current 2.6 on Q950

current 2.6 on Q950

[Riccardo]

Hi all,


after several weeks/months I grabbed a kernel from here


http://people.debian.org/~smarenka/d-i/m68k/images/daily/kernels/

and slapped it on my quadra 950.

Some good news: it boots!

And it hasn't crashed the past 2 hours of moderate use.

Now what is new? Can I test that soemthign was added/fixed? I noticed 
that the clock is still wrong and that hard disk performance is still 
bad. I read there was some activity on the floppy driver, does it work 
on the Q950?


I read:
[42949381.240000] Driver 'sd' needs updating - please use bus_type 
methods
[42949381.330000] sd 0:0:0:0: [sda] 2055035 512-byte hardware sectors: 
(1.05 GB/1003 MiB)
[42949381.350000] sd 0:0:0:0: [sda] Write Protect is off
[42949381.360000] sd 0:0:0:0: [sda] Mode Sense: a3 00 10 08
[42949381.370000] sd 0:0:0:0: [sda] Write cache: enabled, read cache: 
enabled, supports DPO and FUA
[42949381.380000] sd 0:0:0:0: [sda] 2055035 512-byte hardware sectors: 
(1.05 GB/1003 MiB)
[42949381.400000] sd 0:0:0:0: [sda] Write Protect is off
[42949381.410000] sd 0:0:0:0: [sda] Mode Sense: a3 00 10 08
[42949381.420000] sd 0:0:0:0: [sda] Write cache: enabled, read cache: 
enabled, supports DPO and FUA
[42949381.430000]  sda: [mac] sda1 sda2 sda3 sda4 sda5 sda6

driver needs updating ?

bye,
   Riccardo

Re: current 2.6 on Q950

[Brad Boyer]

On Mon, May 11, 2009 at 11:14:41PM +0200, Riccardo wrote:
> after several weeks/months I grabbed a kernel from here
> 
> 
> http://people.debian.org/~smarenka/d-i/m68k/images/daily/kernels/
> 
> and slapped it on my quadra 950.
> 
> Some good news: it boots!

That's good to hear. This is a model that doesn't get a lot of testing.

> And it hasn't crashed the past 2 hours of moderate use.
> 
> Now what is new? Can I test that soemthign was added/fixed? I noticed 
> that the clock is still wrong and that hard disk performance is still 
> bad. I read there was some activity on the floppy driver, does it work 
> on the Q950?

We don't know how to access the hardware clock on the Q900 or Q950. The
other models all had support at one point, but these two have never
worked. In fact, they used to crash because they were trying to use an
access method that didn't work for those systems.

The Q900 and Q950 also don't have the standard access method for the
floppy drives. We know more or less how to access them, but it's a
separate driver. There was a driver in the tree for a while that could
detect disks and eject them properly, but read/write never worked. I
have the documentation laying around, but I never had the time to get
it working. The floppy is the same as the IIfx, which I have. I don't
have either Q9x0 model. They aren't very common. Technically speaking
it would be possible to use the normal swim driver, but it requires
disabling the ADB ports. It would be tough to use the system without
any input devices.

The SCSI is the other thing that is special on these models, so it's
good to hear that it still works. Does the external SCSI port work?
That's something that used to be broken. I think Finn was working
on some of that code to try to get both ports working.

	Brad Boyer
	flar@allandria.com

Re: current 2.6 on Q950

[Finn Thain]

On Mon, 11 May 2009, Riccardo wrote:

> Hi all,
> 
> 
> after several weeks/months I grabbed a kernel from here
> 
> 
> http://people.debian.org/~smarenka/d-i/m68k/images/daily/kernels/
> 
> and slapped it on my quadra 950.
> 
> Some good news: it boots!
> 
> And it hasn't crashed the past 2 hours of moderate use.
> 
> Now what is new?

The patches that I sent were merged into that release (2.6.29). So X11 
works again, and mac_esp supports both SCSI busses on the quadras 9x0.

