[SeaBIOS] [PATCHv2 0/6] Improved multi-platform support

[Ian Campbell]

On Wed, 2013-02-13 at 16:08 +0000, David Woodhouse wrote:
> On Wed, 2013-02-13 at 12:56 +0000, Ian Campbell wrote:
> > 
> > I think it should JustWork. I expect they disabled it either because
> > they intended to eventually use /usr/bin/qemu-foo instead or maybe just
> > as an expedient way of getting the 4.2 package together and because it
> > is a non-default option in 4.2 (and something of a "preview" feature at
> > that).
> 
> OK... I re-enabled it, I now have a /usr/lib/xen/bin/qemu-system-i386
> (not x86_64 but that's fine for now) which *does* appear to have Xen
> support.

Actually the -i386/x86_64 designator is a bit arbitrary/meaningless
since under Xen qemu only implements peripherals and not the actual
processor. Even upstream we only ever build one (I forget which)
regardless of platform and we use the same one regardless of guest type.

> 
> [root at i7 ~]# xl create -c -d  seabios.cfg
> Parsing config from seabios.cfg
> {"domid":null,"config":{"c_info":{"type":"hvm","hap":"<default>","oos":"<default>","ssidref":0,"name":"seabios","uuid":"849a47f8-1333-45e1-b4a6-cde90429cfd4","xsdata":{},"platformdata":{},"poolid":0,"run_hotplug_scripts":"True"},"b_info":{"max_vcpus":0,"avail_vcpus":[],"cpumap":[],"numa_placement":"<default>","tsc_mode":"default","max_memkb":524288,"target_memkb":524288,"video_memkb":-1,"shadow_memkb":4096,"rtc_timeoffset":0,"localtime":"<default>","disable_migrate":"<default>","cpuid":[],"blkdev_start":null,"device_model_version":"qemu_xen","device_model_stubdomain":"<default>","device_model":null,"device_model_ssidref":0,"extra":[],"extra_pv":[],"extra_hvm":[],"sched_params":{"sched":"unknown","weight":-1,"cap":-1,"period":-1,"slice":-1,"latency":-1,"extratime":-1},"ioports":[],"irqs":[],"u":{"firmware":"/home/dwmw2/fedora/xen/f18/xen-4.2.1/tools/firmware/hvmloader/hvmloader","bios":null,"pae":"<default>","apic":"<default>","acpi":"<default>","acpi_s3":"<default>","acpi_s4":"<default>","nx":"<default>","viridian":"<default>","timeoffset":null,"hpet":"<default>","vpt_align":"<default>","timer_mode":null,"nested_hvm":"<default>","nographic":"<default>","vga":{"kind":null},"vnc":{"enable":"<default>","listen":null,"passwd":null,"display":0,"findunused":"<default>"},"keymap":null,"sdl":{"enable":"<default>","opengl":"<default>","display":null,"xauthority":null},"spice":{"enable":"<default>","port":0,"tls_port":0,"host":null,"disable_ticketing":"<default>","passwd":null,"agent_mouse":"<default>"},"gfx_passthru":"<default>","serial":"pty","boot":null,"usb":"<default>","usbdevice":null,"soundhw":null,"xen_platform_pci":"<default>"}},"disks":[],"nics":[],"pcidevs":[],"vfbs":[],"vkbs":[],"on_poweroff":"destroy","on_reboot":"restart","on_watchdog":"destroy","on_crash":"destroy"}}
> xc: info: VIRTUAL MEMORY ARRANGEMENT:
>   Loader:        0000000000100000->00000000001be3a8
>   TOTAL:         0000000000000000->000000001f800000
>   ENTRY ADDRESS: 0000000000100000
> xc: info: PHYSICAL MEMORY ALLOCATION:
>   4KB PAGES: 0x0000000000000200
>   2MB PAGES: 0x00000000000000fb
>   1GB PAGES: 0x0000000000000000
> Daemon running with PID 1383
> xenconsole: Could not read tty from store: No such file or directory
> libxl: error: libxl_exec.c:118:libxl_report_child_exitstatus: console child [0] exited with error status 2
> 
> At this point, the 'xl create' command is running in the background,
> appending this to xl-seabios.log every few seconds...

This usually indicates that either hvmloader or the BIOS is crashing,
often with a triple fault. You should see in xl dmesg:
        "Triple fault on VCPU%d - "
                     "invoking HVM system reset.\n"
        
Actually, that reminds me, you should add "loglvl=all guest_loglvl=all"
to your hypervisor command line.

I think the on_foo = "preserve" stuff I mentioned earlier should stop
the annoying constant looping reboots.

Is this is a Seabios you built yourself (modified or un) or the Fedora
packaged one?

I've attached an hvmloader which I just built from the 4.2 stable branch
and have smoke tested, does it work for you?

> 
> xc: info: VIRTUAL MEMORY ARRANGEMENT:
>   Loader:        0000000000100000->00000000001be3a8
>   TOTAL:         0000000000000000->000000001f800000
>   ENTRY ADDRESS: 0000000000100000
> xc: info: PHYSICAL MEMORY ALLOCATION:
>   4KB PAGES: 0x0000000000000200
>   2MB PAGES: 0x00000000000000fb
>   1GB PAGES: 0x0000000000000000
> Waiting for domain seabios (domid 253) to die [pid 1384]
> Domain 253 has shut down, reason code 1 0x1
> Action for shutdown reason code 1 is restart
> Domain 253 needs to be cleaned up: destroying the domain
> Done. Rebooting now
> 
> I can strace it and see that it *is* invoking qemu and then almost
> immediately sending it a SIGHUP for reasons I don't quite understand.

The domain died so it is tidying up. Probably qemu barely got the chance
to do anything before the guest crashed.

Ian.
-- 
Ian Campbell

"The eleventh commandment was `Thou Shalt Compute' or `Thou Shalt Not
Compute' -- I forget which."
		-- Epigrams in Programming, ACM SIGPLAN Sept. 1982
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hvmloader
Type: application/x-executable
Size: 684488 bytes
Desc: not available
URL: <http://www.seabios.org/pipermail/seabios/attachments/20130214/ad798a0d/attachment-0001.exe>