[PATCH 25/41] staging: lustre: lov: add cl_object_layout_get()

From: James Simmons
Date: Sun Oct 02 2016 - 22:30:32 EST


From: John L. Hammond <john.hammond@xxxxxxxxx>

Add cl_object_layout_get() to return the layout and generation of an
object. Replace some direct accesses to object LSM with calls to this
function.

In ll_getxattr() factor out the LOV xattr specific handling into a new
function ll_getxattr_lov() which calls cl_object_layout_get(). In
ll_listxattr() call ll_getxattr_lov() to determine if a lustre.lov
xattr should be emitted. Add lov_lsm_pack() to generate LOV xattrs
from a LSM.

Remove the unused functions ccc_inode_lsm_{get,put}() and
lov_lsm_get().

Signed-off-by: John L. Hammond <john.hammond@xxxxxxxxx>
Signed-off-by: Jinshan Xiong <jinshan.xiong@xxxxxxxxx>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5814
Reviewed-on: http://review.whamcloud.com/13680
Reviewed-by: Bobi Jam <bobijam@xxxxxxxxxxx>
Reviewed-by: Oleg Drokin <oleg.drokin@xxxxxxxxx>
Signed-off-by: James Simmons <jsimmons@xxxxxxxxxxxxx>
---
drivers/staging/lustre/lustre/include/cl_object.h | 27 +++
.../lustre/lustre/include/lustre/lustre_user.h | 3 +
drivers/staging/lustre/lustre/llite/file.c | 149 +++++++------
drivers/staging/lustre/lustre/llite/lcommon_cl.c | 19 --
.../staging/lustre/lustre/llite/llite_internal.h | 5 -
drivers/staging/lustre/lustre/llite/llite_lib.c | 33 +++-
drivers/staging/lustre/lustre/llite/vvp_internal.h | 13 -
drivers/staging/lustre/lustre/llite/vvp_object.c | 4 +-
drivers/staging/lustre/lustre/llite/xattr.c | 234 ++++++++++----------
drivers/staging/lustre/lustre/lov/lov_internal.h | 2 +
drivers/staging/lustre/lustre/lov/lov_object.c | 51 +++--
drivers/staging/lustre/lustre/lov/lov_pack.c | 99 +++++----
drivers/staging/lustre/lustre/obdclass/cl_object.c | 14 ++
13 files changed, 364 insertions(+), 289 deletions(-)

diff --git a/drivers/staging/lustre/lustre/include/cl_object.h b/drivers/staging/lustre/lustre/include/cl_object.h
index e46a510..b80539d 100644
--- a/drivers/staging/lustre/lustre/include/cl_object.h
+++ b/drivers/staging/lustre/lustre/include/cl_object.h
@@ -301,6 +301,26 @@ enum {
OBJECT_CONF_WAIT = 2
};

+enum {
+ CL_LAYOUT_GEN_NONE = (u32)-2, /* layout lock was cancelled */
+ CL_LAYOUT_GEN_EMPTY = (u32)-1, /* for empty layout */
+};
+
+struct cl_layout {
+ /** the buffer to return the layout in lov_mds_md format. */
+ struct lu_buf cl_buf;
+ /** size of layout in lov_mds_md format. */
+ size_t cl_size;
+ /** Layout generation. */
+ u32 cl_layout_gen;
+ /**
+ * True if this is a released file.
+ * Temporarily added for released file truncate in ll_setattr_raw().
+ * It will be removed later. -Jinshan
+ */
+ bool cl_is_released;
+};
+
/**
* Operations implemented for each cl object layer.
*
@@ -406,6 +426,11 @@ struct cl_object_operations {
int (*coo_fiemap)(const struct lu_env *env, struct cl_object *obj,
struct ll_fiemap_info_key *fmkey,
struct fiemap *fiemap, size_t *buflen);
+ /**
+ * Get layout and generation of the object.
+ */
+ int (*coo_layout_get)(const struct lu_env *env, struct cl_object *obj,
+ struct cl_layout *layout);
};

/**
@@ -2200,6 +2225,8 @@ int cl_object_getstripe(const struct lu_env *env, struct cl_object *obj,
int cl_object_fiemap(const struct lu_env *env, struct cl_object *obj,
struct ll_fiemap_info_key *fmkey, struct fiemap *fiemap,
size_t *buflen);
+int cl_object_layout_get(const struct lu_env *env, struct cl_object *obj,
+ struct cl_layout *cl);

/**
* Returns true, iff \a o0 and \a o1 are slices of the same object.
diff --git a/drivers/staging/lustre/lustre/include/lustre/lustre_user.h b/drivers/staging/lustre/lustre/include/lustre/lustre_user.h
index 043fc1c..80fecba 100644
--- a/drivers/staging/lustre/lustre/include/lustre/lustre_user.h
+++ b/drivers/staging/lustre/lustre/include/lustre/lustre_user.h
@@ -346,6 +346,9 @@ enum ll_lease_type {
#define LOV_ALL_STRIPES 0xffff /* only valid for directories */
#define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */

+#define XATTR_LUSTRE_PREFIX "lustre."
+#define XATTR_LUSTRE_LOV "lustre.lov"
+
#define lov_user_ost_data lov_user_ost_data_v1
struct lov_user_ost_data_v1 { /* per-stripe data structure */
struct ost_id l_ost_oi; /* OST object ID */
diff --git a/drivers/staging/lustre/lustre/llite/file.c b/drivers/staging/lustre/lustre/llite/file.c
index 8fa65a5..73ea446 100644
--- a/drivers/staging/lustre/lustre/llite/file.c
+++ b/drivers/staging/lustre/lustre/llite/file.c
@@ -3187,35 +3187,51 @@ ll_iocontrol_call(struct inode *inode, struct file *file,
int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf)
{
struct ll_inode_info *lli = ll_i2info(inode);
+ struct cl_object *obj = lli->lli_clob;
struct cl_env_nest nest;
struct lu_env *env;
- int result;
+ int rc;

- if (!lli->lli_clob)
+ if (!obj)
return 0;

env = cl_env_nested_get(&nest);
if (IS_ERR(env))
return PTR_ERR(env);

- result = cl_conf_set(env, lli->lli_clob, conf);
- cl_env_nested_put(&nest, env);
+ rc = cl_conf_set(env, obj, conf);
+ if (rc < 0)
+ goto out;

if (conf->coc_opc == OBJECT_CONF_SET) {
struct ldlm_lock *lock = conf->coc_lock;
+ struct cl_layout cl = {
+ .cl_layout_gen = 0,
+ };

LASSERT(lock);
LASSERT(ldlm_has_layout(lock));
- if (result == 0) {
- /* it can only be allowed to match after layout is
- * applied to inode otherwise false layout would be
- * seen. Applying layout should happen before dropping
- * the intent lock.
- */
- ldlm_lock_allow_match(lock);
- }
+
+ /* it can only be allowed to match after layout is
+ * applied to inode otherwise false layout would be
+ * seen. Applying layout should happen before dropping
+ * the intent lock.
+ */
+ ldlm_lock_allow_match(lock);
+
+ rc = cl_object_layout_get(env, obj, &cl);
+ if (rc < 0)
+ goto out;
+
+ CDEBUG(D_VFSTRACE, DFID ": layout version change: %u -> %u\n",
+ PFID(&lli->lli_fid), ll_layout_version_get(lli),
+ cl.cl_layout_gen);
+ ll_layout_version_set(lli, cl.cl_layout_gen);
+ lli->lli_has_smd = lsm_has_objects(conf->u.coc_md->lsm);
}
- return result;
+out:
+ cl_env_nested_put(&nest, env);
+ return rc;
}

/* Fetch layout from MDT with getxattr request, if it's not ready yet */
@@ -3294,7 +3310,7 @@ out:
* in this function.
*/
static int ll_layout_lock_set(struct lustre_handle *lockh, enum ldlm_mode mode,
- struct inode *inode, __u32 *gen, bool reconf)
+ struct inode *inode)
{
struct ll_inode_info *lli = ll_i2info(inode);
struct ll_sb_info *sbi = ll_i2sbi(inode);
@@ -3311,8 +3327,8 @@ static int ll_layout_lock_set(struct lustre_handle *lockh, enum ldlm_mode mode,
LASSERT(lock);
LASSERT(ldlm_has_layout(lock));

- LDLM_DEBUG(lock, "File "DFID"(%p) being reconfigured: %d",
- PFID(&lli->lli_fid), inode, reconf);
+ LDLM_DEBUG(lock, "File " DFID "(%p) being reconfigured",
+ PFID(&lli->lli_fid), inode);

/* in case this is a caching lock and reinstate with new inode */
md_set_lock_data(sbi->ll_md_exp, lockh, inode, NULL);
@@ -3323,15 +3339,8 @@ static int ll_layout_lock_set(struct lustre_handle *lockh, enum ldlm_mode mode,
/* checking lvb_ready is racy but this is okay. The worst case is
* that multi processes may configure the file on the same time.
*/
- if (lvb_ready || !reconf) {
- rc = -ENODATA;
- if (lvb_ready) {
- /* layout_gen must be valid if layout lock is not
- * cancelled and stripe has already set
- */
- *gen = ll_layout_version_get(lli);
- rc = 0;
- }
+ if (lvb_ready) {
+ rc = 0;
goto out;
}

@@ -3347,19 +3356,17 @@ static int ll_layout_lock_set(struct lustre_handle *lockh, enum ldlm_mode mode,
if (lock->l_lvb_data) {
rc = obd_unpackmd(sbi->ll_dt_exp, &md.lsm,
lock->l_lvb_data, lock->l_lvb_len);
- if (rc >= 0) {
- *gen = LL_LAYOUT_GEN_EMPTY;
- if (md.lsm)
- *gen = md.lsm->lsm_layout_gen;
- rc = 0;
- } else {
+ if (rc < 0) {
CERROR("%s: file " DFID " unpackmd error: %d\n",
ll_get_fsname(inode->i_sb, NULL, 0),
PFID(&lli->lli_fid), rc);
+ goto out;
}
+
+ LASSERTF(md.lsm, "lvb_data = %p, lvb_len = %u\n",
+ lock->l_lvb_data, lock->l_lvb_len);
+ rc = 0;
}
- if (rc < 0)
- goto out;

/* set layout to file. Unlikely this will fail as old layout was
* surely eliminated
@@ -3401,20 +3408,7 @@ out:
return rc;
}

-/**
- * This function checks if there exists a LAYOUT lock on the client side,
- * or enqueues it if it doesn't have one in cache.
- *
- * This function will not hold layout lock so it may be revoked any time after
- * this function returns. Any operations depend on layout should be redone
- * in that case.
- *
- * This function should be called before lov_io_init() to get an uptodate
- * layout version, the caller should save the version number and after IO
- * is finished, this function should be called again to verify that layout
- * is not changed during IO time.
- */
-int ll_layout_refresh(struct inode *inode, __u32 *gen)
+static int ll_layout_refresh_locked(struct inode *inode)
{
struct ll_inode_info *lli = ll_i2info(inode);
struct ll_sb_info *sbi = ll_i2sbi(inode);
@@ -3430,17 +3424,6 @@ int ll_layout_refresh(struct inode *inode, __u32 *gen)
};
int rc;

- *gen = ll_layout_version_get(lli);
- if (!(sbi->ll_flags & LL_SBI_LAYOUT_LOCK) || *gen != LL_LAYOUT_GEN_NONE)
- return 0;
-
- /* sanity checks */
- LASSERT(fid_is_sane(ll_inode2fid(inode)));
- LASSERT(S_ISREG(inode->i_mode));
-
- /* take layout lock mutex to enqueue layout lock exclusively. */
- mutex_lock(&lli->lli_layout_mutex);
-
again:
/* mostly layout lock is caching on the local side, so try to match
* it before grabbing layout lock mutex.
@@ -3448,20 +3431,16 @@ again:
mode = ll_take_md_lock(inode, MDS_INODELOCK_LAYOUT, &lockh, 0,
LCK_CR | LCK_CW | LCK_PR | LCK_PW);
if (mode != 0) { /* hit cached lock */
- rc = ll_layout_lock_set(&lockh, mode, inode, gen, true);
+ rc = ll_layout_lock_set(&lockh, mode, inode);
if (rc == -EAGAIN)
goto again;
-
- mutex_unlock(&lli->lli_layout_mutex);
return rc;
}

op_data = ll_prep_md_op_data(NULL, inode, inode, NULL,
0, 0, LUSTRE_OPC_ANY, NULL);
- if (IS_ERR(op_data)) {
- mutex_unlock(&lli->lli_layout_mutex);
+ if (IS_ERR(op_data))
return PTR_ERR(op_data);
- }

/* have to enqueue one */
memset(&it, 0, sizeof(it));
@@ -3485,10 +3464,50 @@ again:
if (rc == 0) {
/* set lock data in case this is a new lock */
ll_set_lock_data(sbi->ll_md_exp, inode, &it, NULL);
- rc = ll_layout_lock_set(&lockh, mode, inode, gen, true);
+ rc = ll_layout_lock_set(&lockh, mode, inode);
if (rc == -EAGAIN)
goto again;
}
+
+ return rc;
+}
+
+/**
+ * This function checks if there exists a LAYOUT lock on the client side,
+ * or enqueues it if it doesn't have one in cache.
+ *
+ * This function will not hold layout lock so it may be revoked any time after
+ * this function returns. Any operations depend on layout should be redone
+ * in that case.
+ *
+ * This function should be called before lov_io_init() to get an uptodate
+ * layout version, the caller should save the version number and after IO
+ * is finished, this function should be called again to verify that layout
+ * is not changed during IO time.
+ */
+int ll_layout_refresh(struct inode *inode, __u32 *gen)
+{
+ struct ll_inode_info *lli = ll_i2info(inode);
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+ int rc;
+
+ *gen = ll_layout_version_get(lli);
+ if (!(sbi->ll_flags & LL_SBI_LAYOUT_LOCK) || *gen != CL_LAYOUT_GEN_NONE)
+ return 0;
+
+ /* sanity checks */
+ LASSERT(fid_is_sane(ll_inode2fid(inode)));
+ LASSERT(S_ISREG(inode->i_mode));
+
+ /* take layout lock mutex to enqueue layout lock exclusively. */
+ mutex_lock(&lli->lli_layout_mutex);
+
+ rc = ll_layout_refresh_locked(inode);
+ if (rc < 0)
+ goto out;
+
+ *gen = ll_layout_version_get(lli);
+out:
mutex_unlock(&lli->lli_layout_mutex);

return rc;
diff --git a/drivers/staging/lustre/lustre/llite/lcommon_cl.c b/drivers/staging/lustre/lustre/llite/lcommon_cl.c
index 64f4aed..bd98ec2 100644
--- a/drivers/staging/lustre/lustre/llite/lcommon_cl.c
+++ b/drivers/staging/lustre/lustre/llite/lcommon_cl.c
@@ -304,22 +304,3 @@ __u32 cl_fid_build_gen(const struct lu_fid *fid)
gen = fid_flatten(fid) >> 32;
return gen;
}
-
-/* lsm is unreliable after hsm implementation as layout can be changed at
- * any time. This is only to support old, non-clio-ized interfaces. It will
- * cause deadlock if clio operations are called with this extra layout refcount
- * because in case the layout changed during the IO, ll_layout_refresh() will
- * have to wait for the refcount to become zero to destroy the older layout.
- *
- * Notice that the lsm returned by this function may not be valid unless called
- * inside layout lock - MDS_INODELOCK_LAYOUT.
- */
-struct lov_stripe_md *ccc_inode_lsm_get(struct inode *inode)
-{
- return lov_lsm_get(ll_i2info(inode)->lli_clob);
-}
-
-inline void ccc_inode_lsm_put(struct inode *inode, struct lov_stripe_md *lsm)
-{
- lov_lsm_put(ll_i2info(inode)->lli_clob, lsm);
-}
diff --git a/drivers/staging/lustre/lustre/llite/llite_internal.h b/drivers/staging/lustre/lustre/llite/llite_internal.h
index e249895..c89e1b8 100644
--- a/drivers/staging/lustre/lustre/llite/llite_internal.h
+++ b/drivers/staging/lustre/lustre/llite/llite_internal.h
@@ -1327,11 +1327,6 @@ static inline void d_lustre_revalidate(struct dentry *dentry)
spin_unlock(&dentry->d_lock);
}

-enum {
- LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
- LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
-};
-
int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
int ll_layout_refresh(struct inode *inode, __u32 *gen);
int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
diff --git a/drivers/staging/lustre/lustre/llite/llite_lib.c b/drivers/staging/lustre/lustre/llite/llite_lib.c
index 6270301..4b53119 100644
--- a/drivers/staging/lustre/lustre/llite/llite_lib.c
+++ b/drivers/staging/lustre/lustre/llite/llite_lib.c
@@ -800,7 +800,7 @@ void ll_lli_init(struct ll_inode_info *lli)
spin_lock_init(&lli->lli_agl_lock);
lli->lli_has_smd = false;
spin_lock_init(&lli->lli_layout_lock);
- ll_layout_version_set(lli, LL_LAYOUT_GEN_NONE);
+ ll_layout_version_set(lli, CL_LAYOUT_GEN_NONE);
lli->lli_clob = NULL;

init_rwsem(&lli->lli_xattrs_list_rwsem);
@@ -1441,14 +1441,33 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import)
* but other attributes must be set
*/
if (S_ISREG(inode->i_mode)) {
- struct lov_stripe_md *lsm;
+ struct cl_layout cl = {
+ .cl_is_released = false,
+ };
+ struct lu_env *env;
+ int refcheck;
__u32 gen;

- ll_layout_refresh(inode, &gen);
- lsm = ccc_inode_lsm_get(inode);
- if (lsm && lsm->lsm_pattern & LOV_PATTERN_F_RELEASED)
- file_is_released = true;
- ccc_inode_lsm_put(inode, lsm);
+ rc = ll_layout_refresh(inode, &gen);
+ if (rc < 0)
+ goto out;
+
+ /*
+ * XXX: the only place we need to know the layout type,
+ * this will be removed by a later patch. -Jinshan
+ */
+ env = cl_env_get(&refcheck);
+ if (IS_ERR(env)) {
+ rc = PTR_ERR(env);
+ goto out;
+ }
+
+ rc = cl_object_layout_get(env, lli->lli_clob, &cl);
+ cl_env_put(env, &refcheck);
+ if (rc < 0)
+ goto out;
+
+ file_is_released = cl.cl_is_released;

if (!hsm_import && attr->ia_valid & ATTR_SIZE) {
if (file_is_released) {
diff --git a/drivers/staging/lustre/lustre/llite/vvp_internal.h b/drivers/staging/lustre/lustre/llite/vvp_internal.h
index 47d035e..a025b35 100644
--- a/drivers/staging/lustre/lustre/llite/vvp_internal.h
+++ b/drivers/staging/lustre/lustre/llite/vvp_internal.h
@@ -323,21 +323,8 @@ static inline struct vvp_lock *cl2vvp_lock(const struct cl_lock_slice *slice)
# define CLOBINVRNT(env, clob, expr) \
((void)sizeof(env), (void)sizeof(clob), (void)sizeof(!!(expr)))

-/**
- * New interfaces to get and put lov_stripe_md from lov layer. This violates
- * layering because lov_stripe_md is supposed to be a private data in lov.
- *
- * NB: If you find you have to use these interfaces for your new code, please
- * think about it again. These interfaces may be removed in the future for
- * better layering.
- */
-struct lov_stripe_md *lov_lsm_get(struct cl_object *clobj);
-void lov_lsm_put(struct cl_object *clobj, struct lov_stripe_md *lsm);
int lov_read_and_clear_async_rc(struct cl_object *clob);

-struct lov_stripe_md *ccc_inode_lsm_get(struct inode *inode);
-void ccc_inode_lsm_put(struct inode *inode, struct lov_stripe_md *lsm);
-
int vvp_io_init(const struct lu_env *env, struct cl_object *obj,
struct cl_io *io);
int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
diff --git a/drivers/staging/lustre/lustre/llite/vvp_object.c b/drivers/staging/lustre/lustre/llite/vvp_object.c
index 3214885..420a649 100644
--- a/drivers/staging/lustre/lustre/llite/vvp_object.c
+++ b/drivers/staging/lustre/lustre/llite/vvp_object.c
@@ -132,7 +132,7 @@ static int vvp_conf_set(const struct lu_env *env, struct cl_object *obj,
CDEBUG(D_VFSTRACE, DFID ": losing layout lock\n",
PFID(&lli->lli_fid));

