Fix: do not rely on DAC_OVERRIDE capability on Linux despite being root

It turns out, for example, that SELinux will prevent this capability
for confined processes running as root.  It then means that sbd cannot
access, as a client, files used for joining two local communication sides
within libqb-arranged IPC mechanism in case those files do not have
permissions to explicitly allow file-based access with credentials
of this client -- which is exactly what happens when the IPC servers
are pacemaker daemons not run as root on their own.

Solution is two-phased:
1. have sbd add respective non-privileged group corresponding to the
   server side of the IPC -- this patch
2. ensure this server side (pacemaker) does allow group-derived access
   (i.e., the access permissions for group are as relaxed as needed,
   umask notwithstanding) -- outside of the sbd's scope

Signed-off-by: Jan Pokorný <[email protected]>

src/sbd-pacemaker.c

@@ -45,6 +45,11 @@
 #include <libgen.h>
 #include <sys/utsname.h>
 
+#ifdef __linux__
+#include <grp.h>  /* getgrnam_r, initgroups */
+#include <pwd.h>  /* getpwuid_r */
+#endif
+
 #include <config.h>
 
 #include <crm_config.h>
@@ -449,6 +454,80 @@ clean_up(int rc)
 	return;
 }
 
+#ifdef __linux__
+#define CLUSTER_GROUP  "haclient"
+/* Try to add well-known cluster stack group (CLUSTER_GROUP) as a supplementary
+   group to this root-privileged process for good measure (see the call site);
+   returns 0 in case of success, respective positive exit status (different
+   from 0, 1, and EXIT_MD_IO_FAIL...EXIT_MD_REQUEST_CRASHDUMP) otherwise. */
+static int
+add_cluster_group()
+{
+    int rc = 0;
+    long gr_limit = -1, pw_limit = -1, limit;
+    char *buf;
+    struct group group, *group_result;
+    struct passwd passwd, *passwd_result;
+    gid_t cluster_gid;
+
+    limit = sysconf(_SC_PAGESIZE);
+    limit = (limit > 0) ? limit : 4096;  /* assume sufficient, just in case */
+    gr_limit = sysconf(_SC_GETGR_R_SIZE_MAX);
+    gr_limit = (gr_limit > 0) ? gr_limit : limit;
+    pw_limit = sysconf(_SC_GETPW_R_SIZE_MAX);
+    limit = (gr_limit >= pw_limit) ? gr_limit : pw_limit;
+
+    if ((buf = malloc(limit)) == NULL) {
+        return 74;  /* EX_IOERR */
+    }
+
+    do {
+        rc = getgrnam_r(CLUSTER_GROUP, &group, buf, limit, &group_result);
+    } while (rc == -1 && errno == EINTR);
+    if (rc == -1 || group_result == NULL) {
+        if (rc == -1) {
+            cl_perror("Unable to get group entry for %s", CLUSTER_GROUP);
+            rc = 69;  /* EX_UNAVAILABLE */
+        } else {
+            cl_log(LOG_ERR, "Unable to get group entry for %s", CLUSTER_GROUP);
+            rc = 78;  /* EX_CONFIG */
+        }
+        goto bail;
+    }
+    cluster_gid = group.gr_gid;
+
+    do {
+        rc = getpwuid_r(0, &passwd, buf, limit, &passwd_result);
+    } while (rc == -1 && errno == EINTR);
+    if (rc == -1 || passwd_result == NULL) {
+        if (rc == -1) {
+            cl_perror("Unable to get passwd entry for UID=0");
+            rc = 69;  /* EX_UNAVAILABLE */
+        } else {
+            cl_log(LOG_ERR, "Unable to get passwd entry for UID=0");
+            rc = 78;  /* EX_CONFIG */
+        }
+        goto bail;
+    }
+
+    /* this is supposed to be root, hence (1) shall succeed and
+       (2) the root shall not gain any new unwanted group-based
+       access over what it already has per its standard supplementary
+       groups even if they have been explicitly dropped by now (not
+       the case now; root is usually just in its own group, too) */
+    rc = initgroups(passwd.pw_name, cluster_gid);
+    if (rc == -1) {
+        cl_perror("Unable to set supplementary group %s", CLUSTER_GROUP);
+        rc = (errno == EPERM) ? 77  /* EX_NOPERM */
+                              : 74  /* EX_IOERR */;
+    }
+
+bail:
+    free(buf);
+    return rc;
+}
+#endif
+
 int
 servant_pcmk(const char *diskname, int mode, const void* argp)
 {
@@ -459,6 +538,19 @@ servant_pcmk(const char *diskname, int mode, const void* argp)
 	set_proc_title("sbd: watcher: Pacemaker");
         setenv("PCMK_watchdog", "true", 1);
 
+#ifdef __linux__
+        /* since we run this as root, we may actually be prevented from
+           accessing hacluster:haclient owned shared IPC mmap'd files of,
+           e.g., pacemakerd-based ("CIB daemon") in some cased where
+           root is not "all powerful" (e.g. under strict SELinux
+           confinement not allowing a DAC_OVERRIDE for any reasons)
+           TODO: first check if CAP_DAC_OVERRIDE is missing? */
+        if ((exit_code = add_cluster_group()) > 0) {
+            cl_log(LOG_CRIT, "Unable to ensure Pacemaker can be watched");
+            clean_up(exit_code);
+        }
+#endif
+
         if(debug == 0) {
             /* We don't want any noisy crm messages */
             set_crm_log_level(LOG_CRIT);