Linux Mac68k v4.1.39 test kernel build is available

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
24 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

I've just uploaded a new kernel build to the Linux Mac68k sourceforge
project: https://sourceforge.net/projects/linux-mac68k/files/

This build has some patches for the mac_scsi and egret drivers that were
backported from mainline v4.11-rc1. For machines that need those drivers
(LC III etc.) this build should work better than previous releases. Please
let me know if you encounter any regressions.

AFAIK, no Macs should need v2.x unless they need the SMC9194 Nubus network
card driver. For some reason that driver only appeared in the v2.2 branch
in the mac68k CVS and never in mainline Linux. It isn't likely to get a
modern rewrite any time soon so I recommend replacing those cards and
running v4.x on any 68k Mac.

--

Reply | Threaded
Open this post in threaded view
|

Re: Linux Mac68k v4.1.39 test kernel build is available

Ingo Jürgensmann-10
Am 17.03.2017 um 01:34 schrieb Finn Thain <[hidden email]>:

> I've just uploaded a new kernel build to the Linux Mac68k sourceforge
> project: https://sourceforge.net/projects/linux-mac68k/files/
>
> This build has some patches for the mac_scsi and egret drivers that were
> backported from mainline v4.11-rc1. For machines that need those drivers
> (LC III etc.) this build should work better than previous releases. Please
> let me know if you encounter any regressions.

What about the esp driver on Amiga (A4000 SCSI, Blizzard 2060 SCSI, Warpengine, ...)? Michael built a test kernel as well some years ago…

--
Ciao...          //        http://blog.windfluechter.net
      Ingo     \X/     XMPP: [hidden email]
       
gpg pubkey:  http://www.juergensmann.de/ij_public_key.asc



Reply | Threaded
Open this post in threaded view
|

Re: Linux Mac68k v4.1.39 test kernel build is available

John Paul Adrian Glaubitz
On 03/17/2017 08:50 PM, Ingo Jürgensmann wrote:
> What about the esp driver on Amiga (A4000 SCSI, Blizzard 2060 SCSI, Warpengine, ...)?
> Michael built a test kernel as well some years ago…

Michael has actually worked on the driver a bit and made improvements. He hasn't
merged it yet though. I think there were some issues left. He was also working
on the driver for the xsurf100, but unfortunately, we never got the card to work
again.

I am planning to get the Amiga up and running again soonish, so that the work
on both drivers can be finished and the code can get hopefully merged upstream.

Adrian

--
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [hidden email]
`. `'   Freie Universitaet Berlin - [hidden email]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Reply | Threaded
Open this post in threaded view
|

Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain
In reply to this post by Finn Thain
On Fri, 17 Mar 2017, I wrote:

> I've just uploaded a new kernel build to the Linux Mac68k sourceforge
> project: https://sourceforge.net/projects/linux-mac68k/files/
>
> This build has some patches for the mac_scsi and egret drivers that were
> backported from mainline v4.11-rc1. For machines that need those drivers
> (LC III etc.) this build should work better than previous releases.
> Please let me know if you encounter any regressions.
>
> AFAIK, no Macs should need v2.x unless they need the SMC9194 Nubus
> network card driver. For some reason that driver only appeared in the
> v2.2 branch in the mac68k CVS and never in mainline Linux. It isn't
> likely to get a modern rewrite any time soon so I recommend replacing
> those cards and running v4.x on any 68k Mac.
>

I've uploaded a second test kernel build. This one includes my patch queue
https://github.com/fthain/linux/commits/mac68k-queue

Apart from these 28 patches this build is exactly the same as the build on
sourceforge described above. It can be downloaded here,

https://www.telegraphics.com.au/~fthain/mac68k-queue/vmlinux-4.1.39-mac_scsi-egret-etc+
https://www.telegraphics.com.au/~fthain/mac68k-queue/linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz

SHA1

12881647716ed02f1c9d862a588b7fc89a3d5c64  vmlinux-4.1.39-mac_scsi-egret-etc+
d7eb947eee6be997d5b0079fd15d98673d5f6a12  linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz

These patches cover a wide range of Mac 68k hardware which makes it
difficult for me to test, So I'm (ab)using the Debian list to try to find
users willing to help improve test coverage prior to patch submission.

Thanks.

--

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

James Hemsing
I can test on a Quadra 700, SE/30, and LC475 once my CD-Rs arrive, if that would be helpful.

