Merge pull request #8408 from umohnani8/sec-opt

Add mask and unmask option to --security-opt

7 files changed, 114 insertions, 22 deletions

diff --git a/cmd/podman/common/specgen.go b/cmd/podman/common/specgen.go
index 0bb6e79e5..e0da142ad 100644
--- a/cmd/podman/common/specgen.go
+++ b/cmd/podman/common/specgen.go
@@ -517,18 +517,22 @@ func FillOutSpecGen(s *specgen.SpecGenerator, c *ContainerCLIOpts, args []string
 			}
 
 			switch con[0] {
-			case "proc-opts":
-				s.ProcOpts = strings.Split(con[1], ",")
+			case "apparmor":
+				s.ContainerSecurityConfig.ApparmorProfile = con[1]
+				s.Annotations[define.InspectAnnotationApparmor] = con[1]
 			case "label":
 				// TODO selinux opts and label opts are the same thing
 				s.ContainerSecurityConfig.SelinuxOpts = append(s.ContainerSecurityConfig.SelinuxOpts, con[1])
 				s.Annotations[define.InspectAnnotationLabel] = strings.Join(s.ContainerSecurityConfig.SelinuxOpts, ",label=")
-			case "apparmor":
-				s.ContainerSecurityConfig.ApparmorProfile = con[1]
-				s.Annotations[define.InspectAnnotationApparmor] = con[1]
+			case "mask":
+				s.ContainerSecurityConfig.Mask = append(s.ContainerSecurityConfig.Mask, strings.Split(con[1], ":")...)
+			case "proc-opts":
+				s.ProcOpts = strings.Split(con[1], ",")
 			case "seccomp":
 				s.SeccompProfilePath = con[1]
 				s.Annotations[define.InspectAnnotationSeccomp] = con[1]
+			case "unmask":
+				s.ContainerSecurityConfig.Unmask = append(s.ContainerSecurityConfig.Unmask, strings.Split(con[1], ":")...)
 			default:
 				return fmt.Errorf("invalid --security-opt 2: %q", opt)
 			}
diff --git a/docs/source/markdown/podman-run.1.md b/docs/source/markdown/podman-run.1.md
index 83aaa33e8..1038906c0 100644
--- a/docs/source/markdown/podman-run.1.md
+++ b/docs/source/markdown/podman-run.1.md
@@ -885,11 +885,16 @@ Security Options
 - **label=level:**_LEVEL_: Set the label level for the container processes
 - **label=filetype:**TYPE_: Set the label file type for the container files
 - **label=disable**: Turn off label separation for the container
+- **mask**=_/path/1:/path/2_: The paths to mask separated by a colon. A masked path
+  cannot be accessed inside the container.
 - **no-new-privileges**: Disable container processes from gaining additional privileges
 - **seccomp=unconfined**: Turn off seccomp confinement for the container
 - **seccomp**=_profile.json_: Allowed syscall list seccomp JSON file to be used as a seccomp filter
 - **proc-opts**=_OPTIONS_ : Comma separated list of options to use for the /proc mount. More details
   for the possible mount options are specified at **proc(5)** man page.
+- **unmask**=_ALL_ or _/path/1:/path/2_: Paths to unmask separated by a colon. If set to **ALL**, it will
+  unmask all the paths that are masked by default.
+  The default masked paths are **/proc/acpi, /proc/kcore, /proc/keys, /proc/latency_stats, /proc/sched_debug, /proc/scsi, /proc/timer_list, /proc/timer_stats, /sys/firmware, and /sys/fs/selinux.**
 
 Note: Labeling can be disabled for all containers by setting **label=false** in the **containers.conf**(5) file.
 
@@ -1479,6 +1484,26 @@ $ podman run --security-opt label=type:svirt_apache_t -i -t centos bash
 
 Note you would have to write policy defining a **svirt_apache_t** type.
 
+To mask additional specific paths in the container, specify the paths
+separated by a colon using the **mask** option with the **--security-opt**
+flag.
+
+```
+$ podman run --security-opt mask=/foo/bar:/second/path fedora bash
+```
+
+To unmask all the paths that are masked by default, set the **unmask** option to
+**ALL**. Or to only unmask specific paths, specify the paths as shown above with
+the **mask** option.
+
+```
+$ podman run --security-opt unmask=ALL fedora bash
+```
+
+```
+$ podman run --security-opt unmask=/foo/bar:/sys/firmware fedora bash
+```
+
 ### Setting device weight
 
 If you want to set _/dev/sda_ device weight to **200**, you can specify the device
diff --git a/pkg/specgen/generate/config_linux.go b/pkg/specgen/generate/config_linux.go
index 2d40dba8f..1808f99b8 100644
--- a/pkg/specgen/generate/config_linux.go
+++ b/pkg/specgen/generate/config_linux.go
@@ -4,13 +4,16 @@ import (
 	"fmt"
 	"io/ioutil"
 	"os"
+	"path"
 	"path/filepath"
 	"strings"
 
 	"github.com/containers/podman/v2/pkg/rootless"
+	"github.com/containers/podman/v2/pkg/util"
 	spec "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/opencontainers/runtime-tools/generate"
 	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
 	"golang.org/x/sys/unix"
 )
 
@@ -137,22 +140,33 @@ func DevicesFromPath(g *generate.Generator, devicePath string) error {
 	return addDevice(g, strings.Join(append([]string{resolvedDevicePath}, devs[1:]...), ":"))
 }
 
-func BlockAccessToKernelFilesystems(privileged, pidModeIsHost bool, g *generate.Generator) {
+func BlockAccessToKernelFilesystems(privileged, pidModeIsHost bool, mask, unmask []string, g *generate.Generator) {
+	defaultMaskPaths := []string{"/proc/acpi",
+		"/proc/kcore",
+		"/proc/keys",
+		"/proc/latency_stats",
+		"/proc/timer_list",
+		"/proc/timer_stats",
+		"/proc/sched_debug",
+		"/proc/scsi",
+		"/sys/firmware",
+		"/sys/fs/selinux",
+		"/sys/dev/block",
+	}
+
+	unmaskAll := false
+	if unmask != nil && unmask[0] == "ALL" {
+		unmaskAll = true
+	}
+
 	if !privileged {
-		for _, mp := range []string{
-			"/proc/acpi",
-			"/proc/kcore",
-			"/proc/keys",
-			"/proc/latency_stats",
-			"/proc/timer_list",
-			"/proc/timer_stats",
-			"/proc/sched_debug",
-			"/proc/scsi",
-			"/sys/firmware",
-			"/sys/fs/selinux",
-			"/sys/dev",
-		} {
-			g.AddLinuxMaskedPaths(mp)
+		if !unmaskAll {
+			for _, mp := range defaultMaskPaths {
+				// check that the path to mask is not in the list of paths to unmask
+				if !util.StringInSlice(mp, unmask) {
+					g.AddLinuxMaskedPaths(mp)
+				}
+			}
 		}
 
 		if pidModeIsHost && rootless.IsRootless() {
@@ -170,6 +184,15 @@ func BlockAccessToKernelFilesystems(privileged, pidModeIsHost bool, g *generate.
 			g.AddLinuxReadonlyPaths(rp)
 		}
 	}
+
+	// mask the paths provided by the user
+	for _, mp := range mask {
+		if !path.IsAbs(mp) && mp != "" {
+			logrus.Errorf("Path %q is not an absolute path, skipping...", mp)
+			continue
+		}
+		g.AddLinuxMaskedPaths(mp)
+	}
 }
 
 // based on getDevices from runc (libcontainer/devices/devices.go)
diff --git a/pkg/specgen/generate/oci.go b/pkg/specgen/generate/oci.go
index 8454458a8..0368ab205 100644
--- a/pkg/specgen/generate/oci.go
+++ b/pkg/specgen/generate/oci.go
@@ -298,7 +298,7 @@ func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runt
 		}
 	}
 
