linux-apus-devel

Re: new parted version available for testing ... (fwd)

[Roman Z. <zi...@li...>]

Hi,

On Fri, 29 Aug 2003, Sven Luther wrote:

> > The block size can be guessed by verifying the checksum of the root block 
> > (e.g. affs does that), but the other values are only in the RDB.
> 
> Seems a nice trick. But it would be a problem if the blocks after the
> one are written only by 0s.

Not if you start with the smallest blocksize, as there is always data 
relativ to the end of the block.

> I guess you could make some guess for the other
> values by testing various values when searching for the rdb.

Not really, to find the root block you already need the reserved value 
(the implementations currently cheat here a bit and assume a small 
reserved value, preferably 2).

> > larger block sizes, but that wasn't successfull, maybe my amiga os version 
> > doesn't support that.
> 
> Ok, thanks, i will have a look. It is GPLed code, right, and there is no
> problem with myself inspiring me from it for writing parted code ?

Yes and no, take whatever you need. :)

bye, Roman

Re: new parted version available for testing ... (fwd)

[Sven L. <sve...@wa...>]

On Fri, Aug 29, 2003 at 02:42:12PM +0200, Roman Zippel wrote:
> Hi,
> 
> On Fri, 29 Aug 2003, Sven Luther wrote:
> 
> > Do someone know if it is possible to autodetect the block size as well
> > as the prealloc and reserved values ?
> 
> The block size can be guessed by verifying the checksum of the root block 
> (e.g. affs does that), but the other values are only in the RDB.

Seems a nice trick. But it would be a problem if the blocks after the
one are written only by 0s.

I guess you could make some guess for the other
values by testing various values when searching for the rdb.

> > Then i will had formatting support for affs and asfs, and it would be
> > possible to create a new bootable system from linux without ever going
> > to amigaos/morphos.
> 
> BTW I have some basic affs tools at
> linux-apus.sf.net/test/affstools-0.1.tar.gz
> It's at least enough to create an affs file system, I played a bit with 
> larger block sizes, but that wasn't successfull, maybe my amiga os version 
> doesn't support that.

Ok, thanks, i will have a look. It is GPLed code, right, and there is no
problem with myself inspiring me from it for writing parted code ?

> > Now, what is the situation of the boot loader on amiga/m68k and
> > amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
> > program we used back then ? If the later is the case, it should be
> > possible to cross compile it, and then to make a debian package with it,
> > installing it in the affs formatted /boot partition or a affs
> > /boot/amiboot or something such.
> 
> Lilo can be be cross compiled, for a complete compile you need a 
> m68k-linux cross compiler, but I included precompiled sources. It works 
> fine for linux-ppc. As Geert mentioned it still needs a small patch for 
> the kernel.

Ok.

Friendly,

Sven Luther

Re: new parted version available for testing ... (fwd)

[Roman Z. <zi...@li...>]

Hi,

On Fri, 29 Aug 2003, Sven Luther wrote:

> Do someone know if it is possible to autodetect the block size as well
> as the prealloc and reserved values ?

The block size can be guessed by verifying the checksum of the root block 
(e.g. affs does that), but the other values are only in the RDB.

> Then i will had formatting support for affs and asfs, and it would be
> possible to create a new bootable system from linux without ever going
> to amigaos/morphos.

BTW I have some basic affs tools at
linux-apus.sf.net/test/affstools-0.1.tar.gz
It's at least enough to create an affs file system, I played a bit with 
larger block sizes, but that wasn't successfull, maybe my amiga os version 
doesn't support that.

> Now, what is the situation of the boot loader on amiga/m68k and
> amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
> program we used back then ? If the later is the case, it should be
> possible to cross compile it, and then to make a debian package with it,
> installing it in the affs formatted /boot partition or a affs
> /boot/amiboot or something such.

Lilo can be be cross compiled, for a complete compile you need a 
m68k-linux cross compiler, but I included precompiled sources. It works 
fine for linux-ppc. As Geert mentioned it still needs a small patch for 
the kernel.

bye, Roman

Re: new parted version available for testing ... (fwd)

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

On Fri, 29 Aug 2003, Sven Luther wrote:
> On Fri, Aug 29, 2003 at 01:26:48PM +0200, Geert Uytterhoeven wrote:
> > On Fri, 29 Aug 2003, Sven Luther wrote:
> > > Now, what is the situation of the boot loader on amiga/m68k and
> > > amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
> > > program we used back then ? If the later is the case, it should be
> > > possible to cross compile it, and then to make a debian package with it,
> > > installing it in the affs formatted /boot partition or a affs
> > > /boot/amiboot or something such.
> > 
> > Amiboot and Amiga-Lilo are part of m68kboot, which is available from
> > cvs.linux-m68k.org.
> 
> So there is Amiga-lilo that you can install from linux ? If so, no need

Yes.

> to do it differently. This is not true for apus though, right ?

APUS support is experimental and needs a kernel patch, IIRC. Roman can tell you
more about it.

> > Recently Roman Zippel made some changes (e.g. preliminary APUS support) to
> > Amiga-Lilo: http://www.xs4all.nl/~zippel/amiga-lilo-2.3.tar.gz
> > 
> > To cross-compile Amiboot, you need to package ADE (Amiga Development
> > Environment) first.
> 
> Yes, altough i guess you could build it on amigaos, and then just use it
> for installing later on.

I guess the Debian autobuilders won't like that ;-)

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: new parted version available for testing ... (fwd)

[Sven L. <sve...@wa...>]

On Fri, Aug 29, 2003 at 01:26:48PM +0200, Geert Uytterhoeven wrote:
> On Fri, 29 Aug 2003, Sven Luther wrote:
> > Now, what is the situation of the boot loader on amiga/m68k and
> > amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
> > program we used back then ? If the later is the case, it should be
> > possible to cross compile it, and then to make a debian package with it,
> > installing it in the affs formatted /boot partition or a affs
> > /boot/amiboot or something such.
> 
> Amiboot and Amiga-Lilo are part of m68kboot, which is available from
> cvs.linux-m68k.org.

So there is Amiga-lilo that you can install from linux ? If so, no need
to do it differently. This is not true for apus though, right ?

> Recently Roman Zippel made some changes (e.g. preliminary APUS support) to
> Amiga-Lilo: http://www.xs4all.nl/~zippel/amiga-lilo-2.3.tar.gz
> 
> To cross-compile Amiboot, you need to package ADE (Amiga Development
> Environment) first.

Yes, altough i guess you could build it on amigaos, and then just use it
for installing later on.

Friendly,

Sven Luther

Re: new parted version available for testing ... (fwd)

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

On Fri, 29 Aug 2003, Sven Luther wrote:
> Now, what is the situation of the boot loader on amiga/m68k and
> amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
> program we used back then ? If the later is the case, it should be
> possible to cross compile it, and then to make a debian package with it,
> installing it in the affs formatted /boot partition or a affs
> /boot/amiboot or something such.

Amiboot and Amiga-Lilo are part of m68kboot, which is available from
cvs.linux-m68k.org.

Recently Roman Zippel made some changes (e.g. preliminary APUS support) to
Amiga-Lilo: http://www.xs4all.nl/~zippel/amiga-lilo-2.3.tar.gz

To cross-compile Amiboot, you need to package ADE (Amiga Development
Environment) first.

Similar things are true for the Atari bootstrap...

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: new parted version available for testing ... (fwd)

[Sven L. <sve...@wa...>]

On Fri, Aug 29, 2003 at 12:12:18AM +0200, Geert Uytterhoeven wrote:
> On Thu, 28 Aug 2003, Roman Zippel wrote:
> > On Thu, 28 Aug 2003, Sven Luther wrote:
> > > > I've put m68k packages for stable and unstable at 
> > > > http://www.xs4all.nl/~zippel/parted/
> > > 
> > > Could you rebuild the new 2.0.0.3 version, please ?
> > > I fixed partition table writing when no amiga partition was previously
> > > found on the disk and incorporated your below changes.
> > 
> > I've updated the stable packages.
> 
> Thanks! This time it seems to work fine on my Amiga (Sven: I'll send you an
> output dump in private email).

Yep, it works nicely. The only problem remaining are the amiga filesystems
detection when using unconventional block sizes, since parted does not
(yet) provide functionality to pass something else from the partition
block to the filesystem code.

Do someone know if it is possible to autodetect the block size as well
as the prealloc and reserved values ?

Then i will had formatting support for affs and asfs, and it would be
possible to create a new bootable system from linux without ever going
to amigaos/morphos.

A yes, i need to add also a way to change partition attributes, like the
boot priority.

Now, what is the situation of the boot loader on amiga/m68k and
amiga/apus ? Do you have a lilo-like thingy, or still the ami/apusboot
program we used back then ? If the later is the case, it should be
possible to cross compile it, and then to make a debian package with it,
installing it in the affs formatted /boot partition or a affs
/boot/amiboot or something such.

Friendly,

Sven Luther

Re: dmasound (wa: Re: Kernel 2.4.21)

[<ami...@vo...>]

>>>>> Looks like a function call through a NULL pointer at
>>>>> dmasound_init+908c/c0a0.
>>>> But what this exactly means? I'm sorry, but I don't understand. I
>>>> have never used any debuging stuff other than printf()/printk()...
>>> dmasound_core calls the low-level chipset-specific routines through
>>> function pointers. If one of them is not set up, it will jump to
>>> address 0, causing a crash.
>> Ok. So why the same code did not crashed on kernel 2.4.18?
> Does this patch fix your problem with dmasound?
> 
> http://linux-m68k-cvs.apia.dhs.org/~geert/linux-m68k-2.4.x-merging/117-d
> masound.diff

The same problem here :-(

-- 
Ondrej Zima <ami...@vo...>
Member of Czech ATO - Amiga Translators Organization

dmasound (wa: Re: Kernel 2.4.21)

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

On Wed, 30 Jul 2003, Marek Szyprowski wrote:
> On 30.07.03, you wrote:
> > On Wed, 30 Jul 2003, Marek Szyprowski wrote:
> >> On 30.07.03, you wrote:
> >>> Looks like a function call through a NULL pointer at
> >>> dmasound_init+908c/c0a0.
> 
> >> But what this exactly means? I'm sorry, but I don't understand. I have
> >> never used any debuging stuff other than printf()/printk()...
> 
> > dmasound_core calls the low-level chipset-specific routines through
> > function pointers. If one of them is not set up, it will jump to address
> > 0, causing a crash.
> 
> Ok. So why the same code did not crashed on kernel 2.4.18?