- ll_layout_version_set(lli, LL_LAYOUT_GEN_NONE);
+ ll_layout_version_set(lli, CL_LAYOUT_GEN_NONE);

/* Clean up page mmap for this inode.
* The reason for us to do this is that if the page has
@@ -164,7 +164,7 @@ static int vvp_conf_set(const struct lu_env *env, struct cl_object *obj,
PFID(&lli->lli_fid), lli->lli_layout_gen);

lli->lli_has_smd = false;
- ll_layout_version_set(lli, LL_LAYOUT_GEN_EMPTY);
+ ll_layout_version_set(lli, CL_LAYOUT_GEN_EMPTY);
}
return 0;
}
diff --git a/drivers/staging/lustre/lustre/llite/xattr.c b/drivers/staging/lustre/lustre/llite/xattr.c
index e070adb..b8b5c32 100644
--- a/drivers/staging/lustre/lustre/llite/xattr.c
+++ b/drivers/staging/lustre/lustre/llite/xattr.c
@@ -353,80 +353,99 @@ static int ll_xattr_get_common(const struct xattr_handler *handler,
OBD_MD_FLXATTR);
}

-static int ll_xattr_get(const struct xattr_handler *handler,
- struct dentry *dentry, struct inode *inode,
- const char *name, void *buffer, size_t size)
+static ssize_t ll_getxattr_lov(struct inode *inode, void *buf, size_t buf_size)
{
- LASSERT(inode);
- LASSERT(name);
+ ssize_t rc;

- CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
- PFID(ll_inode2fid(inode)), inode, name);
-
- if (!strcmp(name, "lov")) {
- struct lov_stripe_md *lsm;
- struct lov_user_md *lump;
- struct lov_mds_md *lmm = NULL;
- struct ptlrpc_request *request = NULL;
- int rc = 0, lmmsize = 0;
+ if (S_ISREG(inode->i_mode)) {
+ struct cl_object *obj = ll_i2info(inode)->lli_clob;
+ struct cl_layout cl = {
+ .cl_buf.lb_buf = buf,
+ .cl_buf.lb_len = buf_size,
+ };
+ struct lu_env *env;
+ int refcheck;
+
+ if (!obj)
+ return -ENODATA;

- ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_GETXATTR, 1);
+ env = cl_env_get(&refcheck);
+ if (IS_ERR(env))
+ return PTR_ERR(env);

- if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
- return -ENODATA;
+ rc = cl_object_layout_get(env, obj, &cl);
+ if (rc < 0)
+ goto out_env;

- lsm = ccc_inode_lsm_get(inode);
- if (!lsm) {
- if (S_ISDIR(inode->i_mode)) {
- rc = ll_dir_getstripe(inode, (void **)&lmm,
- &lmmsize, &request, 0);
- } else {
- rc = -ENODATA;
- }
- } else {
- /* LSM is present already after lookup/getattr call.
- * we need to grab layout lock once it is implemented
- */
- rc = obd_packmd(ll_i2dtexp(inode), &lmm, lsm);
- lmmsize = rc;
+ if (!cl.cl_size) {
+ rc = -ENODATA;
+ goto out_env;
}
- ccc_inode_lsm_put(inode, lsm);

+ rc = cl.cl_size;
+
+ if (!buf_size)
+ goto out_env;
+
+ LASSERT(buf && rc <= buf_size);
+
+ /*
+ * Do not return layout gen for getxattr() since
+ * otherwise it would confuse tar --xattr by
+ * recognizing layout gen as stripe offset when the
+ * file is restored. See LU-2809.
+ */
+ ((struct lov_mds_md *)buf)->lmm_layout_gen = 0;
+out_env:
+ cl_env_put(env, &refcheck);
+
+ return rc;
+ } else if (S_ISDIR(inode->i_mode)) {
+ struct ptlrpc_request *req = NULL;
+ struct lov_mds_md *lmm = NULL;
+ int lmm_size = 0;
+
+ rc = ll_dir_getstripe(inode, (void **)&lmm, &lmm_size,
+ &req, 0);
if (rc < 0)
- goto out;
+ goto out_req;

- if (size == 0) {
- /* used to call ll_get_max_mdsize() forward to get
- * the maximum buffer size, while some apps (such as
- * rsync 3.0.x) care much about the exact xattr value
- * size
- */
- rc = lmmsize;
- goto out;
+ if (!buf_size) {
+ rc = lmm_size;
+ goto out_req;
}

- if (size < lmmsize) {
- CERROR("server bug: replied size %d > %d for %pd (%s)\n",
- lmmsize, (int)size, dentry, name);
+ if (buf_size < lmm_size) {
rc = -ERANGE;
- goto out;
+ goto out_req;
}

- lump = buffer;
- memcpy(lump, lmm, lmmsize);
- /* do not return layout gen for getxattr otherwise it would
- * confuse tar --xattr by recognizing layout gen as stripe
- * offset when the file is restored. See LU-2809.
- */
- lump->lmm_layout_gen = 0;
+ memcpy(buf, lmm, lmm_size);
+ rc = lmm_size;
+out_req:
+ if (req)
+ ptlrpc_req_finished(req);

- rc = lmmsize;
-out:
- if (request)
- ptlrpc_req_finished(request);
- else if (lmm)
- obd_free_diskmd(ll_i2dtexp(inode), &lmm);
return rc;
+ } else {
+ return -ENODATA;
+ }
+}
+
+static int ll_xattr_get(const struct xattr_handler *handler,
+ struct dentry *dentry, struct inode *inode,
+ const char *name, void *buffer, size_t size)
+{
+ LASSERT(inode);
+ LASSERT(name);
+
+ CDEBUG(D_VFSTRACE, "VFS Op:inode=" DFID "(%p), xattr %s\n",
+ PFID(ll_inode2fid(inode)), inode, name);
+
+ if (!strcmp(name, "lov")) {
+ ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_GETXATTR, 1);
+
+ return ll_getxattr_lov(inode, buffer, size);
}

return ll_xattr_get_common(handler, dentry, inode, name, buffer, size);
@@ -435,10 +454,10 @@ out:
ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
struct inode *inode = d_inode(dentry);
- int rc = 0, rc2 = 0;
- struct lov_mds_md *lmm = NULL;
- struct ptlrpc_request *request = NULL;
- int lmmsize;
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+ char *xattr_name;
+ ssize_t rc, rc2;
+ size_t len, rem;

LASSERT(inode);

@@ -450,65 +469,48 @@ ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size)
rc = ll_xattr_list(inode, NULL, XATTR_OTHER_T, buffer, size,
OBD_MD_FLXATTRLS);
if (rc < 0)
- goto out;
-
- if (buffer) {
- struct ll_sb_info *sbi = ll_i2sbi(inode);
- char *xattr_name = buffer;
- int xlen, rem = rc;
-
- while (rem > 0) {
- xlen = strnlen(xattr_name, rem - 1) + 1;
- rem -= xlen;
- if (xattr_type_filter(sbi,
- get_xattr_type(xattr_name)) == 0) {
- /* skip OK xattr type
- * leave it in buffer
- */
- xattr_name += xlen;
- continue;
- }
- /* move up remaining xattrs in buffer
- * removing the xattr that is not OK
- */
- memmove(xattr_name, xattr_name + xlen, rem);
- rc -= xlen;
+ return rc;
+ /*
+ * If we're being called to get the size of the xattr list
+ * (buf_size == 0) then just assume that a lustre.lov xattr
+ * exists.
+ */
+ if (!size)
+ return rc + sizeof(XATTR_LUSTRE_LOV);
+
+ xattr_name = buffer;
+ rem = rc;
+
+ while (rem > 0) {
+ len = strnlen(xattr_name, rem - 1) + 1;
+ rem -= len;
+ if (!xattr_type_filter(sbi, get_xattr_type(xattr_name))) {
+ /* Skip OK xattr type leave it in buffer */
+ xattr_name += len;
+ continue;
}
- }
- if (S_ISREG(inode->i_mode)) {
- if (!ll_i2info(inode)->lli_has_smd)
- rc2 = -1;
- } else if (S_ISDIR(inode->i_mode)) {
- rc2 = ll_dir_getstripe(inode, (void **)&lmm, &lmmsize,
- &request, 0);
+
+ /*
+ * Move up remaining xattrs in buffer
+ * removing the xattr that is not OK
+ */
+ memmove(xattr_name, xattr_name + len, rem);
+ rc -= len;
}

- if (rc2 < 0) {
- rc2 = 0;
- goto out;
- } else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) {
- const int prefix_len = sizeof(XATTR_LUSTRE_PREFIX) - 1;
- const size_t name_len = sizeof("lov") - 1;
- const size_t total_len = prefix_len + name_len + 1;
-
- if (((rc + total_len) > size) && buffer) {
- ptlrpc_req_finished(request);
- return -ERANGE;
- }
+ rc2 = ll_getxattr_lov(inode, NULL, 0);
+ if (rc2 == -ENODATA)
+ return rc;

- if (buffer) {
- buffer += rc;
- memcpy(buffer, XATTR_LUSTRE_PREFIX, prefix_len);
- memcpy(buffer + prefix_len, "lov", name_len);
- buffer[prefix_len + name_len] = '\0';
- }
- rc2 = total_len;
- }
-out:
- ptlrpc_req_finished(request);
- rc = rc + rc2;
+ if (rc2 < 0)
+ return rc2;

- return rc;
+ if (size < rc + sizeof(XATTR_LUSTRE_LOV))
+ return -ERANGE;
+
+ memcpy(buffer + rc, XATTR_LUSTRE_LOV, sizeof(XATTR_LUSTRE_LOV));
+
+ return rc + sizeof(XATTR_LUSTRE_LOV);
}

static const struct xattr_handler ll_user_xattr_handler = {
diff --git a/drivers/staging/lustre/lustre/lov/lov_internal.h b/drivers/staging/lustre/lustre/lov/lov_internal.h
index fffc18c..60397a2 100644
--- a/drivers/staging/lustre/lustre/lov/lov_internal.h
+++ b/drivers/staging/lustre/lustre/lov/lov_internal.h
@@ -176,6 +176,8 @@ int lov_del_target(struct obd_device *obd, __u32 index,
struct obd_uuid *uuidp, int gen);

/* lov_pack.c */
+ssize_t lov_lsm_pack(const struct lov_stripe_md *lsm, void *buf,
+ size_t buf_size);
int lov_packmd(struct obd_export *exp, struct lov_mds_md **lmm,
struct lov_stripe_md *lsm);
int lov_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
diff --git a/drivers/staging/lustre/lustre/lov/lov_object.c b/drivers/staging/lustre/lustre/lov/lov_object.c
index 07bef44..d39724a 100644
--- a/drivers/staging/lustre/lustre/lov/lov_object.c
+++ b/drivers/staging/lustre/lustre/lov/lov_object.c
@@ -75,12 +75,11 @@ struct lov_layout_operations {

static int lov_layout_wait(const struct lu_env *env, struct lov_object *lov);

-void lov_lsm_put(struct cl_object *unused, struct lov_stripe_md *lsm)
+static void lov_lsm_put(struct lov_stripe_md *lsm)
{
if (lsm)
lov_free_memmd(&lsm);
}
-EXPORT_SYMBOL(lov_lsm_put);

/*****************************************************************************
*
@@ -1408,7 +1407,7 @@ obj_put:
cl_object_put(env, subobj);
out:
kvfree(fm_local);
- lov_lsm_put(obj, lsm);
+ lov_lsm_put(lsm);
return rc;
}

@@ -1424,10 +1423,37 @@ static int lov_object_getstripe(const struct lu_env *env, struct cl_object *obj,
return -ENODATA;

rc = lov_getstripe(cl2lov(obj), lsm, lum);
- lov_lsm_put(obj, lsm);
+ lov_lsm_put(lsm);
return rc;
}

+static int lov_object_layout_get(const struct lu_env *env,
+ struct cl_object *obj,
+ struct cl_layout *cl)
+{
+ struct lov_object *lov = cl2lov(obj);
+ struct lov_stripe_md *lsm = lov_lsm_addref(lov);
+ struct lu_buf *buf = &cl->cl_buf;
+ ssize_t rc;
+
+ if (!lsm) {
+ cl->cl_size = 0;
+ cl->cl_layout_gen = CL_LAYOUT_GEN_EMPTY;
+ cl->cl_is_released = false;
+
+ return 0;
+ }
+
+ cl->cl_size = lov_mds_md_size(lsm->lsm_stripe_count, lsm->lsm_magic);
+ cl->cl_layout_gen = lsm->lsm_layout_gen;
+ cl->cl_is_released = lsm_is_released(lsm);
+
+ rc = lov_lsm_pack(lsm, buf->lb_buf, buf->lb_len);
+ lov_lsm_put(lsm);
+
+ return rc < 0 ? rc : 0;
+}
+
static const struct cl_object_operations lov_ops = {
.coo_page_init = lov_page_init,
.coo_lock_init = lov_lock_init,
@@ -1436,6 +1462,7 @@ static const struct cl_object_operations lov_ops = {
.coo_attr_update = lov_attr_update,
.coo_conf_set = lov_conf_set,
.coo_getstripe = lov_object_getstripe,
+ .coo_layout_get = lov_object_layout_get,
.coo_fiemap = lov_object_fiemap,
};

@@ -1488,22 +1515,6 @@ struct lov_stripe_md *lov_lsm_addref(struct lov_object *lov)
return lsm;
}

-struct lov_stripe_md *lov_lsm_get(struct cl_object *clobj)
-{
- struct lu_object *luobj;
- struct lov_stripe_md *lsm = NULL;
-
- if (!clobj)
- return NULL;
-
- luobj = lu_object_locate(&cl_object_header(clobj)->coh_lu,
- &lov_device_type);
- if (luobj)
- lsm = lov_lsm_addref(lu2lov(luobj));
- return lsm;
-}
-EXPORT_SYMBOL(lov_lsm_get);
-
int lov_read_and_clear_async_rc(struct cl_object *clob)
{
struct lu_object *luobj;
diff --git a/drivers/staging/lustre/lustre/lov/lov_pack.c b/drivers/staging/lustre/lustre/lov/lov_pack.c
index be6e985..1156ef9 100644
--- a/drivers/staging/lustre/lustre/lov/lov_pack.c
+++ b/drivers/staging/lustre/lustre/lov/lov_pack.c
@@ -97,6 +97,62 @@ void lov_dump_lmm_v3(int level, struct lov_mds_md_v3 *lmm)
le16_to_cpu(lmm->lmm_stripe_count));
}

+/**
+ * Pack LOV striping metadata for disk storage format (in little
+ * endian byte order).
+ *
+ * This follows the getxattr() conventions. If \a buf_size is zero
+ * then return the size needed. If \a buf_size is too small then
+ * return -ERANGE. Otherwise return the size of the result.
+ */
+ssize_t lov_lsm_pack(const struct lov_stripe_md *lsm, void *buf,
+ size_t buf_size)
+{
+ struct lov_ost_data_v1 *lmm_objects;
+ struct lov_mds_md_v1 *lmmv1 = buf;
+ struct lov_mds_md_v3 *lmmv3 = buf;
+ size_t lmm_size;
+ unsigned int i;
+
+ lmm_size = lov_mds_md_size(lsm->lsm_stripe_count, lsm->lsm_magic);
+ if (!buf_size)
+ return lmm_size;
+
+ if (buf_size < lmm_size)
+ return -ERANGE;
+
+ /*
+ * lmmv1 and lmmv3 point to the same struct and have the
+ * same first fields
+ */
+ lmmv1->lmm_magic = cpu_to_le32(lsm->lsm_magic);
+ lmm_oi_cpu_to_le(&lmmv1->lmm_oi, &lsm->lsm_oi);
+ lmmv1->lmm_stripe_size = cpu_to_le32(lsm->lsm_stripe_size);
+ lmmv1->lmm_stripe_count = cpu_to_le16(lsm->lsm_stripe_count);
+ lmmv1->lmm_pattern = cpu_to_le32(lsm->lsm_pattern);
+ lmmv1->lmm_layout_gen = cpu_to_le16(lsm->lsm_layout_gen);
+
+ if (lsm->lsm_magic == LOV_MAGIC_V3) {
+ CLASSERT(sizeof(lsm->lsm_pool_name) ==
+ sizeof(lmmv3->lmm_pool_name));
+ strlcpy(lmmv3->lmm_pool_name, lsm->lsm_pool_name,
+ sizeof(lmmv3->lmm_pool_name));
+ lmm_objects = lmmv3->lmm_objects;
+ } else {
+ lmm_objects = lmmv1->lmm_objects;
+ }
+
+ for (i = 0; i < lsm->lsm_stripe_count; i++) {
+ struct lov_oinfo *loi = lsm->lsm_oinfo[i];
+
+ ostid_cpu_to_le(&loi->loi_oi, &lmm_objects[i].l_ost_oi);
+ 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;
+}
+
/* Pack LOV object metadata for disk storage. It is packed in LE byte
* order and is opaque to the networking layer.
*
@@ -108,13 +164,8 @@ void lov_dump_lmm_v3(int level, struct lov_mds_md_v3 *lmm)
int lov_obd_packmd(struct lov_obd *lov, struct lov_mds_md **lmmp,
struct lov_stripe_md *lsm)
{
- struct lov_mds_md_v1 *lmmv1;
- struct lov_mds_md_v3 *lmmv3;
__u16 stripe_count;
- struct lov_ost_data_v1 *lmm_objects;
int lmm_size, lmm_magic;
- int i;
- int cplen = 0;

if (lsm) {
lmm_magic = lsm->lsm_magic;
@@ -177,46 +228,10 @@ int lov_obd_packmd(struct lov_obd *lov, struct lov_mds_md **lmmp,
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;

- /* lmmv1 and lmmv3 point to the same struct and have the
- * same first fields
- */
- lmm_oi_cpu_to_le(&lmmv1->lmm_oi, &lsm->lsm_oi);
- lmmv1->lmm_stripe_size = cpu_to_le32(lsm->lsm_stripe_size);
- lmmv1->lmm_stripe_count = cpu_to_le16(stripe_count);
- lmmv1->lmm_pattern = cpu_to_le32(lsm->lsm_pattern);
- lmmv1->lmm_layout_gen = cpu_to_le16(lsm->lsm_layout_gen);
- if (lsm->lsm_magic == LOV_MAGIC_V3) {
- cplen = strlcpy(lmmv3->lmm_pool_name, lsm->lsm_pool_name,
- sizeof(lmmv3->lmm_pool_name));
- if (cplen >= sizeof(lmmv3->lmm_pool_name))
- return -E2BIG;
- 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(ostid_id(&loi->loi_oi) != 0, "lmm_oi "DOSTID
- " stripe %u/%u idx %u\n", POSTID(&lmmv1->lmm_oi),
- i, stripe_count, loi->loi_ost_idx);
- ostid_cpu_to_le(&loi->loi_oi, &lmm_objects[i].l_ost_oi);
- 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;
+ return lov_lsm_pack(lsm, *lmmp, lmm_size);
}

int lov_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
diff --git a/drivers/staging/lustre/lustre/obdclass/cl_object.c b/drivers/staging/lustre/lustre/obdclass/cl_object.c
index 4ad2ee5..dd80b83 100644
--- a/drivers/staging/lustre/lustre/obdclass/cl_object.c
+++ b/drivers/staging/lustre/lustre/obdclass/cl_object.c
@@ -374,6 +374,20 @@ int cl_object_fiemap(const struct lu_env *env, struct cl_object *obj,
}
EXPORT_SYMBOL(cl_object_fiemap);

+int cl_object_layout_get(const struct lu_env *env, struct cl_object *obj,
+ struct cl_layout *cl)
+{
+ struct lu_object_header *top = obj->co_lu.lo_header;
+
+ list_for_each_entry(obj, &top->loh_layers, co_lu.lo_linkage) {
+ if (obj->co_ops->coo_layout_get)
+ return obj->co_ops->coo_layout_get(env, obj, cl);
+ }
+
+ return -EOPNOTSUPP;
+}
+EXPORT_SYMBOL(cl_object_layout_get);
+
/**
* Helper function removing all object locks, and marking object for
* deletion. All object pages must have been deleted at this point.
--
1.7.1