-	BlockAccessToKernelFilesystems(s.Privileged, s.PidNS.IsHost(), &g)
+	BlockAccessToKernelFilesystems(s.Privileged, s.PidNS.IsHost(), s.Mask, s.Unmask, &g)
 
 	for name, val := range s.Env {
 		g.AddProcessEnv(name, val)
diff --git a/pkg/specgen/specgen.go b/pkg/specgen/specgen.go
index fad2406e5..964b89fa4 100644
--- a/pkg/specgen/specgen.go
+++ b/pkg/specgen/specgen.go
@@ -307,6 +307,13 @@ type ContainerSecurityConfig struct {
 	Umask string `json:"umask,omitempty"`
 	// ProcOpts are the options used for the proc mount.
 	ProcOpts []string `json:"procfs_opts,omitempty"`
+	// Mask is the path we want to mask in the container. This masks the paths
+	// given in addition to the default list.
+	// Optional
+	Mask []string `json:"mask,omitempty"`
+	// Unmask is the path we want to unmask in the container. To override
+	// all the default paths that are masked, set unmask=ALL.
+	Unmask []string `json:"unmask,omitempty"`
 }
 
 // ContainerCgroupConfig contains configuration information about a container's
diff --git a/test/e2e/run_test.go b/test/e2e/run_test.go
index 0d65a3e59..efc125d2b 100644
--- a/test/e2e/run_test.go
+++ b/test/e2e/run_test.go
@@ -233,6 +233,39 @@ var _ = Describe("Podman run", func() {
 		return jsonFile
 	}
 
+	It("podman run mask and unmask path test", func() {
+		session := podmanTest.Podman([]string{"run", "-d", "--name=maskCtr1", "--security-opt", "unmask=ALL", "--security-opt", "mask=/proc/acpi", ALPINE, "sleep", "200"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.ExitCode()).To(Equal(0))
+		session = podmanTest.Podman([]string{"exec", "maskCtr1", "ls", "/sys/firmware"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.OutputToString()).To(Not(BeEmpty()))
+		Expect(session.ExitCode()).To(Equal(0))
+		session = podmanTest.Podman([]string{"exec", "maskCtr1", "ls", "/proc/acpi"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.OutputToString()).To(BeEmpty())
+
+		session = podmanTest.Podman([]string{"run", "-d", "--name=maskCtr2", "--security-opt", "unmask=/proc/acpi:/sys/firmware", ALPINE, "sleep", "200"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.ExitCode()).To(Equal(0))
+		session = podmanTest.Podman([]string{"exec", "maskCtr2", "ls", "/sys/firmware"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.OutputToString()).To(Not(BeEmpty()))
+		Expect(session.ExitCode()).To(Equal(0))
+		session = podmanTest.Podman([]string{"exec", "maskCtr2", "ls", "/proc/acpi"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.OutputToString()).To(Not(BeEmpty()))
+		Expect(session.ExitCode()).To(Equal(0))
+
+		session = podmanTest.Podman([]string{"run", "-d", "--name=maskCtr3", "--security-opt", "mask=/sys/power/disk", ALPINE, "sleep", "200"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.ExitCode()).To(Equal(0))
+		session = podmanTest.Podman([]string{"exec", "maskCtr3", "cat", "/sys/power/disk"})
+		session.WaitWithDefaultTimeout()
+		Expect(session.OutputToString()).To(BeEmpty())
+		Expect(session.ExitCode()).To(Equal(0))
+	})
+
 	It("podman run seccomp test", func() {
 		session := podmanTest.Podman([]string{"run", "-it", "--security-opt", strings.Join([]string{"seccomp=", forbidGetCWDSeccompProfile()}, ""), ALPINE, "pwd"})
 		session.WaitWithDefaultTimeout()
diff --git a/test/system/400-unprivileged-access.bats b/test/system/400-unprivileged-access.bats
index 142d7dcd9..20fdd068f 100644
--- a/test/system/400-unprivileged-access.bats
+++ b/test/system/400-unprivileged-access.bats
@@ -118,7 +118,7 @@ EOF
         /proc/scsi
         /sys/firmware
         /sys/fs/selinux
-        /sys/dev
+        /sys/dev/block
     )
 
     # Some of the above may not exist on our host. Find only the ones that do.