State of the PA-RISC/HPPA port (August 2013)

Hello everyone,

this is just a short update on the status of the parisc port from my point of view, including a short summary of what have been done during the last few weeks and what is planned for the near future.

So, what's the status of the Linux kernel for parisc?

Basically, I think all major outstanding kernel patches are now in the upcoming kernel 3.11 tree, and the important ones will show up in the stable kernel 3.10 series soon as well. Kernel 3.9 should be OK as well, but some patches might be missing.

With kernel 3.11 (and 3.10.5 or higher) I expect most instability issues from the past to be gone. There might still be minor issues with userspace segfaults (due to cache issues), but they are rare....

Looking back, the main problems of instability and kernel crashes in the past were assumely due to
a) cache and TLB flushing were unreliable,
b) kernel stack overflows (because of sudden irq processing which used too much stack), and
c) miscompiled code in pa_memcpy() because of C-code which is not valid C...

Testing on my machines showed, that the kernels 3.10.x and 3.11 will boot in UP and SMP mode on nearly all machines. Most notably, we now have support for the C8000 workstation, for which serial ports and framebuffer mode with FireGL cards should work (Thank you Thomas!). Even a patch for the BMC in the C8000 will probably be included latest with kernel 3.12.

Debian:
We have our own debian package repository at
http://ftp.parisc-linux.org/debian-ports/ which currently holds more than 8300 pre-built debian packages (Thank you Dave!!!). Even pre-built debian 3.10 kernel images are now available there. We don't yet have a real debian-installation media yet.
I've prepared a bootable image and some instructions on how to install debian-unstable from our repository, but you should only try this if you are a linux/parisc expert. Brave people will find those instructions here:
http://wiki.parisc-linux.org/Debian_Ports_Installation For all others I'd suggest to wait until we have the debian-installer finished, which is still one of my top-priorities.

Furthermore, we sadly still don't have buildd-server. What really is missing here is someone who is willing to host a parisc server and provide enough direct bandwith to the internet. Giving away real server hardware is probably doable (and those parisc machines luckily aren't expensive any longer on ebay :-)) and even administration would be done by Dave, me or someone else if necessary.
On the other side, Dave has the buildd/wanna-build software installed on his machine. The wanna-build software needs a custom trigger script to handle running on a non local machine. It is needed to populate the source and packages database. Maybe, somebody has already done this and may give him a hint?

Gentoo:
If you don't want to wait for debian, I'd suggest that you try Gentoo instead. For instructions please see their website at
http://www.gentoo.org/main/en/where.xml (architecture hppa).

PARISC website:
Our current parisc website at http://www.parisc-linux.org is still based on CVS, and it's sometimes unreliable (means: sometimes it's not reachable). To solve this issue my plan is to migrate it's content in september to the wiki at kernel.org: https://parisc.wiki.kernel.org/
If someone is interested in doing that - and maybe has some interest and knowledge in how nice web sites should look like, please let me know!

TODO:
There is still much on my todo list, most importantly
- keep pushing kernel patches upstream (https://patchwork.kernel.org/project/linux-parisc/list/)
- build a debian-installer ISO, and
- migrate the website.
For all other things I keep a personal TODO-list at
http://wiki.parisc-linux.org/TODO. Again, volunteers for some of those topics are always welcome!

Last but not least I want to say a thank-you to all people who have contributed to this project during the last weeks. The following list of people is taken purely from the kernel git tree and shows those who have directly or indirectly contributed to parisc kernel source tree. Same thank-you goes of course to all other here unnamed people who have tested stuff, reported issues, fixed userspace or kept on the discussion.

THANK YOU!

Helge

List of contributors to the kernel tree:
----------------------------------------
Alex Ivanov
Al Viro
Bill Pemberton
Bjorn Helgaas
Chen Gang
Cong Wang
Damian Hobson-Garcia
David Howells
Frederic Weisbecker
Geert Uytterhoeven
Guy Martin
Helge Deller
H. Peter Anvin
James Bottomley
Jiang Liu
Jiri Slaby
Joe Perches
John David Anglin
John Stultz
Julia Lawall
Kautuk Consul
Kees Cook
Kyle McMartin
Larry Finger
Mathias Krause
Matt Fleming
Michael S. Tsirkin
Michel Lespinasse
Mike Frysinger
Myron Stowe
Nikitas Angelinas
Paul Bolle
Paul Gortmaker
Peter Huewe
Peter Zijlstra
Rolf Eike Beer
Srivatsa S. Bhat
Thomas Bogendoerfer
Thomas Gleixner
Torben Hohn
Wei Yongjun
Will Deacon
Yinghai Lu
Yong Zhang
Zhao Hongjiang

Thanks for all the work Helge!