Does this patch fix your problem with dmasound?

http://linux-m68k-cvs.apia.dhs.org/~geert/linux-m68k-2.4.x-merging/117-dmasound.diff

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

new parted version available for testing ...

[Sven L. <sve...@wa...>]

Hello everyone.

I have a new parted version available for testing at :

  deb http://people.debian.org/~luther/parted .
 
for apt-get, but you can also get it directly.

It is a full build, but you only need the libparted and parted packages.

This one has full partition table read/write acces, altough the write
support is not yet intesively tested, so please make a copy of your
partition table first. How you do that ? You simply launch parted, use
the print option, and look at the rdb_RDBBlocksHi value of the rdb (609
for me) and then you can do :

  # dd if=/dev/sda of=sda.backup count=609

Which will save the whole partiton table (in my case naturally). You can
restore it with :

  # dd if=sda.backup of=/dev/sda

And everything will be as usual.

There isa lot of debugging information included, but it goes to the
stderr, so you can write it to a log file with :

  # parted 2> parted.log

And if you have any problems, or see wrong stuff in your partition
table, please send it to me.

The code for detecting the filesystem type for ffs, sfs and pfs
filesystem doesn't yet include support for block sizes other than 512.
Support for that would need some changes in parted, but i will be
working on it nextly.

Ffs, sfs and pfs filesystems only support the probing code right now, so
you cannot create new or duplicate or copy existing filesystemes of
these types.

Ok, that is all for this week, please give it a try if you have a spare
harddisk or something, and report to me anything special you notice.

Note to the linux-m68k people, Andrew Clausen said that he (or i) will
add the fixes to the ufs code needed for running on m68k, but it is not
yet in this version, so you have to disable the ufs and maybe reiserfs
filesystems as before. Geert, if i remember well, the amiga filesystem
stuff did also segfault on your m68k box, if this is still true, i would
like to debug this stuff nextly. Also, could someone forward this mail
to the corresponding linux-m68k mailing list, and/or send me the
corresponding address, so i can include it in a next roundmail about
this.

BTW, if someone feels like building parted on morphos i would be happy
to know the result. There doesn't seem to be any fork or something such
in the code, and the low level read/write stuff is either using the
linux arch or the gnu arch. I guess the gnu arch should be able to run
with the ixemul.library, or whatever similar thing we have in morphos
these days. I would do it myself, but i am not really familiar with the
morphos development environment yet.

Friendly,

Sven Luther

Re: Hello everyone ...

[Sven L. <sve...@wa...>]

On Thu, Aug 14, 2003 at 12:36:53PM +0200, Geert Uytterhoeven wrote:
> On Wed, 13 Aug 2003, Sven Luther wrote:
> > On Wed, Aug 13, 2003 at 06:11:52PM +0200, Geert Uytterhoeven wrote:
> > > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > > On Wed, Aug 13, 2003 at 05:34:35PM +0200, Geert Uytterhoeven wrote:
> > > > > BTW, removing the (un)register calls for UFS didn't help, as expected (the
> > > > > assertion just causes UFS not being registered).
> > > > 
> > > > What about removing the affs stuff ?
> > > 
> > > Still the same ;-(
> > 
> > Ok, i have to go today, but i will prepare a new version tomorrow, with
> > more debuging messages. It would be nice to know which function is
> > called when the probe is done. I expect you don't get the real values,
> > because it is a a table with function pointers or so. Could you try to
> > get which of the function pointers is called with gdb or whatever you
> > where using?
> 
> Apparently the last probe function called before the crash is reiserfs_probe(),

So, maybe this has to do with you don't having installed the reiserfs
-dev libraries, which somehow makes sense.

> which seems to corrupt the stack. After disabling the registration of reiserfs
> I get:
> 
> | Disk geometry for /dev/hda: 0.000-1039.500 megabytes
> | Disk label type: amiga
> | Minor    Start       End     Filesystem  Flags
> | 1          0.984      9.351              
> | 2          9.352    259.382              

These two are the MuFS partitions.

> | 3        259.383    291.867  ext2        
> | 4        291.867    324.351  linux-swap  
> | 5        324.352    424.757  ext2        
> | 6        424.758   1039.500  ext2        
> 
> Which is consistent with the data from HdToolBox.
> 
> After re-enabling affs and adding support for MultiUser (patch below), I get:
> 
> | PROBING FOR AN AMIGA PARTITION TABLE.
> | Reserved = 2, Prealloc = 0
> | Cyl = 2112, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> | Block size is 128
> | AMIGA FFS : boot block detected FFS\3 
> | AMIGA FFS : root block should be at 10584 (2016-19151:17136)
> | AMIGA FFS : root-2 block : type = 653206047, subtype = -628857599, sum =
> | -1976770243
> | AMIGA FFS : root-1 block : type = 637044347, subtype = -1881946650, sum =
> | 1885671754
> | AMIGA FFS : root+1 block : type = 0, subtype = 0, sum = 0
> | AMIGA FFS : root+2 block : type = -33552265, subtype = -1, sum = 0
> | AMIGA FFS : root block : type = 2, subtype = 1, sum = 0

Ok, i correctly calculate the root block place here.

