On 8 Jan 2021, at 21:16, Gupta, Sandeep <Sandeep.Gupta@amd.com> wrote:

[AMD Official Use Only - Internal Distribution Only]

Thank you Christophe,

Yes  -  We change .../vendor/src/github.com/intel/govmm/qemu/qemu.go,   "appendBios()"

func (config *Config) appendBios() {
       if config.Bios != "" {
               if config.Knobs.MemEncrypt == true {
                       config.qemuParams = append(config.qemuParams, "-drive")
                       config.qemuParams = append(config.qemuParams, fmt.Sprintf("if=pflash,format=raw,unit=0,file=%s,readonly", config.Bios))
               } else {
                       config.qemuParams = append(config.qemuParams, "-bios")
                       config.qemuParams = append(config.qemuParams, config.Bios)
               }
       }
}

So I see two problems with that change:

1/ I think it depends on using OVMF, not on encrypted memory.

2/ You need the VARS file as discussed earlier.

I believe that a more robust solution would be:

1/ Add an OVMF option alongside "firmware", as well as OVMF_VARS

2/ When OVMF is specified, create a copy of the VARS file, and pass two -drive options, one for OVMF_CODE and one for the copied OVMF_VARS.

3/ When MemEncrypt is true, check that you ahve an OVMF given and no bios.




But then "containerd" does not use this method ?

--Thanking You
--Sandeep


-----Original Message-----
From: Christophe de Dinechin <cdupontd@redhat.com> 
Sent: Thursday, January 7, 2021 12:43 AM
To: Gupta, Sandeep <Sandeep.Gupta@amd.com>
Cc: Fabiano FidĂȘncio <fidencio@redhat.com>; kata-dev@lists.katacontainers.io; Hollingsworth, Brent <brent.hollingsworth@amd.com>
Subject: Re: [kata-dev] Regarding, Kata 2.0.0, kata-runtime, qemu OVMF
Importance: High

[CAUTION: External Email]

On 7 Jan 2021, at 09:01, Gupta, Sandeep <Sandeep.Gupta@amd.com> wrote:

[AMD Official Use Only - Internal Distribution Only]

Hi All,

I may have mixed up my working and not working cases - sorry about 
this, I have finally isolated my working and not-working cases, (for sure) as follows :

This seems consistent with what we deduced from the earlier exchanges.

In short:


     . Not working :

msg="launching /usr/local/bin/qemu-system-x86_64 with: [-name sandbox-csh -uuid 8b5369f5-1439-48d7-9b67-a32f6ef8ff7f -machine q35,accel=kvm,kernel_irqchip -cpu host, -qmp unix:/run/vc/vm/csh/qmp.sock,server,nowait -m 2048M,slots=10,maxmem=129755M -device pci-bridge,bus=pcie.0,id=pci-bridge-0,chassis_nr=1,shpc=on,addr=2,romfile= -device virtio-serial-pci,disable-modern=false,id=serial0,romfile= -device virtconsole,chardev=charconsole0,id=console0 -chardev socket,id=charconsole0,path=/run/vc/vm/csh/console.sock,server,nowait -object rng-random,id=rng0,filename=/dev/urandom -device virtio-rng-pci,rng=rng0,romfile= -device virtserialport,chardev=charch0,id=channel0,name=agent.channel.0 -chardev socket,id=charch0,path=/run/vc/vm/csh/kata.sock,server,nowait -device virtio-9p-pci,disable-modern=false,fsdev=extra-9p-kataShared,mount_tag=kataShared,romfile= -fsdev local,id=extra-9p-kataShared,path=/run/kata-containers/shared/sandboxes/csh/shared,security_model=none,multidevs=remap -rtc base=utc,driftfix=slew,clock=host -global kvm-pit.lost_tick_policy=discard -vga none -no-user-config -nodefaults -nographic --no-reboot -daemonize -object memory-backend-ram,id=dimm1,size=2048M -numa node,memdev=dimm1 -kernel /usr/share/kata-containers/vmlinuz-sev-es-5.6-v3.container -initrd /usr/share/kata-containers/kata-containers-initrd.img -append tsc=reliable no_timer_check rcupdate.rcu_expedited=1 i8042.direct=1 i8042.dumbkbd=1 i8042.nopnp=1 i8042.noaux=1 noreplace-smp reboot=k console=hvc0 console=hvc1 cryptomgr.notests net.ifnames=0 pci=lastbus=0 iommu=off debug panic=1 nr_cpus=240 agent.use_vsock=false root=/dev/vda1 rootflags=data=ordered,errors=remount-ro agent.log=debug initcall_debug -bios /usr/local/share/qemu/OVMF_CODE.fd.kata -pidfile /run/vc/vm/csh/pid -D /run/vc/vm/csh/qemu.log -smp 1,cores=1,threads=1,sockets=240,maxcpus=240]"

So this one uses the -bios option.
-bios /usr/local/share/qemu/OVMF_CODE.fd.kata

I believe you told us this is what 2.0 does, right?



     . Working :

/usr/local/bin/qemu-system-x86_64 -name 
sandbox-91d9f660fe1338d724c0ccd5c755a618d44100e697c82f8753d1051d2fdb30
c1 -uuid e7888d8f-2e65-43d5-bc5a-3a3e853b3e13 -machine 
q35,accel=kvm,kernel_irqchip,nvdimm -cpu host -qmp 
unix:/run/vc/sbs/91d9f660fe1338d724c0ccd5c755a618d44100e697c82f8753d10
51d2fdb30c1/qmp.sock,server,nowait -m 512M,slots=2,maxmem=258779M 
-device 
pci-bridge,bus=pcie.0,id=pci-bridge-0,chassis_nr=1,shpc=on,addr=2 
-device 
virtio-serial-pci,disable-legacy=on,id=serial0,iommu_platform=true 
-device virtconsole,chardev=charconsole0,id=console0 -chardev 
socket,id=charconsole0,path=/run/vc/sbs/91d9f660fe1338d724c0ccd5c755a6
18d44100e697c82f8753d1051d2fdb30c1/console.sock,server,nowait -object 
sev-guest,id=sev0,cbitpos=51,reduced-phys-bits=5 -machine 
memory-encryption=sev0 -device 
virtserialport,chardev=charch0,id=channel0,name=agent.channel.0 
-chardev 
socket,id=charch0,path=/run/vc/sbs/91d9f660fe1338d724c0ccd5c755a618d44
100e697c82f8753d1051d2fdb30c1/kata.sock,server,nowait -device 
virtio-9p-pci,disable-legacy=on,iommu_platform=true,fsdev=extra-9p-kat
aShared,mount_tag=kataShared -fsdev 
local,id=extra-9p-kataShared,path=/run/kata-containers/shared/sandboxe
s/91d9f660fe1338d724c0ccd5c755a618d44100e697c82f8753d1051d2fdb30c1,sec
urity_model=none -netdev 
tap,id=network-0,vhost=on,vhostfds=3:4:5:6:7:8:9:10,fds=11:12:13:14:15
:16:17:18 -device 
driver=virtio-net-pci,netdev=network-0,mac=02:42:ac:11:00:02,disable-l
egacy=on,iommu_platform=true,mq=on,vectors=18 -rtc 
base=utc,driftfix=slew -global kvm-pit.lost_tick_policy=discard -vga 
none -no-user-config -nodefaults -nographic -daemonize -kernel 
/usr/share/kata-containers/vmlinuz-sev-es-5.6-v3.container -initrd 
/usr/share/kata-containers/kata-containers-initrd.img -append 
tsc=reliable no_timer_check rcupdate.rcu_expedited=1 i8042.direct=1 
i8042.dumbkbd=1 i8042.nopnp=1 i8042.noaux=1 noreplace-smp reboot=k 
console=hvc0 console=hvc1 iommu=off cryptomgr.notests net.ifnames=0 
pci=lastbus=0 debug panic=1 initcall_debug nr_cpus=240 
ip=::::::91d9f660fe1338d724c0ccd5c755a618d44100e697c82f8753d1051d2fdb3
0c1::off:: root=/dev/vda1 rootflags=data=ordered,errors=remount-ro 
agent.log=debug initcall_debug -drive 
if=pflash,format=raw,unit=0,file=/usr/local/share/qemu/OVMF_CODE.fd.ka
ta,readonly -smp
1,cores=1,threads=1,sockets=1,maxcpus=240


This one uses -drive
-drive if=pflash,format=raw,unit=0,file=/usr/local/share/qemu/OVMF_CODE.fd.kata,readonly

I believe this was your 1.x demo, right?

Now, this -drive is good, but it is unexpected, because I see nowhere in the normal Kata code that would emit that. This suggests you use a version I don't know about, or maybe that there is for example a patch that was sent to a release but dropped from the 2.0 branch.

Further, please note that Laszlo Ersek seems to imply in a parallel thread that -drive is no longer the right way to do it either. We may need to implement qemu-version dependent options.


--Best Regards,
--Sandeep Gupta


From: Christophe de Dinechin <cdupontd@redhat.com>
Sent: Wednesday, January 6, 2021 7:00 AM
To: Gupta, Sandeep <Sandeep.Gupta@amd.com>
Cc: Fabiano FidĂȘncio <fidencio@redhat.com>; 
kata-dev@lists.katacontainers.io; Hollingsworth, Brent 
<brent.hollingsworth@amd.com>
Subject: Re: [kata-dev] Regarding, Kata 2.0.0, kata-runtime, qemu OVMF

[CAUTION: External Email]
Hi Sandeep,




On 6 Jan 2021, at 15:19, Christophe de Dinechin <cdupontd@redhat.com> wrote:

Hi Sandeep,



On 6 Jan 2021, at 09:51, Christophe de Dinechin <cdupontd@redhat.com> wrote:

Hi Sandeep,


On 5 Jan 2021, at 21:19, Gupta, Sandeep <Sandeep.Gupta@amd.com> wrote:

[AMD Official Use Only - Internal Distribution Only]

Thanks again Christophe,

From my kata-runtime  1.x, setup :

root     2972741 2845780  0 20:13 pts/2    00:00:00 docker run --runtime kata-sev-runtime --cap-add=NET_ADMIN -it apache2-sev


root     2018662 2018416  0 15:01 ?        00:00:16 /usr/local/bin/qemu-system-x86_64 -name sandbox-4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc -uuid fce8220b-780b-4da6-8f5d-77ace1e06ed6 -machine q35,accel=kvm,kernel_irqchip,nvdimm -cpu host -qmp unix:/run/vc/sbs/4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc/qmp.sock,server,nowait -m 512M,slots=2,maxmem=258779M -device pci-bridge,bus=pcie.0,id=pci-bridge-0,chassis_nr=1,shpc=on,addr=2 -device virtio-serial-pci,disable-legacy=on,id=serial0,iommu_platform=true -device virtconsole,chardev=charconsole0,id=console0 -chardev socket,id=charconsole0,path=/run/vc/sbs/4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc/console.sock,server,nowait -object sev-guest,id=sev0,cbitpos=51,reduced-phys-bits=5 -machine memory-encryption=sev0 -device virtserialport,chardev=charch0,id=channel0,name=agent.channel.0 -chardev socket,id=charch0,path=/run/vc/sbs/4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc/kata.sock,server,nowait -device virtio-9p-pci,disable-legacy=on,iommu_platform=true,fsdev=extra-9p-kataShared,mount_tag=kataShared -fsdev local,id=extra-9p-kataShared,path=/run/kata-containers/shared/sandboxes/4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc,security_model=none -netdev tap,id=network-0,vhost=on,vhostfds=3:4:5:6:7:8:9:10,fds=11:12:13:14:15:16:17:18 -device driver=virtio-net-pci,netdev=network-0,mac=02:42:ac:11:00:02,disable-legacy=on,iommu_platform=true,mq=on,vectors=18 -rtc base=utc,driftfix=slew -global kvm-pit.lost_tick_policy=discard -vga none -no-user-config -nodefaults -nographic -daemonize -kernel /usr/share/kata-containers/vmlinuz-sev-es-5.6-v3.container -initrd /usr/share/kata-containers/kata-containers-initrd.img -append tsc=reliable no_timer_check rcupdate.rcu_expedited=1 i8042.direct=1 i8042.dumbkbd=1 i8042.nopnp=1 i8042.noaux=1 noreplace-smp reboot=k console=hvc0 console=hvc1 iommu=off cryptomgr.notests net.ifnames=0 pci=lastbus=0 debug panic=1 initcall_debug nr_cpus=240 ip=::::::4a28f6d66e95a08be407b10e1fbfd6d0bd41d4048b5823029f02a17b7200adbc::off:: root=/dev/vda1 rootflags=data=ordered,errors=remount-ro agent.log=debug initcall_debug -drive if=pflash,format=raw,unit=0,file=/usr/local/share/qemu/OVMF_CODE.fd.kata,readonly -smp 1,cores=1,threads=1,sockets=1,maxcpus=240

As Fabiano pointed out (indirectly), there is no "-bios" option in the 
command line above. So either

I'm sorry, Fabiano pointed out that I somehow had managed to miss the -drive if=pflash,format=raw,unit=0,file=/usr/local/share/qemu/OVMF_CODE.fd.kata,readonly option. Not enough coffee, I guess.

There is something mysterious about this, as I pointed out in https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fkata-containers%2Fkata-containers%2Fissues%2F1231%23issuecomment-755333403&amp;data=04%7C01%7Csandeep.gupta%40amd.com%7C3d884bd4e4904f858e6308d8b2e83de9%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637456057849018749%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=6Gm0u0gxkCbpYwD6eSnpi99BtnGszhvyWVV%2Bzfs2dMc%3D&amp;reserved=0. Basically, govmm does not seem to emit that option.

func (config *Config) appendBios() {
if config.Bios != "" {
config.qemuParams = append(config.qemuParams, "-bios") 
config.qemuParams = append(config.qemuParams, config.Bios) } }

It apparently has always been like this. Do you remember either patching that file, or maybe finding some other way to pass the OVMF arguments to qemu (like using a qemu wrapper)?



Disregard what is below. Instead, now we have an observable command-line difference between 1.x and 2.0. My guess is that this explains why you have trouble with 2.0. Even if my own quick experiments today showed that you can boot an OVMF with -bios, I'm not sure you can get SecureBoot that way, because of the missing vars partition.

I've opened https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fkata-containers%2Fkata-containers%2Fissues%2F1231&amp;data=04%7C01%7Csandeep.gupta%40amd.com%7C3d884bd4e4904f858e6308d8b2e83de9%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637456057849018749%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=tUAcbzpVNvm8iiSX1UuOcLAjwcLTNBNQY1MUEAu2%2Bf0%3D&amp;reserved=0. I will add a comment highlighting the difference.


1) You did that run without having the "firmware = " option. If that 
was the case, please send the command-line with the firmware option

(Hypothesis: maybe you are looking at the wrong configuration file, 
e.g. you have a local /etc/ override; that might explain why it 
"worked" with 1.x, although it would then require investigation to 
understand why SEV was enabled without your firmware)

2) Something else removes the firmware setting or causes qemu to not 
receive the -bios option

When "firmware" is specified, qemu should be invoked with the -bios option, and the corresponding commit dates back to early 2018, so I don't expect any difference between 1.x and 2.0 there.


Cheers,
Christophe




root     2972841 2970388  0 20:13 ?        00:00:00 containerd-shim -namespace moby -workdir /var/lib/docker/containerd/daemon/io.containerd.runtime.v1.linux/moby/54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8 -address /var/run/docker/containerd/containerd.sock -containerd-binary /snap/docker/471/bin/containerd -runtime-root /var/run/docker/runtime-kata-sev-runtime -debug


root     2973736 2972841  0 20:13 ?        00:00:00 /usr/libexec/kata-containers/kata-proxy -listen-socket unix:///run/vc/sbs/54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8/proxy.sock -mux-socket /run/vc/sbs/54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8/kata.sock -log debug -agent-logs-socket /run/vc/sbs/54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8/console.sock


root     2973856 2972841  0 20:13 pts/4    00:00:00 /usr/libexec/kata-containers/kata-shim -agent unix:///run/vc/sbs/54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8/proxy.sock -container 54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8 -exec-id 54db6810ae4ed8ab3e7a378c4e95d73e334797a5a5234978b4f976750f4f64d8 -terminal -log debug


--Best Regards,
--Sandeep Gupta


-----Original Message-----
From: Christophe de Dinechin <cdupontd@redhat.com>
Sent: Tuesday, January 5, 2021 11:25 AM
To: Gupta, Sandeep <Sandeep.Gupta@amd.com>
Cc: Fabiano FidĂȘncio <fidencio@redhat.com>; 
kata-dev@lists.katacontainers.io; Hollingsworth, Brent 
<brent.hollingsworth@amd.com>
Subject: Re: [kata-dev] Regarding, Kata 2.0.0, kata-runtime, qemu OVMF

[CAUTION: External Email]


On 5 Jan 2021, at 17:21, Gupta, Sandeep <Sandeep.Gupta@amd.com> wrote:


Hi Developers,
One more please - for Kata 2.0.0, kata-runtime, qemu OVMF

In /etc/kata-containers/configuration.toml --

======
# WARNING: - any parameter specified here will take priority over the 
default # parameter value of the same name used to start the virtual machine.
# Do not set values here unless you understand the impact of doing so 
as you # may stop the virtual machine from booting.
# To see the list of default parameters, enable hypervisor debug, 
create a # container and look for 'default-kernel-parameters' log entries.
kernel_params = " root=/dev/vda1 rootflags=data=ordered,errors=remount-ro agent.log=debug initcall_debug"

# Path to the firmware.
# If you want that qemu uses the default firmware leave this option 
empty #firmware = "/usr/local/share/qemu/OVMF_CODE.fd.kata"
firmware = ""
==========

If I use the default firmware - I get correct results :

# ctr run --runtime io.containerd.run.kata.v2 -t --rm 
docker.io/library/busybox:latest sh / # / # uname -a Linux 
sandeepg-ubuntu 5.6.0-sev-sev-es-5.6-v3.container #3 SMP Tue Nov 3
06:38:40 UTC 2020 x86_64 GNU/Linux / #

# ps -eaf | grep -i qemu

root      360867       1  0 15:56 ?        00:00:01 /usr/local/bin/qemu-system-x86_64 -name sandbox-sh -uuid 72820914-dcce-4cea-a01f-3cf9851d136d -machine q35,accel=kvm,kernel_irqchip -cpu host,pmu=off -qmp unix:/run/vc/vm/sh/qmp.sock,server,nowait -m 2048M,slots=10,maxmem=129755M -device pci-bridge,bus=pcie.0,id=pci-bridge-0,chassis_nr=1,shpc=on,addr=2,romfile= -device virtio-serial-pci,disable-modern=false,id=serial0,romfile= -device virtconsole,chardev=charconsole0,id=console0 -chardev socket,id=charconsole0,path=/run/vc/vm/sh/console.sock,server,nowait -device virtio-scsi-pci,id=scsi0,disable-modern=false,romfile= -object rng-random,id=rng0,filename=/dev/urandom -device virtio-rng-pci,rng=rng0,romfile= -device vhost-vsock-pci,disable-modern=false,vhostfd=3,id=vsock-1189071429,guest-cid=1189071429,romfile= -device virtio-9p-pci,disable-modern=false,fsdev=extra-9p-kataShared,mount_tag=kataShared,romfile= -fsdev local,id=extra-9p-kataShared,path=/run/kata-containers/shared/sandboxes/sh/shared,security_model=none,multidevs=remap -rtc base=utc,driftfix=slew,clock=host -global kvm-pit.lost_tick_policy=discard -vga none -no-user-config -nodefaults -nographic --no-reboot -daemonize -object memory-backend-ram,id=dimm1,size=2048M -numa node,memdev=dimm1 -kernel /usr/share/kata-containers/vmlinuz-sev-es-5.6-v3.container -initrd /usr/share/kata-containers/kata-containers-initrd.img -append tsc=reliable no_timer_check rcupdate.rcu_expedited=1 i8042.direct=1 i8042.dumbkbd=1 i8042.nopnp=1 i8042.noaux=1 noreplace-smp reboot=k console=hvc0 console=hvc1 cryptomgr.notests net.ifnames=0 pci=lastbus=0 quiet panic=1 nr_cpus=240 scsi_mod.scan=none root=/dev/vda1 rootflags=data=ordered,errors=remount-ro agent.log=debug initcall_debug -pidfile /run/vc/vm/sh/pid -smp 1,cores=1,threads=1,sockets=240,maxcpus=240
root      360966  360958  0 16:05 pts/2    00:00:00 grep --color=auto -i qemu


But if I use :   firmware = "/usr/local/share/qemu/OVMF_CODE.fd.kata"
I get Error:

# ctr run --runtime io.containerd.run.kata.v2 -t --rm 
docker.io/library/busybox:latest s
ctr: failed to launch qemu: exit status 1, error messages from qemu log: qemu: could not load PC BIOS '/usr/local/share/qemu/OVMF_CODE.fd.kata'
: unknown

Any suggestions please.

I did a quick look. The runtime apparently uses the -bios option when the firmware option is passed, not the -drive option as I thought it would. I'm not entirely sure you can boot OVMF that way. I don't have any idea at the moment why it worked with version 1.x. If you could share with me the 1.x qemu command line option with the firmware option, I'm curious to see what it is.

Thanks
Christophe



--Thanking You
--Sandeep