Problem jumping to Boot Loader

[Jeffrey C. Jacobs]

Hello again,

Now that I've got my BIOS booting properly AFAICT, I am running into a 
problem when I jump to my Boot Loader.  I get to the adjusted_boot_notes 
line, and then my Boot Loader is started, but it gets a memory map error 
that I don't understand and fails to load my menu.lst.  This Boot Loader 
(FILO ver 0.5) works with an older version of CoreBoot (specifically a 
version from when it was called LinuxBIOS) but in my current build does 
not seem to work.  You can see my full boot sequence at 
http://coreboot.pastebin.com/f4ebea109 -- the relevant lines start at 
line 1006:

FILO version 0.5 ([e-mail]) Wed Jul 15 13:14:58 EDT 2009
Can't get memory map from firmware. Using hardcoded default.
Press <Enter> for default menu.lst (hda1:/filo/menu.lst), or <Esc> for 
prompt... 2 1 timed out
menu: hda1:/filo/menu.lst
Detected floating bus
No drive detected on IDE channel 0

A proper boot should look like this:

FILO version 0.5 ([e-mail]) Sun Oct 19 14:32:35 EDT 2008
Press <Enter> for default menu.lst (hda1:/filo/menu.lst), or <Esc> for 
prompt... 2 1 timed out
menu: hda1:/filo/menu.lst
hda: LBA48 40GB: IC25N040ATMR04-0                       
hdb: ATAPI: MATSHITACD-RW  CW-8124                 
Mounted ext2fs

Has anyone experienced this before?  Any ideas how to get FILO to 
recognize the mounted hardware?  I'm guessing it's something I missed in 
my CoreBoot modifications since it clearly works in the older version.

Thanks in advance,

Jeffrey.

[Patrick Georgi]

Jeffrey C. Jacobs schrieb:
> This Boot Loader (FILO ver 0.5) works with an older version of
> CoreBoot (specifically a version from when it was called LinuxBIOS)
> but in my current build does not seem to work.  You can see my full
> boot sequence at http://coreboot.pastebin.com/f4ebea109 -- the
> relevant lines start at line 1006:
>
> FILO version 0.5 ([e-mail]) Wed Jul 15 13:14:58 EDT 2009
> Can't get memory map from firmware. Using hardcoded default.

Update your FILO. coreboot now has the capability to move parts of its
tables to high memory, and your FILO won't know about that.

Regards,
Patrick Georgi

[Jeffrey C. Jacobs]

Thanks for the advice, Patrick.

So, I also have the source for GRUB 2, revision 2202 (10 May 2009) that 
I've compiled with the CoreBoot option and further linked with a bunch 
of the GRUB modules to create a payload.elf package.  This, I linked to 
my LinuxBIOS build and it booted fine.  However, when I try it with 
CoreBoot, I can't seem to jump to the boot loader at all; it dies right 
after the line stating the setting for adjusted_boot_notes: 
http://coreboot.pastebin.com/f15277438

I would update my GRUB checkout but every time I try to, I get "svn: 
Server sent unexpected return value (502 Bad Gateway) in response to 
OPTIONS request for 'http://svn.savannah.gnu.org/svn/grub/trunk'" and 
even in verbose mode I can't figure that one out.  But since we're 
talking GRUB 2 and a version within the last 3 months, I think it should 
be compatible with HIMEM.  So why, when I try it with LinuxBIOS, GRUB is 
able to load, but when I do with CoreBoot, it fails to jump to the boot 
loader?

Thanks in advance,

Jeffrey.

Patrick Georgi wrote:
> Jeffrey C. Jacobs schrieb:
> > This Boot Loader (FILO ver 0.5) works with an older version of
> > CoreBoot (specifically a version from when it was called LinuxBIOS)
> > but in my current build does not seem to work.  You can see my full
> > boot sequence at http://coreboot.pastebin.com/f4ebea109 -- the
> > relevant lines start at line 1006:
> >
> > FILO version 0.5 ([e-mail]) Wed Jul 15 13:14:58 EDT 2009
> > Can't get memory map from firmware. Using hardcoded default.
>
> Update your FILO. coreboot now has the capability to move parts of its
> tables to high memory, and your FILO won't know about that.
>
> Regards,
> Patrick Georgi

[Patrick Georgi]

Jeffrey C. Jacobs schrieb:
> I would update my GRUB checkout but every time I try to, I get "svn:
> Server sent unexpected return value (502 Bad Gateway) in response to
> OPTIONS request for 'http://svn.savannah.gnu.org/svn/grub/trunk'" and
> even in verbose mode I can't figure that one out.  But since we're
> talking GRUB 2 and a version within the last 3 months, I think it
> should be compatible with HIMEM.  So why, when I try it with
> LinuxBIOS, GRUB is able to load, but when I do with CoreBoot, it fails
> to jump to the boot loader?

Given that there is absolutely no output by GRUB, that might as well be
another problem - but it might also be high tables support.
I don't know if they support it, as I don't monitor their efforts.
However, I don't think there are any developers left that work on both
codebases, so that feature in coreboot might have happened without them
noticing. As not all boards use high tables yet, they might have missed
the need to implement support for it.

One option for now could be to disable HAVE_HIGH_TABLES in your coreboot
build. Eventually, this config flag will disappear, with high tables
activated in all of coreboot, so this would mostly be something to get
past that road bump _for now_.

You could also try the current, maintained FILO tree. See
http://www.coreboot.org/FILO for information on how to fetch it.

Regards,
Patrick

[Jeffrey C. Jacobs]

Danke vielmals Patrick and many thanks to everyone else,

So, I decided to take your advice about switching (back to) FILO using 
the latest version (svn 103) and latest libpayload (svn 4505) but am 
still having problems recognizing my First Partition on the Primary 
Master IDE device, namely hda1.  Specifically, using the FILO build 
settings specified here: http://coreboot.pastebin.com/f4e644a72

I got the boot sequence found at:

http://coreboot.pastebin.com/f5b707efb

Now, in that sequence, at the very end, where I've highlighted, I get 
the 'boot:' prompt instead of it loading what's in my menu.lst file 
(whose path is highlighted in the first link under the FILO 
configuration options).  Also, note in the configuration that I am 
trying to build FILO in GRUB mode and have specified a prompt of 
"filo-as-grub", not "boot", which you see in my boot sequence.  In fact, 
when I first tried to built the latest FILO, I found it would not build 
because the build/config.h file was generated improperly.  The file was 
missing the CONFIG_PROMPT definition, which I also hardcoded to 
"filo-as-grub".  None the less, when I boot, I still get "boot".

Anyway, since my menu.lst could not be loaded, I typed in the full boot 
command replete with kernel boot parameters, and as you can see I got a 
generic error stating that FILO could not find Device 0 (obviously 
meaning it could not find hda1).  So, I'm not sure if this is a problem 
with FILO, but it seems much more likely I have some errant settings in 
Coreboot which is not allowing FILO to see my drives.  Does anyone have 
any ideas?

Thanks in advance,

Jeffrey.

Patrick Georgi wrote:
> Jeffrey C. Jacobs schrieb:
> > I would update my GRUB checkout but every time I try to, I get "svn:
> > Server sent unexpected return value (502 Bad Gateway) in response to
> > OPTIONS request for 'http://svn.savannah.gnu.org/svn/grub/trunk'" and
> > even in verbose mode I can't figure that one out.  But since we're
> > talking GRUB 2 and a version within the last 3 months, I think it
> > should be compatible with HIMEM.  So why, when I try it with
> > LinuxBIOS, GRUB is able to load, but when I do with CoreBoot, it fails
> > to jump to the boot loader?
>
> Given that there is absolutely no output by GRUB, that might as well be
> another problem - but it might also be high tables support.
> I don't know if they support it, as I don't monitor their efforts.
> However, I don't think there are any developers left that work on both
> codebases, so that feature in coreboot might have happened without them
> noticing. As not all boards use high tables yet, they might have missed
> the need to implement support for it.
>
> One option for now could be to disable HAVE_HIGH_TABLES in your coreboot
> build. Eventually, this config flag will disappear, with high tables
> activated in all of coreboot, so this would mostly be something to get
> past that road bump _for now_.
>
> You could also try the current, maintained FILO tree. See
> http://www.coreboot.org/FILO for information on how to fetch it.
>
> Regards,
> Patrick