On Sun, Mar 19, 2017 at 6:49 PM Finn Thain <[hidden email]> wrote:
On Fri, 17 Mar 2017, I wrote:

> I've just uploaded a new kernel build to the Linux Mac68k sourceforge
> project: https://sourceforge.net/projects/linux-mac68k/files/
>
> This build has some patches for the mac_scsi and egret drivers that were
> backported from mainline v4.11-rc1. For machines that need those drivers
> (LC III etc.) this build should work better than previous releases.
> Please let me know if you encounter any regressions.
>
> AFAIK, no Macs should need v2.x unless they need the SMC9194 Nubus
> network card driver. For some reason that driver only appeared in the
> v2.2 branch in the mac68k CVS and never in mainline Linux. It isn't
> likely to get a modern rewrite any time soon so I recommend replacing
> those cards and running v4.x on any 68k Mac.
>

I've uploaded a second test kernel build. This one includes my patch queue
https://github.com/fthain/linux/commits/mac68k-queue

Apart from these 28 patches this build is exactly the same as the build on
sourceforge described above. It can be downloaded here,

https://www.telegraphics.com.au/~fthain/mac68k-queue/vmlinux-4.1.39-mac_scsi-egret-etc+
https://www.telegraphics.com.au/~fthain/mac68k-queue/linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz

SHA1

12881647716ed02f1c9d862a588b7fc89a3d5c64  vmlinux-4.1.39-mac_scsi-egret-etc+
d7eb947eee6be997d5b0079fd15d98673d5f6a12  linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz

These patches cover a wide range of Mac 68k hardware which makes it
difficult for me to test, So I'm (ab)using the Debian list to try to find
users willing to help improve test coverage prior to patch submission.

Thanks.

--
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

On Mon, 20 Mar 2017, James Hemsing wrote:

> I can test on a Quadra 700, SE/30, and LC475 once my CD-Rs arrive, if
> that would be helpful.

Great! The main areas needing more testing are these:

- ADB
- SONIC ethernet
- Various video modes - 1, 2, 4, 8, 16, 24 and/or 32 bits per pixel
- Device driver initialization

Any regression you notice (particularly during boot) would be of interest.

If you could email to me the kernel output (from 'dmesg', kernel log file
or console capture) this would be much appreciated.

Thanks!
--

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Laurent Vivier-5
In reply to this post by Finn Thain
Le 20/03/2017 à 01:52, Finn Thain a écrit :

> On Fri, 17 Mar 2017, I wrote:
>
>> I've just uploaded a new kernel build to the Linux Mac68k sourceforge
>> project: https://sourceforge.net/projects/linux-mac68k/files/
>>
>> This build has some patches for the mac_scsi and egret drivers that were
>> backported from mainline v4.11-rc1. For machines that need those drivers
>> (LC III etc.) this build should work better than previous releases.
>> Please let me know if you encounter any regressions.
>>
>> AFAIK, no Macs should need v2.x unless they need the SMC9194 Nubus
>> network card driver. For some reason that driver only appeared in the
>> v2.2 branch in the mac68k CVS and never in mainline Linux. It isn't
>> likely to get a modern rewrite any time soon so I recommend replacing
>> those cards and running v4.x on any 68k Mac.
>>
>
> I've uploaded a second test kernel build. This one includes my patch queue
> https://github.com/fthain/linux/commits/mac68k-queue
>
> Apart from these 28 patches this build is exactly the same as the build on
> sourceforge described above. It can be downloaded here,
>
> https://www.telegraphics.com.au/~fthain/mac68k-queue/vmlinux-4.1.39-mac_scsi-egret-etc+
> https://www.telegraphics.com.au/~fthain/mac68k-queue/linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz
>
> SHA1
>
> 12881647716ed02f1c9d862a588b7fc89a3d5c64  vmlinux-4.1.39-mac_scsi-egret-etc+
> d7eb947eee6be997d5b0079fd15d98673d5f6a12  linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz
>
> These patches cover a wide range of Mac 68k hardware which makes it
> difficult for me to test, So I'm (ab)using the Debian list to try to find
> users willing to help improve test coverage prior to patch submission.
Tested on my Quadra 800. ADB/SCSI/Ethernet work fine.

I've also tested with Q800 implementation in QEMU and works fine too.