> Can I test that soemthign was added/fixed? I noticed that the clock is 
> still wrong

I haven't heard anything more since you raised it on this mailing list 
back in October. I never implemented the workaround that was discussed at 
the time because I have more pressing bugs on my todo list, and because 
I'd like to try and figure out how the hardware works before resorting to 
workarounds...

> and that hard disk performance is still bad.

If dmesg says that PDMA is in use on scsi0 and scsi1, then I'm afraid it 
probably isn't going to get any faster. But performance should be 
comparable with NetBSD if not MacOS (though I never benchmarked either).

> I read there was some activity on the floppy driver, does it work on the 
> Q950?
> 
> 
> I read:
> [42949381.240000] Driver 'sd' needs updating - please use bus_type methods
> [42949381.330000] sd 0:0:0:0: [sda] 2055035 512-byte hardware sectors: (1.05 GB/1003 MiB)
> [42949381.350000] sd 0:0:0:0: [sda] Write Protect is off
> [42949381.360000] sd 0:0:0:0: [sda] Mode Sense: a3 00 10 08
> [42949381.370000] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
> [42949381.380000] sd 0:0:0:0: [sda] 2055035 512-byte hardware sectors: (1.05 GB/1003 MiB)
> [42949381.400000] sd 0:0:0:0: [sda] Write Protect is off
> [42949381.410000] sd 0:0:0:0: [sda] Mode Sense: a3 00 10 08
> [42949381.420000] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
> [42949381.430000]  sda: [mac] sda1 sda2 sda3 sda4 sda5 sda6
> 
> driver needs updating ?

The sd driver is universal, it is not m68k specific. The "needs updating" 
warning is inconsequential.

Finn

> 
> bye,
>   Riccardo
> 

Re: current 2.6 on Q950

[Geert Uytterhoeven]

On Mon, May 11, 2009 at 23:14, Riccardo <riccardo@kaffe.org> wrote:
> I read:
> [42949381.240000] Driver 'sd' needs updating - please use bus_type methods

> driver needs updating ?

To be ignored (by us, poor m68k hackers ;-), as it happens on all platforms.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: current 2.6 on Q950

[Riccardo]

Hi,

>> Some good news: it boots!
>
> That's good to hear. This is a model that doesn't get a lot of testing.

It works also quite decently; albeit slow, I was able to compile the 
whole GNUstep core stuff and two applications, that means several hours 
of compilation.

> We don't know how to access the hardware clock on the Q900 or Q950. The
> other models all had support at one point, but these two have never
> worked. In fact, they used to crash because they were trying to use an
> access method that didn't work for those systems.
>
That's unfortunate. I think to remember a disucssion some time ago 
about this, did past version of the 2.2 kernel perhaps get the time 
from the mac bootloader? That would help already.

> it working. The floppy is the same as the IIfx, which I have. I don't
> have either Q9x0 model. They aren't very common. Technically speaking
> it would be possible to use the normal swim driver, but it requires
> disabling the ADB ports. It would be tough to use the system without
> any input devices.
Since my other mac I use linux on is the IIfx, I don't even bother 
asking :)


>
> The SCSI is the other thing that is special on these models, so it's
> good to hear that it still works. Does the external SCSI port work?
> That's something that used to be broken. I think Finn was working
> on some of that code to try to get both ports working.

I have a magneto-optical drive attached externally, I turned it on and 
at reboot I don't see it in dmesg. But it might be that in the external 
scsi bus is disabled from the penguin, I rmember that I needed to put 
in  an option a long time ago.

