LU-13306 mgs: use large NIDS in the nid table on the MGS

[fs/lustre-release.git] / lustre / mgs / mgs_nids.c

diff --git a/lustre/mgs/mgs_nids.c b/lustre/mgs/mgs_nids.c
index d9fdd36..969e06a 100644 (file)
--- a/lustre/mgs/mgs_nids.c
+++ b/lustre/mgs/mgs_nids.c
@@ -111,13 +111,15 @@ static int mgs_nidtbl_read(struct obd_export *exp, struct mgs_nidtbl *tbl,
         */
        list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) {
                int entry_len = sizeof(*entry);
+               int i;
 
                if (tgt->mnt_version < version)
                        continue;
 
                /* write target recover information */
                mti  = &tgt->mnt_mti;
-               LASSERT(mti->mti_nid_count < MTI_NIDS_MAX);
+               if (!target_supports_large_nid(mti))
+                       LASSERT(mti->mti_nid_count < MTI_NIDS_MAX);
                entry_len += mti->mti_nid_count * sizeof(lnet_nid_t);
 
                if (entry_len > unit_size) {
@@ -177,9 +179,36 @@ static int mgs_nidtbl_read(struct obd_export *exp, struct mgs_nidtbl *tbl,
                entry->mne_type      = tgt->mnt_type;
                entry->mne_nid_type  = 0;
                entry->mne_nid_size  = sizeof(lnet_nid_t);
-               entry->mne_nid_count = mti->mti_nid_count;
-               memcpy(entry->u.nids, mti->mti_nids,
-                      mti->mti_nid_count * sizeof(lnet_nid_t));
+               /* We have been sent the newer larger NID format but the
+                * current nidtbl doesn't support it. So filter the NIDs
+                * sent to reject any real larger size NIDS.
+                */
+               if (target_supports_large_nid(mti)) {
+                       entry->mne_nid_count = 0;
+
+                       for (i = 0; i < mti->mti_nid_count; i++) {
+                               struct lnet_nid nid;
+                               int err;
+
+                               err = libcfs_strnid(&nid, mti->mti_nidlist[i]);
+                               if (err < 0)
+                                       GOTO(out, rc = err);
+
+                               /* if the large NID format represents a small
+                                * address space we can still pass it back to
+                                * the older clients.
+                                */
+                               if (nid_is_nid4(&nid)) {
+                                       entry->u.nids[entry->mne_nid_count] =
+                                               lnet_nid_to_nid4(&nid);
+                                       entry->mne_nid_count++;
+                               }
+                       }
+               } else {
+                       entry->mne_nid_count = mti->mti_nid_count;
+                       memcpy(entry->u.nids, mti->mti_nids,
+                              mti->mti_nid_count * sizeof(lnet_nid_t));
+               }
 
                version = tgt->mnt_version;
                rc     += entry_len;
@@ -312,10 +341,10 @@ static int mgs_nidtbl_write(const struct lu_env *env, struct fs_db *fsdb,
        struct mgs_nidtbl_target *tgt;
        bool found = false;
        int type = mti->mti_flags & LDD_F_SV_TYPE_MASK;
+       size_t mti_len = 0;
        int rc = 0;
 
        ENTRY;
-
        type &= ~LDD_F_SV_TYPE_MGS;
        LASSERT(type != 0);
 
@@ -330,8 +359,14 @@ static int mgs_nidtbl_write(const struct lu_env *env, struct fs_db *fsdb,
                        break;
                }
        }
+
+       if (target_supports_large_nid(mti))
+               mti_len = mti->mti_nid_count * LNET_NIDSTR_SIZE;
        if (!found) {
-               OBD_ALLOC_PTR(tgt);
+               size_t len = offsetof(struct mgs_nidtbl_target,
+                                     mnt_mti.mti_nidlist);
+
+               OBD_ALLOC(tgt, len + mti_len);
                if (!tgt)
                        GOTO(out, rc = -ENOMEM);
 
@@ -344,7 +379,9 @@ static int mgs_nidtbl_write(const struct lu_env *env, struct fs_db *fsdb,
        }
 
        tgt->mnt_version = ++tbl->mn_version;
-       tgt->mnt_mti     = *mti;
+       tgt->mnt_mti = *mti;
+       if (target_supports_large_nid(mti))
+               memcpy(tgt->mnt_mti.mti_nidlist, mti->mti_nidlist, mti_len);
 
        list_move_tail(&tgt->mnt_list, &tbl->mn_targets);
 
@@ -362,6 +399,8 @@ out:
 static void mgs_nidtbl_fini_fs(struct fs_db *fsdb)
 {
        struct mgs_nidtbl *tbl = &fsdb->fsdb_nidtbl;
+       size_t len = offsetof(struct mgs_nidtbl_target,
+                             mnt_mti.mti_nidlist);
        LIST_HEAD(head);
 
        mutex_lock(&tbl->mn_lock);
@@ -371,11 +410,14 @@ static void mgs_nidtbl_fini_fs(struct fs_db *fsdb)
 
        while (!list_empty(&head)) {
                struct mgs_nidtbl_target *tgt;
+               size_t mti_len = 0;
 
                tgt = list_first_entry(&head, struct mgs_nidtbl_target,
                                       mnt_list);
+               if (target_supports_large_nid(&tgt->mnt_mti))
+                       mti_len += tgt->mnt_mti.mti_nid_count * LNET_NIDSTR_SIZE;
                list_del(&tgt->mnt_list);
-               OBD_FREE_PTR(tgt);
+               OBD_FREE(tgt, len + mti_len);
        }
 }