libafs: vldb cache timeout option (-volume-ttl)

The unix cache manager caches VLDB information for read-only volumes as
long as a volume callback is held for a read-only volume.  The volume
callback may be held as long as files in the read-only volume are being
accessed.  The cache manager caches VLDB information for read/write
volumes as long as volume level errors (such as VMOVED) are not returned
by a fileserver while accessing files within the volume.

Add a new option to set the maximum amount of time VLDB information will
be cached, even if a callback is still held for a read-only volume, or
no volume errors have been encounted while accessing files in read/write
volumes.

This avoids situations where the vldb information is cached indefinitely
for read-only and read/write volumes.  Instead, the VL servers will be
periodically probed for volume information.

Change-Id: I5f2a57cdaf5cbe7b1bc0440ed6408226cc988fed
Reviewed-on: https://gerrit.openafs.org/11898
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>

diff --git a/doc/man-pages/pod8/afsd.pod b/doc/man-pages/pod8/afsd.pod
index dd48a16..6114a7e 100644 (file)
--- a/doc/man-pages/pod8/afsd.pod
+++ b/doc/man-pages/pod8/afsd.pod
@@ -34,6 +34,7 @@ B<afsd> [B<-afsdb>] [B<-backuptree>]
      [B<-disable-dynamic-vcaches>] 
      S<<< [B<-volumes> <I<number of volume entries>>] >>>
      [B<-waitclose>] [B<-rxmaxfrags> <I<max # of fragments>>]
+     S<<< [B<-volume-ttl> <I<vldb cache timeout>>] >>>
 
 =for html
 </div>
@@ -753,6 +754,21 @@ affected in previous versions of the Cache Manager, to perform synchronous
 writes to the File Server, is now the default behavior. To perform
 asynchronous writes in certain cases, use the B<fs storebehind> command.
 
+=item B<-volume-ttl>
+
+Specifies the maximum amount of time the Cache Manager will cache volume
+information retrieved from VL Servers.
+
+By default, the Cache Manager will cache read-only volume information as long
+as a volume callback is held for a read-only volume. The callback may be held
+as long as files in the read-only volume are being accessed.  The Cache Manager
+will cache read/write volume information as long as volume level errors are not
+returned from fileservers while accessing files within the volume.
+
+Use the B<-volume-ttl> to specify the maximum amount of time in seconds
+volume information will be cached, regardless of connectivity to the
+fileserers. The minimum valid value is 600 seconds (10 minutes).
+
 =back
 
 =head1 EXAMPLES

diff --git a/src/afs/afs.h b/src/afs/afs.h
index bf2273d..d94f328 100644 (file)
--- a/src/afs/afs.h
+++ b/src/afs/afs.h
@@ -1398,6 +1398,10 @@ extern struct brequest afs_brs[NBRS];    /* request structures */
 #define AFS_VOLCHECK_MTPTS     0x4     /* mount point invalidation also */
 #define AFS_VOLCHECK_FORCE     0x8     /* do all forcibly */
 
+/* For volume ttl expiry checks. */
+#define AFS_MIN_VOLUME_TTL 600
+#define AFS_MAX_VOLUME_TTL MAX_AFS_INT32
+
 #endif /* KERNEL */
 
 #define        AFS_FSPORT          ((unsigned short) htons(7000))

diff --git a/src/afs/afs_call.c b/src/afs/afs_call.c
index 3b5641a..bec71c4 100644 (file)
--- a/src/afs/afs_call.c
+++ b/src/afs/afs_call.c
@@ -87,6 +87,8 @@ afs_int32 afs_rx_idledead_rep = AFS_IDLEDEADTIME_REP;
 
 static int afscall_set_rxpck_received = 0;
 
+extern afs_int32 afs_volume_ttl;
+
 /* From afs_util.c */
 extern afs_int32 afs_md5inum;
 
@@ -1317,6 +1319,13 @@ afs_syscall_call(long parm, long parm2, long parm3,
        default:
            code = EINVAL;
        }
+    } else if (parm == AFSOP_SET_VOLUME_TTL) {
+       if ((parm2 < AFS_MIN_VOLUME_TTL) || (parm2 > AFS_MAX_VOLUME_TTL)) {
+           code = EINVAL;
+       } else {
+           afs_volume_ttl = parm2;
+           code = 0;
+       }
     } else {
        code = EINVAL;
     }

diff --git a/src/afs/afs_volume.c b/src/afs/afs_volume.c
index b731a59..ff7cc47 100644 (file)
--- a/src/afs/afs_volume.c
+++ b/src/afs/afs_volume.c
@@ -54,6 +54,7 @@
 #endif /* vlserver error base define */
 
