linux-apus-user

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 21 Jan 2003, Kulwant Bhogal wrote:
> Re: Your e-mail (20/01/2003)
> > On 19 Jan 2003, Kulwant Bhogal wrote:
> >> > So you should try dd if=/dev/sdb of=rdb.dump bs=4k count=1230 .
> >> 
> >> OK, I tried that and I still got "dd: /dev/sdb: Input/Output error"
> 
> > And how long was the resulting file (`ls -l rdb.dump')?
> 
> 0 bytes. 

Hence it really failed in reading from the disk :-(

I did some Googling and apparently Linux supports sector sizes different from
512 bytes, but it failed for 4096 bytes:
http://www.google.com/search?hl=en&lr=&ie=ISO-8859-1&q=%22SCSI+Device+sda%3A+hdwr+sector+%3D+4096+bytes%22

Well, as a last resort I suggest to send a detailed message to
lin...@vg... (please cc linux-apus-devel) and hope something
comes out...

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert,

Re: Your e-mail (20/01/2003)

> On 19 Jan 2003, Kulwant Bhogal wrote:
>> > So you should try dd if=/dev/sdb of=rdb.dump bs=4k count=1230 .
>> 
>> OK, I tried that and I still got "dd: /dev/sdb: Input/Output error"

> And how long was the resulting file (`ls -l rdb.dump')?

0 bytes. 

-rw-r--r--    1    root    root    0    Jan 21 19:29    rdb.dump


When I press return after entering the command, the sdb disk is accessed
once briefly followed by sdc briefly. Then the heartbeat shows
activity once every two seconds and then after a few seconds (approx 8 secs) 
I get the error message. 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 19 Jan 2003, Kulwant Bhogal wrote:
> > So you should try dd if=/dev/sdb of=rdb.dump bs=4k count=1230 .
> 
> OK, I tried that and I still got "dd: /dev/sdb: Input/Output error"

And how long was the resulting file (`ls -l rdb.dump')?

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Michel,

> So you should try dd if=/dev/sdb of=rdb.dump bs=4k count=1230 .

OK, I tried that and I still got "dd: /dev/sdb: Input/Output error"

Regards, 

Kulwant

Re: console, modules and pixel64

[Krystian B. <cah...@st...>]

On Sun, 5 Jan 2003, Lio wrote:
> Hi All,
> I would like to know which gfx driver should I take for a pixel64 gfx card
> (cirrus logic GD5429) ?

Try clgenfb driver. Does pixel64 use some specific bus ? I heard something
about ateo bus but I don't know anytrhing about it.

Regards
Krystian Baclawski

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 10 Jan 2003, Kulwant Bhogal wrote:
> The thing that puzzles me is that the Seagate specs for the drive
> (http://www.seagate.com/support/disc/specs/scsi/st34573lc.html) (lw [mine] is also
> on this page) bear no resemblance to the HDToolbox numbers even after
> accounting for the 4096 blocksize. Examples are number of tracks: 37,505,
> cylinders: 7,501, heads physical: 5, disks (3.5 inch): 3. 
> The number of sectors is not too far out at 8,888,924 (/8 = 1,111,116). 
> (Seagate numbers). 
>  
> HDToolbox states that there are 3 heads. 
> 
> I suspect that HDToolbox does it's own interpretation of the drive geometry
> - which also happens to work (somehow!!). Actually I am surprised that it
> actually all works - well it works as long as you don't want to reformat to a foreign
> (to AmigaOS) disk format!. 
> 
> But for Linux, I am not sure where that leaves us. 

Don't worry. In contrast to IDE devices, SCSI devices are always accessed by
specifying linear sector numbers (for IDE you have CHS and LBA). So it doesn't
matter how many heads you fake to have, as long as you specify the correct
linear sector number it'll work. HdToolBox probably just tries to find optimal
CHS values so you can use as much of the disk as possible (i.e. find C, H, and
S such that (numsectors-C*H*S) is minimal).

BTW, even on IDE devices the CHS values don't correspond to the real number of
heads etc.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 10 Jan 2003, Kulwant Bhogal wrote:
> > Yes from there, take 2.4.13 kernel, it should work.
> 
> > First thing you should have partition formatted with FFS Intl. If affs
> > module isn't compiled into kernel just insert it with insmod.
> > Then mount it somewhere by "mount -t affs /dev/.... /mnt/..."
> 
> Does it have to be FFS Intl.? Most of mine are simply FFS. 
> You assume a lot of knowledge on my part. I will have to research more and
> try out all these suggestions. Probably at the weekend. 

All version of FFS are supported. Except maybe (don't remember) FFS-DC
(Directory Caching), which may be supported read-only.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Michel <mi...@da...>]

On Fre, 2003-01-10 at 01:04, Kulwant Bhogal wrote:
> 
> >> > >   dd if=/dev/sdb of=rdb.dump bs=4k count=xxx
> >> > 
> >> > > Set `xxx' such that `xxx*4096' equals the number of bytes in the
> >> > > first 2 cylinders of the disk (use HdToolBox to find out the
> >> > > cylinder size). Then sent us the output of `hexdump rdb.dump'.
> >> > 
> >> > OK, because I got a count of 615 blocks per cylinder, I used 
> >> > 615*2*4096=5,038,080 as the count. So the commandline I used was:
> >> > 
> >> > # dd if=/dev/sdb of=rdb.dump bs=4k count=5038080
> >                                  ^^^^^^^^^^^^^^^^^^^
> 
> > Is the RDB really 20 gigs though? :)
> 
> Where did you get 20 Gigs from? I only followed Geert's instructions to the
> letter. He said "set 'xxx' such that 'xxx*4096' equals the number of bytes
> in the first two cylinders of the disk and to use HDToolbox to find out the
> cylinder size. 
> 
> So, I got a cylinder size of 615 blocks per cylinder from HDToolbox, I know that there
> are 4096 bytes per block and I am supposed to dump the first two cylinders,
> I get 615*2*4096 which equals 5038080. Which is just over 4.8Mb.

