* GPL HEADER END
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*/
/*
+ * Copyright (c) 2011 Whamcloud, Inc.
+ */
+/*
* This file is part of Lustre, http://www.lustre.org/
* Lustre is a trademark of Sun Microsystems, Inc.
*
#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)
{
-
- 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));
+ 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)
{
- struct lov_ost_data_v1 *lod;
- int i;
-
- CDEBUG(level, "objid "LPX64", magic 0x%08x, pattern %#x\n",
- 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));
+ lov_dump_lmm_common(level, lmm);
CDEBUG(level,"pool_name "LOV_POOLNAMEF"\n", lmm->lmm_pool_name);
-
- 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 {
- CDEBUG(level, "bad stripe_count %u > max_stripe_count %u\n",
- le32_to_cpu(lmm->lmm_stripe_count),
- LOV_V1_INSANE_STRIPE_COUNT);
- }
+ lov_dump_lmm_objects(level, lmm->lmm_objects,
+ le32_to_cpu(lmm->lmm_stripe_count));
}
void lov_dump_lmm(int level, void *lmm)
switch (magic) {
case LOV_MAGIC_V1:
return lov_dump_lmm_v1(level, (struct lov_mds_md_v1 *)(lmm));
- case LOV_MAGIC_JOIN:
- return lov_dump_lmm_join(level, (struct lov_mds_md_join *)(lmm));
case LOV_MAGIC_V3:
return lov_dump_lmm_v3(level, (struct lov_mds_md_v3 *)(lmm));
default:
if (lsm) {
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);
- 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
}
+ if (lsm) {
+ /* 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, lmm_magic,
+ lsm->lsm_stripe_count);
+ lsm->lsm_stripe_count = stripe_count;
+ } else {
+ stripe_count = lsm->lsm_stripe_count;
+ }
+ } else {
+ /* No need to allocate more than maximum supported stripes.
+ * 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 = lov_mds_md_stripecnt(lov->lov_ocd.ocd_max_easize,
+ lmm_magic);
+ if (stripe_count > lov->desc.ld_tgt_count)
+ stripe_count = lov->desc.ld_tgt_count;
+ }
+
/* XXX LOV STACKING call into osc for sizes */
lmm_size = lov_mds_md_size(stripe_count, lmm_magic);
if (*lmmp && !lsm) {
stripe_count = le32_to_cpu((*lmmp)->lmm_stripe_count);
lmm_size = lov_mds_md_size(stripe_count, lmm_magic);
- OBD_FREE(*lmmp, lmm_size);
+ OBD_FREE_LARGE(*lmmp, lmm_size);
*lmmp = NULL;
RETURN(0);
}
if (!*lmmp) {
- OBD_ALLOC(*lmmp, lmm_size);
+ OBD_ALLOC_LARGE(*lmmp, lmm_size);
if (!*lmmp)
RETURN(-ENOMEM);
}
* same first fields
*/
lmmv1->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
- lmmv1->lmm_object_gr = cpu_to_le64(lsm->lsm_object_gr);
+ 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);
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",
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_gr = cpu_to_le64(loi->loi_gr);
+ 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);
}
}
/* Find the max stripecount we should use */
-int lov_get_stripecnt(struct lov_obd *lov, __u32 stripe_count)
+__u32 lov_get_stripecnt(struct lov_obd *lov, __u32 magic, __u32 stripe_count)
{
+ __u32 max_stripes = LOV_MAX_STRIPE_COUNT_OLD;
+
if (!stripe_count)
stripe_count = lov->desc.ld_default_stripe_count;
if (stripe_count > lov->desc.ld_active_tgt_count)
stripe_count = lov->desc.ld_active_tgt_count;
if (!stripe_count)
stripe_count = 1;
- /* for now, we limit the stripe count directly, when bug 4424 is
- * fixed this needs to be somewhat dynamic based on whether ext3
- * can handle larger EA sizes. */
- if (stripe_count > LOV_MAX_STRIPE_COUNT)
- stripe_count = LOV_MAX_STRIPE_COUNT;
+
+ /* stripe count is based on whether ldiskfs can handle
+ * larger EA sizes */
+ if (lov->lov_ocd.ocd_connect_flags & OBD_CONNECT_MAX_EASIZE &&
+ lov->lov_ocd.ocd_max_easize)
+ max_stripes = lov_mds_md_stripecnt(lov->lov_ocd.ocd_max_easize,
+ magic);
+
+ if (stripe_count > max_stripes)
+ stripe_count = max_stripes;
return stripe_count;
}
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);
+ OBD_ALLOC_LARGE(buffer, sz);
if (buffer != NULL) {
int i;
sprintf(buffer+2*i, "%.2X", ((char *)lmm)[i]);
buffer[sz] = '\0';
CERROR("%s\n", buffer);
- OBD_FREE(buffer, sz);
