Re: [Xen-devel] 2.6.32.16: hvm domU boot - oops at add_pin_to_irq_node_nopanic

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Sat, 10 Jul 2010, Bruce Edge wrote:
> So I tried the same hvm boot using the following 3 branches. 
>     xen/next-2.6.32
> Stefano's
>     2.6.34-pvhvm-v4 
>     2.6.34-pvhvm-v6
> 
> The base xen 4.0.1rc3 did not have the patch Stefano posted later in this 
> thread.
> The boot summary is inline and the complete bootlogs are attached. It has a 
> few extra bits required by the .34 kernel, but
> is essentially the same one for all 3 cases.
> 
> The gist of it being:
> CONFIG_XEN=y
> CONFIG_XEN_MAX_DOMAIN_MEMORY=32
> CONFIG_XEN_SAVE_RESTORE=y
> # CONFIG_XEN_DEBUG_FS is not set
> CONFIG_XEN_BLKDEV_FRONTEND=m
> CONFIG_NETXEN_NIC=m
> CONFIG_XEN_NETDEV_FRONTEND=m
> CONFIG_XEN_KBDDEV_FRONTEND=m
> CONFIG_HVC_XEN=y
> CONFIG_XEN_FBDEV_FRONTEND=m
> CONFIG_XEN_BALLOON=y
> CONFIG_XEN_SCRUB_PAGES=y
> CONFIG_XEN_DEV_EVTCHN=m
> CONFIG_XENFS=m
> CONFIG_XEN_COMPAT_XENFS=y
> CONFIG_XEN_SYS_HYPERVISOR=y
> # CONFIG_XEN_PLATFORM_PCI is not set
> 
> The .config is also attached
> 
> 1) xen/next-2.6.32 - Hangs at the end of the following sequence:
> 
> [    0.464686] ACPI: bus type pci registered^M
> [    0.467109] PCI: Using configuration type 1 for base access^M
> [    0.470495] bio: create slab <bio-0> at 0^M
> [    0.523248] ACPI: Interpreter enabled^M
> [    0.525249] ACPI: (supports S0 S3 S4 S5)^M
> [    0.527628] ACPI: Using IOAPIC for interrupt routing^M
> [    0.624323] ACPI: No dock devices found.^M
> [    0.626607] ACPI: PCI Root Bridge [PCI0] (0000:00)^M
> [    0.635170] * Found PM-Timer Bug on the chipset. Due to workarounds for a 
> bug,^M
> [    0.635170] * this clock source is slow. Consider trying other clock 
> sources^M
> [    0.641158] pci 0000:00:01.3: quirk: region 1f40-1f7f claimed by PIIX4 
> ACPI^M
> [    1.115247] ACPI: PCI Interrupt Link [LNKA] (IRQs *5 10 11)^M
> [    1.118966] ACPI: PCI Interrupt Link [LNKB] (IRQs 5 *10 11)^M
> [    1.121346] ACPI: PCI Interrupt Link [LNKC] (IRQs 5 10 *11)^M
> [    1.125128] ACPI: PCI Interrupt Link [LNKD] (IRQs *5 10 11)^M
> <hang>
> 
> 2) 2.6.34-pvhvm-v4 - boots, with a lot of call traces on the serial console:
> 
> [    1.880528] WARNING: at arch/x86/xen/mmu.c:1953 
> xen_hvm_exit_mmap+0x45/0x50()^M
> [    1.884168] Hardware name: HVM domU^M
> [    1.885967] Modules linked in:^M
> [    1.887638] Pid: 48, comm: init Not tainted 2.6.34 #1^M
> [    1.890384] Call Trace:^M
> [    1.891721]  [<ffffffff8105cefb>] warn_slowpath_common+0x7b/0xc0^M
> [    1.894804]  [<ffffffff8105cf54>] warn_slowpath_null+0x14/0x20^M
> [    1.897820]  [<ffffffff81005015>] xen_hvm_exit_mmap+0x45/0x50^M
> [    1.900941]  [<ffffffff8110f7b0>] exit_mmap+0x50/0x190^M
> [    1.903569]  [<ffffffff8105a5e2>] mmput+0x42/0x110^M
> [    1.906059]  [<ffffffff811443a3>] flush_old_exec+0x463/0x620^M
> [    1.908995]  [<ffffffff8118436e>] load_elf_binary+0x38e/0x1d20^M
> [    1.912375]  [<ffffffff81108ba6>] ? follow_page+0x2d6/0x350^M
> [    1.915210]  [<ffffffff81108ba6>] ? follow_page+0x2d6/0x350^M
> [    1.918055]  [<ffffffff8110d7cf>] ? __get_user_pages+0x10f/0x430^M
> [    1.921309]  [<ffffffff8114386c>] ? get_arg_page+0x5c/0xc0^M
> [    1.924127]  [<ffffffff8114528d>] search_binary_handler+0xed/0x310^M
> [    1.927336]  [<ffffffff811458e3>] do_execve+0x333/0x410^M
> [    1.930170]  [<ffffffff812dff58>] ? strncpy_from_user+0x38/0x60^M
> [    1.933261] Clockevents: could not switch to one-shot mode: lapic is not 
> functional.^M
> [    1.937318] Could not switch to high resolution mode on CPU 0^M
> [    1.940399]  [<ffffffff810135da>] sys_execve+0x4a/0x80^M
> [    1.943078]  [<ffffffff8100b48a>] stub_execve+0x6a/0xc0^M
> [    1.945737] ---[ end trace 7264d4229303d265 ]---^M
> 
> but eventually gets to a login and appears functional, but the number of call 
> traces coming out during the boot are not
> comforting.
 
They are harmless even though they are very annoying.
I modified the v6 branch so that this problem should be solved, could
you please clone it form scratch and try again?

> 
> 3) 2.6.34-pvhvm-v6 - Same as v4
> 
> However the root fs of the domU was corrupted after this last test. Possibly 
> the result of using partprobe to get at lvm
> block device partitions and mount them to install new kernels, but I had been 
> doing that for a while with no ill effects.
>

This is really concerning; however considering that you have
CONFIG_XEN_PLATFORM_PCI disabled in your config and blkfront is not
loaded I cannot see how pv on hvm would have caused it. 
Without CONFIG_XEN_PLATFORM_PCI pv drivers don't load so you don't
really have many benefits (or risks); you should enable
CONFIG_XEN_PLATFORM_PCI too in order to have the performance
improvements.
Once you do that you need to change your root= command line option to
root=/dev/xvda1 because the partition naming scheme of blkfront is
xvd*.
If you have data corruptions because of blkfront, check that blkfront
and the emulated IDE disks (/dev/hd*) are not enabled simultaneously,
if they are please let me know.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel