Enable Kata container on ppc64le arch #286

nitkon · 2018-05-07T14:19:37Z

This patch enables kata container using
initrd approach on ppc64le architecture.

Fixes #302

Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

nitkon · 2018-05-07T14:36:51Z

For testing purpose, the kernel for running kata on ppc64le using the initrd approach can be taken from: https://github.com/nitkon/kata/blob/master/vmlinux

devimc · 2018-05-07T15:10:18Z

@nitkon please file and issue and add fixes #XXX in the commit message, where XXX is the number of the issue

devimc · 2018-05-07T15:20:44Z

arch/ppc64le-options.mk

@@ -0,0 +1,21 @@
+# Copyright (c) 2018 Intel Corporation


s/Intel Corporation/IBM/ ?

please use next style for all licence headers

// Copyright (c) 2018 XXXXXXXXXXXXXXXXX // // SPDX-License-Identifier: Apache-2.0 //

devimc · 2018-05-07T15:21:03Z

cli/kata-check_ppc64le.go

@@ -0,0 +1,110 @@
+// Copyright (c) 2018 Intel Corporation


s/Intel Corporation/IBM/ ?

devimc · 2018-05-07T15:21:17Z

virtcontainers/qemu_ppc64le.go

@@ -0,0 +1,199 @@
+//
+// Copyright (c) 2018 Intel Corporation


s/Intel Corporation/IBM/ ?

jodh-intel · 2018-05-08T15:19:22Z

arch/ppc64le-options.mk

+MACHINETYPE := pseries
+KERNELPARAMS :=
+MACHINEACCELERATORS :=
+KERNELPATH := /usr/share/kata-containers/vmlinux.container


Is this line really required? If so, please can you add a comment stating that this architecture must use an uncompressed kernel.

A better approach might be to define a new variable so that you could say something like:

KERNEL_NAME := vmlinux.container

... or even:

KERNEL_TYPE := uncompressed ifeq (uncompressed,$(KERNEL_TYPE)) KERNEL_NAME = vmlinux.container else KERNEL_NAME = vmlinuz.container endif

jodh-intel · 2018-05-08T15:23:01Z

cli/kata-check_amd64.go

+// hostIsVMContainerCapable checks to see if the host is theoretically capable
+// of creating a VM container.
+
+func hostIsVMContainerCapable(details vmContainerCapableDetails) error {


I think this change will break the ARM64 build. Maybe you could leave this function in kata-check.go but rename it to something like genericHostIsVMContainerCapable() and then call it from kata-check_amd64.go and (a new) kata-check_arm64.go?

This also raises a larger issue: our CI currently only tests x86_64. Do you have any thoughts on how we could add PPC64 to our CI to ensure this code doesn't break at some point in the future? If you could handle the ongoing testing of PPC64 that would be great.

/cc @egernst, @bergwolf, @chavafg.

jodh-intel · 2018-05-08T15:25:23Z

virtcontainers/hypervisor.go

@@ -458,6 +458,10 @@ func RunningOnVMM(cpuInfoPath string) (bool, error) {
 		return false, nil
 	}

+	if runtime.GOARCH == "ppc64le" {
+		virtLog.Debugf("Unable to know if the system is running inside a VM")


I think this might be better as an info or warning message. Note that you don't need to use the "f" version of the API as there are no additional args (so virtLog.Info() or virtLog.Warn()).

jodh-intel · 2018-05-08T15:26:07Z

virtcontainers/qemu_arch_base.go

@@ -127,6 +127,9 @@ const (

 	// QemuVirt is the QEMU virt machine type for aarch64
 	QemuVirt = "virt"
+
+	// QemuPpc64le is a QEMU virt machine type for for ppc64le
+	QemuPpc64le = "pseries"


This camel case looks a bit odd to my eyes. If the golang compiler agrees could you make this QemuPPC64le (same comment to some of the code below)?

jodh-intel · 2018-05-08T16:25:09Z

cli/kata-check_ppc64le.go

+
+// kvmIsUsable determines if it will be possible to create a full virtual machine
+// by creating a minimal VM and then deleting it.
+func kvmIsUsable() error {