> | Program received signal SIGSEGV, Segmentation fault.
> | 0xc012e16a in free () from /lib/libc.so.6
> 
> :-(

Err, that should not happen :(((

> For /dev/hdb, which doesn't contain any affs partitions:
> 
> | PROBING FOR AN AMIGA PARTITION TABLE.
> | Reserved = 0, Prealloc = 0
> | Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> | Block size is 128
> | AMIGA FFS : boot block detected FFS\3 
> | AMIGA FFS : root block should be at 287280 (2016-572543:570528)
> | AMIGA FFS : root-2 block : type = 1026176313, subtype = 1598254958, sum =
> | 1968231241
> | AMIGA FFS : root-1 block : type = 1937011297, subtype = 842283008, sum =
> | 302493690
> | AMIGA FFS : root+1 block : type = 741552950, subtype = 741552699, sum =
> | -1510802252
> | AMIGA FFS : root+2 block : type = 1634495599, subtype = 1601200498, sum =
> | -1940758414
> | AMIGA FFS : root block : type = 1599038058, subtype = 976433203, sum =
> | -463838221

The root block is not found, but then you have 0 reserved blocks here.

> | Reserved = 0, Prealloc = 0
> | Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> | Block size is 128
> | Disk geometry for /dev/hdb: 0.000-520.242 megabytes
> | Disk label type: amiga
> | Minor    Start       End     Filesystem  Flags
> | 1          0.984    279.562  ext2        
> | 2        279.562    520.242  ext2       
> 
> Which is OK.

But i suppose that the first partition has a wrong boot block of DOS\3
type, but since libparted does more advanced checking, and doesn't find
the root block, it is correcly detected as ext2.

> --- parted-1.6.5-1.0.0.1/libparted/fs_affs/interface.c.orig	Thu Aug 14 12:16:10 2003
> +++ parted-1.6.5-1.0.0.1/libparted/fs_affs/interface.c	Thu Aug 14 12:20:08 2003
> @@ -55,6 +55,13 @@
>  		case 0x444f5303 :	/* 'DOS\3' : FFS Int. */
>  		case 0x444f5304 :	/* 'DOS\4' : OFS DirCache */
>  		case 0x444f5305 :	/* 'DOS\5' : FFS DirCache */
> +		case 0x6d754653 :	/* 'muFS'  : MultiUser FFS Int. */
> +		case 0x6d754600 :	/* 'muF\0' : MultiUser OFS */
> +		case 0x6d754601 :	/* 'muF\1' : MultiUser FFS */
> +		case 0x6d754602 :	/* 'muF\2' : MultiUser OFS Int. */
> +		case 0x6d754603 :	/* 'muF\3' : MultiUser FFS Int. */
> +		case 0x6d754604 :	/* 'muF\4' : MultiUser OFS DirCache */
> +		case 0x6d754605 :	/* 'muF\5' : MultiUser FFS DirCache */
>  			printf ("AMIGA FFS : boot block detected FFS\\%d \n", block[0]&0xff);
>  			printf ("AMIGA FFS : root block should be at %llu (%llu-%llu:%llu)\n",
>  				root, geom->start, geom->end, geom->length);

Ok thanks, i had added similar code already.

I will try to prepare a new version today, with more extensive
information dumping, and we can go looking at why it segfaults above. It
clearly is in the affs code, so i am responsible for it.

Thanks for testing.

Friendly,

Sven Luther

Re: Hello everyone ...

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

On Thu, 14 Aug 2003, Sven Luther wrote:
> On Thu, Aug 14, 2003 at 12:36:53PM +0200, Geert Uytterhoeven wrote:
> > | PROBING FOR AN AMIGA PARTITION TABLE.
> > | Reserved = 0, Prealloc = 0
> > | Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> > | Block size is 128
> > | AMIGA FFS : boot block detected FFS\3 
> > | AMIGA FFS : root block should be at 287280 (2016-572543:570528)
> > | AMIGA FFS : root-2 block : type = 1026176313, subtype = 1598254958, sum =
> > | 1968231241
> > | AMIGA FFS : root-1 block : type = 1937011297, subtype = 842283008, sum =
> > | 302493690
> > | AMIGA FFS : root+1 block : type = 741552950, subtype = 741552699, sum =
> > | -1510802252
> > | AMIGA FFS : root+2 block : type = 1634495599, subtype = 1601200498, sum =
> > | -1940758414
> > | AMIGA FFS : root block : type = 1599038058, subtype = 976433203, sum =
> > | -463838221
> 
> The root block is not found, but then you have 0 reserved blocks here.
> 
> > | Reserved = 0, Prealloc = 0
> > | Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> > | Block size is 128
> > | Disk geometry for /dev/hdb: 0.000-520.242 megabytes
> > | Disk label type: amiga
> > | Minor    Start       End     Filesystem  Flags
> > | 1          0.984    279.562  ext2        
> > | 2        279.562    520.242  ext2       
> > 
> > Which is OK.
> 
> But i suppose that the first partition has a wrong boot block of DOS\3
> type, but since libparted does more advanced checking, and doesn't find
> the root block, it is correcly detected as ext2.