You can find attached the both dmesg.

Laurent


q800-dmesg-4.1.39.txt (8K) Download Attachment
qemu-dmesg-4.1.39.txt (7K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

On Tue, 21 Mar 2017, Laurent Vivier wrote:

>
> Tested on my Quadra 800. ADB/SCSI/Ethernet work fine.
>
> I've also tested with Q800 implementation in QEMU and works fine too.
>

Thanks Laurent!

This looks like a very convenient way to test certain patches. Did you use
upstream QEMU or your QEMU fork at https://github.com/vivier/qemu-m68k
?

> You can find attached the both dmesg.
>

Interesting... the QEMU log says:

Slot 9:
Board resource not found!

This is a bit surprising because it seems to suggest that
nubus_probe_slot() found a meaningful bytelanes value. That logic got
changed in this patch series.

I wonder whether the vmlinux-4.1.39-mac_scsi-egret+ build on sourceforge
produces the same messages (?)

Thanks again.
--

> Laurent
>
>

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

James Hemsing
FYI netatalk users: Spent a couple of days trying to get netatalk to work on Ubuntu 16.04 with pre OS 9 macs, but it is broken, so I gave up and put 14.04 on a netbook. This was just so I could copy the kernel over without taking anything apart.

dmesg.txt is from a debian 3.1 sid rootfs. 

dmesg2.txt is from the latest aranym rootfs tar from the debian m68k page, tar -C'ed it to a SCSI drive using an Adaptec 2930 on a modern PC. 

ADB/SCSI/Ethernet seem to be working fine on Quadra 700. I will try the SE/30 soon, no idea if it will work at all.

If I reboot the Quadra from Linux into System 7.1, and then launch Linux from Penguin-19 without powering off, the ADB driver will crash and produce a backtrace requiring a reboot. I will take a photo when it happens again.

- James

On Tue, Mar 21, 2017 at 6:53 PM Finn Thain <[hidden email]> wrote:

On Tue, 21 Mar 2017, Laurent Vivier wrote:

>
> Tested on my Quadra 800. ADB/SCSI/Ethernet work fine.
>
> I've also tested with Q800 implementation in QEMU and works fine too.
>

Thanks Laurent!

This looks like a very convenient way to test certain patches. Did you use
upstream QEMU or your QEMU fork at https://github.com/vivier/qemu-m68k
?

> You can find attached the both dmesg.
>

Interesting... the QEMU log says:

Slot 9:
Board resource not found!

This is a bit surprising because it seems to suggest that
nubus_probe_slot() found a meaningful bytelanes value. That logic got
changed in this patch series.

I wonder whether the vmlinux-4.1.39-mac_scsi-egret+ build on sourceforge
produces the same messages (?)

Thanks again.
--

> Laurent
>
>


dmesg2.txt (9K) Download Attachment
dmesg.txt (7K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

On Tue, 28 Mar 2017, James Hemsing wrote:

> FYI netatalk users: Spent a couple of days trying to get netatalk to
> work on Ubuntu 16.04 with pre OS 9 macs, but it is broken, so I gave up
> and put 14.04 on a netbook. This was just so I could copy the kernel
> over without taking anything apart.
>
> dmesg.txt is from a debian 3.1 sid rootfs.
>
> dmesg2.txt is from the latest aranym rootfs tar from the debian m68k
> page, tar -C'ed it to a SCSI drive using an Adaptec 2930 on a modern PC.
>

Thanks for sending these results.

> ADB/SCSI/Ethernet seem to be working fine on Quadra 700. I will try the
> SE/30 soon, no idea if it will work at all.
>

I never tested an SE/30, but I used to test on a Mac II with an '030
accelerator so I'm optimistic that the SE/30 will work too.

> If I reboot the Quadra from Linux into System 7.1, and then launch Linux
> from Penguin-19 without powering off, the ADB driver will crash and
> produce a backtrace requiring a reboot. I will take a photo when it
> happens again.
>

This crash appears to be a regression caused by my github patch queue.

If possible, can you confirm that the 4.1.39-mac_scsi-egret build on
sourceforge is unaffected?

Thanks again!

--

> - James
>

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

James Hemsing
Strangely, I haven't been able to reproduce that ADB error. Launching the kernel is temperamental: I have to open Settings in Penguin and reselect the same kernel so that it will boot and not hang after ABCDEFGH, so it may have been a temporary thing.