 /* Exported variables */
+afs_int32 afs_volume_ttl = 0;
 afs_dcache_id_t volumeInode;   /* Inode for VolumeItems file */
 afs_rwlock_t afs_xvolume;      /** allocation lock for volumes */
 struct volume *afs_freeVolList;
@@ -415,6 +416,12 @@ afs_ResetVolumes(struct server *srvp, struct volume *tv)
  * Dynroot volumes are not setup from vldb queries, so never expire.
  * Read-only volume expiry is tied to the volume callback.
  *
+ * Optionally, invalidate volume information after a fixed timeout.
+ * The vlservers will be periodically probed for volume information.
+ * This avoids a situation where the vldb information is cached
+ * indefinitely as long as files in the volume are accessed (and are
+ * not in the vcache) before the callback expires.
+ *
  * \param tv volume to check
  * \param now current time
  *
@@ -423,8 +430,16 @@ afs_ResetVolumes(struct server *srvp, struct volume *tv)
 static int
 IsExpired(struct volume *tv, afs_int32 now)
 {
-    return (tv->expireTime < (now + 10)) && (tv->states & VRO)
-           && !afs_IsDynrootVolume(tv);
+    if (afs_IsDynrootVolume(tv)) {
+       return 0;
+    }
+    if ((tv->states & VRO) && (tv->expireTime < (now + 10))) {
+       return 1;
+    }
+    if ((afs_volume_ttl != 0) && ((tv->setupTime + afs_volume_ttl) < now)) {
+       return 1;
+    }
+    return 0;
 }
 
 /**

diff --git a/src/afsd/afsd.c b/src/afsd/afsd.c
index 0976099..4692172 100644 (file)
--- a/src/afsd/afsd.c
+++ b/src/afsd/afsd.c
@@ -67,6 +67,7 @@
   *    -splitcache RW/RO ratio for cache.
   *    -rxmaxfrags Max number of UDP fragments per rx packet.
   *    -inumcalc  inode number calculation method; 0=compat, 1=MD5 digest
+  *    -volume-ttl vldb cache timeout in seconds
   *---------------------------------------------------------------------------*/
 
 #include <afsconfig.h>
@@ -284,6 +285,7 @@ int afsd_debug = 0;         /*Are we printing debugging info? */
 static int afsd_CloseSynch = 0;        /*Are closes synchronous or not? */
 static int rxmaxmtu = 0;       /* Are we forcing a limit on the mtu? */
 static int rxmaxfrags = 0;      /* Are we forcing a limit on frags? */
+static int volume_ttl = 0;      /* enable vldb cache timeout support */
 
 #ifdef AFS_SGI62_ENV
 #define AFSD_INO_T ino64_t
@@ -362,6 +364,7 @@ enum optionsList {
     OPT_dynrootsparse,
     OPT_rxmaxfrags,
     OPT_inumcalc,
+    OPT_volume_ttl,
 };
 
 #ifdef MACOS_EVENT_HANDLING
@@ -1905,6 +1908,7 @@ CheckOptions(struct cmd_syndesc *as)
     if (cmd_OptionPresent(as, OPT_inumcalc)) {
        cmd_OptionAsString(as, OPT_inumcalc, &inumcalc);
     }
+    cmd_OptionAsInt(as, OPT_volume_ttl, &volume_ttl);
 
     /* parse cacheinfo file if this is a diskcache */
     if (ParseCacheInfoFile()) {
@@ -2425,6 +2429,14 @@ afsd_run(void)
        afsd_syscall(AFSOP_ROOTVOLUME, rootVolume);
     }
 
+    if (volume_ttl != 0) {
+       if (afsd_verbose)
+           printf("%s: Calling AFSOP_SET_VOLUME_TTL with '%d'\n", rn, volume_ttl);
+       code = afsd_syscall(AFSOP_SET_VOLUME_TTL, volume_ttl);
+       if (code != 0)
+           printf("%s: Error setting volume ttl to %d seconds; code=%d.\n", rn, volume_ttl, code);
+    }
+
     /*
      * Pass the kernel the name of the workstation cache file holding the
      * volume information.
@@ -2624,6 +2636,9 @@ afsd_init(void)
                        "send/receive per Rx packet");
     cmd_AddParmAtOffset(ts, OPT_inumcalc, "-inumcalc", CMD_SINGLE, CMD_OPTIONAL,
                        "Set inode number calculation method");
+    cmd_AddParmAtOffset(ts, OPT_volume_ttl, "-volume-ttl", CMD_SINGLE,
+                       CMD_OPTIONAL,
+                       "Set the vldb cache timeout value in seconds.");
 }
 
 /**
@@ -2704,6 +2719,7 @@ afsd_syscall_populate(struct afsd_syscall_args *args, int syscall, va_list ap)
     case AFSOP_GO:
     case AFSOP_SET_RMTSYS_FLAG:
     case AFSOP_SET_INUMCALC:
+    case AFSOP_SET_VOLUME_TTL:
        params[0] = CAST_SYSCALL_PARAM((va_arg(ap, int)));
        break;
     case AFSOP_SET_THISCELL:

diff --git a/src/config/afs_args.h b/src/config/afs_args.h
index ee98472..4fa286b 100644 (file)
--- a/src/config/afs_args.h
+++ b/src/config/afs_args.h
@@ -55,6 +55,7 @@
 #define AFSOP_SET_RMTSYS_FLAG   44     /* set flag if rmtsys is enabled */
 #define AFSOP_SEED_ENTROPY      45     /* Give the kernel hcrypto entropy */
 #define AFSOP_SET_INUMCALC      46     /* set inode number calculation method */
+#define AFSOP_SET_VOLUME_TTL     47     /* set the vldb cache timeout */
 
 #define AFSOP_RXLISTENER_DAEMON  48    /* starts kernel RX listener */