Workaround for broken kernel autofs mounts

The kernel automounter is mostly broken wrt to containers. Most
notably if you start a new filesystem namespace and then attempt
to unmount any autofs filesystem, it will typically fail with a
weird error message like

  Failed to unmount '/.oldroot/sys/kernel/security':Too many levels of symbolic links

Attempting to detach the autofs mount using umount2(MNT_DETACH)
will also fail with the same error. Therefore if we get any error on
unmount()ing a filesystem from the old root FS when starting a
container, we must immediately break out and detach the entire
old root filesystem (ignoring any mounts below it).

This has the effect of making the old root filesystem inaccessible
to anything inside the container, but at the cost that the mounts
live on in the kernel until the container exits. Given that SystemD
uses autofs by default, we need LXC to be robust this scenario and
thus this tradeoff is worthwhile.

* src/lxc/lxc_container.c: Detach root filesystem if any umount
  operation fails.

diff --git a/src/lxc/lxc_container.c b/src/lxc/lxc_container.c
index 63284e52f5b31bcd425b1f1186664ca85f73d267..1440a097b8b99e60f6a0b2593bfc484a9526ba4e 100644 (file)
--- a/src/lxc/lxc_container.c
+++ b/src/lxc/lxc_container.c
@@ -860,6 +860,9 @@ static int lxcContainerUnmountOldFS(void)
     FILE *procmnt;
     int i;
     char mntbuf[1024];
+    int saveErrno;
+    const char *failedUmount = NULL;
+    int ret = -1;
 
     if (!(procmnt = setmntent("/proc/mounts", "r"))) {
         virReportSystemError(errno, "%s",
@@ -872,17 +875,15 @@ static int lxcContainerUnmountOldFS(void)
             continue;
 
         if (VIR_REALLOC_N(mounts, nmounts+1) < 0) {
-            endmntent(procmnt);
             virReportOOMError();
-            return -1;
+            goto cleanup;
         }
         if (!(mounts[nmounts++] = strdup(mntent.mnt_dir))) {
-            endmntent(procmnt);
             virReportOOMError();
-            return -1;
+            goto cleanup;
         }
+        VIR_DEBUG("Grabbed %s", mntent.mnt_dir);
     }
-    endmntent(procmnt);
 
     if (mounts)
         qsort(mounts, nmounts, sizeof(mounts[0]),
@@ -891,16 +892,42 @@ static int lxcContainerUnmountOldFS(void)
     for (i = 0 ; i < nmounts ; i++) {
         VIR_DEBUG("Umount %s", mounts[i]);
         if (umount(mounts[i]) < 0) {
-            virReportSystemError(errno,
-                                 _("Failed to unmount '%s'"),
-                                 mounts[i]);
-            return -1;
+            char ebuf[1024];
+            failedUmount = mounts[i];
+            saveErrno = errno;
+            VIR_WARN("Failed to unmount '%s', trying to detach root '%s': %s",
+                     failedUmount, mounts[nmounts-1],
+                     virStrerror(errno, ebuf, sizeof(ebuf)));
+            break;
         }
-        VIR_FREE(mounts[i]);
     }
+
+    if (failedUmount) {
+        /* This detaches the old root filesystem */
+        if (umount2(mounts[nmounts-1], MNT_DETACH) < 0) {
+            virReportSystemError(saveErrno,
+                                 _("Failed to unmount '%s' and could not detach old root '%s'"),
+                                 failedUmount, mounts[nmounts-1]);
+            goto cleanup;
+        }
+        /* This unmounts the tmpfs on which the old root filesystem was hosted */
+        if (umount(mounts[nmounts-1]) < 0) {
+            virReportSystemError(saveErrno,
+                                 _("Failed to unmount '%s' and could not unmount old root '%s'"),
+                                 failedUmount, mounts[nmounts-1]);
+            goto cleanup;
+        }
+    }
+
+    ret = 0;
+
+cleanup:
+    for (i = 0 ; i < nmounts ; i++)
+        VIR_FREE(mounts[i]);
+    endmntent(procmnt);
     VIR_FREE(mounts);
 
-    return 0;
+    return ret;
 }