The kernel is working on the SE/30, but it only has 8MB of RAM and has been booting for the past 2 hours.

On Tue, Mar 28, 2017 at 5:34 AM Finn Thain <[hidden email]> wrote:

On Tue, 28 Mar 2017, James Hemsing wrote:

> FYI netatalk users: Spent a couple of days trying to get netatalk to
> work on Ubuntu 16.04 with pre OS 9 macs, but it is broken, so I gave up
> and put 14.04 on a netbook. This was just so I could copy the kernel
> over without taking anything apart.
>
> dmesg.txt is from a debian 3.1 sid rootfs.
>
> dmesg2.txt is from the latest aranym rootfs tar from the debian m68k
> page, tar -C'ed it to a SCSI drive using an Adaptec 2930 on a modern PC.
>

Thanks for sending these results.

> ADB/SCSI/Ethernet seem to be working fine on Quadra 700. I will try the
> SE/30 soon, no idea if it will work at all.
>

I never tested an SE/30, but I used to test on a Mac II with an '030
accelerator so I'm optimistic that the SE/30 will work too.

> If I reboot the Quadra from Linux into System 7.1, and then launch Linux
> from Penguin-19 without powering off, the ADB driver will crash and
> produce a backtrace requiring a reboot. I will take a photo when it
> happens again.
>

This crash appears to be a regression caused by my github patch queue.

If possible, can you confirm that the 4.1.39-mac_scsi-egret build on
sourceforge is unaffected?

Thanks again!

--

> - James
>
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

On Wed, 29 Mar 2017, James Hemsing wrote:

> Strangely, I haven't been able to reproduce that ADB error. Launching
> the kernel is temperamental: I have to open Settings in Penguin and
> reselect the same kernel so that it will boot and not hang after
> ABCDEFGH, so it may have been a temporary thing.
>

Penguin does have issues, but AFAIK wiggling the settings doesn't affect
them. The Penguin docs cover the known issues:
http://www.mac.linux-m68k.org/docs/penguin.php

A hang after "ABCFGHIJK" may indicate an unhandled interrupt (usually from
the ethernet chip). You can avoid that issue by disabling extensions
(restart MacOS with shift key held down) before booting Linux.

> The kernel is working on the SE/30, but it only has 8MB of RAM and has
> been booting for the past 2 hours.
>

With only 16 MHz and 8 MB you really need to treat it like an embedded
system (busybox userland etc).

--

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Laurent Vivier-5
In reply to this post by Finn Thain
Le 22/03/2017 à 01:56, Finn Thain a écrit :

>
> On Tue, 21 Mar 2017, Laurent Vivier wrote:
>
>>
>> Tested on my Quadra 800. ADB/SCSI/Ethernet work fine.
>>
>> I've also tested with Q800 implementation in QEMU and works fine too.
>>
>
> Thanks Laurent!

It is up now for 11 days.

$ uname -a
Linux q800 4.1.39-mac_scsi-egret-etc+ #2 Fri Mar 17 10:35:58 AEDT 2017
m68k GNU/Linux
$ uptime
 12:14:01 up 11 days, 19:41,  2 users,  load average: 0.00, 0.07, 0.07

>
> This looks like a very convenient way to test certain patches. Did you use
> upstream QEMU or your QEMU fork at https://github.com/vivier/qemu-m68k
> ?

You can build one from:

git clone -b q800-dev https://github.com/vivier/qemu-m68k.git

./configure --target-list=m68k-softmmu

./m68k-softmmu/qemu-system-m68k -M q800 \
    -serial none -serial mon:stdio -m 1000M \
    -drive file=./virtm68k-etch.qcow2,format=qcow2 -net
nic,model=dp83932,addr=09:00:07:12:34:57 \
    -net bridge,helper=/usr/libexec/qemu-bridge-helper,br=virbr0 \
    -append "root=/dev/sda2 rw console=ttyS0 console=tty" \
    -kernel vmlinux-4.8.0-2-m68k \
    -initrd initrd.img-4.8.0-2-m68k \
    -g 1024x768x24

>
>> You can find attached the both dmesg.
>>
>
> Interesting... the QEMU log says:
>
> Slot 9:
> Board resource not found!
>
> This is a bit surprising because it seems to suggest that
> nubus_probe_slot() found a meaningful bytelanes value. That logic got
> changed in this patch series.
>
> I wonder whether the vmlinux-4.1.39-mac_scsi-egret+ build on sourceforge
> produces the same messages (?)

