LU-16096 tgt: improve messages for reply_data

Due to the format change in reply_data for WBC, it is not possible
to downgrade the MDT filesystem to an older release after mounting
it with 2.16.0 and later.

On the path to fixing that, improve the error messages printed by
tgt_reply_data_init() so that it is possible to see why the mount
is failing.  The next step is to avoid upgrading the replay_data
file if it is not necessary (no clients mounting with WBC enabled).

Test-Parameters: trivial testlist=conf-sanity env=ONLY=32
Fixes: bbf0017fdea5 ("LU-16096 recovery: upgrade reply data after recovery finish")
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Change-Id: Idd176031fef99b16de9278f05516dc45e817a819
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/49939
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Vitaliy Kuznetsov <vkuznetsov@ddn.com>
Reviewed-by: Qian Yingjin <qian@ddn.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/target/tgt_lastrcvd.c b/lustre/target/tgt_lastrcvd.c
index d0ae469..f8d306c 100644 (file)
--- a/lustre/target/tgt_lastrcvd.c
+++ b/lustre/target/tgt_lastrcvd.c
@@ -2159,8 +2159,8 @@ int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt)
                CDEBUG(D_INFO, "%s: new reply_data file, initializing\n",
                       tgt_name(tgt));
                lrh->lrh_magic = LRH_MAGIC;
-               lrh->lrh_header_size = sizeof(struct lsd_reply_header);
-               lrh->lrh_reply_size = sizeof(struct lsd_reply_data);
+               lrh->lrh_header_size = sizeof(*lrh);
+               lrh->lrh_reply_size = sizeof(*lrd);
                rc = tgt_reply_header_write(env, tgt, lrh);
                if (rc) {
                        CERROR("%s: error writing %s: rc = %d\n",
@@ -2168,7 +2168,8 @@ int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt)
                        GOTO(out, rc);
                }
        } else {
-               __u32 recsz = sizeof(struct lsd_reply_data);
+               __u32 recsz = sizeof(*lrd);
+               const char *lrd_ver = "v2";
 
                rc = tgt_reply_header_read(env, tgt, lrh);
                if (rc) {
@@ -2176,19 +2177,40 @@ int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt)
                               tgt_name(tgt), REPLY_DATA, rc);
                        GOTO(out, rc);
                }
-               if (!(lrh->lrh_magic == LRH_MAGIC &&
-                     lrh->lrh_reply_size == sizeof(struct lsd_reply_data) &&
-                     lrh->lrh_header_size == sizeof(struct lsd_reply_header)) &&
-                   !(lrh->lrh_magic == LRH_MAGIC_V1 &&
-                     lrh->lrh_reply_size == sizeof(struct lsd_reply_data_v1) &&
-                     lrh->lrh_header_size == sizeof(struct lsd_reply_header))) {
-                       CERROR("%s: invalid header in %s\n",
-                              tgt_name(tgt), REPLY_DATA);
-                       GOTO(out, rc = -EINVAL);
-               }
 
-               if (lrh->lrh_magic == LRH_MAGIC_V1)
+               switch (lrh->lrh_magic) {
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 5, 53, 0)
+               /* The old reply_data is replaced on the first mount after
+                * an upgrade, so no need to keep this interop code forever.
+                */
+               case LRH_MAGIC_V1:
                        recsz = sizeof(struct lsd_reply_data_v1);
+                       lrd_ver = "v1";
+
+                       CWARN("%s: %s v1 will be upgraded to new record size\n",
+                             tgt_name(tgt), REPLY_DATA);
+                       fallthrough;
+#endif
+               case LRH_MAGIC_V2:
+                       if (lrh->lrh_header_size != sizeof(*lrh)) {
+                               CERROR("%s: bad %s %s header size: %u != %lu\n",
+                                      tgt_name(tgt), REPLY_DATA, lrd_ver,
+                                      lrh->lrh_header_size, sizeof(*lrh));
+                               GOTO(out, rc = -EINVAL);
+                       }
+                       if (lrh->lrh_reply_size != recsz) {
+                               CERROR("%s: bad %s %s reply size: %u != %u\n",
+                               tgt_name(tgt), REPLY_DATA, lrd_ver,
+                               lrh->lrh_reply_size, recsz);
+                               GOTO(out, rc = -EINVAL);
+                       }
+                       break;
+               default:
+                       CERROR("%s: invalid %s magic: %x != %x/%x\n",
+                              tgt_name(tgt), REPLY_DATA,
+                              lrh->lrh_magic, LRH_MAGIC_V1, LRH_MAGIC_V2);
+                       GOTO(out, rc = -EINVAL);
+               }
 
                hash = cfs_hash_getref(tgt->lut_obd->obd_gen_hash);
                if (hash == NULL)
@@ -2199,7 +2221,7 @@ int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt)
                        GOTO(out, rc = -ENOMEM);
 
                /* Load reply_data from disk */
-               for (idx = 0, off = sizeof(struct lsd_reply_header);
+               for (idx = 0, off = lrh->lrh_header_size;
                     off < reply_data_size; idx++, off += recsz) {
                        rc = tgt_reply_data_read(env, tgt, lrd, off,
                                                 lrh->lrh_magic);