new driver: kubernetes

Tested with `kind` and GKE.

Note: "nodes" shown in `docker buildx ls` are unrelated to Kubernetes "nodes".
Probably buildx should come up with an alternative term.

Usage:

  $ kind create cluster
  $ export KUBECONFIG="$(kind get kubeconfig-path --name="kind")"

  $ docker buildx create --driver kubernetes --driver-opt replicas=3 --use
  $ docker buildx build -t foo --load .

`--load` loads the image into the local Docker.

Driver opts:

  - `image=IMAGE` - Sets the container image to be used for running buildkit.
  - `namespace=NS` - Sets the Kubernetes namespace. Defaults to the current namespace.
  - `replicas=N` - Sets the number of `Pod` replicas. Defaults to 1.
  - `rootless=(true|false)` - Run the container as a non-root user without `securityContext.privileged`. Defaults to false.
  - `loadbalance=(sticky|random)` - Load-balancing strategy. If set to "sticky", the pod is chosen using the hash of the context path. Defaults to "sticky"

Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>

vendor/gopkg.in/yaml.v2/decode.go

@@ -229,6 +229,10 @@ type decoder struct {
 	mapType reflect.Type
 	terrors []string
 	strict  bool
+
+	decodeCount int
+	aliasCount  int
+	aliasDepth  int
 }
 
 var (
@@ -314,7 +318,39 @@ func (d *decoder) prepare(n *node, out reflect.Value) (newout reflect.Value, unm
 	return out, false, false
 }
 
+const (
+	// 400,000 decode operations is ~500kb of dense object declarations, or ~5kb of dense object declarations with 10000% alias expansion
+	alias_ratio_range_low = 400000
+	// 4,000,000 decode operations is ~5MB of dense object declarations, or ~4.5MB of dense object declarations with 10% alias expansion
+	alias_ratio_range_high = 4000000
+	// alias_ratio_range is the range over which we scale allowed alias ratios
+	alias_ratio_range = float64(alias_ratio_range_high - alias_ratio_range_low)
+)
+
+func allowedAliasRatio(decodeCount int) float64 {
+	switch {
+	case decodeCount <= alias_ratio_range_low:
+		// allow 99% to come from alias expansion for small-to-medium documents
+		return 0.99
+	case decodeCount >= alias_ratio_range_high:
+		// allow 10% to come from alias expansion for very large documents
+		return 0.10
+	default:
+		// scale smoothly from 99% down to 10% over the range.
+		// this maps to 396,000 - 400,000 allowed alias-driven decodes over the range.
+		// 400,000 decode operations is ~100MB of allocations in worst-case scenarios (single-item maps).
+		return 0.99 - 0.89*(float64(decodeCount-alias_ratio_range_low)/alias_ratio_range)
+	}
+}
+
 func (d *decoder) unmarshal(n *node, out reflect.Value) (good bool) {
+	d.decodeCount++
+	if d.aliasDepth > 0 {
+		d.aliasCount++
+	}
+	if d.aliasCount > 100 && d.decodeCount > 1000 && float64(d.aliasCount)/float64(d.decodeCount) > allowedAliasRatio(d.decodeCount) {
+		failf("document contains excessive aliasing")
+	}
 	switch n.kind {
 	case documentNode:
 		return d.document(n, out)
@@ -353,7 +389,9 @@ func (d *decoder) alias(n *node, out reflect.Value) (good bool) {
 		failf("anchor '%s' value contains itself", n.value)
 	}
 	d.aliases[n] = true
+	d.aliasDepth++
 	good = d.unmarshal(n.alias, out)
+	d.aliasDepth--
 	delete(d.aliases, n)
 	return good
 }