seccomp: MemoryDenyWriteExecute= should affect both mmap() and mmap2() (#5254)

On i386 we block the old mmap() call entirely, since we cannot properly
filter it. Thankfully it hasn't been used by glibc since quite some
time.

Fixes: #5240

diff --git a/man/systemd.exec.xml b/man/systemd.exec.xml
index bb38ea24674b30252455ac0294df3d7a68ce8134..fd47b0a20a04017f82bb5a3f8754adbd0d3e948c 100644 (file)
--- a/man/systemd.exec.xml
+++ b/man/systemd.exec.xml
@@ -1607,22 +1607,20 @@
         <term><varname>MemoryDenyWriteExecute=</varname></term>
 
         <listitem><para>Takes a boolean argument. If set, attempts to create memory mappings that are writable and
-        executable at the same time, or to change existing memory mappings to become executable, or mapping shared memory
-        segments as executable are prohibited.
-        Specifically, a system call filter is added that rejects
-        <citerefentry><refentrytitle>mmap</refentrytitle><manvolnum>2</manvolnum></citerefentry>
-        system calls with both <constant>PROT_EXEC</constant> and <constant>PROT_WRITE</constant> set,
-        <citerefentry><refentrytitle>mprotect</refentrytitle><manvolnum>2</manvolnum></citerefentry>
-        system calls with <constant>PROT_EXEC</constant> set and
-        <citerefentry><refentrytitle>shmat</refentrytitle><manvolnum>2</manvolnum></citerefentry>
-        system calls with <constant>SHM_EXEC</constant> set. Note that this option is incompatible with programs
-        that generate program code dynamically at runtime, such as JIT execution engines, or programs compiled making
-        use of the code "trampoline" feature of various C compilers. This option improves service security, as it makes
-        harder for software exploits to change running code dynamically.
-        If running in user mode, or in system mode, but without the <constant>CAP_SYS_ADMIN</constant>
-        capability (e.g. setting <varname>User=</varname>), <varname>NoNewPrivileges=yes</varname>
-        is implied.
-        </para></listitem>
+        executable at the same time, or to change existing memory mappings to become executable, or mapping shared
+        memory segments as executable are prohibited.  Specifically, a system call filter is added that rejects
+        <citerefentry><refentrytitle>mmap</refentrytitle><manvolnum>2</manvolnum></citerefentry> system calls with both
+        <constant>PROT_EXEC</constant> and <constant>PROT_WRITE</constant> set,
+        <citerefentry><refentrytitle>mprotect</refentrytitle><manvolnum>2</manvolnum></citerefentry> system calls with
+        <constant>PROT_EXEC</constant> set and
+        <citerefentry><refentrytitle>shmat</refentrytitle><manvolnum>2</manvolnum></citerefentry> system calls with
+        <constant>SHM_EXEC</constant> set. Note that this option is incompatible with programs that generate program
+        code dynamically at runtime, such as JIT execution engines, or programs compiled making use of the code
+        "trampoline" feature of various C compilers. This option improves service security, as it makes harder for
+        software exploits to change running code dynamically. Note that this feature is fully available on x86-64, and
+        partially on x86. Specifically, the <function>shmat()</function> protection is not available on x86. If running
+        in user mode, or in system mode, but without the <constant>CAP_SYS_ADMIN</constant> capability (e.g. setting
+        <varname>User=</varname>), <varname>NoNewPrivileges=yes</varname> is implied.  </para></listitem>
       </varlistentry>
 
       <varlistentry>

diff --git a/src/shared/seccomp-util.c b/src/shared/seccomp-util.c
index 609e0619af9cf533cfdf0aee836f59f9b8ef720d..44706669b4fcf9ea0876eb1936fb98126409da96 100644 (file)
--- a/src/shared/seccomp-util.c
+++ b/src/shared/seccomp-util.c
@@ -1086,27 +1086,81 @@ int seccomp_restrict_realtime(void) {
 }
 
 int seccomp_memory_deny_write_execute(void) {
+
         uint32_t arch;
         int r;
 
         SECCOMP_FOREACH_LOCAL_ARCH(arch) {
                 _cleanup_(seccomp_releasep) scmp_filter_ctx seccomp = NULL;
+                int filter_syscall = 0, block_syscall = 0, shmat_syscall = 0;
 
                 log_debug("Operating on architecture: %s", seccomp_arch_to_string(arch));
 
+                switch (arch) {
+
+                case SCMP_ARCH_X86:
+                        filter_syscall = SCMP_SYS(mmap2);
+                        block_syscall = SCMP_SYS(mmap);
+
+                        /* Note that shmat() isn't available on i386, where the call is multiplexed through ipc(). We
+                         * ignore that here, which means there's still a way to get writable/executable memory, if an
+                         * IPC key is mapped like this on i386. That's a pity, but no total loss. */
+                        break;
+
+                case SCMP_ARCH_X86_64:
+                case SCMP_ARCH_X32:
+                        filter_syscall = SCMP_SYS(mmap);
+                        shmat_syscall = SCMP_SYS(shmat);
+                        break;
+
+                /* Please add more definitions here, if you port systemd to other architectures! */
+
+#if !defined(__i386__) && !defined(__x86_64__)
+#warning "Consider adding the right mmap() syscall definitions here!"
+#endif
+                }
+
+                /* Can't filter mmap() on this arch, then skip it */
+                if (filter_syscall == 0)
+                        continue;
+
                 r = seccomp_init_for_arch(&seccomp, arch, SCMP_ACT_ALLOW);
                 if (r < 0)
                         return r;
 
-                r = seccomp_rule_add_exact(
-                                seccomp,
-                                SCMP_ACT_ERRNO(EPERM),
-                                SCMP_SYS(mmap),
-                                1,
-                                SCMP_A2(SCMP_CMP_MASKED_EQ, PROT_EXEC|PROT_WRITE, PROT_EXEC|PROT_WRITE));
-                if (r < 0) {
-                        log_debug_errno(r, "Failed to add mmap() rule for architecture %s, skipping: %m", seccomp_arch_to_string(arch));
-                        continue;
+                if (filter_syscall != 0)  {
+                        r = seccomp_rule_add_exact(
+                                        seccomp,
+                                        SCMP_ACT_ERRNO(EPERM),
+                                        filter_syscall,
+                                        1,
+                                        SCMP_A2(SCMP_CMP_MASKED_EQ, PROT_EXEC|PROT_WRITE, PROT_EXEC|PROT_WRITE));
+                        if (r < 0) {
+                                _cleanup_free_ char *n = NULL;
+
+                                n = seccomp_syscall_resolve_num_arch(arch, filter_syscall);
+                                log_debug_errno(r, "Failed to add %s() rule for architecture %s, skipping: %m",
+                                                strna(n),
+                                                seccomp_arch_to_string(arch));
+                                continue;
+                        }
+                }
+
+                if (block_syscall != 0) {
+                        r = seccomp_rule_add_exact(
+                                        seccomp,
+                                        SCMP_ACT_ERRNO(EPERM),
+                                        block_syscall,
+                                        0);
+                        if (r < 0) {
+                                _cleanup_free_ char *n = NULL;
+
+                                n = seccomp_syscall_resolve_num_arch(arch, block_syscall);
+                                log_debug_errno(r, "Failed to add %s() rule for architecture %s, skipping: %m",
+                                                strna(n),
+                                                seccomp_arch_to_string(arch));
+                                continue;
+                        }
                 }
 
                 r = seccomp_rule_add_exact(
@@ -1120,15 +1174,17 @@ int seccomp_memory_deny_write_execute(void) {
                         continue;
                 }
 
-                r = seccomp_rule_add_exact(
-                                seccomp,
-                                SCMP_ACT_ERRNO(EPERM),
-                                SCMP_SYS(shmat),
-                                1,
-                                SCMP_A2(SCMP_CMP_MASKED_EQ, SHM_EXEC, SHM_EXEC));
-                if (r < 0) {
-                        log_debug_errno(r, "Failed to add shmat() rule for architecture %s, skipping: %m", seccomp_arch_to_string(arch));
-                        continue;
+                if (shmat_syscall != 0) {
+                        r = seccomp_rule_add_exact(
+                                        seccomp,
+                                        SCMP_ACT_ERRNO(EPERM),
+                                        SCMP_SYS(shmat),
+                                        1,
+                                        SCMP_A2(SCMP_CMP_MASKED_EQ, SHM_EXEC, SHM_EXEC));
+                        if (r < 0) {
+                                log_debug_errno(r, "Failed to add shmat() rule for architecture %s, skipping: %m", seccomp_arch_to_string(arch));
+                                continue;
+                        }
                 }
 
                 r = seccomp_load(seccomp);

diff --git a/src/shared/seccomp-util.h b/src/shared/seccomp-util.h
index 2563fcd38a59c75e9e7fbaf0351fe67ed6140672..bfbfb5ab3d7f083a4ec3217262715cd4a268405b 100644 (file)
--- a/src/shared/seccomp-util.h
+++ b/src/shared/seccomp-util.h
@@ -84,6 +84,13 @@ int seccomp_memory_deny_write_execute(void);
 #define SECCOMP_RESTRICT_ADDRESS_FAMILIES_BROKEN 0
 #endif
 
+/* mmap() blocking is only available on some archs for now */
+#if defined(__x86_64__) || defined(__i386__)
+#define SECCOMP_MEMORY_DENY_WRITE_EXECUTE_BROKEN 0
+#else
+#define SECCOMP_MEMORY_DENY_WRITE_EXECUTE_BROKEN 1
+#endif
+
 extern const uint32_t seccomp_local_archs[];
 
 #define SECCOMP_FOREACH_LOCAL_ARCH(arch) \

diff --git a/src/test/test-seccomp.c b/src/test/test-seccomp.c
index 54e7947c2f1399c05c726dbc5fe6b638b2eb2467..36592388105b3ce66143fd237dbbf3adc8664cca 100644 (file)
--- a/src/test/test-seccomp.c
+++ b/src/test/test-seccomp.c
@@ -384,11 +384,21 @@ static void test_memory_deny_write_execute(void) {
                 assert_se(p != MAP_FAILED);
                 assert_se(munmap(p, page_size()) >= 0);
 
-                seccomp_memory_deny_write_execute();
+                p = mmap(NULL, page_size(), PROT_WRITE|PROT_READ, MAP_PRIVATE|MAP_ANONYMOUS, -1,0);
+                assert_se(p != MAP_FAILED);
+                assert_se(munmap(p, page_size()) >= 0);
 
+                assert_se(seccomp_memory_deny_write_execute() >= 0);
+
+#if SECCOMP_MEMORY_DENY_WRITE_EXECUTE_BROKEN
+                p = mmap(NULL, page_size(), PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_ANONYMOUS, -1,0);
+                assert_se(p != MAP_FAILED);
+                assert_se(munmap(p, page_size()) >= 0);
+#else
                 p = mmap(NULL, page_size(), PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_ANONYMOUS, -1,0);
                 assert_se(p == MAP_FAILED);
                 assert_se(errno == EPERM);
+#endif
 
                 p = mmap(NULL, page_size(), PROT_WRITE|PROT_READ, MAP_PRIVATE|MAP_ANONYMOUS, -1,0);
                 assert_se(p != MAP_FAILED);