Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / staging / lustre / lustre / lov / lov_ea.c
diff --git a/kernel/drivers/staging/lustre/lustre/lov/lov_ea.c b/kernel/drivers/staging/lustre/lustre/lov/lov_ea.c
new file mode 100644 (file)
index 0000000..2bcfaea
--- /dev/null
@@ -0,0 +1,363 @@
+/*
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2012, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lustre/lov/lov_ea.c
+ *
+ * Author: Wang Di <wangdi@clusterfs.com>
+ */
+
+#define DEBUG_SUBSYSTEM S_LOV
+
+#include <asm/div64.h>
+#include "../../include/linux/libcfs/libcfs.h"
+
+#include "../include/obd_class.h"
+#include "../include/lustre/lustre_idl.h"
+
+#include "lov_internal.h"
+
+struct lovea_unpack_args {
+       struct lov_stripe_md *lsm;
+       int                cursor;
+};
+
+static int lsm_lmm_verify_common(struct lov_mds_md *lmm, int lmm_bytes,
+                                __u16 stripe_count)
+{
+       if (stripe_count > LOV_V1_INSANE_STRIPE_COUNT) {
+               CERROR("bad stripe count %d\n", stripe_count);
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+
+       if (lmm_oi_id(&lmm->lmm_oi) == 0) {
+               CERROR("zero object id\n");
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+
+       if (lov_pattern(le32_to_cpu(lmm->lmm_pattern)) != LOV_PATTERN_RAID0) {
+               CERROR("bad striping pattern\n");
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+
+       if (lmm->lmm_stripe_size == 0 ||
+           (le32_to_cpu(lmm->lmm_stripe_size)&(LOV_MIN_STRIPE_SIZE-1)) != 0) {
+               CERROR("bad stripe size %u\n",
+                      le32_to_cpu(lmm->lmm_stripe_size));
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+       return 0;
+}
+
+struct lov_stripe_md *lsm_alloc_plain(__u16 stripe_count, int *size)
+{
+       struct lov_stripe_md *lsm;
+       struct lov_oinfo     *loi;
+       int                i, oinfo_ptrs_size;
+
+       LASSERT(stripe_count <= LOV_MAX_STRIPE_COUNT);
+
+       oinfo_ptrs_size = sizeof(struct lov_oinfo *) * stripe_count;
+       *size = sizeof(struct lov_stripe_md) + oinfo_ptrs_size;
+
+       OBD_ALLOC_LARGE(lsm, *size);
+       if (!lsm)
+               return NULL;
+
+       for (i = 0; i < stripe_count; i++) {
+               OBD_SLAB_ALLOC_PTR_GFP(loi, lov_oinfo_slab, GFP_NOFS);
+               if (loi == NULL)
+                       goto err;
+               lsm->lsm_oinfo[i] = loi;
+       }
+       lsm->lsm_stripe_count = stripe_count;
+       return lsm;
+
+err:
+       while (--i >= 0)
+               OBD_SLAB_FREE(lsm->lsm_oinfo[i], lov_oinfo_slab, sizeof(*loi));
+       OBD_FREE_LARGE(lsm, *size);
+       return NULL;
+}
+
+void lsm_free_plain(struct lov_stripe_md *lsm)
+{
+       __u16 stripe_count = lsm->lsm_stripe_count;
+       int i;
+
+       for (i = 0; i < stripe_count; i++)
+               OBD_SLAB_FREE(lsm->lsm_oinfo[i], lov_oinfo_slab,
+                             sizeof(struct lov_oinfo));
+       OBD_FREE_LARGE(lsm, sizeof(struct lov_stripe_md) +
+                      stripe_count * sizeof(struct lov_oinfo *));
+}
+
+static void lsm_unpackmd_common(struct lov_stripe_md *lsm,
+                               struct lov_mds_md *lmm)
+{
+       /*
+        * This supposes lov_mds_md_v1/v3 first fields are
+        * are the same
+        */
+       lmm_oi_le_to_cpu(&lsm->lsm_oi, &lmm->lmm_oi);
+       lsm->lsm_stripe_size = le32_to_cpu(lmm->lmm_stripe_size);
+       lsm->lsm_pattern = le32_to_cpu(lmm->lmm_pattern);
+       lsm->lsm_layout_gen = le16_to_cpu(lmm->lmm_layout_gen);
+       lsm->lsm_pool_name[0] = '\0';
+}
+
+static void
+lsm_stripe_by_index_plain(struct lov_stripe_md *lsm, int *stripeno,
+                          u64 *lov_off, u64 *swidth)
+{
+       if (swidth)
+               *swidth = (u64)lsm->lsm_stripe_size * lsm->lsm_stripe_count;
+}
+
+static void
+lsm_stripe_by_offset_plain(struct lov_stripe_md *lsm, int *stripeno,
+                          u64 *lov_off, u64 *swidth)
+{
+       if (swidth)
+               *swidth = (u64)lsm->lsm_stripe_size * lsm->lsm_stripe_count;
+}
+
+static int lsm_destroy_plain(struct lov_stripe_md *lsm, struct obdo *oa,
+                            struct obd_export *md_exp)
+{
+       return 0;
+}
+
+/* Find minimum stripe maxbytes value.  For inactive or
+ * reconnecting targets use LUSTRE_STRIPE_MAXBYTES. */
+static void lov_tgt_maxbytes(struct lov_tgt_desc *tgt, __u64 *stripe_maxbytes)
+{
+       struct obd_import *imp = tgt->ltd_obd->u.cli.cl_import;
+
+       if (imp == NULL || !tgt->ltd_active) {
+               *stripe_maxbytes = LUSTRE_STRIPE_MAXBYTES;
+               return;
+       }
+
+       spin_lock(&imp->imp_lock);
+       if (imp->imp_state == LUSTRE_IMP_FULL &&
+           (imp->imp_connect_data.ocd_connect_flags & OBD_CONNECT_MAXBYTES) &&
+           imp->imp_connect_data.ocd_maxbytes > 0) {
+               if (*stripe_maxbytes > imp->imp_connect_data.ocd_maxbytes)
+                       *stripe_maxbytes = imp->imp_connect_data.ocd_maxbytes;
+       } else {
+               *stripe_maxbytes = LUSTRE_STRIPE_MAXBYTES;
+       }
+       spin_unlock(&imp->imp_lock);
+}
+
+static int lsm_lmm_verify_v1(struct lov_mds_md_v1 *lmm, int lmm_bytes,
+                            __u16 *stripe_count)
+{
+       if (lmm_bytes < sizeof(*lmm)) {
+               CERROR("lov_mds_md_v1 too small: %d, need at least %d\n",
+                      lmm_bytes, (int)sizeof(*lmm));
+               return -EINVAL;
+       }
+
+       *stripe_count = le16_to_cpu(lmm->lmm_stripe_count);
+       if (le32_to_cpu(lmm->lmm_pattern) & LOV_PATTERN_F_RELEASED)
+               *stripe_count = 0;
+
+       if (lmm_bytes < lov_mds_md_size(*stripe_count, LOV_MAGIC_V1)) {
+               CERROR("LOV EA V1 too small: %d, need %d\n",
+                      lmm_bytes, lov_mds_md_size(*stripe_count, LOV_MAGIC_V1));
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+
+       return lsm_lmm_verify_common(lmm, lmm_bytes, *stripe_count);
+}
+
+static int lsm_unpackmd_v1(struct lov_obd *lov, struct lov_stripe_md *lsm,
+                          struct lov_mds_md_v1 *lmm)
+{
+       struct lov_oinfo *loi;
+       int i;
+       int stripe_count;
+       __u64 stripe_maxbytes = OBD_OBJECT_EOF;
+
+       lsm_unpackmd_common(lsm, lmm);
+
+       stripe_count = lsm_is_released(lsm) ? 0 : lsm->lsm_stripe_count;
+
+       for (i = 0; i < stripe_count; i++) {
+               /* XXX LOV STACKING call down to osc_unpackmd() */
+               loi = lsm->lsm_oinfo[i];
+               ostid_le_to_cpu(&lmm->lmm_objects[i].l_ost_oi, &loi->loi_oi);
+               loi->loi_ost_idx = le32_to_cpu(lmm->lmm_objects[i].l_ost_idx);
+               loi->loi_ost_gen = le32_to_cpu(lmm->lmm_objects[i].l_ost_gen);
+               if (lov_oinfo_is_dummy(loi))
+                       continue;
+
+               if (loi->loi_ost_idx >= lov->desc.ld_tgt_count) {
+                       CERROR("OST index %d more than OST count %d\n",
+                              loi->loi_ost_idx, lov->desc.ld_tgt_count);
+                       lov_dump_lmm_v1(D_WARNING, lmm);
+                       return -EINVAL;
+               }
+               if (!lov->lov_tgts[loi->loi_ost_idx]) {
+                       CERROR("OST index %d missing\n", loi->loi_ost_idx);
+                       lov_dump_lmm_v1(D_WARNING, lmm);
+                       return -EINVAL;
+               }
+               /* calculate the minimum stripe max bytes */
+               lov_tgt_maxbytes(lov->lov_tgts[loi->loi_ost_idx],
+                                &stripe_maxbytes);
+       }
+
+       lsm->lsm_maxbytes = stripe_maxbytes * lsm->lsm_stripe_count;
+       if (lsm->lsm_stripe_count == 0)
+               lsm->lsm_maxbytes = stripe_maxbytes * lov->desc.ld_tgt_count;
+
+       return 0;
+}
+
+const struct lsm_operations lsm_v1_ops = {
+       .lsm_free           = lsm_free_plain,
+       .lsm_destroy     = lsm_destroy_plain,
+       .lsm_stripe_by_index    = lsm_stripe_by_index_plain,
+       .lsm_stripe_by_offset   = lsm_stripe_by_offset_plain,
+       .lsm_lmm_verify  = lsm_lmm_verify_v1,
+       .lsm_unpackmd      = lsm_unpackmd_v1,
+};
+
+static int lsm_lmm_verify_v3(struct lov_mds_md *lmmv1, int lmm_bytes,
+                            __u16 *stripe_count)
+{
+       struct lov_mds_md_v3 *lmm;
+
+       lmm = (struct lov_mds_md_v3 *)lmmv1;
+
+       if (lmm_bytes < sizeof(*lmm)) {
+               CERROR("lov_mds_md_v3 too small: %d, need at least %d\n",
+                      lmm_bytes, (int)sizeof(*lmm));
+               return -EINVAL;
+       }
+
+       *stripe_count = le16_to_cpu(lmm->lmm_stripe_count);
+       if (le32_to_cpu(lmm->lmm_pattern) & LOV_PATTERN_F_RELEASED)
+               *stripe_count = 0;
+
+       if (lmm_bytes < lov_mds_md_size(*stripe_count, LOV_MAGIC_V3)) {
+               CERROR("LOV EA V3 too small: %d, need %d\n",
+                      lmm_bytes, lov_mds_md_size(*stripe_count, LOV_MAGIC_V3));
+               lov_dump_lmm_common(D_WARNING, lmm);
+               return -EINVAL;
+       }
+
+       return lsm_lmm_verify_common((struct lov_mds_md_v1 *)lmm, lmm_bytes,
+                                    *stripe_count);
+}
+
+static int lsm_unpackmd_v3(struct lov_obd *lov, struct lov_stripe_md *lsm,
+                          struct lov_mds_md *lmmv1)
+{
+       struct lov_mds_md_v3 *lmm;
+       struct lov_oinfo *loi;
+       int i;
+       int stripe_count;
+       __u64 stripe_maxbytes = OBD_OBJECT_EOF;
+       int cplen = 0;
+
+       lmm = (struct lov_mds_md_v3 *)lmmv1;
+
+       lsm_unpackmd_common(lsm, (struct lov_mds_md_v1 *)lmm);
+
+       stripe_count = lsm_is_released(lsm) ? 0 : lsm->lsm_stripe_count;
+
+       cplen = strlcpy(lsm->lsm_pool_name, lmm->lmm_pool_name,
+                       sizeof(lsm->lsm_pool_name));
+       if (cplen >= sizeof(lsm->lsm_pool_name))
+               return -E2BIG;
+
+       for (i = 0; i < stripe_count; i++) {
+               /* XXX LOV STACKING call down to osc_unpackmd() */
+               loi = lsm->lsm_oinfo[i];
+               ostid_le_to_cpu(&lmm->lmm_objects[i].l_ost_oi, &loi->loi_oi);
+               loi->loi_ost_idx = le32_to_cpu(lmm->lmm_objects[i].l_ost_idx);
+               loi->loi_ost_gen = le32_to_cpu(lmm->lmm_objects[i].l_ost_gen);
+               if (lov_oinfo_is_dummy(loi))
+                       continue;
+
+               if (loi->loi_ost_idx >= lov->desc.ld_tgt_count) {
+                       CERROR("OST index %d more than OST count %d\n",
+                              loi->loi_ost_idx, lov->desc.ld_tgt_count);
+                       lov_dump_lmm_v3(D_WARNING, lmm);
+                       return -EINVAL;
+               }
+               if (!lov->lov_tgts[loi->loi_ost_idx]) {
+                       CERROR("OST index %d missing\n", loi->loi_ost_idx);
+                       lov_dump_lmm_v3(D_WARNING, lmm);
+                       return -EINVAL;
+               }
+               /* calculate the minimum stripe max bytes */
+               lov_tgt_maxbytes(lov->lov_tgts[loi->loi_ost_idx],
+                                &stripe_maxbytes);
+       }
+
+       lsm->lsm_maxbytes = stripe_maxbytes * lsm->lsm_stripe_count;
+       if (lsm->lsm_stripe_count == 0)
+               lsm->lsm_maxbytes = stripe_maxbytes * lov->desc.ld_tgt_count;
+
+       return 0;
+}
+
+const struct lsm_operations lsm_v3_ops = {
+       .lsm_free           = lsm_free_plain,
+       .lsm_destroy     = lsm_destroy_plain,
+       .lsm_stripe_by_index    = lsm_stripe_by_index_plain,
+       .lsm_stripe_by_offset   = lsm_stripe_by_offset_plain,
+       .lsm_lmm_verify  = lsm_lmm_verify_v3,
+       .lsm_unpackmd      = lsm_unpackmd_v3,
+};
+
+void dump_lsm(unsigned int level, const struct lov_stripe_md *lsm)
+{
+       CDEBUG(level, "lsm %p, objid " DOSTID ", maxbytes %#llx, magic 0x%08X, stripe_size %u, stripe_count %u, refc: %d, layout_gen %u, pool [" LOV_POOLNAMEF "]\n",
+              lsm,
+              POSTID(&lsm->lsm_oi), lsm->lsm_maxbytes, lsm->lsm_magic,
+              lsm->lsm_stripe_size, lsm->lsm_stripe_count,
+              atomic_read(&lsm->lsm_refc), lsm->lsm_layout_gen,
+              lsm->lsm_pool_name);
+}