X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flov%2Flov_pack.c;h=fea8ccecd17b86fa80a62ce437b7c64d8302c275;hb=0a859380c36ac24871f221b35042f76c56b04438;hp=0f0ff06ee88a6af2cbd3bfcf8f66dd90b3bebb2e;hpb=9a7ce452f70b047399dddd43397f32596acf0859;p=fs%2Flustre-release.git diff --git a/lustre/lov/lov_pack.c b/lustre/lov/lov_pack.c index 0f0ff06..fea8cce 100644 --- a/lustre/lov/lov_pack.c +++ b/lustre/lov/lov_pack.c @@ -1,26 +1,43 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * Copyright (C) 2002, 2003 Cluster File Systems, Inc. - * Author: Andreas Dilger + * GPL HEADER START * - * This file is part of Lustre, http://www.lustre.org. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * Lustre is free software; you can redistribute it and/or - * modify it under the terms of version 2 of the GNU General Public - * License as published by the Free Software Foundation. + * 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. * - * Lustre 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 for more details. + * 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 - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 2008 Sun Microsystems, Inc. All rights reserved + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * lustre/lov/lov_pack.c * * (Un)packing of OST/MDS requests * + * Author: Andreas Dilger */ #define DEBUG_SUBSYSTEM S_LOV @@ -37,44 +54,66 @@ #include "lov_internal.h" -void lov_dump_lmm_v1(int level, struct lov_mds_md_v1 *lmm) +static void lov_dump_lmm_common(int level, void *lmmp) { - struct lov_ost_data_v1 *lod; - int i; - + struct lov_mds_md *lmm = lmmp; + CDEBUG(level, "objid "LPX64", magic 0x%08x, pattern %#x\n", - le64_to_cpu(lmm->lmm_object_id), le32_to_cpu(lmm->lmm_magic), + (__u64)le64_to_cpu(lmm->lmm_object_id), + le32_to_cpu(lmm->lmm_magic), le32_to_cpu(lmm->lmm_pattern)); CDEBUG(level,"stripe_size %u, stripe_count %u\n", le32_to_cpu(lmm->lmm_stripe_size), le32_to_cpu(lmm->lmm_stripe_count)); +} - if (le32_to_cpu(lmm->lmm_stripe_count) <= LOV_V1_INSANE_STRIPE_COUNT) { - for (i = 0, lod = lmm->lmm_objects; - i < (int)le32_to_cpu(lmm->lmm_stripe_count); i++, lod++) - CDEBUG(level, - "stripe %u idx %u subobj "LPX64"/"LPX64"\n", - i, le32_to_cpu(lod->l_ost_idx), - le64_to_cpu(lod->l_object_gr), - le64_to_cpu(lod->l_object_id)); - } else { +static void lov_dump_lmm_objects(int level, struct lov_ost_data *lod, + int stripe_count) +{ + int i; + + if (stripe_count > LOV_V1_INSANE_STRIPE_COUNT) { CDEBUG(level, "bad stripe_count %u > max_stripe_count %u\n", - le32_to_cpu(lmm->lmm_stripe_count), - LOV_V1_INSANE_STRIPE_COUNT); + stripe_count, LOV_V1_INSANE_STRIPE_COUNT); + } + + for (i = 0; i < stripe_count; ++i, ++lod) { + CDEBUG(level, "stripe %u idx %u subobj "LPX64"/"LPX64"\n", i, + le32_to_cpu(lod->l_ost_idx), + (__u64)le64_to_cpu(lod->l_object_seq), + (__u64)le64_to_cpu(lod->l_object_id)); } } -void lov_dump_lmm_join(int level, struct lov_mds_md_join *lmmj) +void lov_dump_lmm_v1(int level, struct lov_mds_md_v1 *lmm) +{ + lov_dump_lmm_common(level, lmm); + lov_dump_lmm_objects(level, lmm->lmm_objects, + le32_to_cpu(lmm->lmm_stripe_count)); +} + +void lov_dump_lmm_v3(int level, struct lov_mds_md_v3 *lmm) { + lov_dump_lmm_common(level, lmm); + CDEBUG(level,"pool_name "LOV_POOLNAMEF"\n", lmm->lmm_pool_name); + lov_dump_lmm_objects(level, lmm->lmm_objects, + le32_to_cpu(lmm->lmm_stripe_count)); +} - CDEBUG(level, "objid "LPX64", magic 0x%08X, pattern %#X\n", - le64_to_cpu(lmmj->lmmj_md.lmm_object_id), - le32_to_cpu(lmmj->lmmj_md.lmm_magic), - le32_to_cpu(lmmj->lmmj_md.lmm_pattern)); - CDEBUG(level,"stripe_size %u, stripe_count %u extent_count %u \n", - le32_to_cpu(lmmj->lmmj_md.lmm_stripe_size), - le32_to_cpu(lmmj->lmmj_md.lmm_stripe_count), - le32_to_cpu(lmmj->lmmj_extent_count)); +void lov_dump_lmm(int level, void *lmm) +{ + int magic; + + magic = ((struct lov_mds_md_v1 *)(lmm))->lmm_magic; + switch (magic) { + case LOV_MAGIC_V1: + return lov_dump_lmm_v1(level, (struct lov_mds_md_v1 *)(lmm)); + case LOV_MAGIC_V3: + return lov_dump_lmm_v3(level, (struct lov_mds_md_v3 *)(lmm)); + default: + CERROR("Cannot recognize lmm_magic %x", magic); + } + return; } #define LMM_ASSERT(test) \ @@ -96,37 +135,59 @@ int lov_packmd(struct obd_export *exp, struct lov_mds_md **lmmp, { struct obd_device *obd = class_exp2obd(exp); struct lov_obd *lov = &obd->u.lov; - struct lov_mds_md *lmm; - int stripe_count = lov->desc.ld_tgt_count; - int lmm_size; + struct lov_mds_md_v1 *lmmv1; + struct lov_mds_md_v3 *lmmv3; + int stripe_count; + struct lov_ost_data_v1 *lmm_objects; + int lmm_size, lmm_magic; int i; ENTRY; if (lsm) { - if (lsm->lsm_magic != LOV_MAGIC) { - CERROR("bad mem LOV MAGIC: 0x%08X != 0x%08X\n", - lsm->lsm_magic, LOV_MAGIC); - RETURN(-EINVAL); - } + lmm_magic = lsm->lsm_magic; + /* If we are just sizing the EA, limit the stripe count * to the actual number of OSTs in this filesystem. */ if (!lmmp) { - stripe_count = lov_get_stripecnt(lov, lsm->lsm_stripe_count); + stripe_count = lov_get_stripecnt(lov, + lsm->lsm_stripe_count); lsm->lsm_stripe_count = stripe_count; } else { stripe_count = lsm->lsm_stripe_count; } + } else { + /* No needs to allocated more than LOV_MAX_STRIPE_COUNT. + * Anyway, this is pretty inaccurate since ld_tgt_count now + * represents max index and we should rely on the actual number + * of OSTs instead */ + stripe_count = min((__u32)LOV_MAX_STRIPE_COUNT, + lov->desc.ld_tgt_count); + + if (lmmp && *lmmp) + lmm_magic = le32_to_cpu((*lmmp)->lmm_magic); + else + /* lsm == NULL and lmmp == NULL */ + lmm_magic = LOV_MAGIC; + } + + if ((lmm_magic != LOV_MAGIC_V1) && + (lmm_magic != LOV_MAGIC_V3)) { + CERROR("bad mem LOV MAGIC: 0x%08X != 0x%08X nor 0x%08X\n", + lmm_magic, LOV_MAGIC_V1, LOV_MAGIC_V3); + RETURN(-EINVAL); + } /* XXX LOV STACKING call into osc for sizes */ - lmm_size = lov_mds_md_size(stripe_count); + lmm_size = lov_mds_md_size(stripe_count, lmm_magic); if (!lmmp) RETURN(lmm_size); if (*lmmp && !lsm) { stripe_count = le32_to_cpu((*lmmp)->lmm_stripe_count); - OBD_FREE(*lmmp, lov_mds_md_size(stripe_count)); + lmm_size = lov_mds_md_size(stripe_count, lmm_magic); + OBD_FREE(*lmmp, lmm_size); *lmmp = NULL; RETURN(0); } @@ -137,28 +198,44 @@ int lov_packmd(struct obd_export *exp, struct lov_mds_md **lmmp, RETURN(-ENOMEM); } - lmm = *lmmp; - lmm->lmm_magic = cpu_to_le32(LOV_MAGIC); /* only write new format */ + CDEBUG(D_INFO, "lov_packmd: LOV_MAGIC 0x%08X, lmm_size = %d \n", + lmm_magic, lmm_size); + + lmmv1 = *lmmp; + lmmv3 = (struct lov_mds_md_v3 *)*lmmp; + if (lmm_magic == LOV_MAGIC_V3) + lmmv3->lmm_magic = cpu_to_le32(LOV_MAGIC_V3); + else + lmmv1->lmm_magic = cpu_to_le32(LOV_MAGIC_V1); if (!lsm) RETURN(lmm_size); - lmm->lmm_object_id = cpu_to_le64(lsm->lsm_object_id); - lmm->lmm_object_gr = cpu_to_le64(lsm->lsm_object_gr); - lmm->lmm_stripe_size = cpu_to_le32(lsm->lsm_stripe_size); - lmm->lmm_stripe_count = cpu_to_le32(stripe_count); - lmm->lmm_pattern = cpu_to_le32(lsm->lsm_pattern); + /* lmmv1 and lmmv3 point to the same struct and have the + * same first fields + */ + lmmv1->lmm_object_id = cpu_to_le64(lsm->lsm_object_id); + lmmv1->lmm_object_seq = cpu_to_le64(lsm->lsm_object_seq); + lmmv1->lmm_stripe_size = cpu_to_le32(lsm->lsm_stripe_size); + lmmv1->lmm_stripe_count = cpu_to_le32(stripe_count); + lmmv1->lmm_pattern = cpu_to_le32(lsm->lsm_pattern); + if (lsm->lsm_magic == LOV_MAGIC_V3) { + strncpy(lmmv3->lmm_pool_name, lsm->lsm_pool_name, + LOV_MAXPOOLNAME); + lmm_objects = lmmv3->lmm_objects; + } else { + lmm_objects = lmmv1->lmm_objects; + } for (i = 0; i < stripe_count; i++) { struct lov_oinfo *loi = lsm->lsm_oinfo[i]; - /* XXX LOV STACKING call down to osc_packmd() to do packing */ LASSERTF(loi->loi_id, "lmm_oid "LPU64" stripe %u/%u idx %u\n", - lmm->lmm_object_id, i, stripe_count, loi->loi_ost_idx); - lmm->lmm_objects[i].l_object_id = cpu_to_le64(loi->loi_id); - lmm->lmm_objects[i].l_object_gr = cpu_to_le64(loi->loi_gr); - lmm->lmm_objects[i].l_ost_gen = cpu_to_le32(loi->loi_ost_gen); - lmm->lmm_objects[i].l_ost_idx = cpu_to_le32(loi->loi_ost_idx); + lmmv1->lmm_object_id, i, stripe_count, loi->loi_ost_idx); + lmm_objects[i].l_object_id = cpu_to_le64(loi->loi_id); + lmm_objects[i].l_object_seq = cpu_to_le64(loi->loi_seq); + lmm_objects[i].l_ost_gen = cpu_to_le32(loi->loi_ost_gen); + lmm_objects[i].l_ost_idx = cpu_to_le32(loi->loi_ost_idx); } RETURN(lmm_size); @@ -188,9 +265,22 @@ static int lov_verify_lmm(void *lmm, int lmm_bytes, int *stripe_count) int rc; if (lsm_op_find(le32_to_cpu(*(__u32 *)lmm)) == NULL) { - CERROR("bad disk LOV MAGIC: 0x%08X; dumping V1 LMM:\n", - le32_to_cpu(*(__u32 *)lmm)); - lov_dump_lmm_v1(D_WARNING, lmm); + char *buffer; + int sz; + + CERROR("bad disk LOV MAGIC: 0x%08X; dumping LMM (size=%d):\n", + le32_to_cpu(*(__u32 *)lmm), lmm_bytes); + sz = lmm_bytes * 2 + 1; + OBD_ALLOC(buffer, sz); + if (buffer != NULL) { + int i; + + for (i = 0; i < lmm_bytes; i++) + sprintf(buffer+2*i, "%.2X", ((char *)lmm)[i]); + buffer[sz] = '\0'; + CERROR("%s\n", buffer); + OBD_FREE(buffer, sz); + } return -EINVAL; } rc = lsm_op_find(le32_to_cpu(*(__u32 *)lmm))->lsm_lmm_verify(lmm, @@ -212,11 +302,12 @@ int lov_alloc_memmd(struct lov_stripe_md **lsmp, int stripe_count, RETURN(-ENOMEM); } - spin_lock_init(&(*lsmp)->lsm_lock); + cfs_spin_lock_init(&(*lsmp)->lsm_lock); (*lsmp)->lsm_magic = magic; (*lsmp)->lsm_stripe_count = stripe_count; (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES * stripe_count; (*lsmp)->lsm_pattern = pattern; + (*lsmp)->lsm_pool_name[0] = '\0'; (*lsmp)->lsm_oinfo[0]->loi_ost_idx = ~0; for (i = 0; i < stripe_count; i++) @@ -290,73 +381,119 @@ int lov_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp, RETURN(lsm_size); } -static int __lov_setstripe(struct obd_export *exp, struct lov_stripe_md **lsmp, +static int __lov_setstripe(struct obd_export *exp, int max_lmm_size, + struct lov_stripe_md **lsmp, struct lov_user_md *lump) { struct obd_device *obd = class_exp2obd(exp); struct lov_obd *lov = &obd->u.lov; - struct lov_user_md lum; + char buffer[sizeof(struct lov_user_md_v3)]; + struct lov_user_md_v3 *lumv3 = (struct lov_user_md_v3 *)&buffer[0]; + struct lov_user_md_v1 *lumv1 = (struct lov_user_md_v1 *)&buffer[0]; + int lmm_magic; int stripe_count; int rc; ENTRY; - rc = copy_from_user(&lum, lump, sizeof(lum)); - if (rc) + if (cfs_copy_from_user(lumv3, lump, sizeof(struct lov_user_md_v1))) RETURN(-EFAULT); - if (lum.lmm_magic != LOV_USER_MAGIC) { - if (lum.lmm_magic == __swab32(LOV_USER_MAGIC)) { - lustre_swab_lov_user_md(&lum); - } else { - CDEBUG(D_IOCTL, "bad userland LOV MAGIC:" - " %#08x != %#08x\n", - lum.lmm_magic, LOV_USER_MAGIC); - RETURN(-EINVAL); - } + lmm_magic = lumv1->lmm_magic; + + if (lmm_magic == __swab32(LOV_USER_MAGIC_V1)) { + lustre_swab_lov_user_md_v1(lumv1); + lmm_magic = LOV_USER_MAGIC_V1; + } else if (lmm_magic == LOV_USER_MAGIC_V3) { + if (cfs_copy_from_user(lumv3, lump, sizeof(*lumv3))) + RETURN(-EFAULT); + } else if (lmm_magic == __swab32(LOV_USER_MAGIC_V3)) { + if (cfs_copy_from_user(lumv3, lump, sizeof(*lumv3))) + RETURN(-EFAULT); + lustre_swab_lov_user_md_v3(lumv3); + lmm_magic = LOV_USER_MAGIC_V3; + } else if (lmm_magic != LOV_USER_MAGIC_V1) { + CDEBUG(D_IOCTL, + "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n", + lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3); + RETURN(-EINVAL); } - if (lum.lmm_pattern == 0) { - lum.lmm_pattern = lov->desc.ld_pattern ? + /* in the rest of the tests, as *lumv1 and lumv3 have the same + * fields, we use lumv1 to avoid code duplication */ + + if (lumv1->lmm_pattern == 0) { + lumv1->lmm_pattern = lov->desc.ld_pattern ? lov->desc.ld_pattern : LOV_PATTERN_RAID0; } - if (lum.lmm_pattern != LOV_PATTERN_RAID0) { + if (lumv1->lmm_pattern != LOV_PATTERN_RAID0) { CDEBUG(D_IOCTL, "bad userland stripe pattern: %#x\n", - lum.lmm_pattern); + lumv1->lmm_pattern); RETURN(-EINVAL); } /* 64kB is the largest common page size we see (ia64), and matches the * check in lfs */ - if (lum.lmm_stripe_size & (LOV_MIN_STRIPE_SIZE - 1)) { + if (lumv1->lmm_stripe_size & (LOV_MIN_STRIPE_SIZE - 1)) { CDEBUG(D_IOCTL, "stripe size %u not multiple of %u, fixing\n", - lum.lmm_stripe_size, LOV_MIN_STRIPE_SIZE); - lum.lmm_stripe_size = LOV_MIN_STRIPE_SIZE; + lumv1->lmm_stripe_size, LOV_MIN_STRIPE_SIZE); + lumv1->lmm_stripe_size = LOV_MIN_STRIPE_SIZE; } - if ((lum.lmm_stripe_offset >= lov->desc.ld_tgt_count) && - (lum.lmm_stripe_offset != (typeof(lum.lmm_stripe_offset))(-1))) { + if ((lumv1->lmm_stripe_offset >= lov->desc.ld_tgt_count) && + (lumv1->lmm_stripe_offset != + (typeof(lumv1->lmm_stripe_offset))(-1))) { CDEBUG(D_IOCTL, "stripe offset %u > number of OSTs %u\n", - lum.lmm_stripe_offset, lov->desc.ld_tgt_count); + lumv1->lmm_stripe_offset, lov->desc.ld_tgt_count); RETURN(-EINVAL); } - stripe_count = lov_get_stripecnt(lov, lum.lmm_stripe_count); + stripe_count = lov_get_stripecnt(lov, lumv1->lmm_stripe_count); + + if (max_lmm_size) { + int max_stripes = (max_lmm_size - + lov_mds_md_size(0, lmm_magic)) / + sizeof(struct lov_ost_data_v1); + if (unlikely(max_stripes < stripe_count)) { + CDEBUG(D_IOCTL, "stripe count reset from %d to %d\n", + stripe_count, max_stripes); + stripe_count = max_stripes; + } + } - if ((__u64)lum.lmm_stripe_size * stripe_count > ~0UL) { - CDEBUG(D_IOCTL, "stripe width %ux%i exeeds %lu bytes\n", - lum.lmm_stripe_size, (int)lum.lmm_stripe_count, ~0UL); - RETURN(-EINVAL); + if (lmm_magic == LOV_USER_MAGIC_V3) { + struct pool_desc *pool; + + pool = lov_find_pool(lov, lumv3->lmm_pool_name); + if (pool != NULL) { + if (lumv3->lmm_stripe_offset != + (typeof(lumv3->lmm_stripe_offset))(-1)) { + rc = lov_check_index_in_pool( + lumv3->lmm_stripe_offset, pool); + if (rc < 0) { + lov_pool_putref(pool); + RETURN(-EINVAL); + } + } + + if (stripe_count > pool_tgt_count(pool)) + stripe_count = pool_tgt_count(pool); + + lov_pool_putref(pool); + } } - rc = lov_alloc_memmd(lsmp, stripe_count, lum.lmm_pattern, LOV_MAGIC); + rc = lov_alloc_memmd(lsmp, stripe_count, lumv1->lmm_pattern, lmm_magic); if (rc >= 0) { - (*lsmp)->lsm_oinfo[0]->loi_ost_idx = lum.lmm_stripe_offset; - (*lsmp)->lsm_stripe_size = lum.lmm_stripe_size; + (*lsmp)->lsm_oinfo[0]->loi_ost_idx = lumv1->lmm_stripe_offset; + (*lsmp)->lsm_stripe_size = lumv1->lmm_stripe_size; + if (lmm_magic == LOV_USER_MAGIC_V3) + strncpy((*lsmp)->lsm_pool_name, lumv3->lmm_pool_name, + LOV_MAXPOOLNAME); rc = 0; } - RETURN(0); + RETURN(rc); } /* Configure object striping information on a new file. @@ -366,8 +503,8 @@ static int __lov_setstripe(struct obd_export *exp, struct lov_stripe_md **lsmp, * lmm_stripe_offset, and lmm_stripe_pattern. lmm_magic must be LOV_MAGIC. * @lsmp is a pointer to an in-core stripe MD that needs to be filled in. */ -int lov_setstripe(struct obd_export *exp, struct lov_stripe_md **lsmp, - struct lov_user_md *lump) +int lov_setstripe(struct obd_export *exp, int max_lmm_size, + struct lov_stripe_md **lsmp, struct lov_user_md *lump) { int rc; mm_segment_t seg; @@ -375,7 +512,7 @@ int lov_setstripe(struct obd_export *exp, struct lov_stripe_md **lsmp, seg = get_fs(); set_fs(KERNEL_DS); - rc = __lov_setstripe(exp, lsmp, lump); + rc = __lov_setstripe(exp, max_lmm_size, lsmp, lump); set_fs(seg); RETURN(rc); } @@ -388,33 +525,40 @@ int lov_setea(struct obd_export *exp, struct lov_stripe_md **lsmp, struct obd_export *oexp; struct lov_obd *lov = &exp->exp_obd->u.lov; obd_id last_id = 0; + struct lov_user_ost_data_v1 *lmm_objects; ENTRY; + + if (lump->lmm_magic == LOV_USER_MAGIC_V3) + lmm_objects = ((struct lov_user_md_v3 *)lump)->lmm_objects; + else + lmm_objects = lump->lmm_objects; + for (i = 0; i < lump->lmm_stripe_count; i++) { __u32 len = sizeof(last_id); - oexp = lov->lov_tgts[lump->lmm_objects[i].l_ost_idx]->ltd_exp; - rc = obd_get_info(oexp, strlen("last_id"), "last_id", - &len, &last_id); + oexp = lov->lov_tgts[lmm_objects[i].l_ost_idx]->ltd_exp; + rc = obd_get_info(oexp, sizeof(KEY_LAST_ID), KEY_LAST_ID, + &len, &last_id, NULL); if (rc) RETURN(rc); - if (lump->lmm_objects[i].l_object_id > last_id) { + if (lmm_objects[i].l_object_id > last_id) { CERROR("Setting EA for object > than last id on " "ost idx %d "LPD64" > "LPD64" \n", - lump->lmm_objects[i].l_ost_idx, - lump->lmm_objects[i].l_object_id, last_id); + lmm_objects[i].l_ost_idx, + lmm_objects[i].l_object_id, last_id); RETURN(-EINVAL); } } - rc = lov_setstripe(exp, lsmp, lump); + rc = lov_setstripe(exp, 0, lsmp, lump); if (rc) RETURN(rc); for (i = 0; i < lump->lmm_stripe_count; i++) { (*lsmp)->lsm_oinfo[i]->loi_ost_idx = - lump->lmm_objects[i].l_ost_idx; - (*lsmp)->lsm_oinfo[i]->loi_id = lump->lmm_objects[i].l_object_id; - (*lsmp)->lsm_oinfo[i]->loi_gr = lump->lmm_objects[i].l_object_gr; + lmm_objects[i].l_ost_idx; + (*lsmp)->lsm_oinfo[i]->loi_id = lmm_objects[i].l_object_id; + (*lsmp)->lsm_oinfo[i]->loi_seq = lmm_objects[i].l_object_seq; } RETURN(0); } @@ -432,9 +576,11 @@ int lov_getstripe(struct obd_export *exp, struct lov_stripe_md *lsm, /* * XXX huge struct allocated on stack. */ - struct lov_user_md lum; + /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */ + struct lov_user_md_v3 lum; struct lov_mds_md *lmmk = NULL; int rc, lmm_size; + int lum_size; mm_segment_t seg; ENTRY; @@ -447,34 +593,74 @@ int lov_getstripe(struct obd_export *exp, struct lov_stripe_md *lsm, */ seg = get_fs(); set_fs(KERNEL_DS); - rc = copy_from_user(&lum, lump, sizeof(lum)); - if (rc) - rc = -EFAULT; - else if (lum.lmm_magic != LOV_USER_MAGIC) - rc = -EINVAL; - else { - rc = lov_packmd(exp, &lmmk, lsm); - if (rc < 0) - RETURN(rc); - lmm_size = rc; - rc = 0; - /* FIXME: Bug 1185 - copy fields properly when structs change */ - CLASSERT(sizeof lum == sizeof *lmmk); - CLASSERT(sizeof lum.lmm_objects[0] == - sizeof lmmk->lmm_objects[0]); - - /* User wasn't expecting this many OST entries */ - if (lum.lmm_stripe_count == 0) { - if (copy_to_user(lump, lmmk, sizeof lum)) - rc = -EFAULT; - } else if (lum.lmm_stripe_count < lmmk->lmm_stripe_count) { - rc = -EOVERFLOW; - } else if (copy_to_user(lump, lmmk, sizeof lum)) - rc = -EFAULT; - - obd_free_diskmd(exp, &lmmk); + /* we only need the header part from user space to get lmm_magic and + * lmm_stripe_count, (the header part is common to v1 and v3) */ + lum_size = sizeof(struct lov_user_md_v1); + if (cfs_copy_from_user(&lum, lump, lum_size)) + GOTO(out_set, rc = -EFAULT); + else if ((lum.lmm_magic != LOV_USER_MAGIC) && + (lum.lmm_magic != LOV_USER_MAGIC_V3)) + GOTO(out_set, rc = -EINVAL); + + if (lum.lmm_stripe_count && + (lum.lmm_stripe_count < lsm->lsm_stripe_count)) { + /* Return right size of stripe to user */ + lum.lmm_stripe_count = lsm->lsm_stripe_count; + rc = cfs_copy_to_user(lump, &lum, lum_size); + GOTO(out_set, rc = -EOVERFLOW); + } + rc = lov_packmd(exp, &lmmk, lsm); + if (rc < 0) + GOTO(out_set, rc); + lmm_size = rc; + rc = 0; + + /* FIXME: Bug 1185 - copy fields properly when structs change */ + /* struct lov_user_md_v3 and struct lov_mds_md_v3 must be the same */ + CLASSERT(sizeof(lum) == sizeof(struct lov_mds_md_v3)); + CLASSERT(sizeof lum.lmm_objects[0] == sizeof lmmk->lmm_objects[0]); + + if ((cpu_to_le32(LOV_MAGIC) != LOV_MAGIC) && + ((lmmk->lmm_magic == cpu_to_le32(LOV_MAGIC_V1)) || + (lmmk->lmm_magic == cpu_to_le32(LOV_MAGIC_V3)))) { + lustre_swab_lov_mds_md(lmmk); + lustre_swab_lov_user_md_objects( + (struct lov_user_ost_data*)lmmk->lmm_objects, + lmmk->lmm_stripe_count); } + if (lum.lmm_magic == LOV_USER_MAGIC) { + /* User request for v1, we need skip lmm_pool_name */ + if (lmmk->lmm_magic == LOV_MAGIC_V3) { + memmove((char*)(&lmmk->lmm_stripe_count) + + sizeof(lmmk->lmm_stripe_count), + ((struct lov_mds_md_v3*)lmmk)->lmm_objects, + lmmk->lmm_stripe_count * + sizeof(struct lov_ost_data_v1)); + lmm_size -= LOV_MAXPOOLNAME; + } + } else { + /* if v3 we just have to update the lum_size */ + lum_size = sizeof(struct lov_user_md_v3); + } + + /* User wasn't expecting this many OST entries */ + if (lum.lmm_stripe_count == 0) + lmm_size = lum_size; + else if (lum.lmm_stripe_count < lmmk->lmm_stripe_count) + GOTO(out_set, rc = -EOVERFLOW); + /* + * Have a difference between lov_mds_md & lov_user_md. + * So we have to re-order the data before copy to user. + */ + lum.lmm_stripe_count = lmmk->lmm_stripe_count; + ((struct lov_user_md*)lmmk)->lmm_stripe_offset = 0; + ((struct lov_user_md*)lmmk)->lmm_stripe_count = lum.lmm_stripe_count; + if (cfs_copy_to_user(lump, lmmk, lmm_size)) + rc = -EFAULT; + + obd_free_diskmd(exp, &lmmk); +out_set: set_fs(seg); RETURN(rc); }