[Patrick Georgi]

Jeffrey C. Jacobs schrieb:
> Danke vielmals Patrick and many thanks to everyone else,
>
> So, I decided to take your advice about switching (back to) FILO using
> the latest version (svn 103) and latest libpayload (svn 4505) but am
> still having problems recognizing my First Partition on the Primary
> Master IDE device, namely hda1.  Specifically, using the FILO build
> settings specified here: http://coreboot.pastebin.com/f4e644a72

How did you change that configuration, simply by editing .config, or by
going through the "make config" routine?
Editing .config doesn't adapt the files in build/, so it should always
be completed with a "make oldconfig" (and then look if it took your
values or overwrote it with something else due to dependencies)
To be extra careful, always run "make clean" after changing the
configuration to force a full build.
> Anyway, since my menu.lst could not be loaded, I typed in the full
> boot command replete with kernel boot parameters, and as you can see I
> got a generic error stating that FILO could not find Device 0
> (obviously meaning it could not find hda1).  So, I'm not sure if this
> is a problem with FILO, but it seems much more likely I have some
> errant settings in Coreboot which is not allowing FILO to see my
> drives.  Does anyone have any ideas?

The device numbering might be weird at times. It's reasonably stable
once you found the right device name, but it can differ from what you
see elsewhere, so try hdb, hdc, and hdd, too.

Patrick

[Jeffrey C. Jacobs]

Patrick et al.,

Thanks again for the help and advice.  Rudolf, I may try to look into 
your solution since I too am trying to boot off an ext3fs system but 
Patrick you have been a great help so far.  I don't as a rule manually 
edit .config scripts, they're just convenient for reviewing settings. 
Actually, I used "make menuconfig", which is what is suggested in the 
FILO README, but that did not seem to generate build/config.h correctly 
so I tried the basic "make config" as you suggested and that seems to 
have generated the correct build.

So I loaded up my CoreBoot+FILO payload and booted and got the result 
listed in: http://coreboot.pastebin.com/f432f39bc

Now, with all those ANSI Escape Codes, I tried to clean up the last bit 
which has my FILO prompt and output: http://coreboot.pastebin.com/f7cb33b5

It's probably so busy because I have all the debugging output turned on. 
  All that said, the bad news is that I still can't boot and it looks 
like it still doesn't recognize my hard drive.  I've highlighted a few 
lines that I think may be relevant.  Firstly, we have:

ERROR: No such CMOS option (boot_devices)
menu: hda1:/boot/filo/menu.lst

I guess we can safely ignore the first line since we've not AFAICT 
entered the IDE probe yet and the second line just reiterates the 
menu.lst location.

There follows a series of probes of memory addresses which are mostly 
uneventful.  The exception is a failure to access the bridge at 01:09.00:

Misconfigured bridge at 01:09.00 skipped.

It seems to reprint this each time it tries to scan that memory area. 
I'm not sure what this represents, but I guess it's okay to ignore for 
the time being.

Finally, we have a better description of my error:

found PCI IDE controller 8086:24cb prog_if=0x8a
primary channel: compatibility mode
skipping 0 native PCI controllers, new index=0
cmd_base=0x1f0 ctrl_base=0x3f4
init_controller: drive 0
Detected floating bus
No drive detected on IDE channel 0
Failed to open IDE.
dev=hda1, path=/boot/filo/menu.lst
Drive 0 does not exist
Failed to open IDE.
Could not open menu.lst file 'hda1:/boot/filo/menu.lst'. Entering 
command line.

So I get this "No drive detected on IDE channel 0", which implies to me 
it's something lower-level that the file system.

So, does anyone have any better idea, given this new information, why 
this version of FILO can't detect my hard drive?