With a 4.8.0-2, I've got:

[    0.180000] NuBus: Scanning NuBus slots.
[    0.180000] Now probing slot 9 at f9ffffff
[    0.180000] Slot 9:
[    0.180000] Board resource not found!
[    0.180000] Now probing slot A at faffffff
[    0.180000] Now probing slot A at fafffffe
[    0.180000] Now probing slot A at fafffffd
[    0.180000] Now probing slot A at fafffffc
[    0.180000] Now probing slot B at fbffffff
[    0.180000] Now probing slot B at fbfffffe
[    0.180000] Now probing slot B at fbfffffd
[    0.180000] Now probing slot B at fbfffffc
[    0.180000] Now probing slot C at fcffffff
[    0.180000] Now probing slot C at fcfffffe
[    0.180000] Now probing slot C at fcfffffd
[    0.180000] Now probing slot C at fcfffffc
[    0.180000] Now probing slot D at fdffffff
[    0.180000] Now probing slot D at fdfffffe
[    0.180000] Now probing slot D at fdfffffd
[    0.180000] Now probing slot D at fdfffffc
[    0.180000] Now probing slot E at feffffff
[    0.180000] Now probing slot E at fefffffe
[    0.180000] Now probing slot E at fefffffd
[    0.180000] Now probing slot E at fefffffc

Slot 9 is the graphic card:

(qemu) info qtree
bus: main-system-bus
  type System
  dev: mac-nubus-bridge, id ""
    mmio 0000000060000000/0000000090000000
    mmio 00000000f0000000/000000000f000000
    bus: nubus-bus.0
      type nubus-bus
      dev: nubus-macfb, id ""
        width = 800 (0x320)
        height = 600 (0x258)
        depth = 16 (0x10)
...

This can be a bad Nubus emulation in qemu.

Thanks,
Laurent

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Stefan Niestegge-2
In reply to this post by Finn Thain


Am 20.03.2017 um 01:52 schrieb Finn Thain:

> On Fri, 17 Mar 2017, I wrote:
>
>> I've just uploaded a new kernel build to the Linux Mac68k sourceforge
>> project: https://sourceforge.net/projects/linux-mac68k/files/
>>
>> This build has some patches for the mac_scsi and egret drivers that were
>> backported from mainline v4.11-rc1. For machines that need those drivers
>> (LC III etc.) this build should work better than previous releases.
>> Please let me know if you encounter any regressions.
>>
>> AFAIK, no Macs should need v2.x unless they need the SMC9194 Nubus
>> network card driver. For some reason that driver only appeared in the
>> v2.2 branch in the mac68k CVS and never in mainline Linux. It isn't
>> likely to get a modern rewrite any time soon so I recommend replacing
>> those cards and running v4.x on any 68k Mac.
>>
>
> I've uploaded a second test kernel build. This one includes my patch queue
> https://github.com/fthain/linux/commits/mac68k-queue
>
> Apart from these 28 patches this build is exactly the same as the build on
> sourceforge described above. It can be downloaded here,
>
> https://www.telegraphics.com.au/~fthain/mac68k-queue/vmlinux-4.1.39-mac_scsi-egret-etc+
> https://www.telegraphics.com.au/~fthain/mac68k-queue/linux-m68k-image-4.1.39-mac_scsi-egret-etc+.tar.gz


Hi there,

we tested this kernel today on a Performa 450 (aka LC3)
ADB works nice, dmesg shows that the kernel detected our
network card (Nubus Slot E, Mac LC Ethernet Adapter, Board_ID 0x2e)
Its an "Asante" card, but ifconfig shows only loopback device.

The modules had been unpacked beforehand.
Is this expected behaviour?

Best regards,
Stefan & Carsten

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain

On Mon, 3 Apr 2017, Stefan Niestegge wrote:

>
> Hi there,
>
> we tested this kernel today on a Performa 450 (aka LC3) ADB works nice,

Thanks for testing!

> dmesg shows that the kernel detected our network card (Nubus Slot E, Mac
> LC Ethernet Adapter, Board_ID 0x2e) Its an "Asante" card, but ifconfig
> shows only loopback device.
>
> The modules had been unpacked beforehand.
>
> Is this expected behaviour?
>

That really depends on your distro's network scripts and their configs.

