schema: Add enumeration to rootfsPropagation values

[zhouhao3]

Signed-off-by: zhouhao zhouhao@cn.fujitsu.com

[zhouhao3]

@wking I think it should be in the config.go file to increase the corresponding value, but I do not know how to name it, is it should be made such a style:
const (
ArchX86 Arch = "SCMP_ARCH_X86"

[wking]

On Thu, Apr 20, 2017 at 10:34:58PM -0700, Zhou Hao wrote: @wking I think it should be in the `config.go` file to increase the corresponding value, but I do not know how to name it, is it should be made such a style: `const (` ` ArchX86 Arch = "SCMP_ARCH_X86"`

I'm not sure what you mean. I'd expect the current [1]: Its value is either slave, private, or shared. should be adjusted to: Runtimes MUST support `slave`, `private`, [whatever other entries are in the enum]. [1]: https://github.com/opencontainers/runtime-spec/blob/cfc95a56767ecc7249bb43ac29bf8e9ee06ad415/config-linux.md#rootfs-mount-propagation

[wking]

You probably want to drop unbindable until #770 lands.

[wking]

#770 has landed, so there's no need to drop unbindable anymore.

[tianon]

I'm +1 on this (as discussed on our maintainer call just now) -- for values like "CPU Quota", the value in our spec is the value as defined in the kernel, so it makes a lot of sense to punt on "valid values" to the kernel docs/implementation.

For values like rootfsPropagation, the value in our spec is a custom string which we've created for the purposes of this spec that maps to some other value (defined by a kernel enum), so in my mind we must enumerate the expected values (and possibly even the kernel mappings unless they're already pretty obvious, which IMO they are in this case).

[tianon]

LGTM

(in other words)

[mrunalp]

LGTM