You made a mistake:

	x * 4096 = 5038080

<=>	x = 5038080 / 4096

<=>	x = 1230

So you should try dd if=/dev/sdb of=rdb.dump bs=4k count=1230 .

> And this also makes sense because there are 2067 cylinders total on 
> the disk so 5038080/2 (because that was for two cylinders) * 2067 = 
> 5206855680 bytes which = 4965.64Mb or 4.849Gb - which is the size of 
> the disk.

No wonder then that trying to read around 20 GB from the disk failed. :)


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert,

> Looks like the driver itself has problems.

So my calculation wasn't wrong was it? (see also other post).

> mount -t affs /dev/<amiga_partition> /mnt

Is the installer menu command no good for mounting partitions? 

> I hope you didn't format that target partition?

No, I'm not that silly. In any case, I was careful, because it is only a
200Mb partition (reserved for OS4 when it eventually comes out :-) which I can afford to lose if things go wrong.

>> I also do not know how big this file is (it may be 0 bytes). What is the
>> way to get a file listing with sizes? 

> ls -l

> (`man ls' or `ls -- help')

I remember typing man xxxx before (where xxxx = some command name) and  it
told me no manuals were installed or something and told me to use
'apt-get install man-db'. But I was going to do this once I had got my installed
partition to boot successfully. I still have to stop the kernel crashing when trying
to set the system clock before I can do so of these other things. 

Thanks to all for all the help so far. 

Kind regards, 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Krystian,

> So try to add "root=/dev/...,rw init=/bin/sh" to kernel-options line then
> linux will start into raw bash shell.

I will try this at some point (need to get some sleep now - at work
tomorrow).

> Uh, I'm a bit puzzled, what's an IIRC ?

IIRC = If I Recall Correctly :-)

> Yes from there, take 2.4.13 kernel, it should work.

> First thing you should have partition formatted with FFS Intl. If affs
> module isn't compiled into kernel just insert it with insmod.
> Then mount it somewhere by "mount -t affs /dev/.... /mnt/..."

Does it have to be FFS Intl.? Most of mine are simply FFS. 
You assume a lot of knowledge on my part. I will have to research more and
try out all these suggestions. Probably at the weekend. 

Thanks for your help. 

Kind regards, 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Michel,

>> > >   dd if=/dev/sdb of=rdb.dump bs=4k count=xxx
>> > 
>> > > Set `xxx' such that `xxx*4096' equals the number of bytes in the
>> > > first 2 cylinders of the disk (use HdToolBox to find out the
>> > > cylinder size). Then sent us the output of `hexdump rdb.dump'.
>> > 
>> > OK, because I got a count of 615 blocks per cylinder, I used 
>> > 615*2*4096=5,038,080 as the count. So the commandline I used was:
>> > 
>> > # dd if=/dev/sdb of=rdb.dump bs=4k count=5038080
>                                  ^^^^^^^^^^^^^^^^^^^

> Is the RDB really 20 gigs though? :)

Where did you get 20 Gigs from? I only followed Geert's instructions to the
letter. He said "set 'xxx' such that 'xxx*4096' equals the number of bytes
in the first two cylinders of the disk and to use HDToolbox to find out the
cylinder size. 

So, I got a cylinder size of 615 blocks per cylinder from HDToolbox, I know that there
are 4096 bytes per block and I am supposed to dump the first two cylinders,
I get 615*2*4096 which equals 5038080. Which is just over 4.8Mb. And this
also makes sense because there are 2067 cylinders total on the disk so
5038080/2 (because that was for two cylinders) * 2067 = 5206855680 bytes
which = 4965.64Mb or 4.849Gb - which is the size of the disk. 

There are 1271205 blocks on the disk and 1271205*4096 also = 4.8Gb. 

The thing that puzzles me is that the Seagate specs for the drive
(http://www.seagate.com/support/disc/specs/scsi/st34573lc.html) (lw [mine] is also
on this page) bear no resemblance to the HDToolbox numbers even after
accounting for the 4096 blocksize. Examples are number of tracks: 37,505,
cylinders: 7,501, heads physical: 5, disks (3.5 inch): 3. 
The number of sectors is not too far out at 8,888,924 (/8 = 1,111,116). 
(Seagate numbers). 
 
HDToolbox states that there are 3 heads. 

I suspect that HDToolbox does it's own interpretation of the drive geometry
- which also happens to work (somehow!!). Actually I am surprised that it
actually all works - well it works as long as you don't want to reformat to a foreign
(to AmigaOS) disk format!. 

But for Linux, I am not sure where that leaves us. 

Kind regards, 

Kulwant
 
  

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Krystian B. <cah...@st...>]

On 8 Jan 2003, Kulwant Bhogal wrote:
> Hello Krystian,
>
> > Hmm. It'd be more useful if you used ksymoops.
>
> Sorry i am a complete Linux newbie, where/how do I use ksymoops? I have never seen this before. Is it a
> bootstrap commandline option?
> Remember I do not get to a Linux comandline prompt when trying to start Linux
> as the kernel crashes while trying to set the system clock. I can only get
> to a Linux Shell when I run the installer (which doesn't crash).

So try to add "root=/dev/...,rw init=/bin/sh" to kernel-options line then
linux will start into raw bash shell.

> > Then those magic numbers
> > would change into kernel symbols. If you want to help you should:
>
> > 1) run kernel with "debug=mem" option
> > 2) after it hangs up run "dmesg >debug.txt" from amigaos
>
> I can try this again, but the last time I tried this, dmesg couldn't find
> anything.

