Support seccomp on musl

Musl-based distributions like Alpine didn't work with seccomp profile
enabled. This was due to musl using different syscalls than glibc under
the cover.

- writev/readv and epoll_pwait have been allowed
- specific ioctl has been allowed to enableprinting the error/exit
  message
- brk has been allowed at runtime
- worker writes to stderr have been allowed (this was also broken on
  glibc)

1 files changed, 19 insertions, 1 deletions

diff --git a/linux_priv.c b/linux_priv.c
index 04155dd..f1ea406 100644
--- a/linux_priv.c
+++ b/linux_priv.c
@@ -2,6 +2,7 @@
 #include <seccomp.h>
 #include <errno.h>
 #include <stdlib.h>
+#include <sys/ioctl.h>
 #include "memcached.h"
 
 // In the future when the system is more tested this could be switched
@@ -16,21 +17,27 @@ void drop_privileges(void) {
 
     int rc = 0;
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(sigreturn), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(rt_sigprocmask), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(futex), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(epoll_wait), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(epoll_pwait), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(accept4), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(accept), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(write), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(writev), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(fstat), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(mmap), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(munmap), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(shmctl), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(exit_group), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(brk), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(ioctl), 1, SCMP_A1(SCMP_CMP_EQ, TIOCGWINSZ));
 
 #ifdef MEMCACHED_DEBUG
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(open), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(fcntl), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(read), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(readv), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(lseek), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(close), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(getpid), 0);
@@ -57,10 +64,13 @@ void drop_worker_privileges(void) {
 
     int rc = 0;
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(sigreturn), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(rt_sigprocmask), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(futex), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(epoll_wait), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(epoll_pwait), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(epoll_ctl), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(read), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(readv), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(mprotect), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(getpeername), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(close), 0);
@@ -70,6 +80,8 @@ void drop_worker_privileges(void) {
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(mremap), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(munmap), 0);
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(recvfrom), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(brk), 0);
+    rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(ioctl), 1, SCMP_A1(SCMP_CMP_EQ, TIOCGWINSZ));
 
     // for spawning the LRU crawler
     rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(clone), 0);
@@ -83,9 +95,9 @@ void drop_worker_privileges(void) {
 
     if (settings.shutdown_command) {
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(tgkill), 0);
+        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(tkill), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(exit_group), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(fstat), 0);
-        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(rt_sigprocmask), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(getpid), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(gettid), 0);
     }
@@ -95,8 +107,14 @@ void drop_worker_privileges(void) {
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(fcntl), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(lseek), 0);
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(write), 0);
+        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(writev), 0);
     } else {
+        // stdout
         rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(write), 1, SCMP_A0(SCMP_CMP_EQ, 1));
+        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(writev), 1, SCMP_A0(SCMP_CMP_EQ, 1));
+        // stderr
+        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(write), 1, SCMP_A0(SCMP_CMP_EQ, 2));
+        rc |= seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(writev), 1, SCMP_A0(SCMP_CMP_EQ, 2));
     }
 
     if (rc != 0) {