seccomp: rework seccomp code, to improve compat with some archs

This substantially reworks the seccomp code, to ensure better
compatibility with some architectures, including i386.

So far we relied on libseccomp's internal handling of the multiple
syscall ABIs supported on Linux. This is problematic however, as it does
not define clear semantics if an ABI is not able to support specific
seccomp rules we install.

This rework hence changes a couple of things:

- We no longer use seccomp_rule_add(), but only
  seccomp_rule_add_exact(), and fail the installation of a filter if the
  architecture doesn't support it.

- We no longer rely on adding multiple syscall architectures to a single filter,
  but instead install a separate filter for each syscall architecture
  supported. This way, we can install a strict filter for x86-64, while
  permitting a less strict filter for i386.

- All high-level filter additions are now moved from execute.c to
  seccomp-util.c, so that we can test them independently of the service
  execution logic.

- Tests have been added for all types of our seccomp filters.

- SystemCallFilters= and SystemCallArchitectures= are now implemented in
  independent filters and installation logic, as they semantically are
  very much independent of each other.

Fixes: #4575

1 files changed, 63 insertions, 50 deletions

diff --git a/src/nspawn/nspawn-seccomp.c b/src/nspawn/nspawn-seccomp.c
index 03a397d30c..72ecc51b16 100644
--- a/src/nspawn/nspawn-seccomp.c
+++ b/src/nspawn/nspawn-seccomp.c
@@ -26,20 +26,21 @@
 #include <seccomp.h>
 #endif
 
+#include "alloc-util.h"
 #include "log.h"
-
+#include "nspawn-seccomp.h"
 #ifdef HAVE_SECCOMP
 #include "seccomp-util.h"
 #endif
-
-#include "nspawn-seccomp.h"
+#include "string-util.h"
 
 #ifdef HAVE_SECCOMP
 
-static int seccomp_add_default_syscall_filter(scmp_filter_ctx ctx,
-                                              uint64_t cap_list_retain) {
-        unsigned i;
-        int r;
+static int seccomp_add_default_syscall_filter(
+                scmp_filter_ctx ctx,
+                uint32_t arch,
+                uint64_t cap_list_retain) {
+
         static const struct {
                 uint64_t capability;
                 int syscall_num;
@@ -111,23 +112,29 @@ static int seccomp_add_default_syscall_filter(scmp_filter_ctx ctx,
                 { CAP_SYS_TIME,   SCMP_SYS(settimeofday)        },
                 { CAP_SYS_TIME,   SCMP_SYS(stime)               },
         };
+        unsigned i;
+        int r, c = 0;
 
         for (i = 0; i < ELEMENTSOF(blacklist); i++) {
                 if (blacklist[i].capability != 0 && (cap_list_retain & (1ULL << blacklist[i].capability)))
                         continue;
 
-                r = seccomp_rule_add(ctx, SCMP_ACT_ERRNO(EPERM), blacklist[i].syscall_num, 0);
-                if (r == -EFAULT)
-                        continue; /* unknown syscall */
-                if (r < 0)
-                        return log_error_errno(r, "Failed to block syscall: %m");
+                r = seccomp_rule_add_exact(ctx, SCMP_ACT_ERRNO(EPERM), blacklist[i].syscall_num, 0);
+                if (r < 0) {
+                        /* If the system call is not known on this architecture, then that's fine, let's ignore it */
+                        _cleanup_free_ char *n = NULL;
+
+                        n = seccomp_syscall_resolve_num_arch(arch, blacklist[i].syscall_num);
+                        log_debug_errno(r, "Failed to add rule for system call %s, ignoring: %m", strna(n));
+                } else
+                        c++;
         }
 
-        return 0;
+        return c;
 }
 
 int setup_seccomp(uint64_t cap_list_retain) {
-        scmp_filter_ctx seccomp;
+        uint32_t arch;
         int r;
 
         if (!is_seccomp_available()) {
@@ -135,45 +142,51 @@ int setup_seccomp(uint64_t cap_list_retain) {
                 return 0;
         }
 
-        r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW);
-        if (r < 0)
-                return log_error_errno(r, "Failed to allocate seccomp object: %m");
-
-        r = seccomp_add_default_syscall_filter(seccomp, cap_list_retain);
-        if (r < 0)
-                goto finish;
-
-        /*
-           Audit is broken in containers, much of the userspace audit
-           hookup will fail if running inside a container. We don't
-           care and just turn off creation of audit sockets.
-
-           This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail
-           with EAFNOSUPPORT which audit userspace uses as indication
-           that audit is disabled in the kernel.
-         */
-
-        r = seccomp_rule_add(
-                        seccomp,
-                        SCMP_ACT_ERRNO(EAFNOSUPPORT),
-                        SCMP_SYS(socket),
-                        2,
-                        SCMP_A0(SCMP_CMP_EQ, AF_NETLINK),
-                        SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT));
-        if (r < 0) {
-                log_error_errno(r, "Failed to add audit seccomp rule: %m");
-                goto finish;
-        }
+        SECCOMP_FOREACH_LOCAL_ARCH(arch) {
+                _cleanup_(seccomp_releasep) scmp_filter_ctx seccomp = NULL;
+                int n;
+
+                log_debug("Operating on architecture: %s", seccomp_arch_to_string(arch));
+
+                r = seccomp_init_for_arch(&seccomp, arch, SCMP_ACT_ALLOW);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to allocate seccomp object: %m");
+
+                n = seccomp_add_default_syscall_filter(seccomp, arch, cap_list_retain);
+                if (n < 0)
+                        return n;
+
+                /*
+                  Audit is broken in containers, much of the userspace audit hookup will fail if running inside a
+                  container. We don't care and just turn off creation of audit sockets.
+
+                  This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail with EAFNOSUPPORT which audit userspace uses
+                  as indication that audit is disabled in the kernel.
+                */
+
+                r = seccomp_rule_add_exact(
+                                seccomp,
+                                SCMP_ACT_ERRNO(EAFNOSUPPORT),
+                                SCMP_SYS(socket),
+                                2,
+                                SCMP_A0(SCMP_CMP_EQ, AF_NETLINK),
+                                SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT));
+                if (r < 0)
+                        log_debug_errno(r, "Failed to add audit seccomp rule, ignoring: %m");
+                else
+                        n++;
+
+                if (n <= 0) /* no rule added? then skip this architecture */
+                        continue;
 
-        r = seccomp_load(seccomp);
-        if (r < 0) {
-                log_error_errno(r, "Failed to install seccomp audit filter: %m");
-                goto finish;
+                r = seccomp_load(seccomp);
+                if (IN_SET(r, -EPERM, -EACCES))
+                        return log_error_errno(r, "Failed to install seccomp audit filter: %m");
+                if (r < 0)
+                        log_debug_errno(r, "Failed to install filter set for architecture %s, skipping: %m", seccomp_arch_to_string(arch));
         }
 
-finish:
-        seccomp_release(seccomp);
-        return r;
+        return 0;
 }
 
 #else