But "ifconfig -a" should reveal all network interfaces, whether configured
by your boot scripts or not. If your network interface card doesn't show
up please send the dmesg output.

BTW, you needn't unpack modules for device drivers as they are all
built-in. But you'll need the modules if you need things like ipv6 or ppp.

--

> Best regards,
> Stefan & Carsten
>
>

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Stefan Niestegge-2


Am 04.04.2017 um 05:47 schrieb Finn Thain:
>
> On Mon, 3 Apr 2017, Stefan Niestegge wrote:

>> we tested this kernel today on a Performa 450 (aka LC3) ADB works nice,
>
> Thanks for testing!
>

You're welcome!

>
> That really depends on your distro's network scripts and their configs.

Well, with one of the last kernels we tried, we got the network up, but
could only login to the Mac through serial console. Our "distro" is
debian, iirc its the tarball Th. Glaser supplied some time ago.
>
> But "ifconfig -a" should reveal all network interfaces, whether configured
> by your boot scripts or not. If your network interface card doesn't show
> up please send the dmesg output.

Nope, only loopback.

I send you the dmesg as attachment.


Best regards,
Stefan

dmesg.log (12K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain
In reply to this post by Laurent Vivier-5

On Sun, 2 Apr 2017, Laurent Vivier wrote:

> Le 22/03/2017 à 01:56, Finn Thain a écrit :
> >
> > On Tue, 21 Mar 2017, Laurent Vivier wrote:
> >
> >>
> >> Tested on my Quadra 800. ADB/SCSI/Ethernet work fine.
> >>
> >> I've also tested with Q800 implementation in QEMU and works fine too.
> >>
> >
> > Thanks Laurent!
>
> It is up now for 11 days.
>
> $ uname -a
> Linux q800 4.1.39-mac_scsi-egret-etc+ #2 Fri Mar 17 10:35:58 AEDT 2017
> m68k GNU/Linux
> $ uptime
>  12:14:01 up 11 days, 19:41,  2 users,  load average: 0.00, 0.07, 0.07
>
Well, this suggests that both the emulator and the kernel are stable.

> >
> > This looks like a very convenient way to test certain patches. Did you use
> > upstream QEMU or your QEMU fork at https://github.com/vivier/qemu-m68k
> > ?
>
> You can build one from:
>
> git clone -b q800-dev https://github.com/vivier/qemu-m68k.git
>
> ./configure --target-list=m68k-softmmu
>
make

> ./m68k-softmmu/qemu-system-m68k -M q800 \
>     -serial none -serial mon:stdio -m 1000M \
>     -drive file=./virtm68k-etch.qcow2,format=qcow2 -net nic,model=dp83932,addr=09:00:07:12:34:57 \
>     -net bridge,helper=/usr/libexec/qemu-bridge-helper,br=virbr0 \
>     -append "root=/dev/sda2 rw console=ttyS0 console=tty" \
>     -kernel vmlinux-4.8.0-2-m68k \
>     -initrd initrd.img-4.8.0-2-m68k \
>     -g 1024x768x24
>

Thanks for these hints.

I created a bridge.conf with "allow virbr0" and then I was able to boot my
usual initramfs without any issues. Too easy! I will try booting an NFS
rootfs when I get time.