Duplicated function. Please can you move the version in cli/kata-check_amd64.go into cli/kata-check.go and maybe prefix with generic.

jodh-intel · 2018-05-08T16:26:38Z

virtcontainers/qemu_ppc64le.go

+}
+
+func (q *qemuPpc64le) cpuModel() string {
+        //fmt.Println("HELLLLO")


Could you remove this line please? 😄

jodh-intel · 2018-05-08T16:28:13Z

virtcontainers/qemu_ppc64le.go

+	return cpuModel
+}
+
+func (q *qemuPpc64le) memoryTopology(memoryMb, hostMemoryMb uint64) govmmQemu.Memory {


This is almost identical to the same function name in virtcontainers/qemu_amd64.go so this could be moved to virtcontainers/qemu.go maybe?

jodh-intel · 2018-05-08T16:28:47Z

virtcontainers/qemu_ppc64le.go

+
+	memory := govmmQemu.Memory{
+		Size: mem,
+		//Slots:  defaultMemSlots,


... Almost identical because of these two lines. Please remove the commented out code if these values should not be set.

jodh-intel · 2018-05-08T16:29:24Z

virtcontainers/qemu_ppc64le.go

+		return nil, err
+	}
+
+	//govmmQemu.SCSIController{}


Commented out line.

jodh-intel · 2018-05-08T16:29:49Z

virtcontainers/qemu_ppc64le.go

+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+


Hi @markdryan - any comments on the qemu code here?

jodh-intel · 2018-05-08T16:31:19Z

Thanks for raising @nitkon! There are a few issues, but most are easily resolved I think.

amshinde · 2018-05-08T17:24:39Z

cli/kata-check_ppc64le.go

+// hostIsVMContainerCapable checks to see if the host is theoretically capable
+// of creating a VM container.
+func hostIsVMContainerCapable(details vmContainerCapableDetails) error {
+		return nil


Can you add comments here explaining why this is nil.

amshinde · 2018-05-08T17:28:39Z

virtcontainers/qemu_ppc64le.go

+	var caps capabilities
+
+	// Only pc machine type supports hotplugging drives
+	if q.machineType == QemuPpc64le {


The comment and the implementation are not in sync. Could you change the comment if the pseries machine type supports this capability.

amshinde · 2018-05-08T17:33:31Z

virtcontainers/qemu_ppc64le.go

+		}
+
+		devices = append(devices,
+			govmmQemu.BridgeDevice{


The other functions for appendBridges assign an address to the bridge as well. Please do that here as well.
Also since the code is identical, can you move it to a common function prefixed with "generic" as @jodh-intel has suggested for other cases.

Also, can you add some unit tests for this architecture?

devimc · 2018-05-15T12:42:23Z

cli/kata-check_ppc64le.go

+// hostIsVMContainerCapable checks to see if the host is theoretically capable
+// of creating a VM container.
+func hostIsVMContainerCapable(details vmContainerCapableDetails) error {
+		return nil


devimc · 2018-05-15T12:45:02Z

virtcontainers/hypervisor.go

@@ -454,7 +454,12 @@ func getHostMemorySizeKb(memInfoPath string) (uint64, error) {
 // RunningOnVMM checks if the system is running inside a VM.
 func RunningOnVMM(cpuInfoPath string) (bool, error) {
 	if runtime.GOARCH == "arm64" {


|| runtime.GOARCH == "ppc64le"

devimc · 2018-05-15T12:45:45Z

virtcontainers/qemu_arch_base.go

@@ -127,6 +127,9 @@ const (

 	// QemuVirt is the QEMU virt machine type for aarch64
 	QemuVirt = "virt"
+
+	// QemuPPC64le is a QEMU virt machine type for for ppc64le
+	QemuPPC64le = "pseries"


variable name should be QemuPseries

devimc · 2018-05-15T12:47:26Z

virtcontainers/qemu_ppc64le.go

+	qemuArchBase
+}
+
+const defaultQemuPath = "/usr/libexec/qemu-kvm"


do you mean /usr/bin/qemu-kvm ?

devimc · 2018-05-15T12:50:41Z

virtcontainers/qemu_ppc64le.go

+	{"iommu", "off"},
+	{"cryptomgr.notests", ""},
+	{"net.ifnames", "0"},
+	{"pci", "lastbus=0"},


this improves boot time of PCIe machines types, does QEMU ppc have PCIe machine types?

