GOTO(out, rc);
}
- rc = ll_prep_inode(&file->f_dentry->d_inode, req, NULL);
+ rc = ll_prep_inode(&file->f_dentry->d_inode, req, NULL, itp);
if (!rc && itp->d.lustre.it_lock_mode)
ll_set_lock_data(sbi->ll_md_exp, file->f_dentry->d_inode,
itp, NULL);
rc = ll_lov_setstripe_ea_info(inode, file, flags, lumv1, lum_size);
if (rc == 0) {
struct lov_stripe_md *lsm;
+ __u32 gen;
+
put_user(0, &lumv1p->lmm_stripe_count);
+
+ ll_layout_refresh(inode, &gen);
lsm = ccc_inode_lsm_get(inode);
rc = obd_iocontrol(LL_IOC_LOV_GETSTRIPE, ll_i2dtexp(inode),
0, lsm, (void *)arg);
RETURN(rc);
}
- rc = ll_prep_inode(&inode, req, NULL);
+ rc = ll_prep_inode(&inode, req, NULL, NULL);
}
out:
ptlrpc_req_finished(req);
result = cl_conf_set(env, lli->lli_clob, conf);
cl_env_nested_put(&nest, env);
+
+ if (conf->coc_opc == OBJECT_CONF_SET) {
+ struct ldlm_lock *lock = conf->coc_lock;
+
+ LASSERT(lock != NULL);
+ 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 shoud happen before dropping
+ * the intent lock. */
+ ldlm_lock_allow_match(lock);
+ }
+ }
RETURN(result);
}
/**
+ * Apply the layout to the inode. Layout lock is held and will be released
+ * in this function.
+ */
+static int ll_layout_lock_set(struct lustre_handle *lockh, ldlm_mode_t mode,
+ struct inode *inode, __u32 *gen, bool reconf)
+{
+ struct ll_inode_info *lli = ll_i2info(inode);
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+ struct ldlm_lock *lock;
+ struct lustre_md md = { NULL };
+ struct cl_object_conf conf;
+ int rc = 0;
+ bool lvb_ready;
+ ENTRY;
+
+ LASSERT(lustre_handle_is_used(lockh));
+
+ lock = ldlm_handle2lock(lockh);
+ LASSERT(lock != NULL);
+ LASSERT(ldlm_has_layout(lock));
+
+ LDLM_DEBUG(lock, "File %p/"DFID" being reconfigured: %d.\n",
+ inode, PFID(&lli->lli_fid), reconf);
+
+ lock_res_and_lock(lock);
+ lvb_ready = !!(lock->l_flags & LDLM_FL_LVB_READY);
+ unlock_res_and_lock(lock);
+ /* 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) {
+ LDLM_LOCK_PUT(lock);
+
+ rc = -ENODATA;
+ if (lvb_ready) {
+ /* layout_gen must be valid if layout lock is not
+ * cancelled and stripe has already set */
+ *gen = lli->lli_layout_gen;
+ rc = 0;
+ }
+ ldlm_lock_decref(lockh, mode);
+ RETURN(rc);
+ }
+
+ /* for layout lock, lmm is returned in lock's lvb.
+ * lvb_data is immutable if the lock is held so it's safe to access it
+ * without res lock. See the description in ldlm_lock_decref_internal()
+ * for the condition to free lvb_data of layout lock */
+ if (lock->l_lvb_data != NULL) {
+ rc = obd_unpackmd(sbi->ll_dt_exp, &md.lsm,
+ lock->l_lvb_data, lock->l_lvb_len);
+ if (rc >= 0) {
+ if (md.lsm != NULL)
+ *gen = md.lsm->lsm_layout_gen;
+ rc = 0;
+ } else {
+ CERROR("%s: file "DFID" unpackmd error: %d\n",
+ ll_get_fsname(inode->i_sb, NULL, 0),
+ PFID(&lli->lli_fid), rc);
+ }
+ }
+ if (rc < 0) {
+ LDLM_LOCK_PUT(lock);
+ ldlm_lock_decref(lockh, mode);
+ RETURN(rc);
+ }
+
+ /* set layout to file. Unlikely this will fail as old layout was
+ * surely eliminated */
+ memset(&conf, 0, sizeof conf);
+ conf.coc_opc = OBJECT_CONF_SET;
+ conf.coc_inode = inode;
+ conf.coc_lock = lock;
+ conf.u.coc_md = &md;
+ rc = ll_layout_conf(inode, &conf);
+ LDLM_LOCK_PUT(lock);
+
+ ldlm_lock_decref(lockh, mode);
+
+ if (md.lsm != NULL)
+ obd_free_memmd(sbi->ll_dt_exp, &md.lsm);
+
+ /* wait for IO to complete if it's still being used. */
+ if (rc == -EBUSY) {
+ CDEBUG(D_INODE, "%s: %p/"DFID" wait for layout reconf.\n",
+ ll_get_fsname(inode->i_sb, NULL, 0),
+ inode, PFID(&lli->lli_fid));
+
+ memset(&conf, 0, sizeof conf);
+ conf.coc_opc = OBJECT_CONF_WAIT;
+ conf.coc_inode = inode;
+ rc = ll_layout_conf(inode, &conf);
+ if (rc == 0)
+ rc = -EAGAIN;
+
+ CDEBUG(D_INODE, "file: "DFID" waiting layout return: %d.\n",
+ PFID(&lli->lli_fid), rc);
+ }
+
+ 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.
