* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
*/
#define DEBUG_SUBSYSTEM S_LMV
-#ifdef __KERNEL__
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
-#include <asm/div64.h>
+#include <linux/math64.h>
#include <linux/seq_file.h>
#include <linux/namei.h>
#include <linux/lustre_intent.h>
-#else
-#include <liblustre.h>
-#endif
#include <obd_support.h>
#include <lustre/lustre_idl.h>
#include <lustre_lib.h>
#include <lustre_net.h>
#include <lustre_dlm.h>
+#include <lustre_mdc.h>
#include <obd_class.h>
#include <lprocfs_status.h>
#include "lmv_internal.h"
-static int lmv_intent_remote(struct obd_export *exp, void *lmm,
- int lmmsize, struct lookup_intent *it,
- const struct lu_fid *parent_fid, int flags,
+static int lmv_intent_remote(struct obd_export *exp, struct lookup_intent *it,
+ const struct lu_fid *parent_fid,
struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
if (body == NULL)
RETURN(-EPROTO);
- LASSERT((body->valid & OBD_MD_MDS));
+ LASSERT((body->mbo_valid & OBD_MD_MDS));
/*
* Unfortunately, we have to lie to MDC/MDS to retrieve
it->d.lustre.it_data = NULL;
}
- LASSERT(fid_is_sane(&body->fid1));
+ LASSERT(fid_is_sane(&body->mbo_fid1));
- tgt = lmv_find_target(lmv, &body->fid1);
+ tgt = lmv_find_target(lmv, &body->mbo_fid1);
if (IS_ERR(tgt))
GOTO(out, rc = PTR_ERR(tgt));
if (op_data == NULL)
GOTO(out, rc = -ENOMEM);
- op_data->op_fid1 = body->fid1;
+ op_data->op_fid1 = body->mbo_fid1;
/* Sent the parent FID to the remote MDT */
if (parent_fid != NULL) {
/* The parent fid is only for remote open to
* see mdt_cross_open */
LASSERT(it->it_op & IT_OPEN);
op_data->op_fid2 = *parent_fid;
- /* Add object FID to op_fid3, in case it needs to check stale
- * (M_CHECK_STALE), see mdc_finish_intent_lock */
- op_data->op_fid3 = body->fid1;
}
op_data->op_bias = MDS_CROSS_REF;
CDEBUG(D_INODE, "REMOTE_INTENT with fid="DFID" -> mds #%d\n",
- PFID(&body->fid1), tgt->ltd_idx);
+ PFID(&body->mbo_fid1), tgt->ltd_idx);
- it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
- rc = md_intent_lock(tgt->ltd_exp, op_data, lmm, lmmsize, it,
- flags, &req, cb_blocking, extra_lock_flags);
+ rc = md_intent_lock(tgt->ltd_exp, op_data, it, &req, cb_blocking,
+ extra_lock_flags);
if (rc)
GOTO(out_free_op_data, rc);
it->d.lustre.it_remote_lock_mode = it->d.lustre.it_lock_mode;
}
- it->d.lustre.it_lock_handle = plock.cookie;
- it->d.lustre.it_lock_mode = pmode;
+ if (pmode) {
+ it->d.lustre.it_lock_handle = plock.cookie;
+ it->d.lustre.it_lock_mode = pmode;
+ }
EXIT;
out_free_op_data:
return rc;
}
+int lmv_revalidate_slaves(struct obd_export *exp, struct mdt_body *mbody,
+ struct lmv_stripe_md *lsm,
+ ldlm_blocking_callback cb_blocking,
+ int extra_lock_flags)
+{
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct mdt_body *body;
+ struct md_op_data *op_data;
+ unsigned long size = 0;
+ unsigned long nlink = 0;
+ obd_time atime = 0;
+ obd_time ctime = 0;
+ obd_time mtime = 0;
+ int i;
+ int rc = 0;
+
+ ENTRY;
+
+ /**
+ * revalidate slaves has some problems, temporarily return,
+ * we may not need that
+ */
+ OBD_ALLOC_PTR(op_data);
+ if (op_data == NULL)
+ RETURN(-ENOMEM);
+
+ /**
+ * Loop over the stripe information, check validity and update them
+ * from MDS if needed.
+ */
+ for (i = 0; i < lsm->lsm_md_stripe_count; i++) {
+ struct lu_fid fid;
+ struct lookup_intent it = { .it_op = IT_GETATTR };
+ struct ptlrpc_request *req = NULL;
+ struct lustre_handle *lockh = NULL;
+ struct lmv_tgt_desc *tgt = NULL;
+ struct inode *inode;
+
+ fid = lsm->lsm_md_oinfo[i].lmo_fid;
+ inode = lsm->lsm_md_oinfo[i].lmo_root;
+
+ /*
+ * Prepare op_data for revalidating. Note that @fid2 shluld be
+ * defined otherwise it will go to server and take new lock
+ * which is not needed here.
+ */
+ memset(op_data, 0, sizeof(*op_data));
+ op_data->op_fid1 = fid;
+ op_data->op_fid2 = fid;
+
+ tgt = lmv_locate_mds(lmv, op_data, &fid);
+ if (IS_ERR(tgt))
+ GOTO(cleanup, rc = PTR_ERR(tgt));
+
+ CDEBUG(D_INODE, "Revalidate slave "DFID" -> mds #%d\n",
+ PFID(&fid), tgt->ltd_idx);
+
+ rc = md_intent_lock(tgt->ltd_exp, op_data, &it, &req,
+ cb_blocking, extra_lock_flags);
+ if (rc < 0)
+ GOTO(cleanup, rc);
+
+ lockh = (struct lustre_handle *)&it.d.lustre.it_lock_handle;
+ if (rc > 0 && req == NULL) {
+ /* slave inode is still valid */
+ CDEBUG(D_INODE, "slave "DFID" is still valid.\n",
+ PFID(&fid));
+ rc = 0;
+ } else {
+ /* refresh slave from server */
+ body = req_capsule_server_get(&req->rq_pill,
+ &RMF_MDT_BODY);
+ LASSERT(body != NULL);
+ if (unlikely(body->mbo_nlink < 2)) {
+ CERROR("%s: nlink %d < 2 corrupt stripe %d "DFID
+ ":" DFID"\n",
+ obd->obd_name, body->mbo_nlink, i,
+ PFID(&lsm->lsm_md_oinfo[i].lmo_fid),
+ PFID(&lsm->lsm_md_oinfo[0].lmo_fid));
+
+ if (req != NULL)
+ ptlrpc_req_finished(req);
+
+ if (it.d.lustre.it_lock_mode && lockh) {
+ ldlm_lock_decref(lockh,
+ it.d.lustre.it_lock_mode);
+ it.d.lustre.it_lock_mode = 0;
+ }
+
+ GOTO(cleanup, rc = -EIO);
+ }
+
+
+ i_size_write(inode, body->mbo_size);
+ set_nlink(inode, body->mbo_nlink);
+ LTIME_S(inode->i_atime) = body->mbo_atime;
+ LTIME_S(inode->i_ctime) = body->mbo_ctime;
+ LTIME_S(inode->i_mtime) = body->mbo_mtime;
+
+ if (req != NULL)
+ ptlrpc_req_finished(req);
+ }
+
+ md_set_lock_data(tgt->ltd_exp, &lockh->cookie, inode, NULL);
+
+ size += i_size_read(inode);
+
+ if (i != 0)
+ nlink += inode->i_nlink - 2;
+ else
+ nlink += inode->i_nlink;
+
+ atime = LTIME_S(inode->i_atime) > atime ?
+ LTIME_S(inode->i_atime) : atime;
+ ctime = LTIME_S(inode->i_ctime) > ctime ?
+ LTIME_S(inode->i_ctime) : ctime;
+ mtime = LTIME_S(inode->i_mtime) > mtime ?
+ LTIME_S(inode->i_mtime) : mtime;
+
+ if (it.d.lustre.it_lock_mode != 0 && lockh != NULL) {
+ ldlm_lock_decref(lockh, it.d.lustre.it_lock_mode);
+ it.d.lustre.it_lock_mode = 0;
+ }
+
+ CDEBUG(D_INODE, "i %d "DFID" size %llu, nlink %u, atime "
+ "%lu, mtime %lu, ctime %lu.\n", i, PFID(&fid),
+ i_size_read(inode), inode->i_nlink,
+ LTIME_S(inode->i_atime), LTIME_S(inode->i_mtime),
+ LTIME_S(inode->i_ctime));
+ }
+
+ /*
+ * update attr of master request.
