Re: [edk2] KVM: MTRR: fix memory type handling if MTRR is completely disabled

[Janusz <januszmk6@gmail.com>]

W dniu 21.09.2015 o 04:51, Xiao Guangrong pisze:
>
> Thanks for your report and analysis, Janusz!
>
> On 09/19/2015 01:48 AM, Janusz wrote:
>> W dniu 18.09.2015 o 12:07, Laszlo Ersek pisze:
>>> On 09/18/15 11:37, Janusz wrote:
>>>> Hello,
>>>>
>>>> I am writting about this patch that was posted by Xiao:
>>>> http://www.spinics.net/lists/kvm/msg119044.html and this:
>>>> http://www.spinics.net/lists/kvm/msg119045.html
>>>> I've tested both kernel 4.2 and 4.3 and problem still exists when I
>>>> use
>>>> OVMF - 100% cpu usage, VM resetting, while it works properly on
>>>> kernel 4.1
>>> My last (still current) request remains "please quirk it". See the end
>>> of <http://thread.gmane.org/gmane.linux.kernel/1952205/focus=1996025>,
>>> and other messages in that subthread.
>> when I saw message from Xiao that he posted patch for it and have seen
>> this code in kernel sources (landed in 4.2-rc3) I though that the status
>> is not "please quirk it" anymore
>>>
>>> I haven't been following kernel development, so maybe the quirk has not
>>> happened. No clue.
>>>
>>> ... "VM resetting" looks something different though; I've been under
>>> the
>>> impression that the pedantic (=unquirked) MTRR configuration didn't
>>> impact other things than speed. Janusz, maybe you could contribute with
>>> a host kernel bisection for the VM reset symptom.
>> To be more exact - VM is mostly not starting or its starting after long
>> time and then resetting it self at random time (but before system boots)
>> or gets very high cpu usage or sometimes boots without problem (the
>> least possible case scenario). When I start VM with -vga std, not with
>> my gpu passthrough - in most of the time I get this:
>> http://pastebin.com/raw.php?i=CKrNsueS
>
> It seems the behaviour is different with previous (previously, it can
> boot
> but slowly), right?
>
> The URL cat not be accessed, i do not know it's your web server issue or
> our networking issue.
>
>>
>> Result of bisect:
>>
>> git bisect start
>> # bad: [d770e558e21961ad6cfdf0ff7df0eb5d7d4f0754] Linux 4.2-rc1
>> git bisect bad d770e558e21961ad6cfdf0ff7df0eb5d7d4f0754
>> # good: [b953c0d234bc72e8489d3bf51a276c5c4ec85345] Linux 4.1
>> git bisect good b953c0d234bc72e8489d3bf51a276c5c4ec85345
>> # good: [4570a37169d4b44d316f40b2ccc681dc93fedc7b] Merge tag
>> 'sound-4.2-rc1' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
>> git bisect good 4570a37169d4b44d316f40b2ccc681dc93fedc7b
>> # good: [8d7804a2f03dbd34940fcb426450c730adf29dae] Merge tag
>> 'driver-core-4.2-rc1' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core
>> git bisect good 8d7804a2f03dbd34940fcb426450c730adf29dae
>> # bad: [78c10e556ed904d5bfbd71e9cadd8ce8f25d6982] Merge branch
>> 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
>> git bisect bad 78c10e556ed904d5bfbd71e9cadd8ce8f25d6982
>> # good: [623f0e137c0fedb81bbf3d88be4ed300eee163da] Staging: lustre: fix
>> space before and after comma in dt_object.c
>> git bisect good 623f0e137c0fedb81bbf3d88be4ed300eee163da
>> # bad: [8c7febe83915332276cab49e89f6580bb963fb9a] Merge tag
>> 'tty-4.2-rc1' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
>> git bisect bad 8c7febe83915332276cab49e89f6580bb963fb9a
>> # good: [bcf5b92d9bbf0b7683199615f0f184e89fa486bc] staging: rtl8192e:
>> Remove rt_hi_throughput::ChnkOp
>> git bisect good bcf5b92d9bbf0b7683199615f0f184e89fa486bc
>> # good: [71eec4836b834b992e0cefeefc8b85efe4cb185b] drivers: PL011: allow
>> avoiding UART enabling/disabling
>> git bisect good 71eec4836b834b992e0cefeefc8b85efe4cb185b
>> # good: [49ef9c850154756cf2fbc50fd3804c44675d4633] staging: comedi:
>> me4000: rename 'thisboard' variables
>> git bisect good 49ef9c850154756cf2fbc50fd3804c44675d4633
>> # good: [2a4462418af771ef9f1f1d1532bcbb8799df842d] tty/serial: kill off
>> set_irq_flags usage
>> git bisect good 2a4462418af771ef9f1f1d1532bcbb8799df842d
>> # good: [6d43b0f482561ab421a91ebf59a51192d66cf8a7] Staging: sm750fb:
>> ddk750_swi2c.c: Insert spaces around operators
>> git bisect good 6d43b0f482561ab421a91ebf59a51192d66cf8a7
>> # good: [84e1eb83d0b9e0969a59b6848d718eaf71e98fcb] MAINTAINERS: tty: add
>> serial docs directory
>> git bisect good 84e1eb83d0b9e0969a59b6848d718eaf71e98fcb
>> # good: [53a20e9e378ecd52f0afa4b60f8f8c81b6f97c27] staging: wilc1000:
>> disable driver due to build warnings
>> git bisect good 53a20e9e378ecd52f0afa4b60f8f8c81b6f97c27
>> # good: [71206b9f8120eb513c621d4f31906577bb658df3] Doc:
>> serial-rs485.txt: update RS485 driver interface
>> git bisect good 71206b9f8120eb513c621d4f31906577bb658df3
>> # bad: [23908db413eccd77084b09c9b0a4451dfb0524c0] Merge tag
>> 'staging-4.2-rc1' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
>> git bisect bad 23908db413eccd77084b09c9b0a4451dfb0524c0
>> # first bad commit: [23908db413eccd77084b09c9b0a4451dfb0524c0] Merge tag
>> 'staging-4.2-rc1' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
>>
>
> The bad commit is:
> commit 23908db413eccd77084b09c9b0a4451dfb0524c0
> Merge: 8d7804a 53a20e9
> Author: Linus Torvalds <torvalds@linux-foundation.org>
> Date:   Fri Jun 26 15:46:08 2015 -0700
>
> Hard to find which one causes the real issue... but i will try to
> reproduce
> it, could you please share your command line?
OPTS="-vga none"
OPTS="$OPTS -drive
if=pflash,format=raw,readonly,file=/home/janusz/uefi/OVMF_CODE.fd"
OPTS="$OPTS -drive if=pflash,format=raw,file=/home/janusz/uefi/OVMF_VARS.fd"
OPTS="$OPTS -enable-kvm -m 10000 -cpu host -smp
8,cores=4,threads=2,sockets=1"
OPTS="$OPTS -device
vfio-pci,host=01:00.0,romfile=/home/janusz/uefi/uefi-vga.bin"
OPTS="$OPTS -device vfio-pci,host=01:00.1"
OPTS="$OPTS -net nic,macaddr=50:E5:49:57:74:E3 -net bridge,vlan=0"
OPTS="$OPTS -soundhw hda"
OPTS="$OPTS -boot d"
OPTS="$OPTS -usb -usbdevice host:09da:000a -usbdevice host:1a2c:0c21"
OPTS="$OPTS -hda /mnt/storage2/win10.img"

QEMU_AUDIO_DRV=pa qemu-system-x86_64 $OPTS

my cpu is i7 6700k

commands for kernel:
intel_iommu=on vfio_iommu_type1.allow_unsafe_interrupts=0

on 4.1 I had to use:
i915.preliminary_hw_support=1 intel_iommu=on
vfio_iommu_type1.allow_unsafe_interrupts=0 intel_iommu=igfx_off because
otherwise I got kernel panic, but on versions of kernel that doesn't
work, with igfx_off my iommu groups were empty and it didn't panic
anymore without it