e03fdc1c 918c devfs fs/devfs
6 e0406da8 6294 TS sched/TS
7 e040d03c 16c TS_DPTBL sched/TS_DPTBL
8 e040d1a8 22a4 swapgeneric misc/swapgeneric
9 e040f44c c96e8 nfs fs/nfs
10 e04d8b34 28bd0 rpcmod strmod/rpcmod
11 e0501704 2d44 tlimod misc/tlimod
12 e0504448 8b2c rpcsec misc/rpcsec
13 e050cf74 135c4 vfe drv/vfe
14 e0520538 4c8c mac misc/mac
15 e05251c4 209f00 ip drv/ip
16 e072f0c4 4de4 md5 misc/md5
17 e0733ea8 2e29c kcf misc/kcf
18 e0762144 1e44 swrand crypto/swrand
19 e0763f88 208c hook misc/hook
20 e0766014 160c neti misc/neti
21 e0767620 3c7c pci drv/pci
22 e076b29c 889c rootnex drv/rootnex
23 e0773b38 75fc nfs_dlboot misc/nfs_dlboot
24 e077b134 3234 strplumb misc/strplumb
25 e077e368 90c mac_ether mac/mac_ether
26 e077ec74 a4d4 dld drv/dld
27 e0789148 8154 dls misc/dls
28 e079129c 434 ip6 drv/ip6
29 e07916d0 44c tcp drv/tcp
30 e0791b1c 45c tcp6 drv/tcp6
31 e0791f78 4a4 udp drv/udp
32 e079241c 454 udp6 drv/udp6
33 e0792870 42c sctp drv/sctp
34 e0792c9c 434 sctp6 drv/sctp6
35 e07930d0 d530 icmp drv/icmp
36 e07a0600 44c icmp6 drv/icmp6
37 e07a0a4c a0cc arp drv/arp
38 e07aab18 6cec timod strmod/timod
39 e07b1804 234 options drv/options
40 e07b1a38 f94 pseudo drv/pseudo
41 e07b29cc 89c clone drv/clone
42 e07b3268 1734 sad drv/sad
43 e07b499c 18e84 dev fs/dev
44 e07cd820 4004 ctfs fs/ctfs
45 0 0 procfs fs/procfs
46 e07d1824 b4e4 tmpfs fs/tmpfs
47 e07dcd08 2a4c objfs fs/objfs
48 e07df754 1e4 consconfig misc/consconfig
49 e07df938 481c consconfig_dacf dacf/consconfig_dacf
50 e07e4154 410c conskbd drv/conskbd
51 e07e8260 55cc kbtrans misc/kbtrans
52 e07ed82c 24e4 consms drv/consms
53 e07efd10 213c wc drv/wc
54 e07f1e4c 68c4 tem misc/tem
55 e07f8710 129c iwscn drv/iwscn
56 e07f99ac 3234 isa drv/isa
57 e07fcbe0 3434 busra misc/busra
58 e0800014 b914 asy drv/asy
59 e080b928 c924 kssl drv/kssl
60 e081824c 67b4 elfexec exec/elfexec
61 e081ea00 b64 cn drv/cn`