I used to use that disk for Lilo testing, too. AmigaOS cannot boot from
partitions that have an unknown ID in the first u32 of the partition, so it had
to be one of the DOS* types.

So yes, you can have a DOS* type in the first u32, but a different file system.

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: Hello everyone ...

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

On Wed, 13 Aug 2003, Sven Luther wrote:
> On Wed, Aug 13, 2003 at 06:11:52PM +0200, Geert Uytterhoeven wrote:
> > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > On Wed, Aug 13, 2003 at 05:34:35PM +0200, Geert Uytterhoeven wrote:
> > > > BTW, removing the (un)register calls for UFS didn't help, as expected (the
> > > > assertion just causes UFS not being registered).
> > > 
> > > What about removing the affs stuff ?
> > 
> > Still the same ;-(
> 
> Ok, i have to go today, but i will prepare a new version tomorrow, with
> more debuging messages. It would be nice to know which function is
> called when the probe is done. I expect you don't get the real values,
> because it is a a table with function pointers or so. Could you try to
> get which of the function pointers is called with gdb or whatever you
> where using?

Apparently the last probe function called before the crash is reiserfs_probe(),
which seems to corrupt the stack. After disabling the registration of reiserfs
I get:

| Disk geometry for /dev/hda: 0.000-1039.500 megabytes
| Disk label type: amiga
| Minor    Start       End     Filesystem  Flags
| 1          0.984      9.351              
| 2          9.352    259.382              
| 3        259.383    291.867  ext2        
| 4        291.867    324.351  linux-swap  
| 5        324.352    424.757  ext2        
| 6        424.758   1039.500  ext2        

Which is consistent with the data from HdToolBox.

After re-enabling affs and adding support for MultiUser (patch below), I get:

| PROBING FOR AN AMIGA PARTITION TABLE.
| Reserved = 2, Prealloc = 0
| Cyl = 2112, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
| Block size is 128
| AMIGA FFS : boot block detected FFS\3 
| AMIGA FFS : root block should be at 10584 (2016-19151:17136)
| AMIGA FFS : root-2 block : type = 653206047, subtype = -628857599, sum =
| -1976770243
| AMIGA FFS : root-1 block : type = 637044347, subtype = -1881946650, sum =
| 1885671754
| AMIGA FFS : root+1 block : type = 0, subtype = 0, sum = 0
| AMIGA FFS : root+2 block : type = -33552265, subtype = -1, sum = 0
| AMIGA FFS : root block : type = 2, subtype = 1, sum = 0
| Program received signal SIGSEGV, Segmentation fault.
| 0xc012e16a in free () from /lib/libc.so.6

:-(

For /dev/hdb, which doesn't contain any affs partitions:

| PROBING FOR AN AMIGA PARTITION TABLE.
| Reserved = 0, Prealloc = 0
| Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
| Block size is 128
| AMIGA FFS : boot block detected FFS\3 
| AMIGA FFS : root block should be at 287280 (2016-572543:570528)
| AMIGA FFS : root-2 block : type = 1026176313, subtype = 1598254958, sum =
| 1968231241
| AMIGA FFS : root-1 block : type = 1937011297, subtype = 842283008, sum =
| 302493690
| AMIGA FFS : root+1 block : type = 741552950, subtype = 741552699, sum =
| -1510802252
| AMIGA FFS : root+2 block : type = 1634495599, subtype = 1601200498, sum =
| -1940758414
| AMIGA FFS : root block : type = 1599038058, subtype = 976433203, sum =
| -463838221
| Reserved = 0, Prealloc = 0
| Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
| Block size is 128
| Disk geometry for /dev/hdb: 0.000-520.242 megabytes
| Disk label type: amiga
| Minor    Start       End     Filesystem  Flags
| 1          0.984    279.562  ext2        
| 2        279.562    520.242  ext2       

Which is OK.

--- parted-1.6.5-1.0.0.1/libparted/fs_affs/interface.c.orig	Thu Aug 14 12:16:10 2003
+++ parted-1.6.5-1.0.0.1/libparted/fs_affs/interface.c	Thu Aug 14 12:20:08 2003
@@ -55,6 +55,13 @@
 		case 0x444f5303 :	/* 'DOS\3' : FFS Int. */
 		case 0x444f5304 :	/* 'DOS\4' : OFS DirCache */
 		case 0x444f5305 :	/* 'DOS\5' : FFS DirCache */
+		case 0x6d754653 :	/* 'muFS'  : MultiUser FFS Int. */
+		case 0x6d754600 :	/* 'muF\0' : MultiUser OFS */
+		case 0x6d754601 :	/* 'muF\1' : MultiUser FFS */
+		case 0x6d754602 :	/* 'muF\2' : MultiUser OFS Int. */
+		case 0x6d754603 :	/* 'muF\3' : MultiUser FFS Int. */
+		case 0x6d754604 :	/* 'muF\4' : MultiUser OFS DirCache */
+		case 0x6d754605 :	/* 'muF\5' : MultiUser FFS DirCache */
 			printf ("AMIGA FFS : boot block detected FFS\\%d \n", block[0]&0xff);
 			printf ("AMIGA FFS : root block should be at %llu (%llu-%llu:%llu)\n",
 				root, geom->start, geom->end, geom->length);


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: Hello everyone ...

[Roman Z. <zi...@li...>]

Hi,

On Wed, 13 Aug 2003, Sven Luther wrote:

> > The dependency to automake1.6 is missing, which might be a problem for 
> 
> The whole autotool thing is a mess anyway, i hope that the maintainers
> does a new package nextly with the new 1.6.6 upstream version, which i
> hope will solve this. There seems to be some problems at :

Argh, I had to touch a few files manually to avoid that configure & co is 
rebuild, as this messes up libtool somehow and the build fails...
I restarted the build now and I hopefully have some packages tomorrow.

bye, Roman

Re: Hello everyone ...

[Roman Z. <zi...@li...>]

Hi,

On Wed, 13 Aug 2003, Sven Luther wrote:

> > > > | A bug has been detected in GNU parted.  Please email a bug report to bug...@gn... containing the version (1.6.5) and the following message:Assertion (sizeof (struct ufs_super_block) == 1380) at ../../../libparted/fs_ufs/ufs.c:308 in function ped_file_system_ufs_init() failed.
> > 
> > What about the failed assertion?
> 
> Mmm, it is in the ufs detection code. Maybe the problem is coming from
> there altough i don't see why this fail for you and not for me.

Different alignment rules. m68k aligns 32bit values at 16bit boundaries, 
struct ufs_super_block needs some padding and a packed attribute wouldn't 
hurt either.

bye, Roman

Re: Hello everyone ...

[Sven L. <sve...@wa...>]

On Wed, Aug 13, 2003 at 05:24:37PM +0200, Geert Uytterhoeven wrote:
> On Wed, 13 Aug 2003, Sven Luther wrote:
> > On Wed, Aug 13, 2003 at 05:16:39PM +0200, Geert Uytterhoeven wrote:
> > > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > > On Wed, Aug 13, 2003 at 05:09:15PM +0200, Geert Uytterhoeven wrote:
> > > > > On Wed, 13 Aug 2003, Geert Uytterhoeven wrote:
> > > > >     hda1 (DH0)         2-  18 (  8 Meg)     muF\53 (0x6d754653), bootable pri 2
> > > > >     hda2 (DH1)        19- 526 (250 Meg)     muF\53 (0x6d754653)
> > > > 
> > > > What are those ?
> > > 
> > > AmigaOS FFS International for MultiUser. The first one is System3.1:, the
> > > second one is App:.
> > 
> > Do you perchance have any documentation on this file system and how to
> > detect it ?
> 
> It's 99% backwards compatible with AFFS.
> 
> However, if a user has reformatted his partition using muFS, the ID in the
> bootblock won't be DOSx, but muFx (x = \0..\5) or muFS. AFFS in Linux knows
> about this.

So the position of the boot and root block are the same, as well as the
root block eyecatcher (type is 2, subtype is 1).

I think the problem comes from ufs not being m68k clean, maybe if it did
know about muFS it would be able to detect it and not crash on the ufs
problem. 

That said, libparted doesn't seem to be really all that ready to handle
multiple filesystem variants.

Friendly,

Sven Luther

Re: Hello everyone ...

[Sven L. <sve...@wa...>]

On Wed, Aug 13, 2003 at 06:11:52PM +0200, Geert Uytterhoeven wrote:
> On Wed, 13 Aug 2003, Sven Luther wrote:
> > On Wed, Aug 13, 2003 at 05:34:35PM +0200, Geert Uytterhoeven wrote:
> > > BTW, removing the (un)register calls for UFS didn't help, as expected (the
> > > assertion just causes UFS not being registered).
> > 
> > What about removing the affs stuff ?
> 
> Still the same ;-(

Ok, i have to go today, but i will prepare a new version tomorrow, with
more debuging messages. It would be nice to know which function is
called when the probe is done. I expect you don't get the real values,
because it is a a table with function pointers or so. Could you try to
get which of the function pointers is called with gdb or whatever you
where using?

Friendly,

Sven Luther

Re: Hello everyone ...

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

On Wed, 13 Aug 2003, Sven Luther wrote:
> On Wed, Aug 13, 2003 at 05:34:35PM +0200, Geert Uytterhoeven wrote:
> > BTW, removing the (un)register calls for UFS didn't help, as expected (the
> > assertion just causes UFS not being registered).
> 
> What about removing the affs stuff ?

Still the same ;-(

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: Hello everyone ...

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

On Wed, 13 Aug 2003, Sven Luther wrote:
> On Wed, Aug 13, 2003 at 05:16:39PM +0200, Geert Uytterhoeven wrote:
> > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > On Wed, Aug 13, 2003 at 05:09:15PM +0200, Geert Uytterhoeven wrote:
> > > > On Wed, 13 Aug 2003, Geert Uytterhoeven wrote:
> > > >     hda1 (DH0)         2-  18 (  8 Meg)     muF\53 (0x6d754653), bootable pri 2
> > > >     hda2 (DH1)        19- 526 (250 Meg)     muF\53 (0x6d754653)
> > > 
> > > What are those ?
> > 
> > AmigaOS FFS International for MultiUser. The first one is System3.1:, the
> > second one is App:.
> 
> Do you perchance have any documentation on this file system and how to
> detect it ?

It's 99% backwards compatible with AFFS.

However, if a user has reformatted his partition using muFS, the ID in the
bootblock won't be DOSx, but muFx (x = \0..\5) or muFS. AFFS in Linux knows
about this.

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: Hello everyone ...

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

On Wed, 13 Aug 2003, Sven Luther wrote:
> On Wed, Aug 13, 2003 at 05:24:37PM +0200, Geert Uytterhoeven wrote:
> > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > On Wed, Aug 13, 2003 at 05:16:39PM +0200, Geert Uytterhoeven wrote:
> > > > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > > > On Wed, Aug 13, 2003 at 05:09:15PM +0200, Geert Uytterhoeven wrote:
> > > > > > On Wed, 13 Aug 2003, Geert Uytterhoeven wrote:
> > > > > >     hda1 (DH0)         2-  18 (  8 Meg)     muF\53 (0x6d754653), bootable pri 2
> > > > > >     hda2 (DH1)        19- 526 (250 Meg)     muF\53 (0x6d754653)
> > > > > 
> > > > > What are those ?
> > > > 
> > > > AmigaOS FFS International for MultiUser. The first one is System3.1:, the
> > > > second one is App:.
> > > 
> > > Do you perchance have any documentation on this file system and how to
> > > detect it ?
> > 
> > It's 99% backwards compatible with AFFS.
> > 
> > However, if a user has reformatted his partition using muFS, the ID in the
> > bootblock won't be DOSx, but muFx (x = \0..\5) or muFS. AFFS in Linux knows
> > about this.
> 
> So the position of the boot and root block are the same, as well as the
> root block eyecatcher (type is 2, subtype is 1).

Yes, it's almost the same as AFFS.

> I think the problem comes from ufs not being m68k clean, maybe if it did
> know about muFS it would be able to detect it and not crash on the ufs
> problem. 

I don't have muFS on the second disk, and there it still crashes.

> That said, libparted doesn't seem to be really all that ready to handle
> multiple filesystem variants.

I have only one file system type on the second disk.

BTW, removing the (un)register calls for UFS didn't help, as expected (the
assertion just causes UFS not being registered).

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: Hello everyone ...

[Sven L. <sve...@wa...>]

On Wed, Aug 13, 2003 at 05:34:35PM +0200, Geert Uytterhoeven wrote:
> On Wed, 13 Aug 2003, Sven Luther wrote:
> > On Wed, Aug 13, 2003 at 05:24:37PM +0200, Geert Uytterhoeven wrote:
> > > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > > On Wed, Aug 13, 2003 at 05:16:39PM +0200, Geert Uytterhoeven wrote:
> > > > > On Wed, 13 Aug 2003, Sven Luther wrote:
> > > > > > On Wed, Aug 13, 2003 at 05:09:15PM +0200, Geert Uytterhoeven wrote:
> > > > > > > On Wed, 13 Aug 2003, Geert Uytterhoeven wrote:
> > > > > > >     hda1 (DH0)         2-  18 (  8 Meg)     muF\53 (0x6d754653), bootable pri 2
> > > > > > >     hda2 (DH1)        19- 526 (250 Meg)     muF\53 (0x6d754653)
> > > > > > 
> > > > > > What are those ?
> > > > > 
> > > > > AmigaOS FFS International for MultiUser. The first one is System3.1:, the
> > > > > second one is App:.
> > > > 
> > > > Do you perchance have any documentation on this file system and how to
> > > > detect it ?
> > > 
> > > It's 99% backwards compatible with AFFS.
> > > 
> > > However, if a user has reformatted his partition using muFS, the ID in the
> > > bootblock won't be DOSx, but muFx (x = \0..\5) or muFS. AFFS in Linux knows
> > > about this.
> > 
> > So the position of the boot and root block are the same, as well as the
> > root block eyecatcher (type is 2, subtype is 1).
> 
> Yes, it's almost the same as AFFS.

Ok.

> > I think the problem comes from ufs not being m68k clean, maybe if it did
> > know about muFS it would be able to detect it and not crash on the ufs
> > problem. 
> 
> I don't have muFS on the second disk, and there it still crashes.
> 
> > That said, libparted doesn't seem to be really all that ready to handle
> > multiple filesystem variants.
> 
> I have only one file system type on the second disk.
> 
> BTW, removing the (un)register calls for UFS didn't help, as expected (the
> assertion just causes UFS not being registered).

What about removing the affs stuff ?

Friendly,

Sven Luther

Re: Hello everyone ...