> >> You can find attached the both dmesg.
> >>
> >
> > Interesting... the QEMU log says:
> >
> > Slot 9:
> > Board resource not found!
> >
> > This is a bit surprising because it seems to suggest that
> > nubus_probe_slot() found a meaningful bytelanes value. That logic got
> > changed in this patch series.
> >
> > I wonder whether the vmlinux-4.1.39-mac_scsi-egret+ build on
> > sourceforge produces the same messages (?)
>
> With a 4.8.0-2, I've got:
>
> [    0.180000] NuBus: Scanning NuBus slots.
> [    0.180000] Now probing slot 9 at f9ffffff
> [    0.180000] Slot 9:
> [    0.180000] Board resource not found!
> [    0.180000] Now probing slot A at faffffff
> [    0.180000] Now probing slot A at fafffffe
> [    0.180000] Now probing slot A at fafffffd
> [    0.180000] Now probing slot A at fafffffc
> [    0.180000] Now probing slot B at fbffffff
> [    0.180000] Now probing slot B at fbfffffe
> [    0.180000] Now probing slot B at fbfffffd
> [    0.180000] Now probing slot B at fbfffffc
> [    0.180000] Now probing slot C at fcffffff
> [    0.180000] Now probing slot C at fcfffffe
> [    0.180000] Now probing slot C at fcfffffd
> [    0.180000] Now probing slot C at fcfffffc
> [    0.180000] Now probing slot D at fdffffff
> [    0.180000] Now probing slot D at fdfffffe
> [    0.180000] Now probing slot D at fdfffffd
> [    0.180000] Now probing slot D at fdfffffc
> [    0.180000] Now probing slot E at feffffff
> [    0.180000] Now probing slot E at fefffffe
> [    0.180000] Now probing slot E at fefffffd
> [    0.180000] Now probing slot E at fefffffc
>
> Slot 9 is the graphic card:
>
> (qemu) info qtree
> bus: main-system-bus
>   type System
>   dev: mac-nubus-bridge, id ""
>     mmio 0000000060000000/0000000090000000
>     mmio 00000000f0000000/000000000f000000
>     bus: nubus-bus.0
>       type nubus-bus
>       dev: nubus-macfb, id ""
>         width = 800 (0x320)
>         height = 600 (0x258)
>         depth = 16 (0x10)
> ...
>
> This can be a bad Nubus emulation in qemu.
>
Maybe not... FWIW, I found the following in an old log file from a
PowerBook Duo 280c. (Perhaps this was with a mini dock fitted...)

NuBus: Scanning NuBus slots.
Slot E, format block at 0xfeffffec
Format block: 00fe 0014  0000 0000  bb60 e2e9  0101 5a93  2bc7 000f
Slot E:
Board resource not found!

Anyway, thanks for confirming that the "not found" message was not a
regression caused by the Nubus patches in my queue.

--

> Thanks,
> Laurent
>
>
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Finn Thain
In reply to this post by Stefan Niestegge-2
On Tue, 4 Apr 2017, Stefan Niestegge wrote:

>
> Well, with one of the last kernels we tried, we got the network up, but
> could only login to the Mac through serial console. Our "distro" is
> debian, iirc its the tarball Th. Glaser supplied some time ago.

Please also try the "4.1.39-mac_scsi-egret" kernel build at
https://sourceforge.net/projects/linux-mac68k/files/ 
That way we can find out if this is a regression in my patch queue. It
would be very helpful to get the dmesg from output from any kernel which
has a device driver that does detect the network card. Thanks!

--

Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Stefan Niestegge-2
Here are the dmesg logs of both kernels.

Am 05.04.2017 um 03:47 schrieb Finn Thain:

> On Tue, 4 Apr 2017, Stefan Niestegge wrote:
>
>>
>> Well, with one of the last kernels we tried, we got the network up, but
>> could only login to the Mac through serial console. Our "distro" is
>> debian, iirc its the tarball Th. Glaser supplied some time ago.
>
> Please also try the "4.1.39-mac_scsi-egret" kernel build at
> https://sourceforge.net/projects/linux-mac68k/files/ 
> That way we can find out if this is a regression in my patch queue. It
> would be very helpful to get the dmesg from output from any kernel which
> has a device driver that does detect the network card. Thanks!
>

dmesg0317.log (12K) Download Attachment
dmesg0316.log (12K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Seeking Mac testers, was Re: Linux Mac68k v4.1.39 test kernel build is available

Stefan Niestegge
In reply to this post by Finn Thain
Here are the dmesg logs of both kernels.

Am 05.04.2017 um 03:47 schrieb Finn Thain:

> On Tue, 4 Apr 2017, Stefan Niestegge wrote:
>
>>
>> Well, with one of the last kernels we tried, we got the network up, but
>> could only login to the Mac through serial console. Our "distro" is
>> debian, iirc its the tarball Th. Glaser supplied some time ago.
>
> Please also try the "4.1.39-mac_scsi-egret" kernel build at
> https://sourceforge.net/projects/linux-mac68k/files/ 
> That way we can find out if this is a regression in my patch queue. It
> would be very helpful to get the dmesg from output from any kernel which
> has a device driver that does detect the network card. Thanks!
>

dmesg0317.log (12K) Download Attachment
dmesg0316.log (12K) Download Attachment
12