*
{
struct ll_inode_info *lli = ll_i2info(inode);
struct ll_sb_info *sbi = ll_i2sbi(inode);
- struct md_op_data *op_data = NULL;
- struct lookup_intent it = { .it_op = IT_LAYOUT };
- struct lustre_handle lockh = { 0 };
+ struct md_op_data *op_data;
+ struct lookup_intent it;
+ struct lustre_handle lockh;
ldlm_mode_t mode;
struct ldlm_enqueue_info einfo = { .ei_type = LDLM_IBITS,
.ei_mode = LCK_CR,
int rc;
ENTRY;
- *gen = 0;
+ *gen = LL_LAYOUT_GEN_ZERO;
if (!(sbi->ll_flags & LL_SBI_LAYOUT_LOCK))
RETURN(0);
/* mostly layout lock is caching on the local side, so try to match
* it before grabbing layout lock mutex. */
- mode = ll_take_md_lock(inode, MDS_INODELOCK_LAYOUT, &lockh,
- LDLM_FL_LVB_READY);
+ mode = ll_take_md_lock(inode, MDS_INODELOCK_LAYOUT, &lockh, 0);
if (mode != 0) { /* hit cached lock */
- /* lsm_layout_gen is started from 0, plus 1 here to distinguish
- * the cases of no layout and first layout. */
- *gen = lli->lli_layout_gen + 1;
+ rc = ll_layout_lock_set(&lockh, mode, inode, gen, false);
+ if (rc == 0)
+ RETURN(0);
- ldlm_lock_decref(&lockh, mode);
- RETURN(0);
+ /* better hold lli_layout_mutex to try again otherwise
+ * it will have starvation problem. */
}
- op_data = ll_prep_md_op_data(NULL, inode, inode, NULL,
- 0, 0, LUSTRE_OPC_ANY, NULL);
- if (IS_ERR(op_data))
- RETURN(PTR_ERR(op_data));
-
/* take layout lock mutex to enqueue layout lock exclusively. */
mutex_lock(&lli->lli_layout_mutex);
- /* try again inside layout mutex */
- mode = ll_take_md_lock(inode, MDS_INODELOCK_LAYOUT, &lockh,
- LDLM_FL_LVB_READY);
+again:
+ /* try again. Maybe somebody else has done this. */
+ mode = ll_take_md_lock(inode, MDS_INODELOCK_LAYOUT, &lockh, 0);
if (mode != 0) { /* hit cached lock */
- *gen = lli->lli_layout_gen + 1;
+ rc = ll_layout_lock_set(&lockh, mode, inode, gen, true);
+ if (rc == -EAGAIN)
+ goto again;
- ldlm_lock_decref(&lockh, mode);
mutex_unlock(&lli->lli_layout_mutex);
- ll_finish_md_op_data(op_data);
- RETURN(0);
+ 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);
+ RETURN(PTR_ERR(op_data));
}
/* have to enqueue one */
+ memset(&it, 0, sizeof(it));
+ it.it_op = IT_LAYOUT;
+ lockh.cookie = 0ULL;
+
+ LDLM_DEBUG_NOLOCK("%s: requeue layout lock for file %p/"DFID".\n",
+ ll_get_fsname(inode->i_sb, NULL, 0), inode,
+ PFID(&lli->lli_fid));
+
rc = md_enqueue(sbi->ll_md_exp, &einfo, &it, op_data, &lockh,
NULL, 0, NULL, 0);
if (it.d.lustre.it_data != NULL)
ptlrpc_req_finished(it.d.lustre.it_data);
it.d.lustre.it_data = NULL;
- if (rc == 0) {
- struct ldlm_lock *lock;
- struct cl_object_conf conf;
- struct lustre_md md = { NULL };
- void *lmm;
- int lmmsize;
+ ll_finish_md_op_data(op_data);
- LASSERT(lustre_handle_is_used(&lockh));
+ mode = it.d.lustre.it_lock_mode;
+ it.d.lustre.it_lock_mode = 0;
+ ll_intent_drop_lock(&it);
+ if (rc == 0) {
/* set lock data in case this is a new lock */
ll_set_lock_data(sbi->ll_md_exp, inode, &it, NULL);
-
- lock = ldlm_handle2lock(&lockh);
- LASSERT(lock != NULL);
-
- /* for IT_LAYOUT lock, lmm is returned in lock's lvb
- * data via completion callback */
- lmm = lock->l_lvb_data;
- lmmsize = lock->l_lvb_len;
- if (lmm != NULL) {
- rc = obd_unpackmd(sbi->ll_dt_exp, &md.lsm,
- lmm, lmmsize);
- if (rc >= 0) {
- if (md.lsm != NULL)
- *gen = md.lsm->lsm_layout_gen + 1;
- rc = 0;
- } else {
- CERROR("file: "DFID" unpackmd error: %d\n",
- PFID(&lli->lli_fid), rc);
- }
- }
- LDLM_LOCK_PUT(lock);
-
- /* set layout to file. This may cause lock expiration as we
- * set layout inside layout ibits lock. */
- memset(&conf, 0, sizeof conf);
- conf.coc_inode = inode;
- conf.u.coc_md = &md;
- ll_layout_conf(inode, &conf);
- /* is this racy? */
- lli->lli_has_smd = md.lsm != NULL;
- if (md.lsm != NULL)
- obd_free_memmd(sbi->ll_dt_exp, &md.lsm);
+ rc = ll_layout_lock_set(&lockh, mode, inode, gen, true);
+ if (rc == -EAGAIN)
+ goto again;
}
- ll_intent_drop_lock(&it);
-
mutex_unlock(&lli->lli_layout_mutex);
- ll_finish_md_op_data(op_data);
RETURN(rc);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff