Unable to boot Solaris 8/9 x86 under Fedora 24

[xtrondo]

qemu-system-x86_64 -version
QEMU emulator version 2.6.2 (qemu-2.6.2-4.fc24), Copyright (c) 2003-2008 Fabrice Bellard

Try several ways without success, I think it was a regression because problem seems to be related with ide fixed on 0.6.0:
- int13 CDROM BIOS fix (aka Solaris x86 install CD fix)
- int15, ah=86 BIOS fix (aka Solaris x86 hardware probe hang up fix)

Solaris 10/11 works without a problem, also booting with "scsi" will circumvent initial problem, but later found problems related with "scsi" cdrom boot and also will not found the "ide" disk device.

qemu-system-i386 -m 712 -drive file=/dev/Virtual_hdd/beryllium0,format=raw -cdrom /repo/Isos/sol-9_905_x86.iso

SunOS Secondary Boot version 3.00

prom_panic: Could not mount filesystem.
Entering boot debugger:
[136419]

Regards,
\\CA,

[John Snow (jnsnow)]

So, if I'm reading you right, Solaris10/11 work just fine, but 8/9 don't -- and have not since qemu version 0.6.0!? From 2004?

I don't have a copy of Solaris9 to test with, so I doubt I can work on trying to reproduce this. Is there any possibility to reproduce a problem on an older, freely available BSD?

[xtrondo (xtrondo)]

yes, 10/11/12 beta work without a problem(and really fast), 8/9 have been
reported to work at least since 0.6.0. with the "ide" fix committed to that
version.
The problem seems related with "ide" emulation and real mode drivers, so I
don't think an older BSD can reproduce.
I will test if an older BSD is also affected by this, I can also provide
you a place for you to get the versions 8/9 x86 iso, if that is ok
in any way.
Many thanks for the time on checking this one.

On Tue, Nov 8, 2016 at 11:13 PM, John Snow <email address hidden>
wrote:

> So, if I'm reading you right, Solaris10/11 work just fine, but 8/9 don't
> -- and have not since qemu version 0.6.0!? From 2004?
>
> I don't have a copy of Solaris9 to test with, so I doubt I can work on
> trying to reproduce this. Is there any possibility to reproduce a
> problem on an older, freely available BSD?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1639394
>
> Title:
> Unable to boot Solaris 8/9 x86 under Fedora 24
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/qemu/+bug/1639394/+subscriptions
>

[John Snow (jnsnow)]

Sorry, I don't understand.

The two fixes you mentioned were committed and released as part of 0.6.0, so does this work with QEMU version 0.6.0 or not?

If it works in 0.6.0, can you tell me the first version where it stopped working? I assume it isn't currently working in 2.6.2, so it broke sometime inbetween, is that correct?

I'm not sure if I am able to download a solaris iso that you've found. However, if you are able to contributes patches upstream, I am willing to review them.

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]

[Madgemade (madgemade-v)]

I also have this problem. On both QEMU 5.2.0 and QEMU 3.1. The situation is still the same as the original report.
For Solaris 8 the following configuration causes an error when the installer attempts to boot:
-m 256M -display gtk -monitor stdio -hda Sol8.qcow2 -cdrom sol8_x86_install.ISO -boot d -machine pc
The error is given as "prom_panic: Could not mount filesystem", the same as reported originally. It does this even when "hda" is not specified.
I can get into the installer by using SCSI devices instead (as alluded to originally) but the SCSI devices are not detected correctly by the installer and it cannot find the boot media.
A slightly different problem occurs with Solaris 7. This configuration (only CDROM) allows the installer to run and it detects the CDROM media correctly: -m 256M -display gtk -monitor stdio -boot a -fda sol-7_boot.img -cdrom sol-7.iso
But when a "hda" hard disk device is introduced the installer does not load, instead hanging on the first stage (where Solaris 8 installer gives the error).
To ensure the ISO was not corrupted or similar I used VirtualBox which works as expected.

[John Snow (jnsnow)]

I have to admit that my time budget for IDE is quite low, so I will be unable to look into this.

If you'd like to help debug it further and you have the time, you can try building QEMU 6.0 (RC0 or so, something quite modern) and enabling the IDE tracing options and trying to boot Solaris as you have been doing, The logs might help illustrate something obviously wrong.

You want to enable IDE traces (but exclude the ones that show the actual IO data), so you need an events file that looks something like this;

ide_ioport_read
ide_ioport_write
ide_status_read
ide_ctrl_write
ide_exec_cmd
ide_cancel_dma_sync_buffered
ide_cancel_dma_sync_remaining
ide_sector_read
ide_sector_write
ide_reset
ide_bus_reset_aio
ide_dma_cb
cd_read_sector_sync
cd_read_sector_cb
cd_read_sector
ide_atapi_cmd_error
ide_atapi_cmd_reply_end
ide_atapi_cmd_reply_end_eot
ide_atapi_cmd_reply_end_bcl
ide_atapi_cmd_reply_end_new
ide_atapi_cmd_check_status
ide_atapi_cmd_read
ide_atapi_cmd
ide_atapi_cmd_read_dma_cb_aio

And use it using `--trace events=ide_events_file`. A full list of trace events can be found here: https://gitlab.com/qemu-project/qemu/-/blob/master/hw/ide/trace-events

If you can trigger it without `-hda`, it would be good to leave it off to help minimize log reports for IDE devices unrelated to problem.

[Lev Kujawski (lkujaw)]

Hello, I believe I have solved the underlying issue with the attached patch. Verified against the behavior of an actual i440FX IDE controller.

See https://lists.nongnu.org/archive/html/qemu-devel/2022-05/msg04229.html .