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

On Wed, 13 Aug 2003, Geert Uytterhoeven wrote:
> | Using /dev/hda
> | (parted) p
> | PROBING FOR AN AMIGA PARTITION TABLE.
> | Reserved = 2, Prealloc = 0
> | Cyl = 2112, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> | Block size is 128
> | Segmentation fault
> 
> If I try my second disk, I get:
> 
> | Using /dev/hdb
> | (parted) p                                                                
> | PROBING FOR AN AMIGA PARTITION TABLE.
> | Reserved = 0, Prealloc = 0
> | Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
> | Block size is 128
> | Segmentation fault
> 
> I'll reboot into AmigaOS so I can tell you the HdToolBox parameters in a few minutes.

  - Quantum Fireball 1080A A1M.

    2112 cylinders 16 heads 63 blocks per track (1008 blocks per cylinder)
    1063944K (1039 Meg)

    hda1 (DH0)         2-  18 (  8 Meg)     muF\53 (0x6d754653), bootable pri 2
    hda2 (DH1)        19- 526 (250 Meg)     muF\53 (0x6d754653)
    hda3 (root)      527- 592 ( 32 Meg)     0x4c4e5800, EXT2
    hda4 (swap)      593- 658 ( 32 Meg)     0x53575000
    hda5 (var)       659- 862 (100 Meg)     0x4c4e5800, EXT2
    hda6 (usr)       863-2111 (614 Meg)     0x4c4e5800, EXT2

  - Conner Peripherals 545M 7AT1

    1957 cylinders 16 heads 63 blocks per track (1008 blocks per cylinder)
    532224K (519 Meg)

    hdb1 (home0)       2- 567 (278 Meg)     0x4c4e5800, EXT2
    hdb2 (home1)     568-1056 (240 Meg)     0x4c4e5800, EXT2, bootable pri 0

So the geometries are correct.

But what about the block size of 128?

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: Hello everyone ...

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

On Wed, 13 Aug 2003, Sven Luther wrote:
> On Wed, Aug 13, 2003 at 04:50:06PM +0200, Geert Uytterhoeven wrote:
> Both of these fail after having read the first partition. Could you get
> an idea of where exactly it hangs ?

Program received signal SIGSEGV, Segmentation fault.
0xc0020e60 in ped_file_system_probe_specific (fs_type=Cannot access memory at address 0x3f714
)
    at ../../libparted/filesys.c:107
107	}
(gdb) bt
#0  0xc0020e60 in ped_file_system_probe_specific (fs_type=Cannot access memory at address 0x3f714
)
    at ../../libparted/filesys.c:107
Cannot access memory at address 0x3f70c
(gdb) 

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: Hello everyone ...

[Sven L. <sve...@wa...>]

On Wed, Aug 13, 2003 at 05:10:25PM +0200, Geert Uytterhoeven wrote:
> On Wed, 13 Aug 2003, Sven Luther wrote:
> > On Wed, Aug 13, 2003 at 04:50:06PM +0200, Geert Uytterhoeven wrote:
> > > | A bug has been detected in GNU parted.  Please email a bug report to bug...@gn... containing the version (1.6.5) and the following message:Assertion (sizeof (struct ufs_super_block) == 1380) at ../../../libparted/fs_ufs/ufs.c:308 in function ped_file_system_ufs_init() failed.
> 
> What about the failed assertion?

Mmm, it is in the ufs detection code. Maybe the problem is coming from
there altough i don't see why this fail for you and not for me. Could
you comment the two register and unregister calls in
ped_file_system_ufs_init and ped_file_system_ufs_done, and try again.

> > BTW, do you know if parted works on m68k with other kind of partition
> > tables (like MBRs for example).
> 
> No idea. I don't have other disks in my Amiga.

Ok.

Friendly,

Sven Luther

Re: Hello everyone ...

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

Build completed!

| A bug has been detected in GNU parted.  Please email a bug report to bug...@gn... containing the version (1.6.5) and the following message:Assertion (sizeof (struct ufs_super_block) == 1380) at ../../../libparted/fs_ufs/ufs.c:308 in function ped_file_system_ufs_init() failed.
| GNU Parted 1.6.5
| Copyright (C) 1998, 1999, 2000, 2001, 2002 Free Software Foundation, Inc.
| This program is free software, covered by the GNU General Public License.
| 
| This program is distributed in the hope that it will be useful, but WITHOUT ANY
| WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
| PARTICULAR PURPOSE.  See the GNU General Public License for more details.
| 
| Using /dev/hda
| (parted) p
| PROBING FOR AN AMIGA PARTITION TABLE.
| Reserved = 2, Prealloc = 0
| Cyl = 2112, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
| Block size is 128
| Segmentation fault

If I try my second disk, I get:

| Using /dev/hdb
| (parted) p                                                                
| PROBING FOR AN AMIGA PARTITION TABLE.
| Reserved = 0, Prealloc = 0
| Cyl = 1057, Sect = 63, Head = 16, Sect*Head = 1008, CylBlocks = 1008
| Block size is 128
| Segmentation fault

I'll reboot into AmigaOS so I can tell you the HdToolBox parameters in a few minutes.

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