+ */
+ CDEBUG(D_INODE, "Return refreshed attrs: size = %lu nlink %lu atime "
+ LPU64 "ctime "LPU64" mtime "LPU64" for " DFID"\n", size, nlink,
+ atime, ctime, mtime, PFID(&lsm->lsm_md_oinfo[0].lmo_fid));
+
+ if (mbody != NULL) {
+ mbody->mbo_atime = atime;
+ mbody->mbo_ctime = ctime;
+ mbody->mbo_mtime = mtime;
+ }
+cleanup:
+ OBD_FREE_PTR(op_data);
+ RETURN(rc);
+}
+
+
/*
* IT_OPEN is intended to open (and create, possible) an object. Parent (pid)
* may be split dir.
*/
int lmv_intent_open(struct obd_export *exp, struct md_op_data *op_data,
- void *lmm, int lmmsize, struct lookup_intent *it,
- int flags, struct ptlrpc_request **reqp,
- ldlm_blocking_callback cb_blocking,
- __u64 extra_lock_flags)
+ struct lookup_intent *it, struct ptlrpc_request **reqp,
+ ldlm_blocking_callback cb_blocking, __u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
int rc;
ENTRY;
- tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ if (it->it_flags & MDS_OPEN_BY_FID) {
+ LASSERT(fid_is_sane(&op_data->op_fid2));
+
+ /* for striped directory, we can't know parent stripe fid
+ * without name, but we can set it to child fid, and MDT
+ * will obtain it from linkea in open in such case. */
+ if (op_data->op_mea1 != NULL)
+ op_data->op_fid1 = op_data->op_fid2;
+
+ tgt = lmv_find_target(lmv, &op_data->op_fid2);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ op_data->op_mds = tgt->ltd_idx;
+ } else {
+ LASSERT(fid_is_sane(&op_data->op_fid1));
+ LASSERT(fid_is_zero(&op_data->op_fid2));
+ LASSERT(op_data->op_name != NULL);
+
+ tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+ }
/* If it is ready to open the file by FID, do not need
* allocate FID at all, otherwise it will confuse MDT */
- if ((it->it_op & IT_CREAT) &&
- !(it->it_flags & MDS_OPEN_BY_FID)) {
+ if ((it->it_op & IT_CREAT) && !(it->it_flags & MDS_OPEN_BY_FID)) {
/*
- * For open with IT_CREATE and for IT_CREATE cases allocate new
- * fid and setup FLD for it.
+ * For lookup(IT_CREATE) cases allocate new fid and setup FLD
+ * for it.
*/
- op_data->op_fid3 = op_data->op_fid2;
- rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
+ rc = lmv_fid_alloc(NULL, exp, &op_data->op_fid2, op_data);
if (rc != 0)
RETURN(rc);
}
" name='%s' -> mds #%d\n", PFID(&op_data->op_fid1),
PFID(&op_data->op_fid2), op_data->op_name, tgt->ltd_idx);
- rc = md_intent_lock(tgt->ltd_exp, op_data, lmm, lmmsize, it, flags,
- reqp, cb_blocking, extra_lock_flags);
+ rc = md_intent_lock(tgt->ltd_exp, op_data, it, reqp, cb_blocking,
+ extra_lock_flags);
if (rc != 0)
RETURN(rc);
/*
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
- /*
- * Not cross-ref case, just get out of here.
- */
- if (likely(!(body->valid & OBD_MD_MDS)))
- RETURN(0);
- /*
- * Okay, MDS has returned success. Probably name has been resolved in
- * remote inode.
- */
- rc = lmv_intent_remote(exp, lmm, lmmsize, it, &op_data->op_fid1, flags,
- reqp, cb_blocking, extra_lock_flags);
- if (rc != 0) {
- LASSERT(rc < 0);
- /*
- * This is possible, that some userspace application will try to
- * open file as directory and we will have -ENOTDIR here. As
- * this is normal situation, we should not print error here,
- * only debug info.
- */
- CDEBUG(D_INODE, "Can't handle remote %s: dir "DFID"("DFID"):"
- "%*s: %d\n", LL_IT2STR(it), PFID(&op_data->op_fid2),
- PFID(&op_data->op_fid1), op_data->op_namelen,
- op_data->op_name, rc);
- RETURN(rc);
+ /* Not cross-ref case, just get out of here. */
+ if (unlikely((body->mbo_valid & OBD_MD_MDS))) {
+ rc = lmv_intent_remote(exp, it, &op_data->op_fid1, reqp,
+ cb_blocking, extra_lock_flags);
+ if (rc != 0)
+ RETURN(rc);
+
+ body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
+ if (body == NULL)
+ RETURN(-EPROTO);
}
RETURN(rc);
/*
* Handler for: getattr, lookup and revalidate cases.
