Whamcloud - gitweb
Branch HEAD
[fs/lustre-release.git] / lustre / llite / dcache.c
index 019ff46..4e6c5a9 100644 (file)
@@ -30,7 +30,9 @@
 #include <lustre_lite.h>
 #include <lustre/lustre_idl.h>
 #include <lustre_dlm.h>
-#include <linux/lustre_version.h>
+#include <lustre_mdc.h>
+//#include <lustre_ver.h>
+//#include <lustre_version.h>
 
 #include "llite_internal.h"
 
@@ -147,6 +149,7 @@ void ll_intent_release(struct lookup_intent *it)
 {
         ENTRY;
 
+        CDEBUG(D_INFO, "intent %p released\n", it);
         ll_intent_drop_lock(it);
 #ifdef LUSTRE_KERNEL_VERSION
         it->it_magic = 0;
@@ -185,6 +188,12 @@ int ll_drop_dentry(struct dentry *dentry)
                 spin_lock(&dcache_lock);
                 return 1;
         }
+        /* disconected dentry can not be find without lookup, because we 
+         * not need his to unhash or mark invalid. */
+        if (dentry->d_flags & DCACHE_DISCONNECTED) {
+                unlock_dentry(dentry);
+                RETURN (0);
+        }
 
 #ifdef LUSTRE_KERNEL_VERSION
         if (!(dentry->d_flags & DCACHE_LUSTRE_INVALID)) {
@@ -200,18 +209,9 @@ int ll_drop_dentry(struct dentry *dentry)
                  * sys_getcwd() could return -ENOENT -bzzz */
 #ifdef LUSTRE_KERNEL_VERSION
                 dentry->d_flags |= DCACHE_LUSTRE_INVALID;
-#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
-                __d_drop(dentry);
-                if (dentry->d_inode) {
-                        /* Put positive dentries to orphan list */
-                        list_add(&dentry->d_hash,
-                                 &ll_i2sbi(dentry->d_inode)->ll_orphan_dentry_list);
-                }
 #endif
-#else
                 if (!dentry->d_inode || !S_ISDIR(dentry->d_inode->i_mode))
                         __d_drop(dentry);
-#endif
 
         }
         unlock_dentry(dentry);
@@ -238,6 +238,11 @@ restart:
         while ((tmp = tmp->next) != head) {
                 struct dentry *dentry = list_entry(tmp, struct dentry, d_alias);
 
+                CDEBUG(D_DENTRY, "dentry in drop %.*s (%p) parent %p "
+                       "inode %p flags %d\n", dentry->d_name.len,
+                       dentry->d_name.name, dentry, dentry->d_parent,
+                       dentry->d_inode, dentry->d_flags);
+
                 if (dentry->d_name.len == 1 && dentry->d_name.name[0] == '/') {
                         CERROR("called on root (?) dentry=%p, inode=%p "
                                "ino=%lu\n", dentry, inode, inode->i_ino);
@@ -255,7 +260,7 @@ restart:
 
                         continue;
                 }
-                
+
                 if (ll_drop_dentry(dentry))
                           goto restart;
         }
@@ -263,8 +268,9 @@ restart:
         EXIT;
 }
 
-int revalidate_it_finish(struct ptlrpc_request *request, int offset,
-                         struct lookup_intent *it, struct dentry *de)
+int ll_revalidate_it_finish(struct ptlrpc_request *request,
+                            int offset, struct lookup_intent *it,
+                            struct dentry *de)
 {
         int rc = 0;
         ENTRY;
@@ -272,10 +278,10 @@ int revalidate_it_finish(struct ptlrpc_request *request, int offset,
         if (!request)
                 RETURN(0);
 
-        if (it_disposition(it, DISP_LOOKUP_NEG))
+        if (it_disposition(it, DISP_LOOKUP_NEG)) 
                 RETURN(-ENOENT);
 
-        rc = ll_prep_inode(ll_i2sbi(de->d_inode)->ll_osc_exp, &de->d_inode,
+        rc = ll_prep_inode(&de->d_inode,
                            request, offset, NULL);
 
         RETURN(rc);
@@ -288,31 +294,31 @@ void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry)
 
         if (it->d.lustre.it_lock_mode && dentry->d_inode != NULL) {
                 struct inode *inode = dentry->d_inode;
+                struct ll_sb_info *sbi = ll_i2sbi(dentry->d_inode);
+
                 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
                        inode, inode->i_ino, inode->i_generation);
-                mdc_set_lock_data(&it->d.lustre.it_lock_handle, inode);
+                md_set_lock_data(sbi->ll_md_exp, &it->d.lustre.it_lock_handle,
+                                 inode);
         }
 
