* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2013, Intel Corporation.
+ * Copyright (c) 2011, 2015, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include <libcfs/list.h>
/* fid_be_to_cpu() */
#include <lustre_fid.h>
-
+#include <lustre_nodemap.h>
#include <lustre_quota.h>
/* context key constructor/destructor: dt_global_key_init, dt_global_key_fini */
.lct_init = dt_global_key_init,
.lct_fini = dt_global_key_fini
};
-EXPORT_SYMBOL(dt_key);
/* no lock is necessary to protect the list, because call-backs
* are added during system startup. Please refer to "struct dt_device".
*/
void dt_txn_callback_add(struct dt_device *dev, struct dt_txn_callback *cb)
{
- cfs_list_add(&cb->dtc_linkage, &dev->dd_txn_callbacks);
+ list_add(&cb->dtc_linkage, &dev->dd_txn_callbacks);
}
EXPORT_SYMBOL(dt_txn_callback_add);
void dt_txn_callback_del(struct dt_device *dev, struct dt_txn_callback *cb)
{
- cfs_list_del_init(&cb->dtc_linkage);
+ list_del_init(&cb->dtc_linkage);
}
EXPORT_SYMBOL(dt_txn_callback_del);
int dt_txn_hook_start(const struct lu_env *env,
struct dt_device *dev, struct thandle *th)
{
- int rc = 0;
- struct dt_txn_callback *cb;
+ int rc = 0;
+ struct dt_txn_callback *cb;
- if (th->th_local)
- return 0;
+ if (th->th_local)
+ return 0;
- cfs_list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
- if (cb->dtc_txn_start == NULL ||
- !(cb->dtc_tag & env->le_ctx.lc_tags))
- continue;
- rc = cb->dtc_txn_start(env, th, cb->dtc_cookie);
- if (rc < 0)
- break;
- }
- return rc;
+ list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
+ struct thandle *dtc_th = th;
+
+ if (cb->dtc_txn_start == NULL ||
+ !(cb->dtc_tag & env->le_ctx.lc_tags))
+ continue;
+
+ /* Usually dt_txn_hook_start is called from bottom device,
+ * and if the thandle has th_top, then we need use top
+ * thandle for the callback in the top thandle layer */
+ if (th->th_top != NULL)
+ dtc_th = th->th_top;
+
+ rc = cb->dtc_txn_start(env, dtc_th, cb->dtc_cookie);
+ if (rc < 0)
+ break;
+ }
+ return rc;
}
EXPORT_SYMBOL(dt_txn_hook_start);
-int dt_txn_hook_stop(const struct lu_env *env, struct thandle *txn)
+int dt_txn_hook_stop(const struct lu_env *env, struct thandle *th)
{
- struct dt_device *dev = txn->th_dev;
- struct dt_txn_callback *cb;
- int rc = 0;
+ struct dt_device *dev = th->th_dev;
+ struct dt_txn_callback *cb;
+ int rc = 0;
- if (txn->th_local)
- return 0;
+ if (th->th_local)
+ return 0;
- cfs_list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
- if (cb->dtc_txn_stop == NULL ||
- !(cb->dtc_tag & env->le_ctx.lc_tags))
- continue;
- rc = cb->dtc_txn_stop(env, txn, cb->dtc_cookie);
- if (rc < 0)
- break;
- }
- return rc;
+ list_for_each_entry(cb, &dev->dd_txn_callbacks, dtc_linkage) {
+ struct thandle *dtc_th = th;
+
+ if (cb->dtc_txn_stop == NULL ||
+ !(cb->dtc_tag & env->le_ctx.lc_tags))
+ continue;
+
+ /* Usually dt_txn_hook_stop is called from bottom device,
+ * and if the thandle has th_top, then we need use top
+ * thandle for the callback in the top thandle layer */
+ if (th->th_top != NULL)
+ dtc_th = th->th_top;
+
+ rc = cb->dtc_txn_stop(env, dtc_th, cb->dtc_cookie);
+ if (rc < 0)
+ break;
+ }
+ return rc;
}
EXPORT_SYMBOL(dt_txn_hook_stop);
-void dt_txn_hook_commit(struct thandle *txn)
+void dt_txn_hook_commit(struct thandle *th)
{
- struct dt_txn_callback *cb;
-
- if (txn->th_local)
- return;
-
- cfs_list_for_each_entry(cb, &txn->th_dev->dd_txn_callbacks,
- dtc_linkage) {
- if (cb->dtc_txn_commit)
- cb->dtc_txn_commit(txn, cb->dtc_cookie);
- }
+ struct dt_txn_callback *cb;
+
+ if (th->th_local)
+ return;
+
+ list_for_each_entry(cb, &th->th_dev->dd_txn_callbacks,
+ dtc_linkage) {
+ /* Right now, the bottom device (OSD) will use this hook
+ * commit to notify OSP, so we do not check and replace
+ * the thandle to top thandle now */
+ if (cb->dtc_txn_commit)
+ cb->dtc_txn_commit(th, cb->dtc_cookie);
+ }
}
EXPORT_SYMBOL(dt_txn_hook_commit);
int dt_device_init(struct dt_device *dev, struct lu_device_type *t)
{
-
- CFS_INIT_LIST_HEAD(&dev->dd_txn_callbacks);
- return lu_device_init(&dev->dd_lu_dev, t);
+ INIT_LIST_HEAD(&dev->dd_txn_callbacks);
+ return lu_device_init(&dev->dd_lu_dev, t);
}
EXPORT_SYMBOL(dt_device_init);
int dt_lookup_dir(const struct lu_env *env, struct dt_object *dir,
const char *name, struct lu_fid *fid)
{
- if (dt_try_as_dir(env, dir))
- return dt_lookup(env, dir, (struct dt_rec *)fid,
- (const struct dt_key *)name, BYPASS_CAPA);
- return -ENOTDIR;
+ if (dt_try_as_dir(env, dir))
+ return dt_lookup(env, dir, (struct dt_rec *)fid,
+ (const struct dt_key *)name);
+ return -ENOTDIR;
}
EXPORT_SYMBOL(dt_lookup_dir);
EXPORT_SYMBOL(dt_locate_at);
/**
- * find a object named \a entry in given \a dfh->dfh_o directory.
+ * find an object named \a entry in given \a dfh->dfh_o directory.
*/
static int dt_find_entry(const struct lu_env *env, const char *entry, void *data)
{
struct dt_thread_info *info = dt_info(env);
struct dt_find_hint *dfh = &info->dti_dfh;
struct dt_object *obj;
- char *local = info->dti_buf;
int result;
dfh->dfh_dt = dt;
dfh->dfh_fid = fid;
- strncpy(local, path, DT_MAX_PATH);
- local[DT_MAX_PATH - 1] = '\0';
+ strlcpy(info->dti_buf, path, sizeof(info->dti_buf));
result = dt->dd_ops->dt_root_get(env, dt, fid);
if (result == 0) {
obj = dt_locate(env, dt, fid);
if (!IS_ERR(obj)) {
dfh->dfh_o = obj;
- result = dt_path_parser(env, local, dt_find_entry, dfh);
+ result = dt_path_parser(env, info->dti_buf,
+ dt_find_entry, dfh);
if (result != 0)
obj = ERR_PTR(result);
else
}
return obj;
}
-EXPORT_SYMBOL(dt_store_resolve);
static struct dt_object *dt_reg_open(const struct lu_env *env,
struct dt_device *dt,
}
return file;
}
-EXPORT_SYMBOL(dt_store_open);
struct dt_object *dt_find_or_create(const struct lu_env *env,
struct dt_device *dt,
int dt_read(const struct lu_env *env, struct dt_object *dt,
struct lu_buf *buf, loff_t *pos)
{
- LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
- return dt->do_body_ops->dbo_read(env, dt, buf, pos, BYPASS_CAPA);
+ LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
+ return dt->do_body_ops->dbo_read(env, dt, buf, pos);
}
EXPORT_SYMBOL(dt_read);
int dt_record_read(const struct lu_env *env, struct dt_object *dt,
struct lu_buf *buf, loff_t *pos)
{
- int rc;
+ ssize_t size;
- LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
+ LASSERTF(dt != NULL, "dt is NULL when we want to read record\n");
- rc = dt->do_body_ops->dbo_read(env, dt, buf, pos, BYPASS_CAPA);
-
- if (rc == buf->lb_len)
- rc = 0;
- else if (rc >= 0)
- rc = -EFAULT;
- return rc;
+ size = dt->do_body_ops->dbo_read(env, dt, buf, pos);
+ if (size < 0)
+ return size;
+ return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
}
EXPORT_SYMBOL(dt_record_read);
int dt_record_write(const struct lu_env *env, struct dt_object *dt,
const struct lu_buf *buf, loff_t *pos, struct thandle *th)
{
- int rc;
+ ssize_t size;
- LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
- LASSERT(th != NULL);
- LASSERT(dt->do_body_ops);
- LASSERT(dt->do_body_ops->dbo_write);
- rc = dt->do_body_ops->dbo_write(env, dt, buf, pos, th, BYPASS_CAPA, 1);
- if (rc == buf->lb_len)
- rc = 0;
- else if (rc >= 0)
- rc = -EFAULT;
- return rc;
+ LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
+ LASSERT(th != NULL);
+ LASSERT(dt->do_body_ops);
+ LASSERT(dt->do_body_ops->dbo_write);
+
+ size = dt->do_body_ops->dbo_write(env, dt, buf, pos, th, 1);
+ if (size < 0)
+ return size;
+ return (size == (ssize_t)buf->lb_len) ? 0 : -EFAULT;
}
EXPORT_SYMBOL(dt_record_write);
vbuf.lb_buf = &version;
vbuf.lb_len = sizeof(version);
- rc = dt_xattr_set(env, o, &vbuf, xname, 0, th, BYPASS_CAPA);
+ rc = dt_xattr_set(env, o, &vbuf, xname, 0, th);
if (rc < 0)
CDEBUG(D_INODE, "Can't set version, rc %d\n", rc);
return;
LASSERT(o);
vbuf.lb_buf = &version;
vbuf.lb_len = sizeof(version);
- rc = dt_xattr_get(env, o, &vbuf, xname, BYPASS_CAPA);
+ rc = dt_xattr_get(env, o, &vbuf, xname);
if (rc != sizeof(version)) {
CDEBUG(D_INODE, "Can't get version, rc %d\n", rc);
version = 0;
};
EXPORT_SYMBOL(dt_quota_slv_features);
+/* nodemap files, nodemap_rec size asserted in nodemap_storage.c */
+const struct dt_index_features dt_nodemap_features = {
+ .dif_flags = DT_IND_UPDATE,
+ .dif_keysize_min = sizeof(__u64), /* 64-bit nodemap/record id */
+ .dif_keysize_max = sizeof(__u64), /* 64-bit nodemap/record id */
+ .dif_recsize_min = sizeof(union nodemap_rec), /* 32 bytes */
+ .dif_recsize_max = sizeof(union nodemap_rec), /* 32 bytes */
+ .dif_ptrsize = 4
+};
+EXPORT_SYMBOL(dt_nodemap_features);
+
/* helper function returning what dt_index_features structure should be used
* based on the FID sequence. This is used by OBD_IDX_READ RPC */
static inline const struct dt_index_features *dt_index_feat_select(__u64 seq,
* \param arg - is a pointer to the idx_info structure
*/
static int dt_index_page_build(const struct lu_env *env, union lu_page *lp,
- int nob, const struct dt_it_ops *iops,
+ size_t nob, const struct dt_it_ops *iops,
struct dt_it *it, __u32 attr, void *arg)
{
struct idx_info *ii = (struct idx_info *)arg;
struct lu_idxpage *lip = &lp->lp_idx;
char *entry;
- int rc, size;
+ size_t size;
+ int rc;
ENTRY;
- /* no support for variable key & record size for now */
- LASSERT((ii->ii_flags & II_FL_VARKEY) == 0);
- LASSERT((ii->ii_flags & II_FL_VARREC) == 0);
+ if (nob < LIP_HDR_SIZE)
+ return -EINVAL;
/* initialize the header of the new container */
memset(lip, 0, LIP_HDR_SIZE);
do {
char *tmp_entry = entry;
struct dt_key *key;
- __u64 hash;
+ __u64 hash;
+ __u16 keysize;
+ __u16 recsize;
/* fetch 64-bit hash value */
hash = iops->store(env, it);
GOTO(out, rc = 0);
}
- if ((ii->ii_flags & II_FL_NOHASH) == 0) {
+ if (!(ii->ii_flags & II_FL_NOHASH)) {
/* client wants to the 64-bit hash value associated with
* each record */
memcpy(tmp_entry, &hash, sizeof(hash));
tmp_entry += sizeof(hash);
}
- /* then the key value */
- LASSERT(iops->key_size(env, it) == ii->ii_keysize);
- key = iops->key(env, it);
