podman cgroup enhancement

currently, setting any sort of resource limit in a pod does nothing. With the newly refactored creation process in c/common, podman ca now set resources at a pod level
meaning that resource related flags can now be exposed to podman pod create.

cgroupfs and systemd are both supported with varying completion. cgroupfs is a much simpler process and one that is virtually complete for all resource types, the flags now just need to be added. systemd on the other hand
has to be handeled via the dbus api meaning that the limits need to be passed as recognized properties to systemd. The properties added so far are the ones that podman pod create supports as well as `cpuset-mems` as this will
be the next flag I work on.

Signed-off-by: Charlie Doern <cdoern@redhat.com>

3 files changed, 26 insertions, 33 deletions

diff --git a/pkg/api/handlers/compat/containers_stats.go b/pkg/api/handlers/compat/containers_stats.go
index d6bc26416..12c5283fc 100644
--- a/pkg/api/handlers/compat/containers_stats.go
+++ b/pkg/api/handlers/compat/containers_stats.go
@@ -12,6 +12,7 @@ import (
 	api "github.com/containers/podman/v4/pkg/api/types"
 	docker "github.com/docker/docker/api/types"
 	"github.com/gorilla/schema"
+	runccgroups "github.com/opencontainers/runc/libcontainer/cgroups"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 )
@@ -133,7 +134,7 @@ streamLabel: // A label to flatten the scope
 		}
 
 		cfg := ctnr.Config()
-		memoryLimit := cgroupStat.Memory.Usage.Limit
+		memoryLimit := cgroupStat.MemoryStats.Usage.Limit
 		if cfg.Spec.Linux != nil && cfg.Spec.Linux.Resources != nil && cfg.Spec.Linux.Resources.Memory != nil && *cfg.Spec.Linux.Resources.Memory.Limit > 0 {
 			memoryLimit = uint64(*cfg.Spec.Linux.Resources.Memory.Limit)
 		}
@@ -144,11 +145,11 @@ streamLabel: // A label to flatten the scope
 				Read:    time.Now(),
 				PreRead: preRead,
 				PidsStats: docker.PidsStats{
-					Current: cgroupStat.Pids.Current,
+					Current: cgroupStat.PidsStats.Current,
 					Limit:   0,
 				},
 				BlkioStats: docker.BlkioStats{
-					IoServiceBytesRecursive: toBlkioStatEntry(cgroupStat.Blkio.IoServiceBytesRecursive),
+					IoServiceBytesRecursive: toBlkioStatEntry(cgroupStat.BlkioStats.IoServiceBytesRecursive),
 					IoServicedRecursive:     nil,
 					IoQueuedRecursive:       nil,
 					IoServiceTimeRecursive:  nil,
@@ -159,14 +160,14 @@ streamLabel: // A label to flatten the scope
 				},
 				CPUStats: CPUStats{
 					CPUUsage: docker.CPUUsage{
-						TotalUsage:        cgroupStat.CPU.Usage.Total,
-						PercpuUsage:       cgroupStat.CPU.Usage.PerCPU,
-						UsageInKernelmode: cgroupStat.CPU.Usage.Kernel,
-						UsageInUsermode:   cgroupStat.CPU.Usage.Total - cgroupStat.CPU.Usage.Kernel,
+						TotalUsage:        cgroupStat.CpuStats.CpuUsage.TotalUsage,
+						PercpuUsage:       cgroupStat.CpuStats.CpuUsage.PercpuUsage,
+						UsageInKernelmode: cgroupStat.CpuStats.CpuUsage.UsageInKernelmode,
+						UsageInUsermode:   cgroupStat.CpuStats.CpuUsage.TotalUsage - cgroupStat.CpuStats.CpuUsage.UsageInKernelmode,
 					},
 					CPU:         stats.CPU,
 					SystemUsage: systemUsage,
-					OnlineCPUs:  uint32(len(cgroupStat.CPU.Usage.PerCPU)),
+					OnlineCPUs:  uint32(len(cgroupStat.CpuStats.CpuUsage.PercpuUsage)),
 					ThrottlingData: docker.ThrottlingData{
 						Periods:          0,
 						ThrottledPeriods: 0,
@@ -175,8 +176,8 @@ streamLabel: // A label to flatten the scope
 				},
 				PreCPUStats: preCPUStats,
 				MemoryStats: docker.MemoryStats{
-					Usage:             cgroupStat.Memory.Usage.Usage,
-					MaxUsage:          cgroupStat.Memory.Usage.Limit,
+					Usage:             cgroupStat.MemoryStats.Usage.Usage,
+					MaxUsage:          cgroupStat.MemoryStats.Usage.Limit,
 					Stats:             nil,
 					Failcnt:           0,
 					Limit:             memoryLimit,
@@ -216,7 +217,7 @@ streamLabel: // A label to flatten the scope
 	}
 }
 
-func toBlkioStatEntry(entries []cgroups.BlkIOEntry) []docker.BlkioStatEntry {
+func toBlkioStatEntry(entries []runccgroups.BlkioStatEntry) []docker.BlkioStatEntry {
 	results := make([]docker.BlkioStatEntry, len(entries))
 	for i, e := range entries {
 		bits, err := json.Marshal(e)
diff --git a/pkg/specgen/generate/container_create.go b/pkg/specgen/generate/container_create.go
index 0dec943d1..6b2e90b22 100644
--- a/pkg/specgen/generate/container_create.go
+++ b/pkg/specgen/generate/container_create.go
@@ -180,10 +180,23 @@ func MakeContainer(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGener
 			if err != nil {
 				return nil, nil, nil, err
 			}
+			resources := runtimeSpec.Linux.Resources
+
+			// resources get overwrritten similarly to pod inheritance, manually assign here if there is a new value
+			marshalRes, err := json.Marshal(resources)
+			if err != nil {
+				return nil, nil, nil, err
+			}
+
 			err = json.Unmarshal(out, runtimeSpec.Linux)
 			if err != nil {
 				return nil, nil, nil, err
 			}
+
+			err = json.Unmarshal(marshalRes, runtimeSpec.Linux.Resources)
+			if err != nil {
+				return nil, nil, nil, err
+			}
 		}
 		if s.ResourceLimits != nil {
 			switch {
diff --git a/pkg/specgen/generate/oci.go b/pkg/specgen/generate/oci.go
index 19f55c9d8..1044854f4 100644
--- a/pkg/specgen/generate/oci.go
+++ b/pkg/specgen/generate/oci.go
@@ -298,8 +298,7 @@ func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runt
 		g.AddAnnotation(key, val)
 	}
 
-	switch {
-	case compatibleOptions.InfraResources == nil && s.ResourceLimits != nil:
+	if s.ResourceLimits != nil {
 		out, err := json.Marshal(s.ResourceLimits)
 		if err != nil {
 			return nil, err
@@ -308,29 +307,9 @@ func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runt
 		if err != nil {
 			return nil, err
 		}
-	case s.ResourceLimits != nil: // if we have predefined resource limits we need to make sure we keep the infra and container limits
-		originalResources, err := json.Marshal(s.ResourceLimits)
-		if err != nil {
-			return nil, err
-		}
-		infraResources, err := json.Marshal(compatibleOptions.InfraResources)
-		if err != nil {
-			return nil, err
-		}
-		err = json.Unmarshal(infraResources, s.ResourceLimits) // put infra's resource limits in the container
-		if err != nil {
-			return nil, err
-		}
-		err = json.Unmarshal(originalResources, s.ResourceLimits) // make sure we did not override anything
-		if err != nil {
-			return nil, err
-		}
 		g.Config.Linux.Resources = s.ResourceLimits
-	default:
-		g.Config.Linux.Resources = compatibleOptions.InfraResources
 	}
 	// Devices
-
 	// set the default rule at the beginning of device configuration
 	if !inUserNS && !s.Privileged {
 		g.AddLinuxResourcesDevice(false, "", nil, nil, "rwm")