@nitkon you didn't answer here.

@devimc : Thanks for pointing out. pci=lastbus=0 is an intel only feature. POWER does not use it. So we can remove this option for ppc64le.
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/admin-guide/kernel-parameters.txt#n3058
{"iommu", "off"} can also be removed.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/admin-guide/kernel-parameters.txt#n1699
{"cryptomgr.notests", ""}, and {"net.ifnames", "0"}, should improve the boot times.

Thanks for the parameters analysis, this was a very complete answer 👍

@jodh-intel : Is it fine to raise a PR to optimize the kernel parameters for ppc64le.

Hi @nitkon - Sure - please do! ;)

devimc · 2018-05-15T12:54:17Z

virtcontainers/qemu_ppc64le.go

+	// NVDIMM device needs memory space 1024MB
+	memoryOffset := 1024
+
+        //  Align hostMemoryMb to 256 MB multiples


uhhmm this is weird, why 256?, have you tried with 128 MB?

Default LMB (Logical Memory Blocksize) in ppc64le arch is 256MB. If we do not adhere to that we get the error: "docker: Error response from daemon: OCI runtime create failed: qemu-system-ppc64: Can't support memory configuration where RAM size 0x80000000 or maxmem size 0x3017400000 isn't aligned to 256 MB: unknown."

make sense, thanks @nitkon

devimc · 2018-05-15T12:57:28Z

virtcontainers/qemu_ppc64le.go

@@ -0,0 +1,192 @@
+// Copyright (c) 2018 IBM


you need to run gofmt -s -w on this file

devimc · 2018-05-15T12:59:06Z

virtcontainers/qemu_ppc64le.go

+        for idx, b := range bridges {
+                t := govmmQemu.PCIBridge
+                if b.Type == pcieBridge {
+                        t = govmmQemu.PCIEBridge


does pseries support pcie ?

@devimc : Yes .

ok, have you compared the boot time with/without pci=lastbus=0 in the kernel command line? probably you will need it

bergwolf · 2018-05-15T14:35:21Z

cli/kata-check.go

+
+        fieldLogger.WithField("device", kvmDevice).Info("device available")
+
+        vm, _, errno := syscall.Syscall(syscall.SYS_IOCTL,


Not related to this patch, but is there any possibility that we have /dev/kvm device but kvm is not enabled? I wonder if the extra step of checking kvm ioctl is necessary. If not, we can remove the only piece of C go from kata runtime.

Actually, yes (kind of). This code was added to protect against the scenario where the user has a different hypervisor running (VMWare, VirtualBox, etc) since they stop qemu from using kvm. See:

Add further checks to "cc-check" command clearcontainers/runtime#848

That's also why kata-check is specified twice in the README, one with sudo and one without since kata-check needs root access to create a KVM VM (which we then destroy), but you can run the command without root (but we can't then detect if another hypervisor is using it):

https://github.com/kata-containers/runtime#hardware-requirements

OK, fair enough. I guess next possibility is to either push the kvm ioctl number to go runtime, or define it in govmm. Then we can still remove the c go code.

bergwolf · 2018-05-15T14:39:57Z

virtcontainers/qemu_ppc64le.go

+	QemuPPC64le: defaultQemuPath,
+}
+
+var kernelParams = []Param{


A lot of these duplicate with kernelParams in qemu_amd64.go, which means we should define a kernelParamsBase in qemu_arch_base.go.

Yes I agree we should define a base of common params being shared across different architectures.

jodh-intel

Thanks for updating @nitkon!

Aside from the inline comments, I think it would be good to create atleast minimal qemu_ppc64el_test.go and qemu_arm64_test.go files.

jodh-intel · 2018-05-16T09:07:25Z

Makefile

@@ -345,6 +352,7 @@ $(GENERATED_FILES): %: %.in Makefile VERSION
 		-e "s|@QEMUPATH@|$(QEMUPATH)|g" \
 		-e "s|@RUNTIME_NAME@|$(TARGET)|g" \
 		-e "s|@MACHINETYPE@|$(MACHINETYPE)|g" \
+		-e "s|@KERNELTYPE@|$(KERNELTYPE)|g" \


I think this line is redundant?

jodh-intel · 2018-05-16T09:08:01Z

arch/ppc64le-options.mk

+# SPDX-License-Identifier: Apache-2.0
+#
+
+# Power Ppc64le settings


The capitalisation looks a bit odd here - should it be "PPC64le"?

jodh-intel · 2018-05-16T10:55:34Z

cli/kata-check_arm64.go

+
+// variables rather than consts to allow tests to modify them
+var (
+	kvmDevice = "/dev/kvm"


This should be defined in kata-check.go only.

jodh-intel · 2018-05-16T10:57:06Z

cli/kata-check_arm64.go

+
+// hostIsVMContainerCapable checks to see if the host is theoretically capable
+// of creating a VM container.
+


Nit: remove blank line.

jodh-intel · 2018-05-16T10:57:28Z

cli/kata-check_amd64.go

-		return false
-	}
+// hostIsVMContainerCapable checks to see if the host is theoretically capable
+// of creating a VM container.


Nit: remove blank line.

jodh-intel · 2018-05-16T11:04:14Z

virtcontainers/qemu_ppc64le.go

+	QemuPseries: defaultQemuPath,
+}
+
+var kernelParams = []Param{


Are all these options really used by ppc64el? If so, I'd suggest moving this to qemu.go so they can be shared by amd64+ppc64el.

jodh-intel · 2018-05-16T11:07:52Z

virtcontainers/qemu_ppc64le.go

+			kernelParamsDebug:     kernelParamsDebug,
+			kernelParams:          kernelParams,
+		},
+	}


This function is missing the block the other arches have:

if config.ImagePath != "" { ... }

In fact, since that block of code should I think be common to all architectures, you could make a new function in qemu.go so that all the newQemuArch() functions can call something like:

return handleImagePath(config HypervisorConfig, q qemuArch)

jodh-intel · 2018-05-16T11:11:31Z

virtcontainers/qemu_ppc64le.go

+	return caps
+}
+
+func (q *qemuPPC64le) bridges(number uint32) []Bridge {


There is quite a bit of commonality in the following functions between amd64+ppc64el:

bridges()

memoryTopology()

appendBridges()

I'd say it's "borderline refactoring territory" but wdyt @sboeuf?

ping @sboeuf.

@nitkon yes you could factorize more for those functions here. Correct me if I'm wrong but the code for ppc64 is a copy of the implementation for amd64 (for the 3 functions @jodh-intel mentioned). bridges() is slightly different because of the type itself, but the semantic is the same.

This looks like the only outstanding issue stopping this from landing.

@jodh-intel : Thanks for pointing it out. Will update my PR accordingly.

@jodh-intel : @sboeuf :
The code for memoryTopology() is not exactly the same for ppc64 and x86. The addition being in ppc64le

hostMemoryMb -= (hostMemoryMb % 256)

The code for appendBridges() has already been refactored.

// appendBridges appends to devices the given bridges func (q *qemuPPC64le) appendBridges(devices []govmmQemu.Device, bridges []Bridge) []govmmQemu.Device { return genericAppendBridges(devices, bridges, q.machineType) }

bridges() can be refactored furthure.