Can anybody help here ?

> > 3) run kernel version that doesn't hang your amiga
>
> Where do I get a different version from? Sourceforge kernels? I find the
> layout a bit confusing. My currently installed version came from a CD ISO
> image that I burnt to CD and is V2.2 IIRC.

Uh, I'm a bit puzzled, what's an IIRC ?

Yes from there, take 2.4.13 kernel, it should work.

> > 4) copy debug.txt file somewhere on your linux partition
>
> How do I copy from the Amigaside to the linux partition. All my attempts so
> far on cross system copies have not been successful.

First thing you should have partition formatted with FFS Intl. If affs
module isn't compiled into kernel just insert it with insmod.
Then mount it somewhere by "mount -t affs /dev/.... /mnt/..."

> Is there an Amiga dev handler (or something) to allow Linux partitions to be
> mounted? I could not find one on Aminet.

No there isn't (as far as I know), you have to do it from linux side.

> > 5) run "ksymoops -V -K -L -m System.map debug.txt >oops.txt" where
> > System.map file have to be in the same version as unstable kernel (each
> > compiled kernel has only one System.map, so if you compiled kernel
> > yourself System.map should be somewhere in source directory, if you
> > downloaded kernel from internet it should be in archive)
>
> Is this on the Linux side? I have to go now, but I will try this tomorrow
> possibly.

Yep, it's linux specific tool. If you have Debian just get it by
"apt-get install ksymoops".

Regards
Krystian Baclawski

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Michel <mi...@da...>]

On Don, 2003-01-09 at 10:33, Geert Uytterhoeven wrote:
> On 8 Jan 2003, Kulwant Bhogal wrote:
> > > If it's the RDSK partition code, we can probably fix it by looking at
> > > your RDB. You can extract the RDB by dumping the first 2 cylinders to a
> > > file: 
> > >   dd if=/dev/sdb of=rdb.dump bs=4k count=xxx
> > 
> > > Set `xxx' such that `xxx*4096' equals the number of bytes in the first 2
> > > cylinders of the disk (use HdToolBox to find out the cylinder size).
> > > Then sent us the output of `hexdump rdb.dump'.
> > 
> > OK, because I got a count of 615 blocks per cylinder, I used 
> > 615*2*4096=5,038,080 as the count. So the commandline I used was:
> > 
> > # dd if=/dev/sdb of=rdb.dump bs=4k count=5038080
                                 ^^^^^^^^^^^^^^^^^^^
> > 
> > After a few moments I got back: 
> > 
> > dd: /dev/sdb: Input/output error
> > 
> > ho hum.
> 
> Looks like the driver itself has problems.

Is the RDB really 20 gigs though? :)


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 8 Jan 2003, Kulwant Bhogal wrote:
> > If it's the RDSK partition code, we can probably fix it by looking at
> > your RDB. You can extract the RDB by dumping the first 2 cylinders to a
> > file: 
> >   dd if=/dev/sdb of=rdb.dump bs=4k count=xxx
> 
> > Set `xxx' such that `xxx*4096' equals the number of bytes in the first 2
> > cylinders of the disk (use HdToolBox to find out the cylinder size).
> > Then sent us the output of `hexdump rdb.dump'.
> 
> OK, because I got a count of 615 blocks per cylinder, I used 
> 615*2*4096=5,038,080 as the count. So the commandline I used was:
> 
> # dd if=/dev/sdb of=rdb.dump bs=4k count=5038080
> 
> After a few moments I got back: 
> 
> dd: /dev/sdb: Input/output error
> 
> ho hum.

Looks like the driver itself has problems.

> The other question I have is, how do I then copy the rdb.dump file to the
> Amiga side for mailing? I tried to mount another Amiga partition (sda2) as

mount -t affs /dev/<amiga_partition> /mnt

> target/amigaside and tried to copy the rdb.dump file (which I could see if I
> did an ls) by using cp rdb.dump target/amigaside, but I do not see the file
> when I get back to AmigaOS. 

I hope you didn't format that target partition?

> I also do not know how big this file is (it may be 0 bytes). What is the way
> to get a file listing with sizes? 

ls -l

(`man ls' or `ls -- help')

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Krystian,

> Hmm. It'd be more useful if you used ksymoops. 

Sorry i am a complete Linux newbie, where/how do I use ksymoops? I have never seen this before. Is it a
bootstrap commandline option? 
Remember I do not get to a Linux comandline prompt when trying to start Linux
as the kernel crashes while trying to set the system clock. I can only get
to a Linux Shell when I run the installer (which doesn't crash).

> Then those magic numbers
> would change into kernel symbols. If you want to help you should:

> 1) run kernel with "debug=mem" option
> 2) after it hangs up run "dmesg >debug.txt" from amigaos

I can try this again, but the last time I tried this, dmesg couldn't find
anything. 

> 3) run kernel version that doesn't hang your amiga

Where do I get a different version from? Sourceforge kernels? I find the
layout a bit confusing. My currently installed version came from a CD ISO
image that I burnt to CD and is V2.2 IIRC. 

> 4) copy debug.txt file somewhere on your linux partition

How do I copy from the Amigaside to the linux partition. All my attempts so
far on cross system copies have not been successful. 

Is there an Amiga dev handler (or something) to allow Linux partitions to be
mounted? I could not find one on Aminet.

> 5) run "ksymoops -V -K -L -m System.map debug.txt >oops.txt" where
> System.map file have to be in the same version as unstable kernel (each
> compiled kernel has only one System.map, so if you compiled kernel
> yourself System.map should be somewhere in source directory, if you
> downloaded kernel from internet it should be in archive)

Is this on the Linux side? I have to go now, but I will try this tomorrow
possibly. 