+ OBD_FREE_LARGE(buffer, sz);
}
return -EINVAL;
}
}
int lov_alloc_memmd(struct lov_stripe_md **lsmp, int stripe_count,
- int pattern, int magic)
+ int pattern, int magic)
{
int i, lsm_size;
ENTRY;
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;
RETURN(rc);
magic = le32_to_cpu(lmm->lmm_magic);
} else {
- stripe_count = lov_get_stripecnt(lov, 0);
magic = LOV_MAGIC;
+ stripe_count = lov_get_stripecnt(lov, magic, 0);
}
/* If we aren't passed an lsmp struct, we just want the size */
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_v3 lumv3;
- struct lov_user_md_v1 *lumv1 = (struct lov_user_md_v1 *)&lumv3;
+ 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;
- if (copy_from_user(&lumv3, lump, sizeof(struct lov_user_md_v1)))
+ if (cfs_copy_from_user(lumv3, lump, sizeof(struct lov_user_md_v1)))
RETURN(-EFAULT);
lmm_magic = lumv1->lmm_magic;
lustre_swab_lov_user_md_v1(lumv1);
lmm_magic = LOV_USER_MAGIC_V1;
} else if (lmm_magic == LOV_USER_MAGIC_V3) {
- if (copy_from_user(&lumv3, lump, sizeof(lumv3)))
+ if (cfs_copy_from_user(lumv3, lump, sizeof(*lumv3)))
RETURN(-EFAULT);
} else if (lmm_magic == __swab32(LOV_USER_MAGIC_V3)) {
- if (copy_from_user(&lumv3, lump, sizeof(lumv3)))
+ if (cfs_copy_from_user(lumv3, lump, sizeof(*lumv3)))
RETURN(-EFAULT);
- lustre_swab_lov_user_md_v3(&lumv3);
+ lustre_swab_lov_user_md_v3(lumv3);
lmm_magic = LOV_USER_MAGIC_V3;
} else if (lmm_magic != LOV_USER_MAGIC_V1) {
CDEBUG(D_IOCTL,
lumv1->lmm_stripe_offset, lov->desc.ld_tgt_count);
RETURN(-EINVAL);
}
- stripe_count = lov_get_stripecnt(lov, lumv1->lmm_stripe_count);
+ stripe_count = lov_get_stripecnt(lov, lmm_magic,
+ 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 (lmm_magic == LOV_USER_MAGIC_V3) {
struct pool_desc *pool;
- pool = lov_find_pool(lov, lumv3.lmm_pool_name);
- if (pool == NULL)
- RETURN(-EINVAL);
-
- 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)
- RETURN(-EINVAL);
+ 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);
}
-
- if (stripe_count > pool_tgt_count(pool))
- stripe_count = pool_tgt_count(pool);
- }
-
- if ((__u64)lumv1->lmm_stripe_size * stripe_count > ~0UL) {
- CDEBUG(D_IOCTL, "stripe width %ux%i exeeds %lu bytes\n",
- lumv1->lmm_stripe_size, (int)lumv1->lmm_stripe_count,
- ~0UL);
- RETURN(-EINVAL);
}
rc = lov_alloc_memmd(lsmp, stripe_count, lumv1->lmm_pattern, lmm_magic);
(*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,
+ strncpy((*lsmp)->lsm_pool_name, lumv3->lmm_pool_name,
LOV_MAXPOOLNAME);
rc = 0;
}
* 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;
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);
}
}
}
- rc = lov_setstripe(exp, lsmp, lump);
+ rc = lov_setstripe(exp, 0, lsmp, lump);
if (rc)
RETURN(rc);
(*lsmp)->lsm_oinfo[i]->loi_ost_idx =
lmm_objects[i].l_ost_idx;
(*lsmp)->lsm_oinfo[i]->loi_id = lmm_objects[i].l_object_id;
- (*lsmp)->lsm_oinfo[i]->loi_gr = lmm_objects[i].l_object_gr;
+ (*lsmp)->lsm_oinfo[i]->loi_seq = lmm_objects[i].l_object_seq;
}
RETURN(0);
}
/* 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 (copy_from_user(&lum, lump, lum_size))
- rc = -EFAULT;
+ 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))
- rc = -EINVAL;
- else {
+ 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 */
- if (lum.lmm_magic == LOV_USER_MAGIC_V3)
- lum_size = sizeof(struct lov_user_md_v3);
+ lum_size = sizeof(struct lov_user_md_v3);
+ }
- rc = lov_packmd(exp, &lmmk, lsm);
- if (rc < 0)
- RETURN(rc);
- lmm_size = rc;
- rc = 0;
+ /* 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;
- /* 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]);
-
- /* User wasn't expecting this many OST entries */
- if (lum.lmm_stripe_count == 0) {
- if (copy_to_user(lump, lmmk, lum_size))
- rc = -EFAULT;
- } else if (lum.lmm_stripe_count < lmmk->lmm_stripe_count) {
- rc = -EOVERFLOW;
- } else if (copy_to_user(lump, lmmk, lmm_size))
- rc = -EFAULT;
-
- obd_free_diskmd(exp, &lmmk);
- }
+ obd_free_diskmd(exp, &lmmk);
+out_set:
set_fs(seg);
RETURN(rc);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff