[Qemu-devel] Q35 FreeBSD install status

[Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

[-- Attachment #1: Type: text/plain, Size: 1206 bytes --]


    I'm looking for some guidance with regards to constructing a VM using a later version of FreeBSD. In reference to a qemu-devel posting:

http://lists.nongnu.org/archive/html/qemu-devel/2012-08/msg03329.html
    I reran with both 1.6.0.rc3 and 1.6.50 the q35 FreeBSD tests with 9.0, 9.1 and 10-CURRENT using the following command line:
qemu-kvm \
   -hda /home/ehv/images/126802.img -enable-kvm -m 1G -monitor stdio -M q35 \
    -vnc 127.0.0.1:1<http://127.0.0.1:1> -vga cirrus -monitor stdio \
    -cdrom /home/ehv/images/FreeBSD-9.0-RELEASE-amd64-dvd1.iso -boot order=d
Note the command line parameter changes necessary for the current versions of qemu. I'm running this on a Fedora 17/3.4.4 host and I've tried specifying SeaBios 1.7.3 with -bios and -acpitable command line options. Note also that both the ide and ahci tests (pc-i440fx-1.6) worked as advertised on this host. My trouble is that the bios does not see either the HDD image nor the cdrom.
    My questions are am I correct to expect the aforementioned q35 test (specifically the command line) to work? Am I missing something obvious? Is there something to look?
    Any insight will be appreciated. Thanks.

    Bret


[-- Attachment #2: Type: text/html, Size: 3687 bytes --]

Re: [Qemu-devel] Q35 FreeBSD install status

[Michael S. Tsirkin]

On Tue, Sep 10, 2013 at 04:16:35PM +0000, Bret Ketchum wrote:
>  
> 
>     I’m looking for some guidance with regards to constructing a VM using a
> later version of FreeBSD. In reference to a qemu-devel posting:
> 
> http://lists.nongnu.org/archive/html/qemu-devel/2012-08/msg03329.html
> 
>     I reran with both 1.6.0.rc3 and 1.6.50 the q35 FreeBSD tests with 9.0, 9.1
> and 10-CURRENT using the following command line:
> 
> qemu-kvm \
> 
>    -hda /home/ehv/images/126802.img -enable-kvm -m 1G -monitor stdio -M q35 \
>     -vnc 127.0.0.1:1 -vga cirrus -monitor stdio \
>     -cdrom /home/ehv/images/FreeBSD-9.0-RELEASE-amd64-dvd1.iso -boot order=d
> 
> Note the command line parameter changes necessary for the current versions of
> qemu. I'm running this on a Fedora 17/3.4.4 host and I've tried specifying
> SeaBios 1.7.3 with -bios and -acpitable command line options. Note also that
> both the ide and ahci tests (pc-i440fx-1.6) worked as advertised on this host.
> My trouble is that the bios does not see either the HDD image nor the cdrom.
> 
>     My questions are am I correct to expect the aforementioned q35 test
> (specifically the command line) to work? Am I missing something obvious? Is
> there something to look?
> 
>     Any insight will be appreciated. Thanks.
> 
>  
> 
>     Bret
> 
>  

-cdrom doesn't work with ahci.
there's a new flag that replaces it, I keep forgetting what it is.
Will look it up tomorrow if no one beats me to it.

-- 
MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

    Great thanks. Any idea why the HD does not show?

-----Original Message-----
From: Michael S. Tsirkin [mailto:mst@redhat.com] 
Sent: Tuesday, September 10, 2013 12:30 PM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
Subject: Re: Q35 FreeBSD install status

On Tue, Sep 10, 2013 at 04:16:35PM +0000, Bret Ketchum wrote:
>  
> 
>     I’m looking for some guidance with regards to constructing a VM 
> using a later version of FreeBSD. In reference to a qemu-devel posting:
> 
> http://lists.nongnu.org/archive/html/qemu-devel/2012-08/msg03329.html
> 
>     I reran with both 1.6.0.rc3 and 1.6.50 the q35 FreeBSD tests with 
> 9.0, 9.1 and 10-CURRENT using the following command line:
> 
> qemu-kvm \
> 
>    -hda /home/ehv/images/126802.img -enable-kvm -m 1G -monitor stdio -M q35 \
>     -vnc 127.0.0.1:1 -vga cirrus -monitor stdio \
>     -cdrom /home/ehv/images/FreeBSD-9.0-RELEASE-amd64-dvd1.iso -boot 
> order=d
> 
> Note the command line parameter changes necessary for the current 
> versions of qemu. I'm running this on a Fedora 17/3.4.4 host and I've 
> tried specifying SeaBios 1.7.3 with -bios and -acpitable command line 
> options. Note also that both the ide and ahci tests (pc-i440fx-1.6) worked as advertised on this host.
> My trouble is that the bios does not see either the HDD image nor the cdrom.
> 
>     My questions are am I correct to expect the aforementioned q35 
> test (specifically the command line) to work? Am I missing something 
> obvious? Is there something to look?
> 
>     Any insight will be appreciated. Thanks.
> 
>  
> 
>     Bret
> 
>  

-cdrom doesn't work with ahci.
there's a new flag that replaces it, I keep forgetting what it is.
Will look it up tomorrow if no one beats me to it.

--
MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Michael S. Tsirkin]

On Tue, Sep 10, 2013 at 05:29:06PM +0000, Bret Ketchum wrote:
> 
>     Great thanks. Any idea why the HD does not show?

There's no HD in your command line.

> -----Original Message-----
> From: Michael S. Tsirkin [mailto:mst@redhat.com] 
> Sent: Tuesday, September 10, 2013 12:30 PM
> To: Bret Ketchum
> Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
> Subject: Re: Q35 FreeBSD install status
> 
> On Tue, Sep 10, 2013 at 04:16:35PM +0000, Bret Ketchum wrote:
> >  
> > 
> >     I’m looking for some guidance with regards to constructing a VM 
> > using a later version of FreeBSD. In reference to a qemu-devel posting:
> > 
> > http://lists.nongnu.org/archive/html/qemu-devel/2012-08/msg03329.html
> > 
> >     I reran with both 1.6.0.rc3 and 1.6.50 the q35 FreeBSD tests with 
> > 9.0, 9.1 and 10-CURRENT using the following command line:
> > 
> > qemu-kvm \
> > 
> >    -hda /home/ehv/images/126802.img -enable-kvm -m 1G -monitor stdio -M q35 \
> >     -vnc 127.0.0.1:1 -vga cirrus -monitor stdio \
> >     -cdrom /home/ehv/images/FreeBSD-9.0-RELEASE-amd64-dvd1.iso -boot 
> > order=d
> > 
> > Note the command line parameter changes necessary for the current 
> > versions of qemu. I'm running this on a Fedora 17/3.4.4 host and I've 
> > tried specifying SeaBios 1.7.3 with -bios and -acpitable command line 
> > options. Note also that both the ide and ahci tests (pc-i440fx-1.6) worked as advertised on this host.
> > My trouble is that the bios does not see either the HDD image nor the cdrom.
> > 
> >     My questions are am I correct to expect the aforementioned q35 
> > test (specifically the command line) to work? Am I missing something 
> > obvious? Is there something to look?
> > 
> >     Any insight will be appreciated. Thanks.
> > 
> >  
> > 
> >     Bret
> > 
> >  
> 
> -cdrom doesn't work with ahci.
> there's a new flag that replaces it, I keep forgetting what it is.
> Will look it up tomorrow if no one beats me to it.
> 
> --
> MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

     As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:

    -device piix4-ide \
    -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
    -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \
    -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
    -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \

    But now the e1000 emulated network device under q35 does not seem to pass interrupts. Using:

    -netdev tap,id=hostnet0,script=/home/ehv/images/qemu-ifup,downscript=/home/ehv/images/qemu-ifdown \
    -device e1000,netdev=hostnet0 \

    Or a user network:

    -netdev user,id=hostnet0,net=192.168.76.0/24,dhcpstart=192.168.76.9 \

    But machine pc-i440fx-1.6 works just fine. Anyone have any thoughts on where to start looking or a possible workaround?




-----Original Message-----
From: Michael S. Tsirkin [mailto:mst@redhat.com] 
Sent: Tuesday, September 10, 2013 12:50 PM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
Subject: Re: Q35 FreeBSD install status

On Tue, Sep 10, 2013 at 05:29:06PM +0000, Bret Ketchum wrote:
> 
>     Great thanks. Any idea why the HD does not show?

There's no HD in your command line.

> -----Original Message-----
> From: Michael S. Tsirkin [mailto:mst@redhat.com]
> Sent: Tuesday, September 10, 2013 12:30 PM
> To: Bret Ketchum
> Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
> Subject: Re: Q35 FreeBSD install status
> 
> On Tue, Sep 10, 2013 at 04:16:35PM +0000, Bret Ketchum wrote:
> >  
> > 
> >     I’m looking for some guidance with regards to constructing a VM 
> > using a later version of FreeBSD. In reference to a qemu-devel posting:
> > 
> > http://lists.nongnu.org/archive/html/qemu-devel/2012-08/msg03329.htm
> > l
> > 
> >     I reran with both 1.6.0.rc3 and 1.6.50 the q35 FreeBSD tests 
> > with 9.0, 9.1 and 10-CURRENT using the following command line:
> > 
> > qemu-kvm \
> > 
> >    -hda /home/ehv/images/126802.img -enable-kvm -m 1G -monitor stdio -M q35 \
> >     -vnc 127.0.0.1:1 -vga cirrus -monitor stdio \
> >     -cdrom /home/ehv/images/FreeBSD-9.0-RELEASE-amd64-dvd1.iso -boot 
> > order=d
> > 
> > Note the command line parameter changes necessary for the current 
> > versions of qemu. I'm running this on a Fedora 17/3.4.4 host and 
> > I've tried specifying SeaBios 1.7.3 with -bios and -acpitable 
> > command line options. Note also that both the ide and ahci tests (pc-i440fx-1.6) worked as advertised on this host.
> > My trouble is that the bios does not see either the HDD image nor the cdrom.
> > 
> >     My questions are am I correct to expect the aforementioned q35 
> > test (specifically the command line) to work? Am I missing something 
> > obvious? Is there something to look?
> > 
> >     Any insight will be appreciated. Thanks.
> > 
> >  
> > 
> >     Bret
> > 
> >  
> 
> -cdrom doesn't work with ahci.
> there's a new flag that replaces it, I keep forgetting what it is.
> Will look it up tomorrow if no one beats me to it.
> 
> --
> MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Michael S. Tsirkin]

On Mon, Sep 16, 2013 at 12:28:41PM +0000, Bret Ketchum wrote:
> 
>      As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:
> 
>     -device piix4-ide \
>     -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
>     -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \
>     -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
>     -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \
> 
>     But now the e1000 emulated network device under q35 does not seem to pass interrupts. Using:
> 
>     -netdev tap,id=hostnet0,script=/home/ehv/images/qemu-ifup,downscript=/home/ehv/images/qemu-ifdown \
>     -device e1000,netdev=hostnet0 \
> 
>     Or a user network:
> 
>     -netdev user,id=hostnet0,net=192.168.76.0/24,dhcpstart=192.168.76.9 \
> 
>     But machine pc-i440fx-1.6 works just fine. Anyone have any
>     thoughts on where to start looking or a possible workaround?

Maybe irq gets asserted, but guest does not have a handler?
Take a look at system log, that might give some hints.


-- 
MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

    Syslog suggests the interface is brought down then up but no interrupt at the irq assigned. Need to enable more debug and make sure the interrupt is wired properly.

-----Original Message-----
From: Michael S. Tsirkin [mailto:mst@redhat.com] 
Sent: Monday, September 16, 2013 7:36 AM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
Subject: Re: Q35 FreeBSD install status

On Mon, Sep 16, 2013 at 12:28:41PM +0000, Bret Ketchum wrote:
> 
>      As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:
> 
>     -device piix4-ide \
>     -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
>     -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \
>     -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
>     -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \
> 
>     But now the e1000 emulated network device under q35 does not seem to pass interrupts. Using:
> 
>     -netdev tap,id=hostnet0,script=/home/ehv/images/qemu-ifup,downscript=/home/ehv/images/qemu-ifdown \
>     -device e1000,netdev=hostnet0 \
> 
>     Or a user network:
> 
>     -netdev user,id=hostnet0,net=192.168.76.0/24,dhcpstart=192.168.76.9 \
> 
>     But machine pc-i440fx-1.6 works just fine. Anyone have any
>     thoughts on where to start looking or a possible workaround?

Maybe irq gets asserted, but guest does not have a handler?
Take a look at system log, that might give some hints.


-- 
MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

     Qemu says the IRQ is 11:

  Bus  0, device   3, function 0:
    Ethernet controller: PCI device 8086:100e
      IRQ 11.
      BAR0: 32 bit memory at 0xfebc0000 [0xfebdffff].
      BAR1: I/O at 0xc040 [0xc07f].
      BAR6: 32 bit memory at 0xffffffffffffffff [0x0003fffe].
      id ""

    FreeBSD thinks the IRQ is 23:
em0@pci0:0:3:0: class=0x020000 card=0x11001af4 chip=0x100e8086 rev=0x03 hdr=0x00
pciconf -r -b pci0:0:4:0 0x3c:0x3d
17 01


-----Original Message-----
From: Bret Ketchum 
Sent: Monday, September 16, 2013 8:03 AM
To: 'Michael S. Tsirkin'
Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
Subject: RE: Q35 FreeBSD install status


    Syslog suggests the interface is brought down then up but no interrupt at the irq assigned. Need to enable more debug and make sure the interrupt is wired properly.

-----Original Message-----
From: Michael S. Tsirkin [mailto:mst@redhat.com] 
Sent: Monday, September 16, 2013 7:36 AM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; kraxel@redhat.com
Subject: Re: Q35 FreeBSD install status

On Mon, Sep 16, 2013 at 12:28:41PM +0000, Bret Ketchum wrote:
> 
>      As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:
> 
>     -device piix4-ide \
>     -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
>     -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \
>     -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
>     -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \
> 
>     But now the e1000 emulated network device under q35 does not seem to pass interrupts. Using:
> 
>     -netdev tap,id=hostnet0,script=/home/ehv/images/qemu-ifup,downscript=/home/ehv/images/qemu-ifdown \
>     -device e1000,netdev=hostnet0 \
> 
>     Or a user network:
> 
>     -netdev user,id=hostnet0,net=192.168.76.0/24,dhcpstart=192.168.76.9 \
> 
>     But machine pc-i440fx-1.6 works just fine. Anyone have any
>     thoughts on where to start looking or a possible workaround?

Maybe irq gets asserted, but guest does not have a handler?
Take a look at system log, that might give some hints.


-- 
MST

Re: [Qemu-devel] Q35 FreeBSD install status

[Gerd Hoffmann]

On Mo, 2013-09-16 at 12:28 +0000, Bret Ketchum wrote:
>      As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:
> 
>     -device piix4-ide \

No need for that one.

>     -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
>     -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \

>     -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
>     -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \

The busses of the q35 ahci controller are named "ide.0" ... "ide.5" (for
sata ports 1..6).  Just use these with -device ide-{hd,cd}.  Each device
needs its own bus, master/slave doesn't exist with sata.

HTH,
  Gerd

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

    Tried that - looks like FreeBSD expects to use MSI for AHCI devices which does not appear to be supported in q35.