- memcpy(tmp_entry, key, ii->ii_keysize);
- tmp_entry += ii->ii_keysize;
+ if (ii->ii_flags & II_FL_VARKEY)
+ keysize = iops->key_size(env, it);
+ else
+ keysize = ii->ii_keysize;
+
+ if (!(ii->ii_flags & II_FL_NOKEY)) {
+ /* then the key value */
+ key = iops->key(env, it);
+ memcpy(tmp_entry, key, keysize);
+ tmp_entry += keysize;
+ }
/* and finally the record */
rc = iops->rec(env, it, (struct dt_rec *)tmp_entry, attr);
lip->lip_nr++;
if (unlikely(lip->lip_nr == 1 && ii->ii_count == 0))
ii->ii_hash_start = hash;
- entry = tmp_entry + ii->ii_recsize;
+
+ if (ii->ii_flags & II_FL_VARREC)
+ recsize = iops->rec_size(env, it, attr);
+ else
+ recsize = ii->ii_recsize;
+
+ entry = tmp_entry + recsize;
nob -= size;
}
return rc;
}
+
/*
* Walk index and fill lu_page containers with key/record pairs
*
* \param obj - is the index object to parse
* \param rdpg - is the lu_rdpg descriptor associated with the transfer
* \param filler - is the callback function responsible for filling a lu_page
- * with key/record pairs in the format wanted by the caller
+ * with key/record pairs in the format wanted by the caller.
+ * If NULL, uses dt_index_page_build
* \param arg - is an opaq argument passed to the filler function
*
* \retval sum (in bytes) of all filled lu_pages
{
struct dt_it *it;
const struct dt_it_ops *iops;
- unsigned int pageidx, nob, nlupgs = 0;
+ size_t pageidx, nob, nlupgs = 0;
int rc;
ENTRY;
LASSERT(rdpg->rp_pages != NULL);
LASSERT(obj->do_index_ops != NULL);
+ if (filler == NULL)
+ filler = dt_index_page_build;
+
nob = rdpg->rp_count;
- if (nob <= 0)
+ if (nob == 0)
RETURN(-EFAULT);
/* Iterate through index and fill containers from @rdpg */
iops = &obj->do_index_ops->dio_it;
LASSERT(iops != NULL);
- it = iops->init(env, obj, rdpg->rp_attrs, BYPASS_CAPA);
+ it = iops->init(env, obj, rdpg->rp_attrs);
if (IS_ERR(it))
RETURN(PTR_ERR(it));
rc = iops->next(env, it);
} else if (rc > 0) {
rc = 0;
+ } else {
+ if (rc == -ENODATA)
+ rc = 0;
+ GOTO(out, rc);
}
/* Fill containers one after the other. There might be multiple
/* fill lu pages */
for (i = 0; i < LU_PAGE_COUNT; i++, lp++, nob -= LU_PAGE_SIZE) {
- rc = filler(env, lp, min_t(int, nob, LU_PAGE_SIZE),
+ rc = filler(env, lp, min_t(size_t, nob, LU_PAGE_SIZE),
iops, it, rdpg->rp_attrs, arg);
if (rc < 0)
break;
kunmap(rdpg->rp_pages[i]);
}
+out:
iops->put(env, it);
iops->fini(env, it);
if (rc >= 0)
- rc = min_t(unsigned int, nlupgs * LU_PAGE_SIZE, rdpg->rp_count);
+ rc = min_t(size_t, nlupgs * LU_PAGE_SIZE, rdpg->rp_count);
RETURN(rc);
}
/* rp_count shouldn't be null and should be a multiple of the container
* size */
- if (rdpg->rp_count <= 0 && (rdpg->rp_count & (LU_PAGE_SIZE - 1)) != 0)
+ if (rdpg->rp_count == 0 || (rdpg->rp_count & (LU_PAGE_SIZE - 1)) != 0)
RETURN(-EFAULT);
- if (fid_seq(&ii->ii_fid) >= FID_SEQ_NORMAL)
- /* we don't support directory transfer via OBD_IDX_READ for the
- * time being */
+ if (!fid_is_quota(&ii->ii_fid) && !fid_is_layout_rbtree(&ii->ii_fid) &&
+ !fid_is_norm(&ii->ii_fid))
RETURN(-EOPNOTSUPP);
- if (!fid_is_quota(&ii->ii_fid) && !fid_is_layout_rbtree(&ii->ii_fid))
- /* Block access to all local files except quota files and
- * layout rbtree. */
- RETURN(-EPERM);
-
/* lookup index object subject to the transfer */
obj = dt_locate(env, dev, &ii->ii_fid);
if (IS_ERR(obj))
}
/* fill ii_flags with supported index features */
- ii->ii_flags &= II_FL_NOHASH;
-
- ii->ii_keysize = feat->dif_keysize_max;
- if ((feat->dif_flags & DT_IND_VARKEY) != 0) {
- /* key size is variable */
- ii->ii_flags |= II_FL_VARKEY;
- /* we don't support variable key size for the time being */
- GOTO(out, rc = -EOPNOTSUPP);
- }
+ ii->ii_flags &= (II_FL_NOHASH | II_FL_NOKEY | II_FL_VARKEY |
+ II_FL_VARREC);
- ii->ii_recsize = feat->dif_recsize_max;
- if ((feat->dif_flags & DT_IND_VARREC) != 0) {
- /* record size is variable */
- ii->ii_flags |= II_FL_VARREC;
- /* we don't support variable record size for the time being */
- GOTO(out, rc = -EOPNOTSUPP);
- }
+ if (!(feat->dif_flags & DT_IND_VARKEY))
+ ii->ii_keysize = feat->dif_keysize_max;
- if ((feat->dif_flags & DT_IND_NONUNQ) != 0)
+ if (!(feat->dif_flags & DT_IND_VARREC))
+ ii->ii_recsize = feat->dif_recsize_max;
+
+ if (feat->dif_flags & DT_IND_NONUNQ)
/* key isn't necessarily unique */
ii->ii_flags |= II_FL_NONUNQ;
}
/* walk the index and fill lu_idxpages with key/record pairs */
- rc = dt_index_walk(env, obj, rdpg, dt_index_page_build ,ii);
+ rc = dt_index_walk(env, obj, rdpg, dt_index_page_build, ii);
if (!fid_is_layout_rbtree(&ii->ii_fid))
dt_read_unlock(env, obj);
}
EXPORT_SYMBOL(dt_index_read);
-#ifdef LPROCFS
-#ifndef HAVE_ONLY_PROCFS_SEQ
-int lprocfs_dt_rd_blksize(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- *eof = 1;
- rc = snprintf(page, count, "%u\n",
- (unsigned) osfs.os_bsize);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_blksize);
-
-int lprocfs_dt_rd_kbytestotal(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- __u32 blk_size = osfs.os_bsize >> 10;
- __u64 result = osfs.os_blocks;
-
- while (blk_size >>= 1)
- result <<= 1;
-
- *eof = 1;
- rc = snprintf(page, count, LPU64"\n", result);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_kbytestotal);
-
-int lprocfs_dt_rd_kbytesfree(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- __u32 blk_size = osfs.os_bsize >> 10;
- __u64 result = osfs.os_bfree;
-
- while (blk_size >>= 1)
- result <<= 1;
-
- *eof = 1;
- rc = snprintf(page, count, LPU64"\n", result);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_kbytesfree);
-
-int lprocfs_dt_rd_kbytesavail(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- __u32 blk_size = osfs.os_bsize >> 10;
- __u64 result = osfs.os_bavail;
-
- while (blk_size >>= 1)
- result <<= 1;
-
- *eof = 1;
- rc = snprintf(page, count, LPU64"\n", result);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_kbytesavail);
-
-int lprocfs_dt_rd_filestotal(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- *eof = 1;
- rc = snprintf(page, count, LPU64"\n", osfs.os_files);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_filestotal);
-
-int lprocfs_dt_rd_filesfree(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct dt_device *dt = data;
- struct obd_statfs osfs;
-
- int rc = dt_statfs(NULL, dt, &osfs);
- if (rc == 0) {
- *eof = 1;
- rc = snprintf(page, count, LPU64"\n", osfs.os_ffree);
- }
-
- return rc;
-}
-EXPORT_SYMBOL(lprocfs_dt_rd_filesfree);
-#endif
-
+#ifdef CONFIG_PROC_FS
int lprocfs_dt_blksize_seq_show(struct seq_file *m, void *v)
{
struct dt_device *dt = m->private;
}
EXPORT_SYMBOL(lprocfs_dt_filesfree_seq_show);
-#endif /* LPROCFS */
+#endif /* CONFIG_PROC_FS */