Hi @nitkon - yes. I think all we really need now is something like the following (untested! :) that the arch-specific qemu functions can call:

func genericMakeBridges(number uint32, bt bridgeType) []Bridge { var bridges []Bridge for i := uint32(0); i < number; i++ { bridges = append(bridges, Bridge{ Type: bt, ID: fmt.Sprintf("%s-bridge-%d", bt, i), Address: make(map[uint32]string), }) } return bridges } func genericMemoryTopology(memoryMb, hostMemoryMb uint64, alignMB uint32) govmmQemu.Memory { // NVDIMM device needs memory space 1024MB memoryOffset := 1024 if alignMB > 0 { hostMemoryMb -= (hostMemoryMb % alignMB) } // add 1G memory space for nvdimm device (vm guest image) memMax := fmt.Sprintf("%dM", hostMemoryMb+uint64(memoryOffset)) mem := fmt.Sprintf("%dM", memoryMb) memory := govmmQemu.Memory{ Size: mem, Slots: defaultMemSlots, MaxMem: memMax, } return memory }

@jodh-intel : Thnx for review comment. I saw this post updating my PR. If you find my changes acceptable then we can keep them else I shall revert to the way u have suggested it.

jodh-intel · 2018-05-18T11:50:39Z

Hi @nitkon - please could you add a note when you've updated the branch again so we can try to get this landed as soon as possible?

codecov · 2018-05-19T13:52:48Z

Codecov Report

Merging #286 into master will increase coverage by 0.02%.
The diff coverage is 89.69%.

@@            Coverage Diff             @@
##           master     #286      +/-   ##
==========================================
+ Coverage   63.74%   63.77%   +0.02%     
==========================================
  Files          87       87              
  Lines        8731     8749      +18     
==========================================
+ Hits         5566     5580      +14     
- Misses       2569     2572       +3     
- Partials      596      597       +1

Impacted Files	Coverage Δ
cli/kata-env.go	`98.57% <0%> (-1.43%)`	⬇️
virtcontainers/hypervisor.go	`72.78% <0%> (ø)`	⬆️
cli/kata-check_amd64.go	`100% <100%> (+6.25%)`	⬆️
virtcontainers/qemu_amd64.go	`92% <100%> (-1.55%)`	⬇️
virtcontainers/qemu_arch_base.go	`77.52% <100%> (+0.52%)`	⬆️
virtcontainers/qemu.go	`19.34% <91.66%> (+7.32%)`	⬆️
cli/kata-check.go	`97.24% <93.54%> (-1.02%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b3b0612...e14eab0. Read the comment docs.

nitkon · 2018-05-19T13:58:15Z

Hi @jodh-intel : I have updated the PR.

devimc · 2018-05-22T12:37:43Z

@jodh-intel please take a look
@nitkon please rebase it

katabuilder · 2018-05-22T18:28:54Z

PSS Measurement:
Qemu: 141956 KB
Proxy: 4565 KB
Shim: 10871 KB

Memory inside container:
Total Memory: 2045972 KB
Free Memory: 2012952 KB

nitkon · 2018-05-23T05:31:08Z

Rebased my PR.

jodh-intel

This is looking good. My only concerns are:

the duplication in the qemu_${arch}.go.
how this changes affects ARM64.

/cc @sboeuf, @bergwolf.

jodh-intel · 2018-05-21T08:14:31Z

virtcontainers/qemu_ppc64le.go

+
+const defaultQemuMachineType = QemuPseries
+
+const defaultQemuMachineOptions = "accel=kvm,usb=off"


USB is also specified by arm64. I wonder if we should also disable for x86_64 @devimc, @mcastelino, @grahamwhaley?

@jodh-intel I don't see that option for x86_64
https://github.com/qemu/qemu/blob/master/include/hw/i386/pc.h

Yep - looks like -usb is required to explicitly enable on x86_64. But we disable it anyway 😄:

https://github.com/kata-containers/packaging/blob/master/scripts/configure-hypervisor.sh#L293..L295

jodh-intel · 2018-05-21T08:15:33Z

virtcontainers/qemu_ppc64le.go

+	return caps
+}
+
+func (q *qemuPPC64le) bridges(number uint32) []Bridge {


ping @sboeuf.

jodh-intel · 2018-05-23T16:06:20Z

/cc @WeiZhang555.

jodh-intel · 2018-05-24T13:58:19Z

Hi @nitkon - please can you try running the unit tests for this PR on a PPC64el system:

$ cd $GOPATH/src/github.com/kata-containers/runtime
$ go get -d github.com/kata-containers/tests
$ script -fec 'KATA_DEV_MODE=true ../tests/.ci/go-test.sh'

I think you'll find that you need to atleast create a cli/kata-check_data_ppc64le_test.go.

jodh-intel · 2018-05-24T17:00:47Z

@nitkon - once you've got the tests passing, and we've agreed on whether to do that bit of refactor (let's wait for @sboeuf's view), I'm happy to ack this btw ;)

nitkon · 2018-05-25T19:01:03Z

@jodh-intel : Thanks for the review. I have updated my PR.

=== RUN   TestWriteFileErrNoPath
--- PASS: TestWriteFileErrNoPath (0.00s)
=== RUN   TestVersion
--- PASS: TestVersion (0.00s)
PASS
coverage: 85.3% of statements
ok  	github.com/kata-containers/runtime/cli	3.460s	coverage: 85.3% of statements_

katabuilder · 2018-05-25T19:34:43Z

PSS Measurement:
Qemu: 141846 KB
Proxy: 4605 KB
Shim: 10847 KB

Memory inside container:
Total Memory: 2045972 KB
Free Memory: 2011092 KB

sboeuf

@nitkon nice PR, but I have a request about the format. Could you split this PR into 3 distincts commits so that it is easier to review and easier to understand all changes that have been done.
First commit being the factorization of some functions for the purpose of being reused later by ppc64 arch.
Second commit introducing the qemu ppc64 implementation in virtcontainers, and relying on the new factorized functions.
Last commit being the CLI part leveraging the new support for ppc64.

nitkon · 2018-05-28T16:50:18Z

@sboeuf : I have updated my PR as suggested.

katabuilder · 2018-05-28T17:41:53Z

PSS Measurement:
Qemu: 145972 KB
Proxy: 6779 KB
Shim: 10691 KB

Memory inside container:
Total Memory: 2045972 KB
Free Memory: 2013296 KB

jodh-intel · 2018-05-29T15:44:22Z

Thanks for fixing the tests @nitkon.

@bergwolf, @sboeuf - ptal.

devimc · 2018-05-29T19:24:04Z

nice patch @nitkon !! lgtm 😄

sboeuf

@nitkon this looks better ! Still a few comments/questions not addressed I think.
Also, would be good to factorize the implementation duplication between amd64 and ppc64.

jodh-intel · 2018-05-30T08:47:59Z

Hi @nitkon - there is still the issue of how we ensure the PPC64el build doesn't break after this PR lands.

Can you help the Kata Project in any way with PPC64el CI resources? We recently noticed that the ARM64 build broke as we don't yet have ARM64 CI support (but it's in progress - see #349). We don't want a repeat of that and as such are very keen to ensure that as new architectures are added, our CI can test them all.

jodh-intel · 2018-05-30T08:48:28Z

/cc @egernst, @grahamwhaley, @chavafg, @gnawux.

nitkon · 2018-05-30T15:21:59Z

@jodh-intel : We have ppc64le support in travis-ci.org. So we need to enhance the runtime/.travis.yml file to run on ppc64le as well.

jodh-intel · 2018-05-30T15:28:05Z

Hi @nitkon - great to hear! However, please not that we're actually trying to move away from Travis since, atleast for the public service that offers x86_64 services, it doesn't provide the virtualisation extensions we need to run the full suite of tests. See:

amshinde · 2018-05-30T18:50:08Z

lgtm!

Fixes #302 Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

katabuilder · 2018-05-31T17:51:19Z

PSS Measurement:
Qemu: 161699 KB
Proxy: 6784 KB
Shim: 8818 KB

Memory inside container:
Total Memory: 2045972 KB
Free Memory: 1999484 KB

sboeuf · 2018-05-31T19:00:47Z

@nitkon thanks for your work reworking this PR, this LGTM now.
@jodh-intel please check if you're fine with latest changes and merge if you've nothing to add :)

jodh-intel · 2018-06-01T08:58:33Z

still...

lgtm

Thanks @nitkon for all your work, persistence, and patience! 😄

Please feel free to raise a PR for the Travis changes you mentioned to ensure this new architecture continues to pass all the CI tests.

…d-ns namespace: Add check to make sure PID namespace is not received

devimc reviewed May 7, 2018

View reviewed changes

cli/kata-check_ppc64le.go Outdated

@@ -0,0 +1,110 @@

// Copyright (c) 2018 Intel Corporation

Copy link

devimc May 7, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

s/Intel Corporation/IBM/ ?

devimc reviewed May 7, 2018

View reviewed changes

virtcontainers/qemu_ppc64le.go Outdated

@@ -0,0 +1,199 @@

//

// Copyright (c) 2018 Intel Corporation

Copy link

devimc May 7, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

s/Intel Corporation/IBM/ ?

jodh-intel suggested changes May 8, 2018

View reviewed changes

amshinde reviewed May 8, 2018

View reviewed changes

devimc reviewed May 15, 2018

View reviewed changes

virtcontainers/qemu_ppc64le.go Outdated

qemuArchBase

}

const defaultQemuPath = "/usr/libexec/qemu-kvm"

Copy link

devimc May 15, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do you mean /usr/bin/qemu-kvm ?

devimc reviewed May 15, 2018

View reviewed changes

virtcontainers/qemu_ppc64le.go

@@ -0,0 +1,192 @@

// Copyright (c) 2018 IBM

Copy link

devimc May 15, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

you need to run gofmt -s -w on this file

devimc reviewed May 15, 2018

View reviewed changes

bergwolf reviewed May 15, 2018

View reviewed changes

jodh-intel reviewed May 16, 2018

View reviewed changes

jodh-intel reviewed May 23, 2018

View reviewed changes

sboeuf suggested changes May 25, 2018

View reviewed changes

jodh-intel approved these changes May 29, 2018

View reviewed changes

sboeuf reviewed May 29, 2018

View reviewed changes

jodh-intel mentioned this pull request May 30, 2018

Kata-env and Kata-check are not architecture-independent #349

Closed

jodh-intel mentioned this pull request May 30, 2018

ARM: get the ARM CIs up kata-containers/ci#30

Closed

nitkon added 4 commits May 31, 2018 17:58

virtcontainers/cli: refactor code

4276c0c

Fixes #302 Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

virtcontainers: Get qemu suppport for ppc64le

baa553d

Fixes #302 Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

cli: Leverage the new support for ppc64le

12e4dbe

Fixes #302 Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

runtime: Add testcases for ppc64le and arm64

e14eab0

Fixes #302 Signed-off-by: Nitesh Konkar niteshkonkar@in.ibm.com

sboeuf approved these changes May 31, 2018

View reviewed changes

jodh-intel merged commit 2400978 into kata-containers:master Jun 1, 2018

zklei pushed a commit to zklei/runtime that referenced this pull request Jun 13, 2019

Merge pull request kata-containers#286 from amshinde/add-check-for-pi…

bc973da

…d-ns namespace: Add check to make sure PID namespace is not received


		fieldLogger.WithField("device", kvmDevice).Info("device available")

		vm, _, errno := syscall.Syscall(syscall.SYS_IOCTL,


		// hostIsVMContainerCapable checks to see if the host is theoretically capable
		// of creating a VM container.


		const defaultQemuMachineType = QemuPseries

		const defaultQemuMachineOptions = "accel=kvm,usb=off"

Enable Kata container on ppc64le arch #286

Enable Kata container on ppc64le arch #286

Conversation

nitkon commented May 7, 2018 • edited Loading

nitkon commented May 7, 2018

devimc commented May 7, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jodh-intel commented May 8, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jodh-intel left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jodh-intel commented May 18, 2018

codecov bot commented May 19, 2018 • edited Loading

Codecov Report

nitkon commented May 19, 2018

devimc commented May 22, 2018

katabuilder commented May 22, 2018

nitkon commented May 23, 2018

jodh-intel left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jodh-intel commented May 23, 2018

jodh-intel commented May 24, 2018

nitkon commented May 7, 2018 •

edited

Loading

codecov bot commented May 19, 2018 •

edited

Loading