Fix cpu architecture detection issue on linux/arm

[fangn2]

This PR is to fix cpu architecture detection issues #4109, #21, #1146 when images/containers in ARM arch were built/executed on x86 host. The root cause of the issue is the getCPUVariant function looks for "Cpu architecture" field in /proc/cpuinfo to determine the actual ARM variant. The function will fail if the ARM containers/images ran in a user-mode emulated environment(like qemu) on a x86 host whose /proc/cpuinfo doesn't have "Cpu architecture" field and whose content is shared with ARM containers.
The PR adds a fall back func to get cpu variant from uname machine field through a system call if the previous method failed.

This PR is inspired by the closed PR.
This PR also splits different platforms(Linux and ! Linux) into different go files.

Test method: local unit test(we don't have the test infrastructure(x86 host with ARM qemu) required in CI
Test environment setup:

1. On x86_64 host

$ uname -a
Linux compute.internal 5.10.135-122.509.amzn2.x86_64 SMP Thu Aug 11 22:41:16 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

3. create an arm64 arch container with qemu

$ docker run --rm -v ~/containerd:/home/containerd -it --platform arm64 golang:1.19-buster
root@c2cedcd6f6e8:/go# uname -a
Linux c2cedcd6f6e8 5.10.135-122.509.amzn2.x86_64 #1 SMP Thu Aug 11 22:41:16 UTC 2022 aarch64 GNU/Linux

4. Run unit test for TestCPUVariant prior to the PR change

root@a2670a5d75ca:/home/containerd/platforms# go test -v -run TestCPUVariant
=== RUN   TestCPUVariant
ERRO[0000] failure getting variant                       error="getCPUInfo for pattern: Cpu architecture: not found"
    cpuinfo_test.go:39: could not get valid variant as expected: [v8 v7 v6 v5 v4 v3]
--- FAIL: TestCPUVariant (0.03s)
FAIL
exit status 1
FAIL	github.com/containerd/containerd/platforms	0.130s

5. Run unit test after the PR change

root@a2670a5d75ca:/home/containerd/platforms# go test -v -run TestCPUVariant
=== RUN   TestCPUVariant
    cpuinfo_test.go:34: got valid variant as expected: "v8" = "v8"
--- PASS: TestCPUVariant (0.03s)
PASS
ok  	github.com/containerd/containerd/platforms	0.128s

[k8s-ci-robot]

Hi @fangn2. Thanks for your PR.

I'm waiting for a containerd member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[kzys]

Please follow https://github.com/containerd/project/blob/main/CONTRIBUTING.md#commit-messages regarding commit messages. Some of the information you put in the PR should be in the commit message too.

[henry118]

A more generic question:

Isn't the proc filesystem supposed to provide an interface to kernel data structures? Even under QEMU isn't it expected to expose the cpuinfo of the emulated guest OS?

Just wonder if we are actually trying to work around an issue that should be fixed at some lower level. Or did I misunderstand something?

[kzys]

Looks almost good!

[kzys]

The change looks good. Can you clean up the commit history?

[fangn2]

Done, commit history cleaned up.

[kzys]

Looks good to me. I'm fine merging the two commits. If we revert the first commit, we have to revert the second as well, so they are not logically separated. But not a blocker.

[roshanr95]

Any (rough) idea when this will make it into a release? Seems like it didn't make 1.6.15...

[fangn2]

It's part of the coming 1.7 release.

I am not sure if maintainers are ok to backport this change to 1.6. Open for suggestions.