         /* drop lookup or getattr locks immediately */
         if (it->it_op == IT_LOOKUP || it->it_op == IT_GETATTR) {
-#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
                 /* on 2.6 there are situation when several lookups and
                  * revalidations may be requested during single operation.
                  * therefore, we don't release intent here -bzzz */
                 ll_intent_drop_lock(it);
-#else
-                ll_intent_release(it);
-#endif
         }
 }
 
 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft)
 {
         struct lookup_intent *it = *itp;
-#if defined(LUSTRE_KERNEL_VERSION)&&(LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
+#if defined(LUSTRE_KERNEL_VERSION)
         if (it) {
-                LASSERTF(it->it_magic == INTENT_MAGIC, "bad intent magic: %x\n",
-                         it->it_magic);
+                LASSERTF(it->it_magic == INTENT_MAGIC, 
+                         "%p has bad intent magic: %x\n",
+                         it, it->it_magic);
         }
 #endif
 
@@ -328,10 +334,11 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
                      struct lookup_intent *it)
 {
         int rc;
-        struct mdc_op_data op_data;
+        struct md_op_data *op_data;
         struct ptlrpc_request *req = NULL;
         struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
         struct obd_export *exp;
+        struct inode *parent;
 
         ENTRY;
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,intent=%s\n", de->d_name.name,
@@ -357,7 +364,7 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
                 RETURN(rc);
         }
 
-        exp = ll_i2mdcexp(de->d_inode);
+        exp = ll_i2mdexp(de->d_inode);
 
         /* Never execute intents for mount points.
          * Attributes will be fixed up in ll_inode_revalidate_it */
@@ -366,28 +373,44 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
 
         /* Root of the lustre tree. Always valid.
          * Attributes will be fixed up in ll_inode_revalidate_it */
-        if (de->d_name.name[0] == '/' && de->d_name.len == 1)
+        if (de == de->d_sb->s_root)
                 RETURN(1);
 
         OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_REVALIDATE_PAUSE, 5);
         ll_frob_intent(&it, &lookup_it);
         LASSERT(it);
 
-        ll_prepare_mdc_op_data(&op_data, de->d_parent->d_inode, de->d_inode,
-                               de->d_name.name, de->d_name.len, 0);
+        parent = de->d_parent->d_inode;
+
+        if (it->it_op & IT_CREAT) {
+                op_data = ll_prep_md_op_data(NULL, parent, NULL,
+                                             de->d_name.name, de->d_name.len,
+                                             0, LUSTRE_OPC_CREATE, NULL);
+        } else {
+                op_data = ll_prep_md_op_data(NULL, parent, de->d_inode,
+                                             de->d_name.name, de->d_name.len,
+                                             0, LUSTRE_OPC_ANY, NULL);
+        }
+        if (IS_ERR(op_data))
+                RETURN(PTR_ERR(op_data));
+
 
         if ((it->it_op == IT_OPEN) && de->d_inode) {
                 struct inode *inode = de->d_inode;
                 struct ll_inode_info *lli = ll_i2info(inode);
                 struct obd_client_handle **och_p;
                 __u64 *och_usecount;
-                /* We used to check for MDS_INODELOCK_OPEN here, but in fact
+                
+                /*
+                 * We used to check for MDS_INODELOCK_OPEN here, but in fact
                  * just having LOOKUP lock is enough to justify inode is the
                  * same. And if inode is the same and we have suitable
                  * openhandle, then there is no point in doing another OPEN RPC
-                 * just to throw away newly received openhandle.
-                 * There are no security implications too, if file owner or
-                 * access mode is change, LOOKUP lock is revoked */
+                 * just to throw away newly received openhandle.  There are no
+                 * security implications too, if file owner or access mode is
+                 * change, LOOKUP lock is revoked.
+                 */
+
 
                 if (it->it_flags & FMODE_WRITE) {
                         och_p = &lli->lli_mds_write_och;
@@ -395,7 +418,7 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
                 } else if (it->it_flags & FMODE_EXEC) {
                         och_p = &lli->lli_mds_exec_och;
                         och_usecount = &lli->lli_open_fd_exec_count;
-                 } else {
+                } else {
                         och_p = &lli->lli_mds_read_och;
                         och_usecount = &lli->lli_open_fd_read_count;
                 }
@@ -405,7 +428,7 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
                 down(&lli->lli_och_sem);
                 if (*och_p) { /* Everything is open already, do nothing */
                         /*(*och_usecount)++;  Do not let them steal our open
-                                              handle from under us */
+                          handle from under us */
                         /* XXX The code above was my original idea, but in case
                            we have the handle, but we cannot use it due to later
                            checks (e.g. O_CREAT|O_EXCL flags set), nobody
@@ -414,6 +437,7 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
                            if it would be, we'll reopen the open request to
                            MDS later during file open path */
                         up(&lli->lli_och_sem);
+                        ll_finish_md_op_data(op_data);
                         RETURN(1);
                 } else {
                         up(&lli->lli_och_sem);
@@ -422,10 +446,13 @@ int ll_revalidate_it(struct dentry *de, int lookup_flags,
 