On the glibc front I don't have as good news. We shipped 2.18 without all of the hppa patches, but keep a patchset up to date, and I'm merging in some changes from Dave.

The glibc testsuite results are really bad, but hopefully I'll poke at those over the coming year.

I'm using gentoo, and taking patches form the gentoo people and feeding those upstream as they are reported (I still have a few to go through).

Cheers,
Carlos.

Thanks to all of you working on that.

I currently do not have problems with glibc, at least I don't see any (Gentoo, using their 2.14 version). What I see are problems in binutils, although AFAICT those are not really HPPA-specific. One easy test is to build a recent version of CMake: if you specify no compiler flags bootstrapping will work, as I inserted -Wl,--unique=.text.* into the LDFLAGS there. But if you pass -O2 if will not build (binutils 2.23.[12]). If you remove the linker flag it will not work without -O2. I can't build newer kernels for my C8000 with frame pointers as it will crash ld with a NULL deref somewhere, older kernels work (probably smaller).

But otherwise my machine became pretty stable in the last time, that is really good ;)

Greetings,

Eike

Attached results from last Debian 2.17 build.

Have kernel patch installed that fixed tst-mqueue2. However, I think there are still issues with writev syscall.

We might knock out quite a few if we found the cause of cancel test failures.

Dave
--
John David Anglin [email protected]

#
# Testsuite failures, someone should be working towards

# fixing these! They are listed here for the purpose of
# regression testing during builds.
# Format: <Failed test>, Error <Make error code> [(ignored)]
#
annexc.out, Error 1 (ignored)
check-execstack.out, Error 1
check-localplt.out, Error 1
check-textrel.out, Error 1
run-conformtest.out, Error 1 (ignored)
test-double.out, Error 1
test-float.out, Error 1
test-idouble.out, Error 1
test-ifloat.out, Error 1
tst-addr1.out, Error 1
tst-attr3.out, Error 1
tst-audit2.out, Error 139
tst-backtrace2.out, Error 1
tst-backtrace3.out, Error 1
tst-backtrace4.out, Error 1
tst-backtrace5.out, Error 1
tst-backtrace6.out, Error 1
tst-cancel22.out, Error 1
tst-cancel23.out, Error 1
tst-cancel4.out, Error 1
tst-cancel5.out, Error 1
tst-cancelx20.out, Error 1
tst-cancelx21.out, Error 1
tst-cancelx4.out, Error 1
tst-cancelx5.out, Error 1
tst-cleanupx4.out, Error 1
tst-cond25.out, Error 1
tst-cpuclock2.out, Error 1
tst-getcpu.out, Error 1
tst-makecontext.out, Error 255
tst-mqueue8x.out, Error 1
tst-trim1.out, Error 1

Hi Helge,

Le 3 ao=FBt 2013 =E0 22:41, Helge Deller <[email protected]> a =E9crit :

missing

ough

probably

on ebay

e if

I think this is not exactly the issue. As I told you before, you are we= lcome to set up the A500 (since the rp3440 is broken, and I don't have = the means to repair it) hosted at ESIEE (which has plenty of bw) as an = autobuild. I think the problem has more to do with finding someone know= ledgeable about the process AND willing to go through the pain of said = setup. I for one will not volunteer, for the reasons I already detailed=
in previous emails ;-)

HTH

T-Bone

--=20
Thibaut VAR=C8NE
http://hacks.slashdirt.org/--
To unsubscribe from this list: send the line "unsubscribe linux-parisc"=
in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Hi Thibaut,

Sure, I didn't forgot your offer, and yes, it's a lot of work... Nevertheless, it doesn't hurt if we have more build servers as well.

Thanks!
Helge
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc"=
in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

)

e

I have buildd running on my rp3440 and also have installed the wanna-=20 build
database infrastructure. What is lacking are the trigger scripts for =20 wanna-build to
handle the remote debian source and parisc package archives. There is = =20
a local
trigger script.

So at the moment, the simplest solution appears to be mirroring the =20 archives locally.
I'm a bit worried about how much net bandwidth this will require.

All the packages to set up buildd and wanna-build are in the parisc =20 archive. It's
not that much work. The work in running a buildd server is in =20 monitoring
it and resolving package build issues. I'm not seeing the ports =20 access issue being
resolved unless we can demonstrate a working system.

Dave
--
John David Anglin [email protected]

--
To unsubscribe from this list: send the line "unsubscribe linux-parisc"=
in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

04.08.2013, 00:41, "Helge Deller" <[email protected]>:

Helge,

Thanks for the image and an updated guide. Now i was able to deploy from stratch in minutes (not counting the time spent on debootstrap process). No more pokering with Gentoo live cd for doing debootstrap, like i did before :-)

Guys, thank you for your hard work!