Kind regards, 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert,

> If it's the RDSK partition code, we can probably fix it by looking at
> your RDB. You can extract the RDB by dumping the first 2 cylinders to a
> file: 
>   dd if=/dev/sdb of=rdb.dump bs=4k count=xxx

> Set `xxx' such that `xxx*4096' equals the number of bytes in the first 2
> cylinders of the disk (use HdToolBox to find out the cylinder size).
> Then sent us the output of `hexdump rdb.dump'.

OK, because I got a count of 615 blocks per cylinder, I used 
615*2*4096=5,038,080 as the count. So the commandline I used was:

# dd if=/dev/sdb of=rdb.dump bs=4k count=5038080

After a few moments I got back: 

dd: /dev/sdb: Input/output error

ho hum.
The other question I have is, how do I then copy the rdb.dump file to the
Amiga side for mailing? I tried to mount another Amiga partition (sda2) as
target/amigaside and tried to copy the rdb.dump file (which I could see if I
did an ls) by using cp rdb.dump target/amigaside, but I do not see the file
when I get back to AmigaOS. 
I also do not know how big this file is (it may be 0 bytes). What is the way
to get a file listing with sizes? 
Dir doesn't seem to work - I get not found. I could see it in one of the bin
directories, but trying to run that came back with some sort of error trying
to load shared libraries. 
Apologies for all the simpleton questions, but as you can probably guess, I
am very new to LINUX. I am moderately proficient with AmigaOS, MSDOS and
Windows and quite basic on emulated 68k Macintosh but this is my first foray
into LINUX. 

Kind regards, 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 7 Jan 2003, Kulwant Bhogal wrote:
> I noticed during the initialisation of the IDE partition that the disk light
> was on and that the power light level varied (as it does when the audio
> filter is being turned on and off or the machine has crashed). Is this normal?
> It can be a little unnerving at first.

That's the heartbeat feature :-) The more busy your machine is, the faster it
blinks.

> >> Partition check: 
> >> 
> >> sda: RDSK sda1 sda2 sda3 sda4 sda5 sda6 sda7
> >> 
> >> sdb: sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
> >> sector 0
> >> unable to read partition table.
> 
> > Either the SCSI disk driver or the RDSK partition code cannot handle
> > 4096-byte sectors.
> 
> I hope it can be fixed, because otherwise, I will have the hassle of trying
> to get them converted somehow. 

If it's the RDSK partition code, we can probably fix it by looking at your RDB.
You can extract the RDB by dumping the first 2 cylinders to a file:

  dd if=/dev/sdb of=rdb.dump bs=4k count=xxx

Set `xxx' such that `xxx*4096' equals the number of bytes in the first 2
cylinders of the disk (use HdToolBox to find out the cylinder size).
Then sent us the output of `hexdump rdb.dump'.

> > So the driver doesn't seem to handle 4K sectors. Anyone on lkml who has a
> > definitive answer about that?
> 
> Haven't seen a response yet. 

Yes, unfortunate.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Krystian B. <cah...@st...>]

Hi !

> Hello all,
>
> OK, tried starting Linux again and it got past the bit where it hung before
> and instead crashed after trying to initialise the hardware clock. Can I
> disable this update by using a startup switch to the bootstrap command?
>
> The error message was something like this (as much as I could scribble
> down):
>
> Setting the System clock using the hardware clock as reference...
> NIP: c00dca04     XER: 20000000
> LR: c00dc9f8    REGS: c5c8dd60
> TRAP 0300
> MSR: 00001072    EE: 0    PR: 0     FP: 0 ME
> call backtrace
>
> blaa blaa
>
> kernel access of bad area pc c00daa04    IP    c00dca04
> lr c00dc9f8    address 70 tsk hwclock/25
>
> and it said kernel panic somewhere and rebooting in 180 secs.

Hmm. It'd be more useful if you used ksymoops. Then those magic numbers
would change into kernel symbols. If you want to help you should:

1) run kernel with "debug=mem" option
2) after it hangs up run "dmesg >debug.txt" from amigaos
3) run kernel version that doesn't hang your amiga
4) copy debug.txt file somewhere on your linux partition
5) run "ksymoops -V -K -L -m System.map debug.txt >oops.txt" where
System.map file have to be in the same version as unstable kernel (each
compiled kernel has only one System.map, so if you compiled kernel
yourself System.map should be somewhere in source directory, if you
downloaded kernel from internet it should be in archive)
6) send your oops.txt file, some hardware info and of course linux
kernel version to linux-apus-devel list

Kind Regards
Krystian Baclawski

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello all,

OK, tried starting Linux again and it got past the bit where it hung before
and instead crashed after trying to initialise the hardware clock. Can I
disable this update by using a startup switch to the bootstrap command? 

The error message was something like this (as much as I could scribble
down): 

Setting the System clock using the hardware clock as reference...
NIP: c00dca04     XER: 20000000
LR: c00dc9f8    REGS: c5c8dd60
TRAP 0300
MSR: 00001072    EE: 0    PR: 0     FP: 0 ME
call backtrace

blaa blaa

kernel access of bad area pc c00daa04    IP    c00dca04
lr c00dc9f8    address 70 tsk hwclock/25

and it said kernel panic somewhere and rebooting in 180 secs.

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert, All,

>> Warning: Could not read block 0: Attempt to read block from filesystem
>> resulted in short read.

> Oops...

Not fatal then?

>> Writing inode tables: []1/39
>> 
>> At which point the machine appeared to hang (I left it for several
>> minutes). It would accept a carriage return (I would get a new line) but
>> I didn't get the command prompt (#) back - even after a Ctrl-C. So I
>> rebooted.  