 do_lock:
         it->it_create_mode &= ~current->fs->umask;
-
-        rc = mdc_intent_lock(exp, &op_data, NULL, 0, it, lookup_flags,
-                             &req, ll_mdc_blocking_ast, 0);
-        /* If req is NULL, then mdc_intent_lock only tried to do a lock match;
+        it->it_flags |= O_CHECK_STALE;
+        rc = md_intent_lock(exp, op_data, NULL, 0, it,
+                            lookup_flags,
+                            &req, ll_md_blocking_ast, 0);
+        it->it_flags &= ~O_CHECK_STALE;
+        ll_finish_md_op_data(op_data);
+        /* If req is NULL, then md_intent_lock only tried to do a lock match;
          * if all was well, it will return 1 if it found locks, 0 otherwise. */
         if (req == NULL && rc >= 0) {
                 if (!rc)
@@ -442,11 +469,13 @@ do_lock:
         }
 
 revalidate_finish:
-        rc = revalidate_it_finish(req, DLM_REPLY_REC_OFF, it, de);
+        rc = ll_revalidate_it_finish(req, DLM_REPLY_REC_OFF, it, de);
         if (rc != 0) {
-                ll_intent_release(it);
+                if (rc != -ESTALE && rc != -ENOENT)
+                        ll_intent_release(it);
                 GOTO(out, rc = 0);
         }
+
         if ((it->it_op & IT_OPEN) && de->d_inode && 
             !S_ISREG(de->d_inode->i_mode) && 
             !S_ISDIR(de->d_inode->i_mode)) {
@@ -463,7 +492,7 @@ revalidate_finish:
         __d_rehash(de, 0);
         spin_unlock(&dcache_lock);
 
- out:
+out:
         /* We do not free request as it may be reused during following lookup
          * (see comment in mdc/mdc_locks.c::mdc_intent_lock()), request will
          * be freed in ll_lookup_it or in ll_intent_release. But if
@@ -474,16 +503,16 @@ revalidate_finish:
 #ifdef LUSTRE_KERNEL_VERSION
                 ll_unhash_aliases(de->d_inode);
                 /* done in ll_unhash_aliases()
-                dentry->d_flags |= DCACHE_LUSTRE_INVALID; */
+                   dentry->d_flags |= DCACHE_LUSTRE_INVALID; */
 #else
                 /* We do not want d_invalidate to kill all child dentries too */
                 d_drop(de);
 #endif
         } else {
                 CDEBUG(D_DENTRY, "revalidated dentry %.*s (%p) parent %p "
-                               "inode %p refc %d\n", de->d_name.len,
-                               de->d_name.name, de, de->d_parent, de->d_inode,
-                               atomic_read(&de->d_count));
+                       "inode %p refc %d\n", de->d_name.len,
+                       de->d_name.name, de, de->d_parent, de->d_inode,
+                       atomic_read(&de->d_count));
                 ll_lookup_finish_locks(it, de);
 #ifdef LUSTRE_KERNEL_VERSION
                 lock_dentry(de);
@@ -492,38 +521,59 @@ revalidate_finish:
 #endif
         }
         RETURN(rc);
-/* This part is here to combat evil-evil race in real_lookup on 2.6 kernels.
- * The race details are: We enter do_lookup() looking for some name,
- * there is nothing in dcache for this name yet and d_lookup() returns NULL.
- * We proceed to real_lookup(), and while we do this, another process does
- * open on the same file we looking up (most simple reproducer), open succeeds
- * and the dentry is added. Now back to us. In real_lookup() we do d_lookup()
- * again and suddenly find the dentry, so we call d_revalidate on it, but there
- * is no lock, so without this code we would return 0, but unpatched
- * real_lookup just returns -ENOENT in such a case instead of retrying the
- * lookup. Once this is dealt with in real_lookup(), all of this ugly mess
- * can go and we can just check locks in ->d_revalidate without doing any
- * RPCs ever. */
+        
+        /*
+         * This part is here to combat evil-evil race in real_lookup on 2.6
+         * kernels.  The race details are: We enter do_lookup() looking for some
+         * name, there is nothing in dcache for this name yet and d_lookup()
+         * returns NULL.  We proceed to real_lookup(), and while we do this,
+         * another process does open on the same file we looking up (most simple
+         * reproducer), open succeeds and the dentry is added. Now back to
+         * us. In real_lookup() we do d_lookup() again and suddenly find the
+         * dentry, so we call d_revalidate on it, but there is no lock, so
+         * without this code we would return 0, but unpatched real_lookup just
+         * returns -ENOENT in such a case instead of retrying the lookup. Once
+         * this is dealt with in real_lookup(), all of this ugly mess can go and
+         * we can just check locks in ->d_revalidate without doing any RPCs
+         * ever.
+         */
 do_lookup:
         if (it != &lookup_it) {
+                /* MDS_INODELOCK_UPDATE needed for IT_GETATTR case. */
+                if (it->it_op == IT_GETATTR)
+                        lookup_it.it_op = IT_GETATTR;
                 ll_lookup_finish_locks(it, de);
                 it = &lookup_it;
         }
-        /*do real lookup here */
-        ll_prepare_mdc_op_data(&op_data, de->d_parent->d_inode, NULL,
-                               de->d_name.name, de->d_name.len, 0);
-        rc = mdc_intent_lock(exp, &op_data, NULL, 0,  it, 0, &req,
-                             ll_mdc_blocking_ast, 0);
+        
+        /* Do real lookup here. */
+        op_data = ll_prep_md_op_data(NULL, parent, NULL, de->d_name.name,
+                                     de->d_name.len, 0, (it->it_op & IT_CREAT ?
+                                                         LUSTRE_OPC_CREATE :
+                                                         LUSTRE_OPC_ANY), NULL);
+        if (IS_ERR(op_data))
+                RETURN(PTR_ERR(op_data));
+
+        rc = md_intent_lock(exp, op_data, NULL, 0,  it, 0, &req,
+                            ll_md_blocking_ast, 0);
         if (rc >= 0) {
-                struct mds_body *mds_body = lustre_msg_buf(req->rq_repmsg,
+                struct mdt_body *mdt_body = lustre_msg_buf(req->rq_repmsg,
                                                            DLM_REPLY_REC_OFF,
-                                                           sizeof(*mds_body));
+                                                           sizeof(*mdt_body));
+                struct lu_fid fid = {.f_seq = 0, .f_oid = 0, .f_ver = 0};
+                
+                if (de->d_inode)
+                        fid = *ll_inode2fid(de->d_inode);
+
                 /* see if we got same inode, if not - return error */
-                if(!memcmp(&op_data.fid2, &mds_body->fid1,
-                           sizeof(op_data.fid2)))
+                if (lu_fid_eq(&fid, &mdt_body->fid1)) {
+                        ll_finish_md_op_data(op_data);
+                        op_data = NULL;
                         goto revalidate_finish;
+                }
                 ll_intent_release(it);
         }
+        ll_finish_md_op_data(op_data);
         GOTO(out, rc = 0);
 }
 
@@ -533,6 +583,7 @@ do_lookup:
         struct ll_sb_info *sbi = ll_i2sbi(inode);
         struct ll_dentry_data *ldd = ll_d2d(de);
         struct obd_client_handle *handle;
+        struct obd_capa *oc;
         int rc = 0;
         ENTRY;
         LASSERT(ldd);
@@ -556,9 +607,9 @@ do_lookup:
         unlock_kernel();
 
         handle = (flag) ? &ldd->lld_mnt_och : &ldd->lld_cwd_och;
-        rc = obd_pin(sbi->ll_mdc_exp, inode->i_ino, inode->i_generation,
-                     inode->i_mode & S_IFMT, handle, flag);
-
+        oc = ll_mdscapa_get(inode);
+        rc = obd_pin(sbi->ll_md_exp, ll_inode2fid(inode), oc, handle, flag);
+        capa_put(oc);
         if (rc) {
                 lock_kernel();
                 memset(handle, 0, sizeof(*handle));
@@ -606,12 +657,11 @@ do_lookup:
                 return;
         }
 
-        rc = obd_unpin(sbi->ll_mdc_exp, &handle, flag);
+        rc = obd_unpin(sbi->ll_md_exp, &handle, flag);
         EXIT;
         return;
 }
 
-#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
 #ifdef LUSTRE_KERNEL_VERSION
 static int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd)
 {
@@ -695,14 +745,9 @@ out_it:
         RETURN(rc);
 }
 #endif
-#endif
 
 struct dentry_operations ll_d_ops = {
-#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
         .d_revalidate = ll_revalidate_nd,
-#else
-        .d_revalidate_it = ll_revalidate_it,
-#endif
         .d_release = ll_release,
         .d_delete = ll_ddelete,
 #ifdef LUSTRE_KERNEL_VERSION