-----Original Message-----
From: Gerd Hoffmann [mailto:kraxel@redhat.com] 
Sent: Monday, September 16, 2013 7:55 AM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; mst@redhat.com
Subject: Re: [Qemu-devel] Q35 FreeBSD install status

On Mo, 2013-09-16 at 12:28 +0000, Bret Ketchum wrote:
>      As a workaround until I can uncover the qemu/FreeBSD AHCI/SATA issue, I simply create a legacy IDE to hang the CD and HD from:
> 
>     -device piix4-ide \

No need for that one.

>     -drive if=none,file=/home/ehv/images/FreeBSD-9.1-RELEASE-amd64-dvd1.iso,id=drive-ide0-0-0 \
>     -device ide-cd,bus=ide.0,drive=drive-ide0-0-0,id=ide0-0-0 \

>     -drive if=none,file=/home/ehv/images/r060501.img,format=raw,id=drive-ide0-0-1 \
>     -device ide-hd,bus=ide.0,drive=drive-ide0-0-1,id=ide0-0-1 \

The busses of the q35 ahci controller are named "ide.0" ... "ide.5" (for sata ports 1..6).  Just use these with -device ide-{hd,cd}.  Each device needs its own bus, master/slave doesn't exist with sata.

HTH,
  Gerd

Re: [Qemu-devel] Q35 FreeBSD install status

[Gerd Hoffmann]

On Mo, 2013-09-16 at 12:57 +0000, Bret Ketchum wrote:
>     Tried that - looks like FreeBSD expects to use MSI for AHCI devices which does not appear to be supported in q35.

It is supported and linux has no problems using ahci with msi.

Given that e1000 seems to have interrupt problems too I wouldn't be
surprised if both issues have the same root cause.  Might be something
in the seabios acpi tables.  Or a bug in FreeBSD.

cheers,
  Gerd

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

    The emulated e1000 uses legacy interrupts. I'm wondering if someone can help me understand the mptable (used by FreeBSD to route interrupts and defined in Intel's MP spec):

Bus:            Bus ID  Type
                 0       PCI
                 1       ISA

I/O APICs:      APIC ID Version State           Address
                 0       0x11    usable          0xfec00000

I/O Ints:       Type    Polarity    Trigger     Bus ID   IRQ    APIC ID    PIN#
                INT     active-hi    conforms        0    31:A        2    10
                INT     active-hi    conforms        0     3:A        2    11


     The interrupts associated with the PCI bus where e1000 hangs suggest there is an APIC with an ID of  2 but there is only one APIC (with an ID of 0). 

-----Original Message-----
From: Gerd Hoffmann [mailto:kraxel@redhat.com] 
Sent: Monday, September 16, 2013 10:15 AM
To: Bret Ketchum
Cc: Qemu-devel@nongnu.org; mst@redhat.com
Subject: Re: [Qemu-devel] Q35 FreeBSD install status

On Mo, 2013-09-16 at 12:57 +0000, Bret Ketchum wrote:
>     Tried that - looks like FreeBSD expects to use MSI for AHCI devices which does not appear to be supported in q35.

It is supported and linux has no problems using ahci with msi.

Given that e1000 seems to have interrupt problems too I wouldn't be surprised if both issues have the same root cause.  Might be something in the seabios acpi tables.  Or a bug in FreeBSD.

cheers,
  Gerd

Re: [Qemu-devel] Q35 FreeBSD install status

[Michael S. Tsirkin]

Could be MPT is not parsed correctly?
I see that we put 0 in the ID in bios code.
Could you dump the MPT and send it on list?