*/
-int lmv_intent_lookup(struct obd_export *exp, struct md_op_data *op_data,
- void *lmm, int lmmsize, struct lookup_intent *it,
- int flags, struct ptlrpc_request **reqp,
- ldlm_blocking_callback cb_blocking,
- __u64 extra_lock_flags)
+static int
+lmv_intent_lookup(struct obd_export *exp, struct md_op_data *op_data,
+ struct lookup_intent *it, struct ptlrpc_request **reqp,
+ ldlm_blocking_callback cb_blocking,
+ __u64 extra_lock_flags)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt = NULL;
- struct mdt_body *body;
- int rc = 0;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt = NULL;
+ struct mdt_body *body;
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
+ int rc = 0;
ENTRY;
+ /* If it returns ERR_PTR(-EBADFD) then it is an unknown hash type
+ * it will try all stripes to locate the object */
tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
- if (IS_ERR(tgt))
+ if (IS_ERR(tgt) && (PTR_ERR(tgt) != -EBADFD))
RETURN(PTR_ERR(tgt));
+ /* Both migrating dir and unknown hash dir need to try
+ * all of sub-stripes */
+ if (lsm != NULL && !lmv_is_known_hash_type(lsm)) {
+ struct lmv_oinfo *oinfo;
+
+ oinfo = &lsm->lsm_md_oinfo[0];
+
+ op_data->op_fid1 = oinfo->lmo_fid;
+ op_data->op_mds = oinfo->lmo_mds;
+ tgt = lmv_get_target(lmv, oinfo->lmo_mds, NULL);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+ }
+
if (!fid_is_sane(&op_data->op_fid2))
fid_zero(&op_data->op_fid2);
CDEBUG(D_INODE, "LOOKUP_INTENT with fid1="DFID", fid2="DFID
- ", name='%s' -> mds #%d\n", PFID(&op_data->op_fid1),
- PFID(&op_data->op_fid2),
+ ", name='%s' -> mds #%d lsm=%p lsm_magic=%x\n",
+ PFID(&op_data->op_fid1), PFID(&op_data->op_fid2),
op_data->op_name ? op_data->op_name : "<NULL>",
- tgt->ltd_idx);
+ tgt->ltd_idx, lsm, lsm == NULL ? -1 : lsm->lsm_md_magic);
op_data->op_bias &= ~MDS_CROSS_REF;
- rc = md_intent_lock(tgt->ltd_exp, op_data, lmm, lmmsize, it,
- flags, reqp, cb_blocking, extra_lock_flags);
+ rc = md_intent_lock(tgt->ltd_exp, op_data, it, reqp, cb_blocking,
+ extra_lock_flags);
+ if (rc < 0)
+ RETURN(rc);
- if (rc < 0 || *reqp == NULL)
+ if (*reqp == NULL) {
+ /* If RPC happens, lsm information will be revalidated
+ * during update_inode process (see ll_update_lsm_md) */
+ if (op_data->op_mea2 != NULL) {
+ rc = lmv_revalidate_slaves(exp, NULL, op_data->op_mea2,
+ cb_blocking,
+ extra_lock_flags);
+ if (rc != 0)
+ RETURN(rc);
+ }
RETURN(rc);
+ } else if (it_disposition(it, DISP_LOOKUP_NEG) && lsm != NULL &&
+ lmv_need_try_all_stripes(lsm)) {
+ /* For migrating and unknown hash type directory, it will
+ * try to target the entry on other stripes */
+ int stripe_index;
+
+ for (stripe_index = 1;
+ stripe_index < lsm->lsm_md_stripe_count &&
+ it_disposition(it, DISP_LOOKUP_NEG); stripe_index++) {
+ struct lmv_oinfo *oinfo;
+
+ /* release the previous request */
+ ptlrpc_req_finished(*reqp);
+ it->d.lustre.it_data = NULL;
+ *reqp = NULL;
+
+ oinfo = &lsm->lsm_md_oinfo[stripe_index];
+ tgt = lmv_find_target(lmv, &oinfo->lmo_fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ CDEBUG(D_INODE, "Try other stripes " DFID"\n",
+ PFID(&oinfo->lmo_fid));
+
+ op_data->op_fid1 = oinfo->lmo_fid;
+ it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
+ rc = md_intent_lock(tgt->ltd_exp, op_data, it, reqp,
+ cb_blocking, extra_lock_flags);
+ if (rc != 0)
+ RETURN(rc);
+ }
+ }
/*
* MDS has returned success. Probably name has been resolved in
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
- /* Not cross-ref case, just get out of here. */
- if (likely(!(body->valid & OBD_MD_MDS)))
- RETURN(0);
- rc = lmv_intent_remote(exp, lmm, lmmsize, it, NULL, flags, reqp,
- cb_blocking, extra_lock_flags);
+ /* Not cross-ref case, just get out of here. */
+ if (unlikely((body->mbo_valid & OBD_MD_MDS))) {
+ rc = lmv_intent_remote(exp, it, NULL, reqp, cb_blocking,
+ extra_lock_flags);
+ if (rc != 0)
+ RETURN(rc);
+ body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
+ if (body == NULL)
+ RETURN(-EPROTO);
+ }
RETURN(rc);
}
int lmv_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
- void *lmm, int lmmsize, struct lookup_intent *it,
- int flags, struct ptlrpc_request **reqp,
- ldlm_blocking_callback cb_blocking,
+ struct lookup_intent *it, struct ptlrpc_request **reqp,
+ ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
- struct obd_device *obd = exp->exp_obd;
- int rc;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ int rc;
+ ENTRY;
- LASSERT(it != NULL);
- LASSERT(fid_is_sane(&op_data->op_fid1));
+ LASSERT(it != NULL);
+ LASSERT(fid_is_sane(&op_data->op_fid1));
- CDEBUG(D_INODE, "INTENT LOCK '%s' for '%*s' on "DFID"\n",
- LL_IT2STR(it), op_data->op_namelen, op_data->op_name,
- PFID(&op_data->op_fid1));
+ CDEBUG(D_INODE, "INTENT LOCK '%s' for "DFID" '%.*s' on "DFID"\n",
+ LL_IT2STR(it), PFID(&op_data->op_fid2), op_data->op_namelen,
+ op_data->op_name, PFID(&op_data->op_fid1));
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
-
- if (it->it_op & (IT_LOOKUP | IT_GETATTR | IT_LAYOUT))
- rc = lmv_intent_lookup(exp, op_data, lmm, lmmsize, it,
- flags, reqp, cb_blocking,
- extra_lock_flags);
- else if (it->it_op & IT_OPEN)
- rc = lmv_intent_open(exp, op_data, lmm, lmmsize, it,
- flags, reqp, cb_blocking,
- extra_lock_flags);
- else
- LBUG();
- RETURN(rc);
+ rc = lmv_check_connect(obd);
+ if (rc)
+ RETURN(rc);
+
+ if (it->it_op & (IT_LOOKUP | IT_GETATTR | IT_LAYOUT))
+ rc = lmv_intent_lookup(exp, op_data, it, reqp, cb_blocking,
+ extra_lock_flags);
+ else if (it->it_op & IT_OPEN)
+ rc = lmv_intent_open(exp, op_data, it, reqp, cb_blocking,
+ extra_lock_flags);
+ else
+ LBUG();
+
+ if (rc < 0) {
+ struct lustre_handle lock_handle;
+
+ if (it->d.lustre.it_lock_mode != 0) {
+ lock_handle.cookie = it->d.lustre.it_lock_handle;
+ ldlm_lock_decref(&lock_handle,
+ it->d.lustre.it_lock_mode);
+ }
+
+ it->d.lustre.it_lock_handle = 0;
+ it->d.lustre.it_lock_mode = 0;
+
+ if (it->d.lustre.it_remote_lock_mode != 0) {
+ lock_handle.cookie = it->d.lustre.it_remote_lock_handle;
+ ldlm_lock_decref(&lock_handle,
+ it->d.lustre.it_remote_lock_mode);
+ }
+
+ it->d.lustre.it_remote_lock_handle = 0;
+ it->d.lustre.it_remote_lock_mode = 0;
+ }
+
+ RETURN(rc);
}