Do you think the machine was doing anything useful when I interrupted it? I
don't think the hard drive activity light was flashing but I'm not 100%
certain. 

Anyway, I have now set-up the Linux partitions on the 420Mb IDE and the installation
program recognises them and they are now initialised and activated. I have
installed the base system on the hda2 root but I could find no command in
the installation menu to "configure the base system". (I am following the
Debian installation instructions from http://abone.turk.net/yurdaerd (they
seemed the most up to date and the author's system is closely matched with
my own). 

I noticed during the initialisation of the IDE partition that the disk light
was on and that the power light level varied (as it does when the audio
filter is being turned on and off or the machine has crashed). Is this normal?
It can be a little unnerving at first.

Does it sound right that I cannot use the menu items for "Install the
OS kernel and Modules". Numerous variations of the instructions from various
places say that this option does not work on PPC Amigas yet and that was my
experience also - however I want to make sure that this is really the case
and that I wasn't doing something wrong. 

The trouble is that after installing the base system, which completed ok,
after a reboot, the linux_start script hangs after the Amiga side filesystems
have been inhibited and before Linux actually boots up. 

My bootstrap line for starting Linux looks like this (the contents of the linux_start script): 

bootstrap --apus -k linux.bin root=/dev/hda2 60nsram nobats video=pm2fb:mode:1024x768-75


The rest of this message is about the scsi problems..

>> Vendor: SEAGATE              Model ST34573LW
>> Type: Direct Access            Rev: 6246        ANSI SCSI Revision 02
>> 
>> SCSI device sda: hdwr sector=512 bytes. Sectors=35566000 [17366MB]
>> [17.4GB] SCSI Device sdb: hdwr sector = 4096 bytes. Sectors = 1271205
>> [4964MB][5.0GB] SCSI Device sdc: hdwr sector = 4096 bytes. Sectors =
>> 1271205 [4964MB][5.0GB] SCSI Device sdd: hdwr sector = 512 bytes.
>> Sectors = 17773524 [8678MB][8.7GB]

> So those disks are recognized correctly as having 4096-byte sectors.

Yes. Strange isn't it. 

>> Partition check: 
>> 
>> sda: RDSK sda1 sda2 sda3 sda4 sda5 sda6 sda7
>> 
>> sdb: sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
>> sector 0
>> unable to read partition table.

> Either the SCSI disk driver or the RDSK partition code cannot handle
> 4096-byte sectors.

I hope it can be fixed, because otherwise, I will have the hassle of trying
to get them converted somehow. 

> So the driver doesn't seem to handle 4K sectors. Anyone on lkml who has a
> definitive answer about that?

Haven't seen a response yet. 

> I've had them living on one disk for many many years. But I have to admit
> all my disks are < 4 GB.

Saw an email from John Lord who has been using a 10 Gig IDE shared between
Amiga and Linux for 18 months with no ill effects. Sounds promising. 

>> I included the bits about the speeds because on the Amiga side, my
>> drives accept 10MHz synchronous. Why does LINUX manage only 5MHz? Does
>> this get better with a proper installation? 

> I don't know.

Does anyone else? 

Kind regards, 

Kulwant

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[John L. <ch...@la...>]

Hi Geert

On 07-Jan-03, you wrote:

> So the driver doesn't seem to handle 4K sectors. Anyone on lkml who
> has a definitive answer about that?

>> > The partitions look fine (cfr. your other email). So having Linux
>> > and Amiga partitions shared on the 18 GB disk should work fine.

>> Should? That doesn't sound like 100%. I I think I will try the 420Mb
>> Connor first, now that I have the data from it backed up. Something
>> makes me slightly uneasy about sharing AmigaOS native and foreign
>> partitions on the same disk especially when the disk is >4Gb which in
>> itself requires OS patches to get it recognised and working. Just
>> being cautious :-).

> I've had them living on one disk for many many years. But I have to
> admit all my disks are < 4 GB.

I have been running a 10 Gig IDE drive as both my AmigaOS AND Linux boot
for about 18 months or so. ATM I can say that I've not lost any data on
the miggy partitions because of a fault on the linux ones. And I have
had errors on the linux side that made me run fsdsk manually to repair
it.

And all the partitions AmigaOS AND linux have a block size of 1024.


Regards
-- 
John

Chaos often breeds life, when order breeds habit. - Henry B. Adams

mailto:ch...@la... or mailto:lor...@uk...
http://www.lordofchaos.co.uk ICQ 110130385 
Nick = Chaos A1200 BlizzardPPC 240mhz 060/50 BlizzardVision OS3.9

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 7 Jan 2003, Kulwant Bhogal wrote:
> > mke2fs /dev/<DEVICE>
> 
> Ok, I tried mke2fs dev/sdc because mke2fs /dev/sdc1 kinda obviously didn't
> work because LINUX can't see the partitions. And this is what I got: 
> 
> # mke2fs /dev/sdc
> 
> mke2fs 1.27 (8-Mar-2002)
> /dev/sdc is entire device, not just one partition!
> proceed anyway (y,n) y
> filesystem label= 
> OS type LINUX
> Block size = 4096 (log=2)
> Fragment size = 4096 bytes
> 636480 inodes, 1271205 blocks
> 63560 blocks (5.00%) reserved for the super use.
> First data block=0
> 39 block groups (I think that was a 9 - I can't read my own writing!)
> 32768 blocks per group, 32768 fragments per group 16320 inodes per group
> superblock backups stored on blocks: 
> 32708, 98304, 163840, 229376, 294912, 819200, 884736
> Warning: Could not read block 0: Attempt to read block from filesystem
> resulted in short read.

Oops...

> Writing inode tables: []1/39
> 
> ( [] was basically a black block cursor). 
> 
> At which point the machine appeared to hang (I left it for several minutes). It would accept a carriage return
> (I would get a new line) but I didn't get the command prompt (#) back - even
> after a Ctrl-C. So I rebooted. 
> 
> >> >   2. Does Linux recognize the blocksize during disk probing? E.g. for
> >> > my disks      it says
> >> >      | SCSI device sda: 8910423 512-byte hdwr sectors (4562 MB)
> >> >      | SCSI device sdb: 6281856 512-byte hdwr sectors (3216 MB)
> >> >      during startup.
> > Use `dmesg | less' or `dmesg | more'.
> 
> Here is an excerpt of the relevant bits of the dmesg output: 
> 
> Total memory 96Mb. 
> linux version 2.2.10 (root@pismo)
> (gcc version 2.95.2 20000220
> (Debian GNU/Linux)) #1 Wed Jan 24 00:44:26 CET 2001
> .
> . lots of other output about various other devices and drivers etc - all
> . looked ok.
> .
> SCSI: 1 host
> SCSI0: target 0 accepting period 200ns offset 8 5.00MHz Synchronous SCSI
> SCSI 0: setting target 0 to period 200ns offset 8 5.00MHz synchronous SCSI
> 
> Vendor: COMPAQ         Model DGHS18Y 
> Type: Direct Access     Rev 03F1                 ANSI SCSI Revision 03
> 
> Detected SCSI disk sda at SCSI 0, Channel 0, id 0, lun 0
> 
> Vendor: SEAGATE              Model ST34573LW
> Type: Direct Access            Rev: 6246        ANSI SCSI Revision 02
> 
> SCSI device sda: hdwr sector=512 bytes. Sectors=35566000 [17366MB] [17.4GB]
> SCSI Device sdb: hdwr sector = 4096 bytes. Sectors = 1271205 [4964MB][5.0GB]
> SCSI Device sdc: hdwr sector = 4096 bytes. Sectors = 1271205 [4964MB][5.0GB]
> SCSI Device sdd: hdwr sector = 512 bytes. Sectors = 17773524 [8678MB][8.7GB]

So those disks are recognized correctly as having 4096-byte sectors.

> Partition check: 
> 
> sda: RDSK sda1 sda2 sda3 sda4 sda5 sda6 sda7
> 
> sdb: sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
> sector 0
> unable to read partition table.

Either the SCSI disk driver or the RDSK partition code cannot handle 4096-byte
sectors.

> sdc: sd.cBad block number/count requestedscsidisk I/O error: dev 08:20,
> sector 0
> unable to read partition table.
> 
> sdd: RDSK sdd1 sdd2 sdd3 sdd4 sdd5 
> 
> hda: RDSK hda1 hda2 hda3
> 
> EXT2 - fs warning: checktime reached, running e2fsck is recommended.
> VFS: Mounted root (ext2 filesystem)
> 
> CFS: Disk change detected on device sr(11,0)
> attempt to access beyond end of device
> 0b:00: rw=0, want=33, limit=2
> isofs_read_super: bread failed, 
> dev=ob:00, iso_blknum=16. block=32
> sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
> sector 0
> sdc: sd.cBad block number/count requestedscsidisk I/O error: dev 08:20,
> sector 0
> unable to read partition table.
> 
> 
> I did run e2fsck on dev/sdc and I basically got the same sort of error about
> sector 0, short read and the partition being "zero length?" etc.

So the driver doesn't seem to handle 4K sectors. Anyone on lkml who has a
definitive answer about that?

> > The partitions look fine (cfr. your other email). So having Linux and
> > Amiga partitions shared on the 18 GB disk should work fine.
> 
> Should? That doesn't sound like 100%. I I think I will try the 420Mb Connor
> first, now that I have the data from it backed up. Something makes me
> slightly uneasy about sharing AmigaOS native and foreign partitions on the same disk
> especially when the disk is >4Gb which in itself requires OS patches to get it
> recognised and working. Just being cautious :-).

I've had them living on one disk for many many years. But I have to admit all
my disks are < 4 GB.

> I included the bits about the speeds because on the Amiga side, my drives
> accept 10MHz synchronous. Why does LINUX manage only 5MHz? Does this get
> better with a proper installation? 

I don't know.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert,

Thanks for your email.

> mke2fs /dev/<DEVICE>

Ok, I tried mke2fs dev/sdc because mke2fs /dev/sdc1 kinda obviously didn't
work because LINUX can't see the partitions. And this is what I got: 

# mke2fs /dev/sdc

mke2fs 1.27 (8-Mar-2002)
/dev/sdc is entire device, not just one partition!
proceed anyway (y,n) y
filesystem label= 
OS type LINUX
Block size = 4096 (log=2)
Fragment size = 4096 bytes
636480 inodes, 1271205 blocks
63560 blocks (5.00%) reserved for the super use.
First data block=0
39 block groups (I think that was a 9 - I can't read my own writing!)
32768 blocks per group, 32768 fragments per group 16320 inodes per group
superblock backups stored on blocks: 
32708, 98304, 163840, 229376, 294912, 819200, 884736
Warning: Could not read block 0: Attempt to read block from filesystem
resulted in short read.
Writing inode tables: []1/39

( [] was basically a black block cursor). 

At which point the machine appeared to hang (I left it for several minutes). It would accept a carriage return
(I would get a new line) but I didn't get the command prompt (#) back - even
after a Ctrl-C. So I rebooted. 

>> >   2. Does Linux recognize the blocksize during disk probing? E.g. for
>> > my disks      it says
>> >      | SCSI device sda: 8910423 512-byte hdwr sectors (4562 MB)
>> >      | SCSI device sdb: 6281856 512-byte hdwr sectors (3216 MB)
>> >      during startup.
> Use `dmesg | less' or `dmesg | more'.

Here is an excerpt of the relevant bits of the dmesg output: 

Total memory 96Mb. 
linux version 2.2.10 (root@pismo)
(gcc version 2.95.2 20000220
(Debian GNU/Linux)) #1 Wed Jan 24 00:44:26 CET 2001
.
. lots of other output about various other devices and drivers etc - all
. looked ok.
.
SCSI: 1 host
SCSI0: target 0 accepting period 200ns offset 8 5.00MHz Synchronous SCSI
SCSI 0: setting target 0 to period 200ns offset 8 5.00MHz synchronous SCSI

Vendor: COMPAQ         Model DGHS18Y 
Type: Direct Access     Rev 03F1                 ANSI SCSI Revision 03

Detected SCSI disk sda at SCSI 0, Channel 0, id 0, lun 0

Vendor: SEAGATE              Model ST34573LW
Type: Direct Access            Rev: 6246        ANSI SCSI Revision 02

SCSI device sda: hdwr sector=512 bytes. Sectors=35566000 [17366MB] [17.4GB]
SCSI Device sdb: hdwr sector = 4096 bytes. Sectors = 1271205 [4964MB][5.0GB]
SCSI Device sdc: hdwr sector = 4096 bytes. Sectors = 1271205 [4964MB][5.0GB]
SCSI Device sdd: hdwr sector = 512 bytes. Sectors = 17773524 [8678MB][8.7GB]

Partition check: 

sda: RDSK sda1 sda2 sda3 sda4 sda5 sda6 sda7

sdb: sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
sector 0
unable to read partition table.

sdc: sd.cBad block number/count requestedscsidisk I/O error: dev 08:20,
sector 0
unable to read partition table.

sdd: RDSK sdd1 sdd2 sdd3 sdd4 sdd5 

hda: RDSK hda1 hda2 hda3

EXT2 - fs warning: checktime reached, running e2fsck is recommended.
VFS: Mounted root (ext2 filesystem)

CFS: Disk change detected on device sr(11,0)
attempt to access beyond end of device
0b:00: rw=0, want=33, limit=2
isofs_read_super: bread failed, 
dev=ob:00, iso_blknum=16. block=32
sd.cBad block number/count requestedscsidisk I/O error: dev 08:10,
sector 0
sdc: sd.cBad block number/count requestedscsidisk I/O error: dev 08:20,
sector 0
unable to read partition table.


I did run e2fsck on dev/sdc and I basically got the same sort of error about
sector 0, short read and the partition being "zero length?" etc.

> The partitions look fine (cfr. your other email). So having Linux and
> Amiga partitions shared on the 18 GB disk should work fine.

Should? That doesn't sound like 100%. I I think I will try the 420Mb Connor
first, now that I have the data from it backed up. Something makes me
slightly uneasy about sharing AmigaOS native and foreign partitions on the same disk
especially when the disk is >4Gb which in itself requires OS patches to get it
recognised and working. Just being cautious :-).

I included the bits about the speeds because on the Amiga side, my drives
accept 10MHz synchronous. Why does LINUX manage only 5MHz? Does this get
better with a proper installation? 

And I found the dmesg text output really slow on my Permidia2 based BVision. Is all this
stuff running on the PPC exclusively? I was expecting blazing speeds....(my
PPC is the 240MHz one). The output scrolled as slow as a full screen CLI with
68k CPU caches turned off. the 50MHz 68060 performs better than that on a
normal 1024x768x24bit screen under AmigaOS...

Kind regards, 

Kulwant










 

Re: Ahah, 4096 blocksize disk is not recognised under Linux?

[Geert U. <ge...@li...>]

On 6 Jan 2003, Kulwant Bhogal wrote:
> > So now you saw the `Launch shell' entry as well? :-)
> 
> Yes, but I didn't know what the syntax for the format command should be. But

mke2fs /dev/<DEVICE>

> I won't be using that anyway now until I create some space for the Linux
> partitions on a "normal" 512 blocksize disk. Unless of course there is a
> Linux command that I can use to do the low level format :-)
> 
> >> Partitions list
> >> 
> >> The following partitions have been detected: 
> >> 
> >> Device        Mounted On            Type
> >> /dev/sda1    --Not mounted--    FFS
> >> /dev/sda2    --Not mounted--    FFS
> >> /dev/sda3    --Not mounted--    FFS
> >> /dev/sda4    --Not mounted--    FFS
> >> /dev/sda5    --Not mounted--    FFS
> >> /dev/sda6    --Not mounted--    FFS
> >> /dev/sda7    --Not mounted--    FFS
> >> 
> >> Disk /dev/sdb has no partitons defined
> >> Disk /dev/sdc has no partitions defined
> 
> > Aha, so it could be a problem with the partitioning support. Were these
> > disks partitioned under AmigaOS using HDToolBox? If yes, chances are high
> > that the Amiga RDB (Rigid Disk Block) support doesn't handle 4096-byte
> > blocks yet. 
> 
> Yes the disks were partitioned under AmigaOS using HDToolbox. I do know that
> the Amiga had trouble setting the blocksize to 512, I can set it to 4096 and
> above, but nothing less than 4096. To do this the disks would have to be low
> level formatted with a blocksize of 512. Once a disk has a blocksize of 512,
> the Amiga supports all blocksize from 512 up to 16384 (I think), but the
> low level blocksize is critical. Unfortunately none of the low level format
> commands on the Amiga have succeeded in resetting the block size to 512 :-( 

So probably the Amiga RDB support in Linux doesn't handle 4096-byte blocksizes.