Riccardo

Re: current 2.6 on Q950

[Riccardo]

Hi,

On Tuesday, May 12, 2009, at 02:11  PM, Geert Uytterhoeven wrote:

> On Mon, May 11, 2009 at 23:14, Riccardo <riccardo@kaffe.org> wrote:
>> I read:
>> [42949381.240000] Driver 'sd' needs updating - please use bus_type 
>> methods
>
>> driver needs updating ?
>
> To be ignored (by us, poor m68k hackers ;-), as it happens on all 
> platforms.
>

I just noticed this on on my x86 box too after my latest debian 
update... so while x86 has it, we can stay safe :)

Riccardo

Re: current 2.6 on Q950

[Riccardo]

Hi,

> The patches that I sent were merged into that release (2.6.29). So X11
> works again, and mac_esp supports both SCSI busses on the quadras 9x0.
>

very cool, I extra checked and I still had the option to disable the 
scsi BUS in penguin.
Indeed I see:
[42949375.000000] mac_esp: using PDMA for controller 0
[42949378.080000] mac_esp: using PDMA for controller 1

My external MO disk doesn't get recognized though. I'll check that it 
still works on MacOS though.

X11 doesn't work for me, but I'll write separately about that.

> I haven't heard anything more since you raised it on this mailing list
> back in October. I never implemented the workaround that was discussed 
> at
> the time because I have more pressing bugs on my todo list, and because
> I'd like to try and figure out how the hardware works before resorting 
> to
> workarounds...

Oh sure, also because once a workaround gets implemented, the 
motivation to do the real thing gets even less. The real access would 
be very good together with EMILE.

>
>> and that hard disk performance is still bad.
>
> If dmesg says that PDMA is in use on scsi0 and scsi1, then I'm afraid 
> it
> probably isn't going to get any faster. But performance should be
> comparable with NetBSD if not MacOS (though I never benchmarked 
> either).
>

I just run bonnie:

Version 1.03e       ------Sequential Output------ --Sequential Input- 
--Random-
                     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- 
--Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  
/sec %CP
elendil        120M   113  78  1323  72   754  66   123  89  1607  73  
99.3  35
                     ------Sequential Create------ --------Random 
Create--------
                     -Create-- --Read--- -Delete-- -Create-- --Read--- 
-Delete--
               files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  
/sec %CP
                  16    27  88  5272  99  1450  99    30  93  6229  99   
105  78

char operations are very slow (cpu bound? some libraries are 
inefficient?) , but thi could have other reasons. I see that block 
write is 1323 K/s and bloc reaad is 1.6, that is slower than the mac, 
but not abysmal, I don't remember this exact configuration, but 2-3 M/s 
were average for those old macs. IIRC the Q950 was a bit faster than 
the others on the internal chain, but it is clear there that the driver 
performs reasonably now.

Riccardo

Re: current 2.6 on Q950

[Finn Thain]

On Thu, 14 May 2009, Riccardo wrote:

> Hi,
> 
> > The patches that I sent were merged into that release (2.6.29). So X11
> > works again, and mac_esp supports both SCSI busses on the quadras 9x0.
> >
> 
> very cool, I extra checked and I still had the option to disable the scsi BUS
> in penguin.

That option is ignored by the current driver.

> Indeed I see:
> [42949375.000000] mac_esp: using PDMA for controller 0
> [42949378.080000] mac_esp: using PDMA for controller 1

Good. PDMA is significantly faster than PIO (used in 2.6.26 and earlier).

> My external MO disk doesn't get recognized though. I'll check that it 
> still works on MacOS though.
> 
> X11 doesn't work for me, but I'll write separately about that.
> 
> > I haven't heard anything more since you raised it on this mailing list 
> > back in October. I never implemented the workaround that was discussed 
> > at the time because I have more pressing bugs on my todo list, and 
> > because I'd like to try and figure out how the hardware works before 
> > resorting to workarounds...
> 
> Oh sure, also because once a workaround gets implemented, the motivation 
> to do the real thing gets even less. The real access would be very good 
> together with EMILE.
>
[snip]

Well, we established that the workaround would be OK with either Penguin 
or Emile. But the workaround offers no way to update the RTC, which means 
more clock drift between reboots worse than with a working RTC driver, 
which makes network time more attractive (and as I said in October, clock 
batteries are problematic to begin with, and for me this necessitates 
network time anyway). So that's why it is a low priority for me.

Finn