Thanks in Advance!

Jeffrey.

Patrick Georgi wrote:
> Jeffrey C. Jacobs schrieb:
> > Danke vielmals Patrick and many thanks to everyone else,
> >
> > So, I decided to take your advice about switching (back to) FILO using
> > the latest version (svn 103) and latest libpayload (svn 4505) but am
> > still having problems recognizing my First Partition on the Primary
> > Master IDE device, namely hda1.  Specifically, using the FILO build
> > settings specified here: http://coreboot.pastebin.com/f4e644a72
>
> How did you change that configuration, simply by editing .config, or by
> going through the "make config" routine?
> Editing .config doesn't adapt the files in build/, so it should always
> be completed with a "make oldconfig" (and then look if it took your
> values or overwrote it with something else due to dependencies)
> To be extra careful, always run "make clean" after changing the
> configuration to force a full build.
> > Anyway, since my menu.lst could not be loaded, I typed in the full
> > boot command replete with kernel boot parameters, and as you can see I
> > got a generic error stating that FILO could not find Device 0
> > (obviously meaning it could not find hda1).  So, I'm not sure if this
> > is a problem with FILO, but it seems much more likely I have some
> > errant settings in Coreboot which is not allowing FILO to see my
> > drives.  Does anyone have any ideas?
>
> The device numbering might be weird at times. It's reasonably stable
> once you found the right device name, but it can differ from what you
> see elsewhere, so try hdb, hdc, and hdd, too.
>
> Patrick

[Patrick Georgi]

On Thu, 06 Aug 2009 11:53:52 -0400, "Jeffrey C. Jacobs"
 wrote:
> There follows a series of probes of memory addresses which are mostly 
> uneventful.  The exception is a failure to access the bridge at 01:09.00:
>
> Misconfigured bridge at 01:09.00 skipped.
>
> It seems to reprint this each time it tries to scan that memory area. 
> I'm not sure what this represents, but I guess it's okay to ignore for 
> the time being.

This means that there is some bridge that claims is new client bus has the
bus id 0. That leads to an endless loop (as any bridge is a child of bus 0
directly or indirectly)
> Detected floating bus

This indicates that you're using the "old" IDE driver. You could also try
CONFIG_IDE_NEW_DISK, which is another IDE driver that sometimes helps and
sometimes doesn't. It's worth a try.

Regards,
Patrick

[Jeffrey C. Jacobs]

Patrick Georgi wrote:
> On Thu, 06 Aug 2009 11:53:52 -0400, "Jeffrey C. Jacobs" wrote:
> > Detected floating bus
>
> This indicates that you're using the "old" IDE driver. You could also try
> CONFIG_IDE_NEW_DISK, which is another IDE driver that sometimes helps and
> sometimes doesn't. It's worth a try.

Hmm.  Okay, so I updated to the latest coreboot v2 just in case and then 
since that option is for FILO I ran make config and activated the new 
disk driver and made clean and then all.  None the less, I still get:

found PCI IDE controller 8086:24cb prog_if=0x8a
primary channel: compatibility mode
skipping 0 native PCI controllers, new index=0
Failed to open IDE.
Could not open menu.lst file 'hda1:/boot/filo/menu.lst'. Entering 
command line.

Alas.  Do you have any other ideas, Patrick?  Or anyone else?

Thanks in advance,

Jeffrey.

[ron minnich]

A simple question which I can not tell from your boot line. Is this
disk partitioned?

ron

[Jeffrey C. Jacobs]

ron minnich wrote:
> A simple question which I can not tell from your boot line. Is this
> disk partitioned?

Sorry that was confusing; yes, it is partitioned using standard MBR 
partition tables (fdisk) and the first partition on the disk is an 
ext3fs partition containing my kernel and initrd.

Jeffrey.

[Rudolf Marek]

Hi,

Maybe it is a filesystem issue. I had some problems with ext3 volumes. Why don;t 
you try SeaBIOS payload instead and boot with classic grub?

It would show if it is this or another issue.

Rudolf