> >   2. Does Linux recognize the blocksize during disk probing? E.g. for my
> > disks      it says
> >      | SCSI device sda: 8910423 512-byte hdwr sectors (4562 MB)
> >      | SCSI device sdb: 6281856 512-byte hdwr sectors (3216 MB)
> >      during startup.
> 
> Will I see this on the text that flashes up when the Penguin appears at the
> top left of the screen and a whole load of white text on a black background
> appears? Trouble is, this stuff looks really interesting, but dissappears

Yes, that's it.

> before I have a chance to read any of it :-( Is there any way I can probe
> again from a shell? 

Use `dmesg | less' or `dmesg | more'.

> I have tried debug=mem, but dmesg or bootmesg never finds anything after a reboot :-(

May not work on APUS, don't know.

> > Linux has its own support for disks, which handles the >4 GB case just
> > fine. Don't know about the Amiga partitioning support, though.
> 
> Does that mean Linux and Amiga partitions shared on a 18Gb disk are a no no? 
> 
> > BTW, since your Compaq is much larger than 4 GB, what does
> > `cat /proc/partitions' say?
> 
> I will try this out andlet you know. 

The partitions look fine (cfr. your other email). So having Linux and Amiga
partitions shared on the 18 GB disk should work fine.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

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: Ahah, 4096 blocksize disk is not recognised under Linux?

[Kulwant B. <kul...@bt...>]

Hello Geert,

> So now you saw the `Launch shell' entry as well? :-)

Yes, but I didn't know what the syntax for the format command should be. But
I won't be using that anyway now until I create some space for the Linux
partitions on a "normal" 512 blocksize disk. Unless of course there is a
Linux command that I can use to do the low level format :-)

>> Partitions list
>> 
>> The following partitions have been detected: 
>> 
>> Device        Mounted On            Type
>> /dev/sda1    --Not mounted--    FFS
>> /dev/sda2    --Not mounted--    FFS
>> /dev/sda3    --Not mounted--    FFS
>> /dev/sda4    --Not mounted--    FFS
>> /dev/sda5    --Not mounted--    FFS
>> /dev/sda6    --Not mounted--    FFS
>> /dev/sda7    --Not mounted--    FFS
>> 
>> Disk /dev/sdb has no partitons defined
>> Disk /dev/sdc has no partitions defined

> Aha, so it could be a problem with the partitioning support. Were these
> disks partitioned under AmigaOS using HDToolBox? If yes, chances are high
> that the Amiga RDB (Rigid Disk Block) support doesn't handle 4096-byte
> blocks yet. 

Yes the disks were partitioned under AmigaOS using HDToolbox. I do know that
the Amiga had trouble setting the blocksize to 512, I can set it to 4096 and
above, but nothing less than 4096. To do this the disks would have to be low
level formatted with a blocksize of 512. Once a disk has a blocksize of 512,
the Amiga supports all blocksize from 512 up to 16384 (I think), but the
low level blocksize is critical. Unfortunately none of the low level format
commands on the Amiga have succeeded in resetting the block size to 512 :-( 

>> /dev/sdd1    --Not mounted--    FFS
>> /dev/sdd2    --Not mounted--    FFS
>> /dev/sdd3    --Not mounted--    FFS
>> /dev/sdd4    --Not mounted--    FFS
>> /dev/sdd5    --Not mounted--    FFS
>> 
>> /dev/hda1    --Not mounted--    FFS
>> /dev/hda2    --Not mounted--    FFS
>> /dev/hda3    --Not mounted--    FFS
>> 
>> where:
>> 
>> sda is a Compaq 18Gb with a 512 blocksize
>> sdb is a Seagate Baracudda 4.8Gb with a 4096 blocksize
>> sdc is a Seagate Baracudda 4.8Gb with a 4096 blocksize
>> sdd is a Compaq 9Gb with a 512blocksize and
>> hda is a Connor 420Mb with a 512 blocksize
>> 
>> And all of the disks have a number of partitions each. 
>> 
>> sdc is the drive I have set up the Linux partitions on. 
>> 
>> All the SCSI drives are second hand and the Seagates came with the 4096
>> blocksize and no Amiga program has been able to low level format them
>> with a 512 blocksize. Only a Seagate tool (which runs on PC only) is
>> available for this job apparently. 

> Just thinking of somethings:
>   1. Is there a jumper on the drive to choose the blocksize?

None that I am aware of, just the ones one would expect on a  SCSI drive - 
SCSI ID, Motor start & delay options, Write Protect, Parity check,
SingleEnded I/O & Terminator power. 

>   2. Does Linux recognize the blocksize during disk probing? E.g. for my
> disks      it says
>      | SCSI device sda: 8910423 512-byte hdwr sectors (4562 MB)
>      | SCSI device sdb: 6281856 512-byte hdwr sectors (3216 MB)
>      during startup.

Will I see this on the text that flashes up when the Penguin appears at the
top left of the screen and a whole load of white text on a black background
appears? Trouble is, this stuff looks really interesting, but dissappears
before I have a chance to read any of it :-( Is there any way I can probe
again from a shell? 

I have tried debug=mem, but dmesg or bootmesg never finds anything after a reboot :-(

> Should be safe. I never managed to corrupt my AmigaOS partitions from
> within Linux. I know others did, but that was due to bugs in the linux
> AFFS support, which were solved a long time ago.

Sounds reassuring. 

> Linux has its own support for disks, which handles the >4 GB case just
> fine. Don't know about the Amiga partitioning support, though.

Does that mean Linux and Amiga partitions shared on a 18Gb disk are a no no? 

> BTW, since your Compaq is much larger than 4 GB, what does
> `cat /proc/partitions' say?

I will try this out andlet you know. 

Kind regards, 

Kulwant