On Tue, Sep 17, 2013 at 03:25:38PM +0000, Bret Ketchum wrote:
> 
>     The emulated e1000 uses legacy interrupts. I'm wondering if someone can help me understand the mptable (used by FreeBSD to route interrupts and defined in Intel's MP spec):
> 
> Bus:            Bus ID  Type
>                  0       PCI
>                  1       ISA
> 
> I/O APICs:      APIC ID Version State           Address
>                  0       0x11    usable          0xfec00000
> 
> I/O Ints:       Type    Polarity    Trigger     Bus ID   IRQ    APIC ID    PIN#
>                 INT     active-hi    conforms        0    31:A        2    10
>                 INT     active-hi    conforms        0     3:A        2    11
> 
> 
>      The interrupts associated with the PCI bus where e1000 hangs suggest there is an APIC with an ID of  2 but there is only one APIC (with an ID of 0). 
> 
> -----Original Message-----
> From: Gerd Hoffmann [mailto:kraxel@redhat.com] 
> Sent: Monday, September 16, 2013 10:15 AM
> To: Bret Ketchum
> Cc: Qemu-devel@nongnu.org; mst@redhat.com
> Subject: Re: [Qemu-devel] Q35 FreeBSD install status
> 
> On Mo, 2013-09-16 at 12:57 +0000, Bret Ketchum wrote:
> >     Tried that - looks like FreeBSD expects to use MSI for AHCI devices which does not appear to be supported in q35.
> 
> It is supported and linux has no problems using ahci with msi.
> 
> Given that e1000 seems to have interrupt problems too I wouldn't be surprised if both issues have the same root cause.  Might be something in the seabios acpi tables.  Or a bug in FreeBSD.
> 
> cheers,
>   Gerd
> 
> 
> 

Re: [Qemu-devel] Q35 FreeBSD install status

[Bret Ketchum]

    For those following along. The mptable I presented was an old table, updating to SeaBIOS 1.7.3 fixed that but still no interrupts. However, it appears FreeBSD has trouble parsing the ACPI tables presented via q35-acpi-dsdt.aml as it does not create the sysresource devices associated with ACPI. A work around until I uncover the parsing issue it to revert to using the mptable to wire PCI interrupts via 'debug.acpi.disabled=pci_link'. FreeBSD will whine about failing to force attach of GSIs but thus far all seems to be well with the world. I've yet to check but I suspect the parsing issue is also root cause of uhci/ehci missing interrupts.

-----Original Message-----
From: Michael S. Tsirkin [mailto:mst@redhat.com] 
Sent: Tuesday, September 17, 2013 11:44 AM
To: Bret Ketchum
Cc: Gerd Hoffmann; Qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] Q35 FreeBSD install status


Could be MPT is not parsed correctly?
I see that we put 0 in the ID in bios code.
Could you dump the MPT and send it on list?

On Tue, Sep 17, 2013 at 03:25:38PM +0000, Bret Ketchum wrote:
> 
>     The emulated e1000 uses legacy interrupts. I'm wondering if someone can help me understand the mptable (used by FreeBSD to route interrupts and defined in Intel's MP spec):
> 
> Bus:            Bus ID  Type
>                  0       PCI
>                  1       ISA
> 
> I/O APICs:      APIC ID Version State           Address
>                  0       0x11    usable          0xfec00000
> 
> I/O Ints:       Type    Polarity    Trigger     Bus ID   IRQ    APIC ID    PIN#
>                 INT     active-hi    conforms        0    31:A        2    10
>                 INT     active-hi    conforms        0     3:A        2    11
> 
> 
>      The interrupts associated with the PCI bus where e1000 hangs suggest there is an APIC with an ID of  2 but there is only one APIC (with an ID of 0). 
> 
> -----Original Message-----
> From: Gerd Hoffmann [mailto:kraxel@redhat.com] 
> Sent: Monday, September 16, 2013 10:15 AM
> To: Bret Ketchum
> Cc: Qemu-devel@nongnu.org; mst@redhat.com
> Subject: Re: [Qemu-devel] Q35 FreeBSD install status
> 
> On Mo, 2013-09-16 at 12:57 +0000, Bret Ketchum wrote:
> >     Tried that - looks like FreeBSD expects to use MSI for AHCI devices which does not appear to be supported in q35.
> 
> It is supported and linux has no problems using ahci with msi.
> 
> Given that e1000 seems to have interrupt problems too I wouldn't be surprised if both issues have the same root cause.  Might be something in the seabios acpi tables.  Or a bug in FreeBSD.
> 
> cheers,
>   Gerd
> 
> 
>