From 36f20c6df75d599393d79c7feb6283b20913e3d5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 25 Nov 2005 17:09:57 -0500 Subject: NFSv4: Fix buggy nfs_wait_on_sequence() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 0675f3215e..5ef4c57618 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -644,12 +644,15 @@ void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t f struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) { + struct rpc_sequence *sequence = counter->sequence; struct nfs_seqid *new; new = kmalloc(sizeof(*new), GFP_KERNEL); if (new != NULL) { new->sequence = counter; - INIT_LIST_HEAD(&new->list); + spin_lock(&sequence->lock); + list_add_tail(&new->list, &sequence->list); + spin_unlock(&sequence->lock); } return new; } @@ -658,12 +661,10 @@ void nfs_free_seqid(struct nfs_seqid *seqid) { struct rpc_sequence *sequence = seqid->sequence->sequence; - if (!list_empty(&seqid->list)) { - spin_lock(&sequence->lock); - list_del(&seqid->list); - spin_unlock(&sequence->lock); - } - rpc_wake_up_next(&sequence->wait); + spin_lock(&sequence->lock); + list_del(&seqid->list); + spin_unlock(&sequence->lock); + rpc_wake_up(&sequence->wait); kfree(seqid); } @@ -722,11 +723,10 @@ int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) if (sequence->list.next == &seqid->list) goto out; spin_lock(&sequence->lock); - if (!list_empty(&sequence->list)) { + if (sequence->list.next != &seqid->list) { rpc_sleep_on(&sequence->wait, task, NULL, NULL); status = -EAGAIN; - } else - list_add(&seqid->list, &sequence->list); + } spin_unlock(&sequence->lock); out: return status; -- cgit v1.2.2 From ff6040667ad5a21fa1090e02941ecefb94ebe32c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 25 Nov 2005 17:10:01 -0500 Subject: NFSv4: Fix typo in lock caching When caching locks due to holding a file delegation, we must always check against local locks before sending anything to the server. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 21482b2518..60e0dd800c 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3071,15 +3071,15 @@ static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock struct nfs4_client *clp = state->owner->so_client; int status; - down_read(&clp->cl_sem); /* Is this a delegated open? */ - if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { + if (NFS_I(state->inode)->delegation_state != 0) { /* Yes: cache locks! */ status = do_vfs_lock(request->fl_file, request); /* ...but avoid races with delegation recall... */ if (status < 0 || test_bit(NFS_DELEGATED_STATE, &state->flags)) - goto out; + return status; } + down_read(&clp->cl_sem); status = nfs4_set_lock_state(state, request); if (status != 0) goto out; -- cgit v1.2.2 From b37b03b7051493c9f9a6b336c9c0f81334885b7d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 25 Nov 2005 17:10:06 -0500 Subject: NFS: Fix a spinlock recursion inside nfs_update_inode() In cases where the server has gone insane, nfs_update_inode() may end up calling nfs_invalidate_inode(), which again calls stuff that takes the inode->i_lock that we're already holding. In addition, given the sort of things we have in NFS these days that need to be cleaned up on inode release, I'm not sure we should ever be calling make_bad_inode(). Fix up spinlock recursion, and limit nfs_invalidate_inode() to clearing the caches, and marking the inode as being stale. Thanks to Steve Dickson for spotting this. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 6391d89642..aaab1a5ac4 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -643,14 +643,11 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt) /* * Invalidate the local caches */ -void -nfs_zap_caches(struct inode *inode) +static void nfs_zap_caches_locked(struct inode *inode) { struct nfs_inode *nfsi = NFS_I(inode); int mode = inode->i_mode; - spin_lock(&inode->i_lock); - NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode); NFS_ATTRTIMEO_UPDATE(inode) = jiffies; @@ -659,7 +656,12 @@ nfs_zap_caches(struct inode *inode) nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE; else nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE; +} +void nfs_zap_caches(struct inode *inode) +{ + spin_lock(&inode->i_lock); + nfs_zap_caches_locked(inode); spin_unlock(&inode->i_lock); } @@ -676,16 +678,13 @@ static void nfs_zap_acl_cache(struct inode *inode) } /* - * Invalidate, but do not unhash, the inode + * Invalidate, but do not unhash, the inode. + * NB: must be called with inode->i_lock held! */ -static void -nfs_invalidate_inode(struct inode *inode) +static void nfs_invalidate_inode(struct inode *inode) { - umode_t save_mode = inode->i_mode; - - make_bad_inode(inode); - inode->i_mode = save_mode; - nfs_zap_caches(inode); + set_bit(NFS_INO_STALE, &NFS_FLAGS(inode)); + nfs_zap_caches_locked(inode); } struct nfs_find_desc { @@ -1528,14 +1527,13 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n", __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode); #endif + out_err: /* * No need to worry about unhashing the dentry, as the * lookup validation will know that the inode is bad. * (But we fall through to invalidate the caches.) */ nfs_invalidate_inode(inode); - out_err: - set_bit(NFS_INO_STALE, &NFS_FLAGS(inode)); return -ESTALE; } -- cgit v1.2.2 From 223db122bfccd463751d8b0c09a638abee03681d Mon Sep 17 00:00:00 2001 From: Steve Dickson Date: Wed, 30 Nov 2005 09:25:33 -0500 Subject: NFS: Fix cache consistency regression Make sure cache_change_attribute is initialized to jiffies so when the mtime changes on directory, the directory will be refreshed. Signed-off by: Steve Dickson Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index aaab1a5ac4..b551b19aa6 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -2066,6 +2066,7 @@ static struct inode *nfs_alloc_inode(struct super_block *sb) return NULL; nfsi->flags = 0UL; nfsi->cache_validity = 0UL; + nfsi->cache_change_attribute = jiffies; #ifdef CONFIG_NFS_V3_ACL nfsi->acl_access = ERR_PTR(-EAGAIN); nfsi->acl_default = ERR_PTR(-EAGAIN); -- cgit v1.2.2 From 24aa1fe6779eaddb3e0b1b802585dcf6faf9cc44 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 3 Dec 2005 15:20:07 -0500 Subject: NFS: Fix a few further cache consistency regressions Steve Dickson writes: Doing the following: 1. On server: $ mkdir ~/t $ echo Hello > ~/t/tmp 2. On client, wait for a string to appear in this file: $ until grep -q foo t/tmp ; do echo -n . ; sleep 1 ; done 3. On server, create a *new* file with the same name containing that string: $ mv ~/t/tmp ~/t/tmp.old; echo foo > ~/t/tmp will show how the client will never (and I mean never ;-) ) see the updated file. The problem is that we do not update nfsi->cache_change_attribute when the file changes on the server (we only update it when our client makes the changes). This again means that functions like nfs_check_verifier() will fail to register when the parent directory has changed and should trigger a dentry lookup revalidation. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 54 ++++++++++++++++++++---------------------------------- 1 file changed, 20 insertions(+), 34 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index b551b19aa6..afd75d0463 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -54,7 +54,7 @@ #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1) static void nfs_invalidate_inode(struct inode *); -static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long); +static int nfs_update_inode(struct inode *, struct nfs_fattr *); static struct inode *nfs_alloc_inode(struct super_block *sb); static void nfs_destroy_inode(struct inode *); @@ -1080,8 +1080,6 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) int status = -ESTALE; struct nfs_fattr fattr; struct nfs_inode *nfsi = NFS_I(inode); - unsigned long verifier; - unsigned long cache_validity; dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n", inode->i_sb->s_id, (long long)NFS_FILEID(inode)); @@ -1106,8 +1104,6 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) } } - /* Protect against RPC races by saving the change attribute */ - verifier = nfs_save_change_attribute(inode); status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr); if (status != 0) { dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n", @@ -1122,7 +1118,7 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) } spin_lock(&inode->i_lock); - status = nfs_update_inode(inode, &fattr, verifier); + status = nfs_update_inode(inode, &fattr); if (status) { spin_unlock(&inode->i_lock); dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n", @@ -1130,20 +1126,11 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) (long long)NFS_FILEID(inode), status); goto out; } - cache_validity = nfsi->cache_validity; - nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE; - - /* - * We may need to keep the attributes marked as invalid if - * we raced with nfs_end_attr_update(). - */ - if (time_after_eq(verifier, nfsi->cache_change_attribute)) - nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); spin_unlock(&inode->i_lock); nfs_revalidate_mapping(inode, inode->i_mapping); - if (cache_validity & NFS_INO_INVALID_ACL) + if (nfsi->cache_validity & NFS_INO_INVALID_ACL) nfs_zap_acl_cache(inode); dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n", @@ -1346,10 +1333,8 @@ int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr) return 0; spin_lock(&inode->i_lock); nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE; - if (nfs_verify_change_attribute(inode, fattr->time_start)) - nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); if (time_after(fattr->time_start, nfsi->last_updated)) - status = nfs_update_inode(inode, fattr, fattr->time_start); + status = nfs_update_inode(inode, fattr); else status = nfs_check_inode_attributes(inode, fattr); @@ -1375,10 +1360,7 @@ int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr) nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS; goto out; } - status = nfs_update_inode(inode, fattr, fattr->time_start); - if (time_after_eq(fattr->time_start, nfsi->cache_change_attribute)) - nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE); - nfsi->cache_change_attribute = jiffies; + status = nfs_update_inode(inode, fattr); out: spin_unlock(&inode->i_lock); return status; @@ -1396,12 +1378,12 @@ out: * * A very similar scenario holds for the dir cache. */ -static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier) +static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) { struct nfs_inode *nfsi = NFS_I(inode); loff_t cur_isize, new_isize; unsigned int invalid = 0; - int data_unstable; + int data_stable; dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n", __FUNCTION__, inode->i_sb->s_id, inode->i_ino, @@ -1432,8 +1414,9 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign nfsi->last_updated = jiffies; /* Are we racing with known updates of the metadata on the server? */ - data_unstable = ! (nfs_verify_change_attribute(inode, verifier) || - (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)); + data_stable = nfs_verify_change_attribute(inode, fattr->time_start); + if (data_stable) + nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); /* Check if our cached file size is stale */ new_isize = nfs_size_to_loff_t(fattr->size); @@ -1442,7 +1425,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign /* Do we perhaps have any outstanding writes? */ if (nfsi->npages == 0) { /* No, but did we race with nfs_end_data_update()? */ - if (time_after_eq(verifier, nfsi->cache_change_attribute)) { + if (data_stable) { inode->i_size = new_isize; invalid |= NFS_INO_INVALID_DATA; } @@ -1451,6 +1434,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign inode->i_size = new_isize; invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; } + nfsi->cache_change_attribute = jiffies; dprintk("NFS: isize change on server for file %s/%ld\n", inode->i_sb->s_id, inode->i_ino); } @@ -1460,8 +1444,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime)); dprintk("NFS: mtime change on server for file %s/%ld\n", inode->i_sb->s_id, inode->i_ino); - if (!data_unstable) - invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; + invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; + nfsi->cache_change_attribute = jiffies; } if ((fattr->valid & NFS_ATTR_FATTR_V4) @@ -1469,15 +1453,15 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign dprintk("NFS: change_attr change on server for file %s/%ld\n", inode->i_sb->s_id, inode->i_ino); nfsi->change_attr = fattr->change_attr; - if (!data_unstable) - invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; + invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; + nfsi->cache_change_attribute = jiffies; } /* If ctime has changed we should definitely clear access+acl caches */ if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) { - if (!data_unstable) - invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; + invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime)); + nfsi->cache_change_attribute = jiffies; } memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime)); @@ -1515,6 +1499,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))) invalid &= ~NFS_INO_INVALID_DATA; + if (data_stable) + invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE); if (!nfs_have_delegation(inode, FMODE_READ)) nfsi->cache_validity |= invalid; -- cgit v1.2.2 From bb713d6d38f7be4f4e7d790cddb1b076e7da6699 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 3 Dec 2005 15:20:14 -0500 Subject: NFS: use set_page_writeback() in the appropriate places Ensure that we use set_page_writeback() in the appropriate places to help the VM in keeping its page radix_tree in sync. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 8f71e766cc..3107908e5f 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -189,6 +189,7 @@ static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode, (long long)NFS_FILEID(inode), count, (long long)(page_offset(page) + offset)); + set_page_writeback(page); nfs_begin_data_update(inode); do { if (count < wsize) @@ -221,6 +222,7 @@ static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode, io_error: nfs_end_data_update(inode); + end_page_writeback(page); nfs_writedata_free(wdata); return written ? written : result; } @@ -929,7 +931,7 @@ static int nfs_flush_multi(struct list_head *head, struct inode *inode, int how) atomic_set(&req->wb_complete, requests); ClearPageError(page); - SetPageWriteback(page); + set_page_writeback(page); offset = 0; nbytes = req->wb_bytes; do { @@ -992,7 +994,7 @@ static int nfs_flush_one(struct list_head *head, struct inode *inode, int how) nfs_list_remove_request(req); nfs_list_add_request(req, &data->pages); ClearPageError(req->wb_page); - SetPageWriteback(req->wb_page); + set_page_writeback(req->wb_page); *pages++ = req->wb_page; count += req->wb_bytes; } -- cgit v1.2.2 From 5ba7cc4801ae0fe74b6e0160f008521ae71d9f5d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 3 Dec 2005 15:20:17 -0500 Subject: NFS: Fix post-op attribute revalidation... - Missing nfs_mark_for_revalidate in nfs_proc_link() - Missing nfs_mark_for_revalidate in nfs_rename() Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 3 +++ fs/nfs/proc.c | 1 + 2 files changed, 4 insertions(+) (limited to 'fs/nfs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 7370583b61..c0d1a21457 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1287,6 +1287,7 @@ dentry->d_parent->d_name.name, dentry->d_name.name); nfs_begin_data_update(dentry->d_inode); error = NFS_PROTO(dir)->rename(dir, &dentry->d_name, dir, &qsilly); + nfs_mark_for_revalidate(dentry->d_inode); nfs_end_data_update(dentry->d_inode); } else error = NFS_PROTO(dir)->rename(dir, &dentry->d_name, @@ -1334,6 +1335,7 @@ static int nfs_safe_remove(struct dentry *dentry) /* The VFS may want to delete this inode */ if (error == 0) inode->i_nlink--; + nfs_mark_for_revalidate(inode); nfs_end_data_update(inode); } else error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); @@ -1556,6 +1558,7 @@ go_ahead: nfs_begin_data_update(old_inode); error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name, new_dir, &new_dentry->d_name); + nfs_mark_for_revalidate(old_inode); nfs_end_data_update(old_inode); nfs_end_data_update(new_dir); nfs_end_data_update(old_dir); diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index a48a003242..e1e3ca5d74 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -375,6 +375,7 @@ nfs_proc_link(struct inode *inode, struct inode *dir, struct qstr *name) dprintk("NFS call link %s\n", name->name); status = rpc_call(NFS_CLIENT(inode), NFSPROC_LINK, &arg, NULL, 0); + nfs_mark_for_revalidate(inode); nfs_mark_for_revalidate(dir); dprintk("NFS reply link: %d\n", status); return status; -- cgit v1.2.2 From 3b6efee9231e12fce09c94930bfc59f66f18d662 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 3 Dec 2005 15:20:21 -0500 Subject: NFSv4: Fix an Oops in the synchronous write path - Missing initialisation of attribute bitmask in _nfs4_proc_write() - On success, _nfs4_proc_write() must return number of bytes written. - Missing post_op_update_inode() in _nfs4_proc_write() - Missing initialisation of attribute bitmask in _nfs4_proc_commit() - Missing post_op_update_inode() in _nfs4_proc_commit() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 60e0dd800c..f988a9417b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1506,10 +1506,15 @@ static int _nfs4_proc_write(struct nfs_write_data *wdata) dprintk("NFS call write %d @ %Ld\n", wdata->args.count, (long long) wdata->args.offset); + wdata->args.bitmask = server->attr_bitmask; + wdata->res.server = server; nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, rpcflags); dprintk("NFS reply write: %d\n", status); - return status; + if (status < 0) + return status; + nfs_post_op_update_inode(inode, fattr); + return wdata->res.count; } static int nfs4_proc_write(struct nfs_write_data *wdata) @@ -1540,9 +1545,13 @@ static int _nfs4_proc_commit(struct nfs_write_data *cdata) dprintk("NFS call commit %d @ %Ld\n", cdata->args.count, (long long) cdata->args.offset); + cdata->args.bitmask = server->attr_bitmask; + cdata->res.server = server; nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, 0); dprintk("NFS reply commit: %d\n", status); + if (status >= 0) + nfs_post_op_update_inode(inode, fattr); return status; } -- cgit v1.2.2 From 29884df0d89c1df0dec3449405bc41569bb44800 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 13 Dec 2005 16:13:54 -0500 Subject: NFS: Fix another O_DIRECT race Ensure we call unmap_mapping_range() and sync dirty pages to disk before doing an NFS direct write. Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 24 ++++++------------------ fs/nfs/file.c | 23 ++++------------------- fs/nfs/inode.c | 28 +++++++++++++++++++++++----- 3 files changed, 33 insertions(+), 42 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index b497c71384..0792288176 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -678,15 +678,9 @@ nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t if (!count) goto out; - if (mapping->nrpages) { - retval = filemap_fdatawrite(mapping); - if (retval == 0) - retval = nfs_wb_all(inode); - if (retval == 0) - retval = filemap_fdatawait(mapping); - if (retval) - goto out; - } + retval = nfs_sync_mapping(mapping); + if (retval) + goto out; retval = nfs_direct_read(inode, ctx, &iov, pos, 1); if (retval > 0) @@ -764,15 +758,9 @@ nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, if (!count) goto out; - if (mapping->nrpages) { - retval = filemap_fdatawrite(mapping); - if (retval == 0) - retval = nfs_wb_all(inode); - if (retval == 0) - retval = filemap_fdatawait(mapping); - if (retval) - goto out; - } + retval = nfs_sync_mapping(mapping); + if (retval) + goto out; retval = nfs_direct_write(inode, ctx, &iov, pos, 1); if (mapping->nrpages) diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 57d3e77d97..eb5cd4c3bb 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -433,11 +433,7 @@ static int do_unlk(struct file *filp, int cmd, struct file_lock *fl) * Flush all pending writes before doing anything * with locks.. */ - filemap_fdatawrite(filp->f_mapping); - down(&inode->i_sem); - nfs_wb_all(inode); - up(&inode->i_sem); - filemap_fdatawait(filp->f_mapping); + nfs_sync_mapping(filp->f_mapping); /* NOTE: special case * If we're signalled while cleaning up locks on process exit, we @@ -465,15 +461,8 @@ static int do_setlk(struct file *filp, int cmd, struct file_lock *fl) * Flush all pending writes before doing anything * with locks.. */ - status = filemap_fdatawrite(filp->f_mapping); - if (status == 0) { - down(&inode->i_sem); - status = nfs_wb_all(inode); - up(&inode->i_sem); - if (status == 0) - status = filemap_fdatawait(filp->f_mapping); - } - if (status < 0) + status = nfs_sync_mapping(filp->f_mapping); + if (status != 0) goto out; lock_kernel(); @@ -497,11 +486,7 @@ static int do_setlk(struct file *filp, int cmd, struct file_lock *fl) * Make sure we clear the cache whenever we try to get the lock. * This makes locking act as a cache coherency point. */ - filemap_fdatawrite(filp->f_mapping); - down(&inode->i_sem); - nfs_wb_all(inode); /* we may have slept */ - up(&inode->i_sem); - filemap_fdatawait(filp->f_mapping); + nfs_sync_mapping(filp->f_mapping); nfs_zap_caches(inode); out: rpc_clnt_sigunmask(NFS_CLIENT(inode), &oldset); diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index afd75d0463..432f41cd75 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -640,6 +640,27 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt) return 0; } +/** + * nfs_sync_mapping - helper to flush all mmapped dirty data to disk + */ +int nfs_sync_mapping(struct address_space *mapping) +{ + int ret; + + if (mapping->nrpages == 0) + return 0; + unmap_mapping_range(mapping, 0, 0, 0); + ret = filemap_fdatawrite(mapping); + if (ret != 0) + goto out; + ret = filemap_fdatawait(mapping); + if (ret != 0) + goto out; + ret = nfs_wb_all(mapping->host); +out: + return ret; +} + /* * Invalidate the local caches */ @@ -1179,11 +1200,8 @@ void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping) struct nfs_inode *nfsi = NFS_I(inode); if (nfsi->cache_validity & NFS_INO_INVALID_DATA) { - if (S_ISREG(inode->i_mode)) { - if (filemap_fdatawrite(mapping) == 0) - filemap_fdatawait(mapping); - nfs_wb_all(inode); - } + if (S_ISREG(inode->i_mode)) + nfs_sync_mapping(mapping); invalidate_inode_pages2(mapping); spin_lock(&inode->i_lock); -- cgit v1.2.2 From 0800c5f7a499a8961c3868e95cc4cf1715166457 Mon Sep 17 00:00:00 2001 From: ASANO Masahiro Date: Thu, 22 Dec 2005 13:24:54 +0900 Subject: [PATCH] fix posix lock on NFS NFS client prevents mandatory lock, but there is a flaw on it; Locks are possibly left if the mode is changed while locking. This permits unlocking even if the mandatory lock bits are set. Signed-off-by: ASANO Masahiro Signed-off-by: Linus Torvalds --- fs/nfs/file.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/file.c b/fs/nfs/file.c index eb5cd4c3bb..7a79fbe9f5 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -509,7 +509,8 @@ static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) return -EINVAL; /* No mandatory locks over NFS */ - if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID) + if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID && + fl->fl_type != F_UNLCK) return -ENOLCK; if (IS_GETLK(cmd)) -- cgit v1.2.2 From 14c850212ed8f8cbb5972ad6b8812e08a0bc901c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 27 Dec 2005 02:43:12 -0200 Subject: [INET_SOCK]: Move struct inet_sock & helper functions to net/inet_sock.h To help in reducing the number of include dependencies, several files were touched as they were getting needed headers indirectly for stuff they use. Thanks also to Alan Menegotto for pointing out that net/dccp/proto.c had linux/dccp.h include twice. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- fs/nfs/callback.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs/nfs') diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c index f2ca782aba..30cae36028 100644 --- a/fs/nfs/callback.c +++ b/fs/nfs/callback.c @@ -14,6 +14,9 @@ #include #include #include + +#include + #include "nfs4_fs.h" #include "callback.h" -- cgit v1.2.2 From abd3e641d5ef9f836ab2f2b04d80b8619b051531 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:02 +0100 Subject: NFS: Work correctly with single-page ->writepage() calls Ensure that we always initiate flushing of data before we exit a single-page ->writepage() call. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 3107908e5f..95d00f9132 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -232,19 +232,16 @@ static int nfs_writepage_async(struct nfs_open_context *ctx, unsigned int offset, unsigned int count) { struct nfs_page *req; - int status; req = nfs_update_request(ctx, inode, page, offset, count); - status = (IS_ERR(req)) ? PTR_ERR(req) : 0; - if (status < 0) - goto out; + if (IS_ERR(req)) + return PTR_ERR(req); /* Update file length */ nfs_grow_file(page, offset, count); /* Set the PG_uptodate flag? */ nfs_mark_uptodate(page, offset, count); nfs_unlock_request(req); - out: - return status; + return 0; } static int wb_priority(struct writeback_control *wbc) @@ -304,11 +301,8 @@ do_it: lock_kernel(); if (!IS_SYNC(inode) && inode_referenced) { err = nfs_writepage_async(ctx, inode, page, 0, offset); - if (err >= 0) { - err = 0; - if (wbc->for_reclaim) - nfs_flush_inode(inode, 0, 0, FLUSH_STABLE); - } + if (!wbc->for_writepages) + nfs_flush_inode(inode, 0, 0, wb_priority(wbc)); } else { err = nfs_writepage_sync(ctx, inode, page, 0, offset, priority); -- cgit v1.2.2 From 963d8fe53339128ee46a7701f2e36305f0ccff8c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:04 +0100 Subject: RPC: Clean up RPC task structure Shrink the RPC task structure. Instead of storing separate pointers for task->tk_exit and task->tk_release, put them in a structure. Also pass the user data pointer as a parameter instead of passing it via task->tk_calldata. This enables us to nest callbacks. Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 1 - fs/nfs/nfs3proc.c | 44 +++++++++++++--------- fs/nfs/nfs4proc.c | 107 +++++++++++++++++++++++++++++++----------------------- fs/nfs/proc.c | 28 +++++++++----- fs/nfs/read.c | 10 ++--- fs/nfs/unlink.c | 19 +++++----- fs/nfs/write.c | 21 +++-------- 7 files changed, 126 insertions(+), 104 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 0792288176..a834423942 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -269,7 +269,6 @@ static void nfs_direct_read_schedule(struct nfs_direct_req *dreq, data->task.tk_cookie = (unsigned long) inode; data->task.tk_calldata = data; - data->task.tk_release = nfs_readdata_release; data->complete = nfs_direct_read_result; lock_kernel(); diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index 92c870d19c..c172a75846 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -732,19 +732,23 @@ nfs3_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, extern u32 *nfs3_decode_dirent(u32 *, struct nfs_entry *, int); -static void -nfs3_read_done(struct rpc_task *task) +static void nfs3_read_done(struct rpc_task *task, void *calldata) { - struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata; + struct nfs_read_data *data = calldata; if (nfs3_async_handle_jukebox(task)) return; /* Call back common NFS readpage processing */ if (task->tk_status >= 0) nfs_refresh_inode(data->inode, &data->fattr); - nfs_readpage_result(task); + nfs_readpage_result(task, calldata); } +static const struct rpc_call_ops nfs3_read_ops = { + .rpc_call_done = nfs3_read_done, + .rpc_release = nfs_readdata_release, +}; + static void nfs3_proc_read_setup(struct nfs_read_data *data) { @@ -762,23 +766,26 @@ nfs3_proc_read_setup(struct nfs_read_data *data) flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0); /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs3_read_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_read_ops, data); rpc_call_setup(task, &msg, 0); } -static void -nfs3_write_done(struct rpc_task *task) +static void nfs3_write_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data; + struct nfs_write_data *data = calldata; if (nfs3_async_handle_jukebox(task)) return; - data = (struct nfs_write_data *)task->tk_calldata; if (task->tk_status >= 0) nfs_post_op_update_inode(data->inode, data->res.fattr); - nfs_writeback_done(task); + nfs_writeback_done(task, calldata); } +static const struct rpc_call_ops nfs3_write_ops = { + .rpc_call_done = nfs3_write_done, + .rpc_release = nfs_writedata_release, +}; + static void nfs3_proc_write_setup(struct nfs_write_data *data, int how) { @@ -806,23 +813,26 @@ nfs3_proc_write_setup(struct nfs_write_data *data, int how) flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs3_write_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_write_ops, data); rpc_call_setup(task, &msg, 0); } -static void -nfs3_commit_done(struct rpc_task *task) +static void nfs3_commit_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data; + struct nfs_write_data *data = calldata; if (nfs3_async_handle_jukebox(task)) return; - data = (struct nfs_write_data *)task->tk_calldata; if (task->tk_status >= 0) nfs_post_op_update_inode(data->inode, data->res.fattr); - nfs_commit_done(task); + nfs_commit_done(task, calldata); } +static const struct rpc_call_ops nfs3_commit_ops = { + .rpc_call_done = nfs3_commit_done, + .rpc_release = nfs_commit_release, +}; + static void nfs3_proc_commit_setup(struct nfs_write_data *data, int how) { @@ -840,7 +850,7 @@ nfs3_proc_commit_setup(struct nfs_write_data *data, int how) flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs3_commit_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_commit_ops, data); rpc_call_setup(task, &msg, 0); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f988a9417b..3d5d3c07d6 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -196,14 +196,12 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf /* Helper for asynchronous RPC calls */ static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin, - rpc_action tk_exit, void *calldata) + const struct rpc_call_ops *tk_ops, void *calldata) { struct rpc_task *task; - if (!(task = rpc_new_task(clnt, tk_exit, RPC_TASK_ASYNC))) + if (!(task = rpc_new_task(clnt, RPC_TASK_ASYNC, tk_ops, calldata))) return -ENOMEM; - - task->tk_calldata = calldata; task->tk_action = tk_begin; rpc_execute(task); return 0; @@ -867,10 +865,10 @@ struct nfs4_closedata { struct nfs_fattr fattr; }; -static void nfs4_free_closedata(struct nfs4_closedata *calldata) +static void nfs4_free_closedata(void *data) { - struct nfs4_state *state = calldata->state; - struct nfs4_state_owner *sp = state->owner; + struct nfs4_closedata *calldata = data; + struct nfs4_state_owner *sp = calldata->state->owner; nfs4_put_open_state(calldata->state); nfs_free_seqid(calldata->arg.seqid); @@ -878,9 +876,9 @@ static void nfs4_free_closedata(struct nfs4_closedata *calldata) kfree(calldata); } -static void nfs4_close_done(struct rpc_task *task) +static void nfs4_close_done(struct rpc_task *task, void *data) { - struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; + struct nfs4_closedata *calldata = data; struct nfs4_state *state = calldata->state; struct nfs_server *server = NFS_SERVER(calldata->inode); @@ -904,7 +902,6 @@ static void nfs4_close_done(struct rpc_task *task) } } nfs_refresh_inode(calldata->inode, calldata->res.fattr); - nfs4_free_closedata(calldata); } static void nfs4_close_begin(struct rpc_task *task) @@ -918,10 +915,8 @@ static void nfs4_close_begin(struct rpc_task *task) .rpc_cred = state->owner->so_cred, }; int mode = 0, old_mode; - int status; - status = nfs_wait_on_sequence(calldata->arg.seqid, task); - if (status != 0) + if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0) return; /* Recalculate the new open mode in case someone reopened the file * while we were waiting in line to be scheduled. @@ -937,9 +932,8 @@ static void nfs4_close_begin(struct rpc_task *task) spin_unlock(&calldata->inode->i_lock); spin_unlock(&state->owner->so_lock); if (mode == old_mode || test_bit(NFS_DELEGATED_STATE, &state->flags)) { - nfs4_free_closedata(calldata); - task->tk_exit = NULL; - rpc_exit(task, 0); + /* Note: exit _without_ calling nfs4_close_done */ + task->tk_action = NULL; return; } nfs_fattr_init(calldata->res.fattr); @@ -949,6 +943,11 @@ static void nfs4_close_begin(struct rpc_task *task) rpc_call_setup(task, &msg, 0); } +static const struct rpc_call_ops nfs4_close_ops = { + .rpc_call_done = nfs4_close_done, + .rpc_release = nfs4_free_closedata, +}; + /* * It is possible for data to be read/written from a mem-mapped file * after the sys_close call (which hits the vfs layer as a flush). @@ -982,7 +981,7 @@ int nfs4_do_close(struct inode *inode, struct nfs4_state *state) calldata->res.server = server; status = nfs4_call_async(server->client, nfs4_close_begin, - nfs4_close_done, calldata); + &nfs4_close_ops, calldata); if (status == 0) goto out; @@ -2125,10 +2124,9 @@ static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, return err; } -static void -nfs4_read_done(struct rpc_task *task) +static void nfs4_read_done(struct rpc_task *task, void *calldata) { - struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata; + struct nfs_read_data *data = calldata; struct inode *inode = data->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) { @@ -2138,9 +2136,14 @@ nfs4_read_done(struct rpc_task *task) if (task->tk_status > 0) renew_lease(NFS_SERVER(inode), data->timestamp); /* Call back common NFS readpage processing */ - nfs_readpage_result(task); + nfs_readpage_result(task, calldata); } +static const struct rpc_call_ops nfs4_read_ops = { + .rpc_call_done = nfs4_read_done, + .rpc_release = nfs_readdata_release, +}; + static void nfs4_proc_read_setup(struct nfs_read_data *data) { @@ -2160,14 +2163,13 @@ nfs4_proc_read_setup(struct nfs_read_data *data) flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0); /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs4_read_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_read_ops, data); rpc_call_setup(task, &msg, 0); } -static void -nfs4_write_done(struct rpc_task *task) +static void nfs4_write_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; + struct nfs_write_data *data = calldata; struct inode *inode = data->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) { @@ -2179,9 +2181,14 @@ nfs4_write_done(struct rpc_task *task) nfs_post_op_update_inode(inode, data->res.fattr); } /* Call back common NFS writeback processing */ - nfs_writeback_done(task); + nfs_writeback_done(task, calldata); } +static const struct rpc_call_ops nfs4_write_ops = { + .rpc_call_done = nfs4_write_done, + .rpc_release = nfs_writedata_release, +}; + static void nfs4_proc_write_setup(struct nfs_write_data *data, int how) { @@ -2214,14 +2221,13 @@ nfs4_proc_write_setup(struct nfs_write_data *data, int how) flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs4_write_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_write_ops, data); rpc_call_setup(task, &msg, 0); } -static void -nfs4_commit_done(struct rpc_task *task) +static void nfs4_commit_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; + struct nfs_write_data *data = calldata; struct inode *inode = data->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) { @@ -2231,9 +2237,14 @@ nfs4_commit_done(struct rpc_task *task) if (task->tk_status >= 0) nfs_post_op_update_inode(inode, data->res.fattr); /* Call back common NFS writeback processing */ - nfs_commit_done(task); + nfs_commit_done(task, calldata); } +static const struct rpc_call_ops nfs4_commit_ops = { + .rpc_call_done = nfs4_commit_done, + .rpc_release = nfs_commit_release, +}; + static void nfs4_proc_commit_setup(struct nfs_write_data *data, int how) { @@ -2255,7 +2266,7 @@ nfs4_proc_commit_setup(struct nfs_write_data *data, int how) flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs4_commit_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_commit_ops, data); rpc_call_setup(task, &msg, 0); } @@ -2263,11 +2274,10 @@ nfs4_proc_commit_setup(struct nfs_write_data *data, int how) * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special * standalone procedure for queueing an asynchronous RENEW. */ -static void -renew_done(struct rpc_task *task) +static void nfs4_renew_done(struct rpc_task *task, void *data) { struct nfs4_client *clp = (struct nfs4_client *)task->tk_msg.rpc_argp; - unsigned long timestamp = (unsigned long)task->tk_calldata; + unsigned long timestamp = (unsigned long)data; if (task->tk_status < 0) { switch (task->tk_status) { @@ -2284,6 +2294,10 @@ renew_done(struct rpc_task *task) spin_unlock(&clp->cl_lock); } +static const struct rpc_call_ops nfs4_renew_ops = { + .rpc_call_done = nfs4_renew_done, +}; + int nfs4_proc_async_renew(struct nfs4_client *clp) { @@ -2294,7 +2308,7 @@ nfs4_proc_async_renew(struct nfs4_client *clp) }; return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT, - renew_done, (void *)jiffies); + &nfs4_renew_ops, (void *)jiffies); } int @@ -2866,15 +2880,16 @@ static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata) } } -static void nfs4_locku_complete(struct nfs4_unlockdata *calldata) +static void nfs4_locku_complete(void *data) { + struct nfs4_unlockdata *calldata = data; complete(&calldata->completion); nfs4_locku_release_calldata(calldata); } -static void nfs4_locku_done(struct rpc_task *task) +static void nfs4_locku_done(struct rpc_task *task, void *data) { - struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; + struct nfs4_unlockdata *calldata = data; nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid); switch (task->tk_status) { @@ -2890,10 +2905,8 @@ static void nfs4_locku_done(struct rpc_task *task) default: if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) { rpc_restart_call(task); - return; } } - nfs4_locku_complete(calldata); } static void nfs4_locku_begin(struct rpc_task *task) @@ -2911,14 +2924,18 @@ static void nfs4_locku_begin(struct rpc_task *task) if (status != 0) return; if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) { - nfs4_locku_complete(calldata); - task->tk_exit = NULL; - rpc_exit(task, 0); + /* Note: exit _without_ running nfs4_locku_done */ + task->tk_action = NULL; return; } rpc_call_setup(task, &msg, 0); } +static const struct rpc_call_ops nfs4_locku_ops = { + .rpc_call_done = nfs4_locku_done, + .rpc_release = nfs4_locku_complete, +}; + static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) { struct nfs4_unlockdata *calldata; @@ -2963,7 +2980,7 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * init_completion(&calldata->completion); status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin, - nfs4_locku_done, calldata); + &nfs4_locku_ops, calldata); if (status == 0) wait_for_completion_interruptible(&calldata->completion); do_vfs_lock(request->fl_file, request); diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index e1e3ca5d74..6145e82b45 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -547,10 +547,9 @@ nfs_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, extern u32 * nfs_decode_dirent(u32 *, struct nfs_entry *, int); -static void -nfs_read_done(struct rpc_task *task) +static void nfs_read_done(struct rpc_task *task, void *calldata) { - struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata; + struct nfs_read_data *data = calldata; if (task->tk_status >= 0) { nfs_refresh_inode(data->inode, data->res.fattr); @@ -560,9 +559,14 @@ nfs_read_done(struct rpc_task *task) if (data->args.offset + data->args.count >= data->res.fattr->size) data->res.eof = 1; } - nfs_readpage_result(task); + nfs_readpage_result(task, calldata); } +static const struct rpc_call_ops nfs_read_ops = { + .rpc_call_done = nfs_read_done, + .rpc_release = nfs_readdata_release, +}; + static void nfs_proc_read_setup(struct nfs_read_data *data) { @@ -580,20 +584,24 @@ nfs_proc_read_setup(struct nfs_read_data *data) flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0); /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs_read_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs_read_ops, data); rpc_call_setup(task, &msg, 0); } -static void -nfs_write_done(struct rpc_task *task) +static void nfs_write_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; + struct nfs_write_data *data = calldata; if (task->tk_status >= 0) nfs_post_op_update_inode(data->inode, data->res.fattr); - nfs_writeback_done(task); + nfs_writeback_done(task, calldata); } +static const struct rpc_call_ops nfs_write_ops = { + .rpc_call_done = nfs_write_done, + .rpc_release = nfs_writedata_release, +}; + static void nfs_proc_write_setup(struct nfs_write_data *data, int how) { @@ -614,7 +622,7 @@ nfs_proc_write_setup(struct nfs_write_data *data, int how) flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; /* Finalize the task. */ - rpc_init_task(task, NFS_CLIENT(inode), nfs_write_done, flags); + rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs_write_ops, data); rpc_call_setup(task, &msg, 0); } diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 5f20eafba8..21486242c3 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -42,9 +42,8 @@ mempool_t *nfs_rdata_mempool; #define MIN_POOL_READ (32) -void nfs_readdata_release(struct rpc_task *task) +void nfs_readdata_release(void *data) { - struct nfs_read_data *data = (struct nfs_read_data *)task->tk_calldata; nfs_readdata_free(data); } @@ -220,9 +219,6 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, NFS_PROTO(inode)->read_setup(data); data->task.tk_cookie = (unsigned long)inode; - data->task.tk_calldata = data; - /* Release requests */ - data->task.tk_release = nfs_readdata_release; dprintk("NFS: %4d initiated read call (req %s/%Ld, %u bytes @ offset %Lu)\n", data->task.tk_pid, @@ -452,9 +448,9 @@ static void nfs_readpage_result_full(struct nfs_read_data *data, int status) * This is the callback from RPC telling us whether a reply was * received or some error occurred (timeout or socket shutdown). */ -void nfs_readpage_result(struct rpc_task *task) +void nfs_readpage_result(struct rpc_task *task, void *calldata) { - struct nfs_read_data *data = (struct nfs_read_data *)task->tk_calldata; + struct nfs_read_data *data = calldata; struct nfs_readargs *argp = &data->args; struct nfs_readres *resp = &data->res; int status = task->tk_status; diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c index d639d172d5..1494484ba8 100644 --- a/fs/nfs/unlink.c +++ b/fs/nfs/unlink.c @@ -116,10 +116,9 @@ nfs_async_unlink_init(struct rpc_task *task) * * Do the directory attribute update. */ -static void -nfs_async_unlink_done(struct rpc_task *task) +static void nfs_async_unlink_done(struct rpc_task *task, void *calldata) { - struct nfs_unlinkdata *data = (struct nfs_unlinkdata *)task->tk_calldata; + struct nfs_unlinkdata *data = calldata; struct dentry *dir = data->dir; struct inode *dir_i; @@ -141,13 +140,17 @@ nfs_async_unlink_done(struct rpc_task *task) * We need to call nfs_put_unlinkdata as a 'tk_release' task since the * rpc_task would be freed too. */ -static void -nfs_async_unlink_release(struct rpc_task *task) +static void nfs_async_unlink_release(void *calldata) { - struct nfs_unlinkdata *data = (struct nfs_unlinkdata *)task->tk_calldata; + struct nfs_unlinkdata *data = calldata; nfs_put_unlinkdata(data); } +static const struct rpc_call_ops nfs_unlink_ops = { + .rpc_call_done = nfs_async_unlink_done, + .rpc_release = nfs_async_unlink_release, +}; + /** * nfs_async_unlink - asynchronous unlinking of a file * @dentry: dentry to unlink @@ -179,10 +182,8 @@ nfs_async_unlink(struct dentry *dentry) data->count = 1; task = &data->task; - rpc_init_task(task, clnt, nfs_async_unlink_done , RPC_TASK_ASYNC); - task->tk_calldata = data; + rpc_init_task(task, clnt, RPC_TASK_ASYNC, &nfs_unlink_ops, data); task->tk_action = nfs_async_unlink_init; - task->tk_release = nfs_async_unlink_release; spin_lock(&dentry->d_lock); dentry->d_flags |= DCACHE_NFSFS_RENAMED; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 95d00f9132..80bc4ea1b8 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -104,9 +104,8 @@ static inline void nfs_commit_free(struct nfs_write_data *p) mempool_free(p, nfs_commit_mempool); } -static void nfs_writedata_release(struct rpc_task *task) +void nfs_writedata_release(void *wdata) { - struct nfs_write_data *wdata = (struct nfs_write_data *)task->tk_calldata; nfs_writedata_free(wdata); } @@ -871,9 +870,6 @@ static void nfs_write_rpcsetup(struct nfs_page *req, data->task.tk_priority = flush_task_priority(how); data->task.tk_cookie = (unsigned long)inode; - data->task.tk_calldata = data; - /* Release requests */ - data->task.tk_release = nfs_writedata_release; dprintk("NFS: %4d initiated write call (req %s/%Ld, %u bytes @ offset %Lu)\n", data->task.tk_pid, @@ -1131,9 +1127,9 @@ static void nfs_writeback_done_full(struct nfs_write_data *data, int status) /* * This function is called when the WRITE call is complete. */ -void nfs_writeback_done(struct rpc_task *task) +void nfs_writeback_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; + struct nfs_write_data *data = calldata; struct nfs_writeargs *argp = &data->args; struct nfs_writeres *resp = &data->res; @@ -1200,9 +1196,8 @@ void nfs_writeback_done(struct rpc_task *task) #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) -static void nfs_commit_release(struct rpc_task *task) +void nfs_commit_release(void *wdata) { - struct nfs_write_data *wdata = (struct nfs_write_data *)task->tk_calldata; nfs_commit_free(wdata); } @@ -1238,9 +1233,6 @@ static void nfs_commit_rpcsetup(struct list_head *head, data->task.tk_priority = flush_task_priority(how); data->task.tk_cookie = (unsigned long)inode; - data->task.tk_calldata = data; - /* Release requests */ - data->task.tk_release = nfs_commit_release; dprintk("NFS: %4d initiated commit call\n", data->task.tk_pid); } @@ -1277,10 +1269,9 @@ nfs_commit_list(struct list_head *head, int how) /* * COMMIT call returned */ -void -nfs_commit_done(struct rpc_task *task) +void nfs_commit_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = (struct nfs_write_data *)task->tk_calldata; + struct nfs_write_data *data = calldata; struct nfs_page *req; int res = 0; -- cgit v1.2.2 From 4ce70ada1ff1d0b80916ec9ec5764ce44a50a54f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:05 +0100 Subject: SUNRPC: Further cleanups Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 21 +++++++++++---------- fs/nfs/unlink.c | 13 +++++-------- 2 files changed, 16 insertions(+), 18 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3d5d3c07d6..368b75b3bc 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -195,14 +195,13 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf } /* Helper for asynchronous RPC calls */ -static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin, +static int nfs4_call_async(struct rpc_clnt *clnt, const struct rpc_call_ops *tk_ops, void *calldata) { struct rpc_task *task; if (!(task = rpc_new_task(clnt, RPC_TASK_ASYNC, tk_ops, calldata))) return -ENOMEM; - task->tk_action = tk_begin; rpc_execute(task); return 0; } @@ -882,6 +881,8 @@ static void nfs4_close_done(struct rpc_task *task, void *data) struct nfs4_state *state = calldata->state; struct nfs_server *server = NFS_SERVER(calldata->inode); + if (RPC_ASSASSINATED(task)) + return; /* hmm. we are done with the inode, and in the process of freeing * the state_owner. we keep this around to process errors */ @@ -904,9 +905,9 @@ static void nfs4_close_done(struct rpc_task *task, void *data) nfs_refresh_inode(calldata->inode, calldata->res.fattr); } -static void nfs4_close_begin(struct rpc_task *task) +static void nfs4_close_prepare(struct rpc_task *task, void *data) { - struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; + struct nfs4_closedata *calldata = data; struct nfs4_state *state = calldata->state; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE], @@ -944,6 +945,7 @@ static void nfs4_close_begin(struct rpc_task *task) } static const struct rpc_call_ops nfs4_close_ops = { + .rpc_call_prepare = nfs4_close_prepare, .rpc_call_done = nfs4_close_done, .rpc_release = nfs4_free_closedata, }; @@ -980,8 +982,7 @@ int nfs4_do_close(struct inode *inode, struct nfs4_state *state) calldata->res.fattr = &calldata->fattr; calldata->res.server = server; - status = nfs4_call_async(server->client, nfs4_close_begin, - &nfs4_close_ops, calldata); + status = nfs4_call_async(server->client, &nfs4_close_ops, calldata); if (status == 0) goto out; @@ -2909,9 +2910,9 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) } } -static void nfs4_locku_begin(struct rpc_task *task) +static void nfs4_locku_prepare(struct rpc_task *task, void *data) { - struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; + struct nfs4_unlockdata *calldata = data; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU], .rpc_argp = &calldata->arg, @@ -2932,6 +2933,7 @@ static void nfs4_locku_begin(struct rpc_task *task) } static const struct rpc_call_ops nfs4_locku_ops = { + .rpc_call_prepare = nfs4_locku_prepare, .rpc_call_done = nfs4_locku_done, .rpc_release = nfs4_locku_complete, }; @@ -2979,8 +2981,7 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * atomic_set(&calldata->refcount, 2); init_completion(&calldata->completion); - status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin, - &nfs4_locku_ops, calldata); + status = nfs4_call_async(NFS_SERVER(inode)->client, &nfs4_locku_ops, calldata); if (status == 0) wait_for_completion_interruptible(&calldata->completion); do_vfs_lock(request->fl_file, request); diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c index 1494484ba8..a65c7b53d5 100644 --- a/fs/nfs/unlink.c +++ b/fs/nfs/unlink.c @@ -87,10 +87,9 @@ nfs_copy_dname(struct dentry *dentry, struct nfs_unlinkdata *data) * We delay initializing RPC info until after the call to dentry_iput() * in order to minimize races against rename(). */ -static void -nfs_async_unlink_init(struct rpc_task *task) +static void nfs_async_unlink_init(struct rpc_task *task, void *calldata) { - struct nfs_unlinkdata *data = (struct nfs_unlinkdata *)task->tk_calldata; + struct nfs_unlinkdata *data = calldata; struct dentry *dir = data->dir; struct rpc_message msg = { .rpc_cred = data->cred, @@ -147,6 +146,7 @@ static void nfs_async_unlink_release(void *calldata) } static const struct rpc_call_ops nfs_unlink_ops = { + .rpc_call_prepare = nfs_async_unlink_init, .rpc_call_done = nfs_async_unlink_done, .rpc_release = nfs_async_unlink_release, }; @@ -160,7 +160,6 @@ nfs_async_unlink(struct dentry *dentry) { struct dentry *dir = dentry->d_parent; struct nfs_unlinkdata *data; - struct rpc_task *task; struct rpc_clnt *clnt = NFS_CLIENT(dir->d_inode); int status = -ENOMEM; @@ -181,15 +180,13 @@ nfs_async_unlink(struct dentry *dentry) nfs_deletes = data; data->count = 1; - task = &data->task; - rpc_init_task(task, clnt, RPC_TASK_ASYNC, &nfs_unlink_ops, data); - task->tk_action = nfs_async_unlink_init; + rpc_init_task(&data->task, clnt, RPC_TASK_ASYNC, &nfs_unlink_ops, data); spin_lock(&dentry->d_lock); dentry->d_flags |= DCACHE_NFSFS_RENAMED; spin_unlock(&dentry->d_lock); - rpc_sleep_on(&nfs_delete_queue, task, NULL, NULL); + rpc_sleep_on(&nfs_delete_queue, &data->task, NULL, NULL); status = 0; out: return status; -- cgit v1.2.2 From 44c288732fdbd7e38460d156a40d29590bf93bce Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:06 +0100 Subject: NFSv4: stateful NFSv4 RPC call interface The NFSv4 model requires us to complete all RPC calls that might establish state on the server whether or not the user wants to interrupt it. We may also need to schedule new work (including new RPC calls) in order to cancel the new state. The asynchronous RPC model will allow us to ensure that RPC calls always complete, but in order to allow for "synchronous" RPC, we want to add the ability to wait for completion. The waits are, of course, interruptible. Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index a834423942..ae2be07441 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -268,7 +268,6 @@ static void nfs_direct_read_schedule(struct nfs_direct_req *dreq, NFS_PROTO(inode)->read_setup(data); data->task.tk_cookie = (unsigned long) inode; - data->task.tk_calldata = data; data->complete = nfs_direct_read_result; lock_kernel(); -- cgit v1.2.2 From 06f814a3ad0ddfe19e6e4f44e3da5d490547faf3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:07 +0100 Subject: NFSv4: Make locku use the new RPC "wait on completion" interface. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 65 ++++++++++++++++++++++++++++++------------------------- 1 file changed, 36 insertions(+), 29 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 368b75b3bc..c7bec43192 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -206,6 +206,17 @@ static int nfs4_call_async(struct rpc_clnt *clnt, return 0; } +static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task) +{ + sigset_t oldset; + int ret; + + rpc_clnt_sigmask(task->tk_client, &oldset); + ret = rpc_wait_for_completion_task(task); + rpc_clnt_sigunmask(task->tk_client, &oldset); + return ret; +} + static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) { struct inode *inode = state->inode; @@ -2867,31 +2878,23 @@ struct nfs4_unlockdata { struct nfs_lockres res; struct nfs4_lock_state *lsp; struct nfs_open_context *ctx; - atomic_t refcount; - struct completion completion; }; -static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata) -{ - if (atomic_dec_and_test(&calldata->refcount)) { - nfs_free_seqid(calldata->luargs.seqid); - nfs4_put_lock_state(calldata->lsp); - put_nfs_open_context(calldata->ctx); - kfree(calldata); - } -} - -static void nfs4_locku_complete(void *data) +static void nfs4_locku_release_calldata(void *data) { struct nfs4_unlockdata *calldata = data; - complete(&calldata->completion); - nfs4_locku_release_calldata(calldata); + nfs_free_seqid(calldata->luargs.seqid); + nfs4_put_lock_state(calldata->lsp); + put_nfs_open_context(calldata->ctx); + kfree(calldata); } static void nfs4_locku_done(struct rpc_task *task, void *data) { struct nfs4_unlockdata *calldata = data; + if (RPC_ASSASSINATED(task)) + return; nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid); switch (task->tk_status) { case 0: @@ -2935,7 +2938,7 @@ static void nfs4_locku_prepare(struct rpc_task *task, void *data) static const struct rpc_call_ops nfs4_locku_ops = { .rpc_call_prepare = nfs4_locku_prepare, .rpc_call_done = nfs4_locku_done, - .rpc_release = nfs4_locku_complete, + .rpc_release = nfs4_locku_release_calldata, }; static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) @@ -2944,26 +2947,28 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp; - int status; + struct rpc_task *task; + int status = 0; /* Is this a delegated lock? */ if (test_bit(NFS_DELEGATED_STATE, &state->flags)) - return do_vfs_lock(request->fl_file, request); + goto out; status = nfs4_set_lock_state(state, request); if (status != 0) - return status; + goto out; lsp = request->fl_u.nfs4_fl.owner; /* We might have lost the locks! */ if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) - return 0; + goto out; + status = -ENOMEM; calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); if (calldata == NULL) - return -ENOMEM; + goto out; calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); if (calldata->luargs.seqid == NULL) { kfree(calldata); - return -ENOMEM; + goto out; } calldata->luargs.stateid = &lsp->ls_stateid; calldata->arg.fh = NFS_FH(inode); @@ -2978,14 +2983,16 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * /* Ensure we don't close file until we're done freeing locks! */ calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data); - atomic_set(&calldata->refcount, 2); - init_completion(&calldata->completion); - - status = nfs4_call_async(NFS_SERVER(inode)->client, &nfs4_locku_ops, calldata); - if (status == 0) - wait_for_completion_interruptible(&calldata->completion); + task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_locku_ops, calldata); + if (!IS_ERR(task)) { + status = nfs4_wait_for_completion_rpc_task(task); + rpc_release_task(task); + } else { + status = PTR_ERR(task); + nfs4_locku_release_calldata(calldata); + } +out: do_vfs_lock(request->fl_file, request); - nfs4_locku_release_calldata(calldata); return status; } -- cgit v1.2.2 From e56e0b78eb1097a8e06512b9ed4be94d7538e7ac Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:08 +0100 Subject: NFSv4: Allocate OPEN call RPC arguments using kmalloc() Cleanup in preparation for making OPEN calls interruptible by the user. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 213 ++++++++++++++++++++++++++++++------------------------ 1 file changed, 117 insertions(+), 96 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index c7bec43192..4a5cc84021 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -194,6 +194,76 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf spin_unlock(&inode->i_lock); } +struct nfs4_opendata { + struct nfs_openargs o_arg; + struct nfs_openres o_res; + struct nfs_fattr f_attr; + struct nfs_fattr dir_attr; + struct dentry *dentry; + struct dentry *dir; + struct nfs4_state_owner *owner; + struct iattr attrs; +}; + +static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry, + struct nfs4_state_owner *sp, int flags, + const struct iattr *attrs) +{ + struct dentry *parent = dget_parent(dentry); + struct inode *dir = parent->d_inode; + struct nfs_server *server = NFS_SERVER(dir); + struct nfs4_opendata *p; + + p = kzalloc(sizeof(*p), GFP_KERNEL); + if (p == NULL) + goto err; + p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + if (p->o_arg.seqid == NULL) + goto err_free; + p->dentry = dget(dentry); + p->dir = parent; + p->owner = sp; + atomic_inc(&sp->so_count); + p->o_arg.fh = NFS_FH(dir); + p->o_arg.open_flags = flags, + p->o_arg.clientid = server->nfs4_state->cl_clientid; + p->o_arg.id = sp->so_id; + p->o_arg.name = &dentry->d_name; + p->o_arg.server = server; + p->o_arg.bitmask = server->attr_bitmask; + p->o_arg.claim = NFS4_OPEN_CLAIM_NULL; + p->o_res.f_attr = &p->f_attr; + p->o_res.dir_attr = &p->dir_attr; + p->o_res.server = server; + nfs_fattr_init(&p->f_attr); + nfs_fattr_init(&p->dir_attr); + if (flags & O_EXCL) { + u32 *s = (u32 *) p->o_arg.u.verifier.data; + s[0] = jiffies; + s[1] = current->pid; + } else if (flags & O_CREAT) { + p->o_arg.u.attrs = &p->attrs; + memcpy(&p->attrs, attrs, sizeof(p->attrs)); + } + return p; +err_free: + kfree(p); +err: + dput(parent); + return NULL; +} + +static void nfs4_opendata_free(struct nfs4_opendata *p) +{ + if (p != NULL) { + nfs_free_seqid(p->o_arg.seqid); + nfs4_put_state_owner(p->owner); + dput(p->dir); + dput(p->dentry); + kfree(p); + } +} + /* Helper for asynchronous RPC calls */ static int nfs4_call_async(struct rpc_clnt *clnt, const struct rpc_call_ops *tk_ops, void *calldata) @@ -314,57 +384,45 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state struct nfs4_state_owner *sp = state->owner; struct inode *inode = dentry->d_inode; struct nfs_server *server = NFS_SERVER(inode); - struct dentry *parent = dget_parent(dentry); - struct nfs_openargs arg = { - .fh = NFS_FH(parent->d_inode), - .clientid = server->nfs4_state->cl_clientid, - .name = &dentry->d_name, - .id = sp->so_id, - .server = server, - .bitmask = server->attr_bitmask, - .claim = NFS4_OPEN_CLAIM_DELEGATE_CUR, - }; - struct nfs_openres res = { - .server = server, - }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR], - .rpc_argp = &arg, - .rpc_resp = &res, .rpc_cred = sp->so_cred, }; + struct nfs4_opendata *opendata; int status = 0; if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) goto out; if (state->state == 0) goto out; - arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + opendata = nfs4_opendata_alloc(dentry, sp, state->state, NULL); status = -ENOMEM; - if (arg.seqid == NULL) + if (opendata == NULL) goto out; - arg.open_flags = state->state; - memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data)); + opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR; + msg.rpc_argp = &opendata->o_arg; + msg.rpc_resp = &opendata->o_res; + memcpy(opendata->o_arg.u.delegation.data, state->stateid.data, + sizeof(opendata->o_arg.u.delegation.data)); status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs_increment_open_seqid(status, arg.seqid); + nfs_increment_open_seqid(status, opendata->o_arg.seqid); if (status != 0) goto out_free; - if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) { + if(opendata->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM) { status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode), - sp, &res.stateid, arg.seqid); + sp, &opendata->o_res.stateid, opendata->o_arg.seqid); if (status != 0) goto out_free; } nfs_confirm_seqid(&sp->so_seqid, 0); if (status >= 0) { - memcpy(state->stateid.data, res.stateid.data, + memcpy(state->stateid.data, opendata->o_res.stateid.data, sizeof(state->stateid.data)); clear_bit(NFS_DELEGATED_STATE, &state->flags); } out_free: - nfs_free_seqid(arg.seqid); + nfs4_opendata_free(opendata); out: - dput(parent); return status; } @@ -506,21 +564,8 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st struct dentry *parent = dget_parent(dentry); struct inode *dir = parent->d_inode; struct inode *inode = state->inode; - struct nfs_server *server = NFS_SERVER(dir); struct nfs_delegation *delegation = NFS_I(inode)->delegation; - struct nfs_fattr f_attr, dir_attr; - struct nfs_openargs o_arg = { - .fh = NFS_FH(dir), - .open_flags = state->state, - .name = &dentry->d_name, - .bitmask = server->attr_bitmask, - .claim = NFS4_OPEN_CLAIM_NULL, - }; - struct nfs_openres o_res = { - .f_attr = &f_attr, - .dir_attr = &dir_attr, - .server = server, - }; + struct nfs4_opendata *opendata; int status = 0; if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) { @@ -531,38 +576,38 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st set_bit(NFS_DELEGATED_STATE, &state->flags); goto out; } - o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); status = -ENOMEM; - if (o_arg.seqid == NULL) + opendata = nfs4_opendata_alloc(dentry, sp, state->state, NULL); + if (opendata == NULL) goto out; - nfs_fattr_init(&f_attr); - nfs_fattr_init(&dir_attr); - status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); + status = _nfs4_proc_open(dir, sp, &opendata->o_arg, &opendata->o_res); if (status != 0) goto out_nodeleg; /* Check if files differ */ - if ((f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT)) + if ((opendata->f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT)) goto out_stale; /* Has the file handle changed? */ - if (nfs_compare_fh(&o_res.fh, NFS_FH(inode)) != 0) { + if (nfs_compare_fh(&opendata->o_res.fh, NFS_FH(inode)) != 0) { /* Verify if the change attributes are the same */ - if (f_attr.change_attr != NFS_I(inode)->change_attr) + if (opendata->f_attr.change_attr != NFS_I(inode)->change_attr) goto out_stale; - if (nfs_size_to_loff_t(f_attr.size) != inode->i_size) + if (nfs_size_to_loff_t(opendata->f_attr.size) != inode->i_size) goto out_stale; /* Lets just pretend that this is the same file */ - nfs_copy_fh(NFS_FH(inode), &o_res.fh); - NFS_I(inode)->fileid = f_attr.fileid; + nfs_copy_fh(NFS_FH(inode), &opendata->o_res.fh); + NFS_I(inode)->fileid = opendata->f_attr.fileid; } - memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid)); - if (o_res.delegation_type != 0) { + memcpy(&state->stateid, &opendata->o_res.stateid, sizeof(state->stateid)); + if (opendata->o_res.delegation_type != 0) { if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) - nfs_inode_set_delegation(inode, sp->so_cred, &o_res); + nfs_inode_set_delegation(inode, sp->so_cred, + &opendata->o_res); else - nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res); + nfs_inode_reclaim_delegation(inode, sp->so_cred, + &opendata->o_res); } out_nodeleg: - nfs_free_seqid(o_arg.seqid); + nfs4_opendata_free(opendata); clear_bit(NFS_DELEGATED_STATE, &state->flags); out: dput(parent); @@ -706,21 +751,8 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st struct nfs_server *server = NFS_SERVER(dir); struct nfs4_client *clp = server->nfs4_state; struct inode *inode = NULL; + struct nfs4_opendata *opendata; int status; - struct nfs_fattr f_attr, dir_attr; - struct nfs_openargs o_arg = { - .fh = NFS_FH(dir), - .open_flags = flags, - .name = &dentry->d_name, - .server = server, - .bitmask = server->attr_bitmask, - .claim = NFS4_OPEN_CLAIM_NULL, - }; - struct nfs_openres o_res = { - .f_attr = &f_attr, - .dir_attr = &dir_attr, - .server = server, - }; /* Protect against reboot recovery conflicts */ down_read(&clp->cl_sem); @@ -729,45 +761,34 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n"); goto out_err; } - if (flags & O_EXCL) { - u32 *p = (u32 *) o_arg.u.verifier.data; - p[0] = jiffies; - p[1] = current->pid; - } else - o_arg.u.attrs = sattr; - /* Serialization for the sequence id */ + opendata = nfs4_opendata_alloc(dentry, sp, flags, sattr); + if (opendata == NULL) + goto err_put_state_owner; - o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); - if (o_arg.seqid == NULL) - return -ENOMEM; - nfs_fattr_init(&f_attr); - nfs_fattr_init(&dir_attr); - status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); + status = _nfs4_proc_open(dir, sp, &opendata->o_arg, &opendata->o_res); if (status != 0) - goto out_err; + goto err_opendata_free; status = -ENOMEM; - inode = nfs_fhget(dir->i_sb, &o_res.fh, &f_attr); + inode = nfs_fhget(dir->i_sb, &opendata->o_res.fh, &opendata->f_attr); if (!inode) - goto out_err; + goto err_opendata_free; state = nfs4_get_open_state(inode, sp); if (!state) - goto out_err; - update_open_stateid(state, &o_res.stateid, flags); - if (o_res.delegation_type != 0) - nfs_inode_set_delegation(inode, cred, &o_res); - nfs_free_seqid(o_arg.seqid); + goto err_opendata_free; + update_open_stateid(state, &opendata->o_res.stateid, flags); + if (opendata->o_res.delegation_type != 0) + nfs_inode_set_delegation(inode, cred, &opendata->o_res); + nfs4_opendata_free(opendata); nfs4_put_state_owner(sp); up_read(&clp->cl_sem); *res = state; return 0; +err_opendata_free: + nfs4_opendata_free(opendata); +err_put_state_owner: + nfs4_put_state_owner(sp); out_err: - if (sp != NULL) { - if (state != NULL) - nfs4_put_open_state(state); - nfs_free_seqid(o_arg.seqid); - nfs4_put_state_owner(sp); - } /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */ up_read(&clp->cl_sem); if (inode != NULL) -- cgit v1.2.2 From 24ac23ab88df5b21b5b2df8cde748bf99b289099 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:11 +0100 Subject: NFSv4: Convert open() into an asynchronous RPC call OPEN is a stateful operation, so we must ensure that it always completes. In order to allow users to interrupt the operation, we need to make the RPC call asynchronous, and then wait on completion (or cancel). Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 193 ++++++++++++++++++++++++++++++++++++------------------ fs/nfs/nfs4xdr.c | 3 - 2 files changed, 130 insertions(+), 66 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 4a5cc84021..aed8701c1a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -195,6 +195,7 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf } struct nfs4_opendata { + atomic_t count; struct nfs_openargs o_arg; struct nfs_openres o_res; struct nfs_fattr f_attr; @@ -203,6 +204,8 @@ struct nfs4_opendata { struct dentry *dir; struct nfs4_state_owner *owner; struct iattr attrs; + int rpc_status; + int cancelled; }; static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry, @@ -220,6 +223,7 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry, p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); if (p->o_arg.seqid == NULL) goto err_free; + atomic_set(&p->count, 1); p->dentry = dget(dentry); p->dir = parent; p->owner = sp; @@ -255,7 +259,7 @@ err: static void nfs4_opendata_free(struct nfs4_opendata *p) { - if (p != NULL) { + if (p != NULL && atomic_dec_and_test(&p->count)) { nfs_free_seqid(p->o_arg.seqid); nfs4_put_state_owner(p->owner); dput(p->dir); @@ -305,6 +309,26 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, spin_unlock(&state->owner->so_lock); } +static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data) +{ + struct inode *inode; + struct nfs4_state *state = NULL; + + if (!(data->f_attr.valid & NFS_ATTR_FATTR)) + goto out; + inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr); + if (inode == NULL) + goto out; + state = nfs4_get_open_state(inode, data->owner); + if (state == NULL) + goto put_inode; + update_open_stateid(state, &data->o_res.stateid, data->o_arg.open_flags); +put_inode: + iput(inode); +out: + return state; +} + /* * OPEN_RECLAIM: * reclaim state on the server after a reboot. @@ -473,41 +497,105 @@ static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *f return status; } -static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, struct nfs_openargs *o_arg, struct nfs_openres *o_res) +static void nfs4_open_prepare(struct rpc_task *task, void *calldata) { - struct nfs_server *server = NFS_SERVER(dir); + struct nfs4_opendata *data = calldata; + struct nfs4_state_owner *sp = data->owner; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN], - .rpc_argp = o_arg, - .rpc_resp = o_res, + .rpc_argp = &data->o_arg, + .rpc_resp = &data->o_res, .rpc_cred = sp->so_cred, }; - int status; + + if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0) + return; + /* Update sequence id. */ + data->o_arg.id = sp->so_id; + data->o_arg.clientid = sp->so_client->cl_clientid; + rpc_call_setup(task, &msg, 0); +} - /* Update sequence id. The caller must serialize! */ - o_arg->id = sp->so_id; - o_arg->clientid = sp->so_client->cl_clientid; +static void nfs4_open_done(struct rpc_task *task, void *calldata) +{ + struct nfs4_opendata *data = calldata; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - if (status == 0) { - /* OPEN on anything except a regular file is disallowed in NFSv4 */ - switch (o_res->f_attr->mode & S_IFMT) { + data->rpc_status = task->tk_status; + if (RPC_ASSASSINATED(task)) + return; + if (task->tk_status == 0) { + switch (data->o_res.f_attr->mode & S_IFMT) { case S_IFREG: break; case S_IFLNK: - status = -ELOOP; + data->rpc_status = -ELOOP; break; case S_IFDIR: - status = -EISDIR; + data->rpc_status = -EISDIR; break; default: - status = -ENOTDIR; + data->rpc_status = -ENOTDIR; } } + nfs_increment_open_seqid(data->rpc_status, data->o_arg.seqid); +} + +static void nfs4_open_release(void *calldata) +{ + struct nfs4_opendata *data = calldata; + struct nfs4_state *state = NULL; + + /* If this request hasn't been cancelled, do nothing */ + if (data->cancelled == 0) + goto out_free; + /* In case of error, no cleanup! */ + if (data->rpc_status != 0) + goto out_free; + /* In case we need an open_confirm, no cleanup! */ + if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM) + goto out_free; + nfs_confirm_seqid(&data->owner->so_seqid, 0); + state = nfs4_opendata_to_nfs4_state(data); + if (state != NULL) + nfs4_close_state(state, data->o_arg.open_flags); +out_free: + nfs4_opendata_free(data); +} + +static const struct rpc_call_ops nfs4_open_ops = { + .rpc_call_prepare = nfs4_open_prepare, + .rpc_call_done = nfs4_open_done, + .rpc_release = nfs4_open_release, +}; - nfs_increment_open_seqid(status, o_arg->seqid); +/* + * Note: On error, nfs4_proc_open will free the struct nfs4_opendata + */ +static int _nfs4_proc_open(struct nfs4_opendata *data) +{ + struct inode *dir = data->dir->d_inode; + struct nfs_server *server = NFS_SERVER(dir); + struct nfs_openargs *o_arg = &data->o_arg; + struct nfs_openres *o_res = &data->o_res; + struct rpc_task *task; + int status; + + atomic_inc(&data->count); + task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_ops, data); + if (IS_ERR(task)) { + nfs4_opendata_free(data); + return PTR_ERR(task); + } + status = nfs4_wait_for_completion_rpc_task(task); + if (status != 0) { + data->cancelled = 1; + smp_wmb(); + } else + status = data->rpc_status; + rpc_release_task(task); if (status != 0) - goto out; + return status; + if (o_arg->open_flags & O_CREAT) { update_changeattr(dir, &o_res->cinfo); nfs_post_op_update_inode(dir, o_res->dir_attr); @@ -515,15 +603,14 @@ static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, stru nfs_refresh_inode(dir, o_res->dir_attr); if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { status = _nfs4_proc_open_confirm(server->client, &o_res->fh, - sp, &o_res->stateid, o_arg->seqid); + data->owner, &o_res->stateid, o_arg->seqid); if (status != 0) - goto out; + return status; } - nfs_confirm_seqid(&sp->so_seqid, 0); + nfs_confirm_seqid(&data->owner->so_seqid, 0); if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) - status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); -out: - return status; + return server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); + return 0; } static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags) @@ -562,14 +649,15 @@ out: static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) { struct dentry *parent = dget_parent(dentry); - struct inode *dir = parent->d_inode; struct inode *inode = state->inode; struct nfs_delegation *delegation = NFS_I(inode)->delegation; struct nfs4_opendata *opendata; + struct nfs4_state *newstate; + int openflags = state->state & (FMODE_READ|FMODE_WRITE); int status = 0; if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) { - status = _nfs4_do_access(inode, sp->so_cred, state->state); + status = _nfs4_do_access(inode, sp->so_cred, openflags); if (status < 0) goto out; memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid)); @@ -577,27 +665,15 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st goto out; } status = -ENOMEM; - opendata = nfs4_opendata_alloc(dentry, sp, state->state, NULL); + opendata = nfs4_opendata_alloc(dentry, sp, openflags, NULL); if (opendata == NULL) goto out; - status = _nfs4_proc_open(dir, sp, &opendata->o_arg, &opendata->o_res); + status = _nfs4_proc_open(opendata); if (status != 0) goto out_nodeleg; - /* Check if files differ */ - if ((opendata->f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT)) + newstate = nfs4_opendata_to_nfs4_state(opendata); + if (newstate != state) goto out_stale; - /* Has the file handle changed? */ - if (nfs_compare_fh(&opendata->o_res.fh, NFS_FH(inode)) != 0) { - /* Verify if the change attributes are the same */ - if (opendata->f_attr.change_attr != NFS_I(inode)->change_attr) - goto out_stale; - if (nfs_size_to_loff_t(opendata->f_attr.size) != inode->i_size) - goto out_stale; - /* Lets just pretend that this is the same file */ - nfs_copy_fh(NFS_FH(inode), &opendata->o_res.fh); - NFS_I(inode)->fileid = opendata->f_attr.fileid; - } - memcpy(&state->stateid, &opendata->o_res.stateid, sizeof(state->stateid)); if (opendata->o_res.delegation_type != 0) { if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) nfs_inode_set_delegation(inode, sp->so_cred, @@ -606,6 +682,8 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st nfs_inode_reclaim_delegation(inode, sp->so_cred, &opendata->o_res); } +out_close_state: + nfs4_close_state(newstate, openflags); out_nodeleg: nfs4_opendata_free(opendata); clear_bit(NFS_DELEGATED_STATE, &state->flags); @@ -618,7 +696,7 @@ out_stale: nfs4_drop_state_owner(sp); d_drop(dentry); /* Should we be trying to close that stateid? */ - goto out_nodeleg; + goto out_close_state; } static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) @@ -656,7 +734,7 @@ static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *sta } /* - * Returns an nfs4_state + an extra reference to the inode + * Returns a referenced nfs4_state if there is an open delegation on the file */ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res) { @@ -709,7 +787,6 @@ out_ok: nfs4_put_state_owner(sp); up_read(&nfsi->rwsem); up_read(&clp->cl_sem); - igrab(inode); *res = state; return 0; out_err: @@ -742,7 +819,7 @@ static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, st } /* - * Returns an nfs4_state + an referenced inode + * Returns a referenced nfs4_state */ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res) { @@ -750,7 +827,6 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st struct nfs4_state *state = NULL; struct nfs_server *server = NFS_SERVER(dir); struct nfs4_client *clp = server->nfs4_state; - struct inode *inode = NULL; struct nfs4_opendata *opendata; int status; @@ -765,20 +841,16 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st if (opendata == NULL) goto err_put_state_owner; - status = _nfs4_proc_open(dir, sp, &opendata->o_arg, &opendata->o_res); + status = _nfs4_proc_open(opendata); if (status != 0) goto err_opendata_free; status = -ENOMEM; - inode = nfs_fhget(dir->i_sb, &opendata->o_res.fh, &opendata->f_attr); - if (!inode) - goto err_opendata_free; - state = nfs4_get_open_state(inode, sp); - if (!state) + state = nfs4_opendata_to_nfs4_state(opendata); + if (state == NULL) goto err_opendata_free; - update_open_stateid(state, &opendata->o_res.stateid, flags); if (opendata->o_res.delegation_type != 0) - nfs_inode_set_delegation(inode, cred, &opendata->o_res); + nfs_inode_set_delegation(state->inode, cred, &opendata->o_res); nfs4_opendata_free(opendata); nfs4_put_state_owner(sp); up_read(&clp->cl_sem); @@ -791,8 +863,6 @@ err_put_state_owner: out_err: /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */ up_read(&clp->cl_sem); - if (inode != NULL) - iput(inode); *res = NULL; return status; } @@ -1066,7 +1136,7 @@ nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) d_add(dentry, NULL); return (struct dentry *)state; } - res = d_add_unique(dentry, state->inode); + res = d_add_unique(dentry, igrab(state->inode)); if (res != NULL) dentry = res; nfs4_intent_set_file(nd, dentry, state); @@ -1078,7 +1148,6 @@ nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, st { struct rpc_cred *cred; struct nfs4_state *state; - struct inode *inode; cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0); if (IS_ERR(cred)) @@ -1102,9 +1171,7 @@ nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, st } goto out_drop; } - inode = state->inode; - iput(inode); - if (inode == dentry->d_inode) { + if (state->inode == dentry->d_inode) { nfs4_intent_set_file(nd, dentry, state); return 1; } @@ -1633,7 +1700,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, status = PTR_ERR(state); goto out; } - d_instantiate(dentry, state->inode); + d_instantiate(dentry, igrab(state->inode)); if (flags & O_EXCL) { struct nfs_fattr fattr; status = nfs4_do_setattr(NFS_SERVER(dir), &fattr, diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index fbbace8a30..db2bcf722f 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1499,9 +1499,6 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena }; int status; - status = nfs_wait_on_sequence(args->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); -- cgit v1.2.2 From cdd4e68b5f0ed12c64b3e2be83655d2a47588a74 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:12 +0100 Subject: NFSv4: Make open_confirm() asynchronous too Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 101 +++++++++++++++++++++++++++++++++++++++++------------- fs/nfs/nfs4xdr.c | 7 ++-- 2 files changed, 80 insertions(+), 28 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index aed8701c1a..8154f25794 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -57,7 +57,8 @@ #define NFS4_POLL_RETRY_MIN (1*HZ) #define NFS4_POLL_RETRY_MAX (15*HZ) -static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid); +struct nfs4_opendata; +static int _nfs4_proc_open_confirm(struct nfs4_opendata *data); static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); @@ -198,6 +199,8 @@ struct nfs4_opendata { atomic_t count; struct nfs_openargs o_arg; struct nfs_openres o_res; + struct nfs_open_confirmargs c_arg; + struct nfs_open_confirmres c_res; struct nfs_fattr f_attr; struct nfs_fattr dir_attr; struct dentry *dentry; @@ -249,6 +252,9 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry, p->o_arg.u.attrs = &p->attrs; memcpy(&p->attrs, attrs, sizeof(p->attrs)); } + p->c_arg.fh = &p->o_res.fh; + p->c_arg.stateid = &p->o_res.stateid; + p->c_arg.seqid = p->o_arg.seqid; return p; err_free: kfree(p); @@ -433,8 +439,7 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state if (status != 0) goto out_free; if(opendata->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM) { - status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode), - sp, &opendata->o_res.stateid, opendata->o_arg.seqid); + status = _nfs4_proc_open_confirm(opendata); if (status != 0) goto out_free; } @@ -472,28 +477,79 @@ int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) return err; } -static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid) +static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata) { - struct nfs_open_confirmargs arg = { - .fh = fh, - .seqid = seqid, - .stateid = *stateid, - }; - struct nfs_open_confirmres res; - struct rpc_message msg = { - .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM], - .rpc_argp = &arg, - .rpc_resp = &res, - .rpc_cred = sp->so_cred, + struct nfs4_opendata *data = calldata; + struct rpc_message msg = { + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM], + .rpc_argp = &data->c_arg, + .rpc_resp = &data->c_res, + .rpc_cred = data->owner->so_cred, }; + rpc_call_setup(task, &msg, 0); +} + +static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata) +{ + struct nfs4_opendata *data = calldata; + + data->rpc_status = task->tk_status; + if (RPC_ASSASSINATED(task)) + return; + if (data->rpc_status == 0) + memcpy(data->o_res.stateid.data, data->c_res.stateid.data, + sizeof(data->o_res.stateid.data)); + nfs_increment_open_seqid(data->rpc_status, data->c_arg.seqid); + nfs_confirm_seqid(&data->owner->so_seqid, data->rpc_status); +} + +static void nfs4_open_confirm_release(void *calldata) +{ + struct nfs4_opendata *data = calldata; + struct nfs4_state *state = NULL; + + /* If this request hasn't been cancelled, do nothing */ + if (data->cancelled == 0) + goto out_free; + /* In case of error, no cleanup! */ + if (data->rpc_status != 0) + goto out_free; + nfs_confirm_seqid(&data->owner->so_seqid, 0); + state = nfs4_opendata_to_nfs4_state(data); + if (state != NULL) + nfs4_close_state(state, data->o_arg.open_flags); +out_free: + nfs4_opendata_free(data); +} + +static const struct rpc_call_ops nfs4_open_confirm_ops = { + .rpc_call_prepare = nfs4_open_confirm_prepare, + .rpc_call_done = nfs4_open_confirm_done, + .rpc_release = nfs4_open_confirm_release, +}; + +/* + * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata + */ +static int _nfs4_proc_open_confirm(struct nfs4_opendata *data) +{ + struct nfs_server *server = NFS_SERVER(data->dir->d_inode); + struct rpc_task *task; int status; - status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR); - /* Confirm the sequence as being established */ - nfs_confirm_seqid(&sp->so_seqid, status); - nfs_increment_open_seqid(status, seqid); - if (status >= 0) - memcpy(stateid, &res.stateid, sizeof(*stateid)); + atomic_inc(&data->count); + task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_confirm_ops, data); + if (IS_ERR(task)) { + nfs4_opendata_free(data); + return PTR_ERR(task); + } + status = nfs4_wait_for_completion_rpc_task(task); + if (status != 0) { + data->cancelled = 1; + smp_wmb(); + } else + status = data->rpc_status; + rpc_release_task(task); return status; } @@ -602,8 +658,7 @@ static int _nfs4_proc_open(struct nfs4_opendata *data) } else nfs_refresh_inode(dir, o_res->dir_attr); if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { - status = _nfs4_proc_open_confirm(server->client, &o_res->fh, - data->owner, &o_res->stateid, o_arg->seqid); + status = _nfs4_proc_open_confirm(data); if (status != 0) return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index db2bcf722f..2ba9906f2a 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -964,9 +964,9 @@ static int encode_open_confirm(struct xdr_stream *xdr, const struct nfs_open_con { uint32_t *p; - RESERVE_SPACE(8+sizeof(arg->stateid.data)); + RESERVE_SPACE(8+sizeof(arg->stateid->data)); WRITE32(OP_OPEN_CONFIRM); - WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); + WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data)); WRITE32(arg->seqid->sequence->counter); return 0; @@ -1535,9 +1535,6 @@ static int nfs4_xdr_enc_open_confirm(struct rpc_rqst *req, uint32_t *p, struct n }; int status; - status = nfs_wait_on_sequence(args->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); -- cgit v1.2.2 From e761692381f294ea079d2e869fcd7c0afc79e394 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:13 +0100 Subject: NFSv4: Make nfs4_state track O_RDWR, O_RDONLY and O_WRONLY separately A closer reading of RFC3530 reveals that OPEN_DOWNGRADE must always specify a access modes that have been the argument of a previous OPEN operation. IOW: doing OPEN(O_RDWR) and then OPEN_DOWNGRADE(O_WRONLY) is forbidden unless the user called OPEN(O_WRONLY) In order to fix that, we really need to track the three possible open states separately. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 5 +++-- fs/nfs/nfs4proc.c | 34 ++++++++++++++++++++++------------ fs/nfs/nfs4state.c | 31 +++++++++++++++++-------------- 3 files changed, 42 insertions(+), 28 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index b7f262dcb6..4ad5981f93 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -182,8 +182,9 @@ struct nfs4_state { nfs4_stateid stateid; - unsigned int nreaders; - unsigned int nwriters; + unsigned int n_rdonly; + unsigned int n_wronly; + unsigned int n_rdwr; int state; /* State on the server (R,W, or RW) */ atomic_t count; }; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 8154f25794..e494cc2ea9 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -297,6 +297,20 @@ static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task) return ret; } +static inline void update_open_stateflags(struct nfs4_state *state, mode_t open_flags) +{ + switch (open_flags) { + case FMODE_WRITE: + state->n_wronly++; + break; + case FMODE_READ: + state->n_rdonly++; + break; + case FMODE_READ|FMODE_WRITE: + state->n_rdwr++; + } +} + static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) { struct inode *inode = state->inode; @@ -306,10 +320,7 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, spin_lock(&state->owner->so_lock); spin_lock(&inode->i_lock); memcpy(&state->stateid, stateid, sizeof(state->stateid)); - if ((open_flags & FMODE_WRITE)) - state->nwriters++; - if (open_flags & FMODE_READ) - state->nreaders++; + update_open_stateflags(state, open_flags); nfs4_state_set_mode_locked(state, state->state | open_flags); spin_unlock(&inode->i_lock); spin_unlock(&state->owner->so_lock); @@ -822,10 +833,7 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred err = -ENOENT; if ((state->state & open_flags) == open_flags) { spin_lock(&inode->i_lock); - if (open_flags & FMODE_READ) - state->nreaders++; - if (open_flags & FMODE_WRITE) - state->nwriters++; + update_open_stateflags(state, open_flags); spin_unlock(&inode->i_lock); goto out_ok; } else if (state->state != 0) @@ -1082,10 +1090,12 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) spin_lock(&state->owner->so_lock); spin_lock(&calldata->inode->i_lock); mode = old_mode = state->state; - if (state->nreaders == 0) - mode &= ~FMODE_READ; - if (state->nwriters == 0) - mode &= ~FMODE_WRITE; + if (state->n_rdwr == 0) { + if (state->n_rdonly == 0) + mode &= ~FMODE_READ; + if (state->n_wronly == 0) + mode &= ~FMODE_WRITE; + } nfs4_state_set_mode_locked(state, mode); spin_unlock(&calldata->inode->i_lock); spin_unlock(&state->owner->so_lock); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 5ef4c57618..41a5f1a8a9 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -349,14 +349,9 @@ nfs4_alloc_open_state(void) { struct nfs4_state *state; - state = kmalloc(sizeof(*state), GFP_KERNEL); + state = kzalloc(sizeof(*state), GFP_KERNEL); if (!state) return NULL; - state->state = 0; - state->nreaders = 0; - state->nwriters = 0; - state->flags = 0; - memset(state->stateid.data, 0, sizeof(state->stateid.data)); atomic_set(&state->count, 1); INIT_LIST_HEAD(&state->lock_states); spin_lock_init(&state->state_lock); @@ -475,15 +470,23 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) /* Protect against nfs4_find_state() */ spin_lock(&owner->so_lock); spin_lock(&inode->i_lock); - if (mode & FMODE_READ) - state->nreaders--; - if (mode & FMODE_WRITE) - state->nwriters--; + switch (mode & (FMODE_READ | FMODE_WRITE)) { + case FMODE_READ: + state->n_rdonly--; + break; + case FMODE_WRITE: + state->n_wronly--; + break; + case FMODE_READ|FMODE_WRITE: + state->n_rdwr--; + } oldstate = newstate = state->state; - if (state->nreaders == 0) - newstate &= ~FMODE_READ; - if (state->nwriters == 0) - newstate &= ~FMODE_WRITE; + if (state->n_rdwr == 0) { + if (state->n_rdonly == 0) + newstate &= ~FMODE_READ; + if (state->n_wronly == 0) + newstate &= ~FMODE_WRITE; + } if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { nfs4_state_set_mode_locked(state, newstate); oldstate = newstate; -- cgit v1.2.2 From 864472e9b8fa76ffaad17dfcb84d79e16df6828c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:15 +0100 Subject: NFSv4: Make open recovery track O_RDWR, O_RDONLY and O_WRONLY correctly When recovering from a delegation recall or a network partition, we need to replay open(O_RDWR), open(O_RDONLY) and open(O_WRONLY) separately. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 276 +++++++++++++++++++++++++++++------------------------- fs/nfs/nfs4xdr.c | 11 ++- 2 files changed, 156 insertions(+), 131 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index e494cc2ea9..3ecb7da220 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -58,7 +58,7 @@ #define NFS4_POLL_RETRY_MAX (15*HZ) struct nfs4_opendata; -static int _nfs4_proc_open_confirm(struct nfs4_opendata *data); +static int _nfs4_proc_open(struct nfs4_opendata *data); static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); @@ -346,32 +346,108 @@ out: return state; } +static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state) +{ + struct nfs_inode *nfsi = NFS_I(state->inode); + struct nfs_open_context *ctx; + + spin_lock(&state->inode->i_lock); + list_for_each_entry(ctx, &nfsi->open_files, list) { + if (ctx->state != state) + continue; + get_nfs_open_context(ctx); + spin_unlock(&state->inode->i_lock); + return ctx; + } + spin_unlock(&state->inode->i_lock); + return ERR_PTR(-ENOENT); +} + +static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, mode_t openflags, nfs4_stateid *stateid) +{ + int ret; + + opendata->o_arg.open_flags = openflags; + ret = _nfs4_proc_open(opendata); + if (ret != 0) + return ret; + memcpy(stateid->data, opendata->o_res.stateid.data, + sizeof(stateid->data)); + return 0; +} + +static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state) +{ + nfs4_stateid stateid; + struct nfs4_state *newstate; + int mode = 0; + int delegation = 0; + int ret; + + /* memory barrier prior to reading state->n_* */ + smp_rmb(); + if (state->n_rdwr != 0) { + ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &stateid); + if (ret != 0) + return ret; + mode |= FMODE_READ|FMODE_WRITE; + if (opendata->o_res.delegation_type != 0) + delegation = opendata->o_res.delegation_type; + smp_rmb(); + } + if (state->n_wronly != 0) { + ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &stateid); + if (ret != 0) + return ret; + mode |= FMODE_WRITE; + if (opendata->o_res.delegation_type != 0) + delegation = opendata->o_res.delegation_type; + smp_rmb(); + } + if (state->n_rdonly != 0) { + ret = nfs4_open_recover_helper(opendata, FMODE_READ, &stateid); + if (ret != 0) + return ret; + mode |= FMODE_READ; + } + clear_bit(NFS_DELEGATED_STATE, &state->flags); + if (mode == 0) + return 0; + if (opendata->o_res.delegation_type == 0) + opendata->o_res.delegation_type = delegation; + opendata->o_arg.open_flags |= mode; + newstate = nfs4_opendata_to_nfs4_state(opendata); + if (newstate != NULL) { + if (opendata->o_res.delegation_type != 0) { + struct nfs_inode *nfsi = NFS_I(newstate->inode); + int delegation_flags = 0; + if (nfsi->delegation) + delegation_flags = nfsi->delegation->flags; + if (!(delegation_flags & NFS_DELEGATION_NEED_RECLAIM)) + nfs_inode_set_delegation(newstate->inode, + opendata->owner->so_cred, + &opendata->o_res); + else + nfs_inode_reclaim_delegation(newstate->inode, + opendata->owner->so_cred, + &opendata->o_res); + } + nfs4_close_state(newstate, opendata->o_arg.open_flags); + } + if (newstate != state) + return -ESTALE; + return 0; +} + /* * OPEN_RECLAIM: * reclaim state on the server after a reboot. */ -static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) +static int _nfs4_do_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) { - struct inode *inode = state->inode; - struct nfs_server *server = NFS_SERVER(inode); - struct nfs_delegation *delegation = NFS_I(inode)->delegation; - struct nfs_openargs o_arg = { - .fh = NFS_FH(inode), - .id = sp->so_id, - .open_flags = state->state, - .clientid = server->nfs4_state->cl_clientid, - .claim = NFS4_OPEN_CLAIM_PREVIOUS, - .bitmask = server->attr_bitmask, - }; - struct nfs_openres o_res = { - .server = server, /* Grrr */ - }; - struct rpc_message msg = { - .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR], - .rpc_argp = &o_arg, - .rpc_resp = &o_res, - .rpc_cred = sp->so_cred, - }; + struct nfs_delegation *delegation = NFS_I(state->inode)->delegation; + struct nfs4_opendata *opendata; + int delegation_type = 0; int status; if (delegation != NULL) { @@ -381,38 +457,27 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st set_bit(NFS_DELEGATED_STATE, &state->flags); return 0; } - o_arg.u.delegation_type = delegation->type; + delegation_type = delegation->type; } - o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); - if (o_arg.seqid == NULL) + opendata = nfs4_opendata_alloc(dentry, sp, 0, NULL); + if (opendata == NULL) return -ENOMEM; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* Confirm the sequence as being established */ - nfs_confirm_seqid(&sp->so_seqid, status); - nfs_increment_open_seqid(status, o_arg.seqid); - if (status == 0) { - memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid)); - if (o_res.delegation_type != 0) { - nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res); - /* Did the server issue an immediate delegation recall? */ - if (o_res.do_recall) - nfs_async_inode_return_delegation(inode, &o_res.stateid); - } - } - nfs_free_seqid(o_arg.seqid); - clear_bit(NFS_DELEGATED_STATE, &state->flags); - /* Ensure we update the inode attributes */ - NFS_CACHEINV(inode); + opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS; + opendata->o_arg.fh = NFS_FH(state->inode); + nfs_copy_fh(&opendata->o_res.fh, opendata->o_arg.fh); + opendata->o_arg.u.delegation_type = delegation_type; + status = nfs4_open_recover(opendata, state); + nfs4_opendata_free(opendata); return status; } -static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) +static int nfs4_do_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) { struct nfs_server *server = NFS_SERVER(state->inode); struct nfs4_exception exception = { }; int err; do { - err = _nfs4_open_reclaim(sp, state); + err = _nfs4_do_open_reclaim(sp, state, dentry); if (err != -NFS4ERR_DELAY) break; nfs4_handle_exception(server, err, &exception); @@ -420,50 +485,36 @@ static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *sta return err; } +static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) +{ + struct nfs_open_context *ctx; + int ret; + + ctx = nfs4_state_find_open_context(state); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + ret = nfs4_do_open_reclaim(sp, state, ctx->dentry); + put_nfs_open_context(ctx); + return ret; +} + static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) { struct nfs4_state_owner *sp = state->owner; - struct inode *inode = dentry->d_inode; - struct nfs_server *server = NFS_SERVER(inode); - struct rpc_message msg = { - .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR], - .rpc_cred = sp->so_cred, - }; struct nfs4_opendata *opendata; - int status = 0; + int ret; if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) - goto out; - if (state->state == 0) - goto out; - opendata = nfs4_opendata_alloc(dentry, sp, state->state, NULL); - status = -ENOMEM; + return 0; + opendata = nfs4_opendata_alloc(dentry, sp, 0, NULL); if (opendata == NULL) - goto out; + return -ENOMEM; opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR; - msg.rpc_argp = &opendata->o_arg; - msg.rpc_resp = &opendata->o_res; memcpy(opendata->o_arg.u.delegation.data, state->stateid.data, sizeof(opendata->o_arg.u.delegation.data)); - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs_increment_open_seqid(status, opendata->o_arg.seqid); - if (status != 0) - goto out_free; - if(opendata->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM) { - status = _nfs4_proc_open_confirm(opendata); - if (status != 0) - goto out_free; - } - nfs_confirm_seqid(&sp->so_seqid, 0); - if (status >= 0) { - memcpy(state->stateid.data, opendata->o_res.stateid.data, - sizeof(state->stateid.data)); - clear_bit(NFS_DELEGATED_STATE, &state->flags); - } -out_free: + ret = nfs4_open_recover(opendata, state); nfs4_opendata_free(opendata); -out: - return status; + return ret; } int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) @@ -580,6 +631,8 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) /* Update sequence id. */ data->o_arg.id = sp->so_id; data->o_arg.clientid = sp->so_client->cl_clientid; + if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) + msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR]; rpc_call_setup(task, &msg, 0); } @@ -714,55 +767,31 @@ out: */ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) { - struct dentry *parent = dget_parent(dentry); struct inode *inode = state->inode; struct nfs_delegation *delegation = NFS_I(inode)->delegation; struct nfs4_opendata *opendata; - struct nfs4_state *newstate; int openflags = state->state & (FMODE_READ|FMODE_WRITE); - int status = 0; + int ret; if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) { - status = _nfs4_do_access(inode, sp->so_cred, openflags); - if (status < 0) - goto out; + ret = _nfs4_do_access(inode, sp->so_cred, openflags); + if (ret < 0) + return ret; memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid)); set_bit(NFS_DELEGATED_STATE, &state->flags); - goto out; + return 0; } - status = -ENOMEM; opendata = nfs4_opendata_alloc(dentry, sp, openflags, NULL); if (opendata == NULL) - goto out; - status = _nfs4_proc_open(opendata); - if (status != 0) - goto out_nodeleg; - newstate = nfs4_opendata_to_nfs4_state(opendata); - if (newstate != state) - goto out_stale; - if (opendata->o_res.delegation_type != 0) { - if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) - nfs_inode_set_delegation(inode, sp->so_cred, - &opendata->o_res); - else - nfs_inode_reclaim_delegation(inode, sp->so_cred, - &opendata->o_res); + return -ENOMEM; + ret = nfs4_open_recover(opendata, state); + if (ret == -ESTALE) { + /* Invalidate the state owner so we don't ever use it again */ + nfs4_drop_state_owner(sp); + d_drop(dentry); } -out_close_state: - nfs4_close_state(newstate, openflags); -out_nodeleg: nfs4_opendata_free(opendata); - clear_bit(NFS_DELEGATED_STATE, &state->flags); -out: - dput(parent); - return status; -out_stale: - status = -ESTALE; - /* Invalidate the state owner so we don't ever use it again */ - nfs4_drop_state_owner(sp); - d_drop(dentry); - /* Should we be trying to close that stateid? */ - goto out_close_state; + return ret; } static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) @@ -781,22 +810,15 @@ static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_ static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state) { - struct nfs_inode *nfsi = NFS_I(state->inode); struct nfs_open_context *ctx; - int status; + int ret; - spin_lock(&state->inode->i_lock); - list_for_each_entry(ctx, &nfsi->open_files, list) { - if (ctx->state != state) - continue; - get_nfs_open_context(ctx); - spin_unlock(&state->inode->i_lock); - status = nfs4_do_open_expired(sp, state, ctx->dentry); - put_nfs_open_context(ctx); - return status; - } - spin_unlock(&state->inode->i_lock); - return -ENOENT; + ctx = nfs4_state_find_open_context(state); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + ret = nfs4_do_open_expired(sp, state, ctx->dentry); + put_nfs_open_context(ctx); + return ret; } /* diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 2ba9906f2a..3100172822 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1552,19 +1552,19 @@ static int nfs4_xdr_enc_open_noattr(struct rpc_rqst *req, uint32_t *p, struct nf { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; - status = nfs_wait_on_sequence(args->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); if (status) goto out; status = encode_open(&xdr, args); + if (status) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -3825,6 +3825,9 @@ static int nfs4_xdr_dec_open_noattr(struct rpc_rqst *rqstp, uint32_t *p, struct if (status) goto out; status = decode_open(&xdr, res); + if (status) + goto out; + decode_getfattr(&xdr, res->f_attr, res->server); out: return status; } -- cgit v1.2.2 From 911d1aaf26fc4d771174d98fcab710a44e2a5fa0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:16 +0100 Subject: NFSv4: locking XDR cleanup Get rid of some unnecessary intermediate structures Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 192 +++++++++++++++++++++++------------------------------- fs/nfs/nfs4xdr.c | 131 +++++++++++++++++++++---------------- 2 files changed, 155 insertions(+), 168 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3ecb7da220..857125705b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2935,43 +2935,17 @@ nfs4_set_lock_task_retry(unsigned long timeout) return timeout; } -static inline int -nfs4_lck_type(int cmd, struct file_lock *request) -{ - /* set lock type */ - switch (request->fl_type) { - case F_RDLCK: - return IS_SETLKW(cmd) ? NFS4_READW_LT : NFS4_READ_LT; - case F_WRLCK: - return IS_SETLKW(cmd) ? NFS4_WRITEW_LT : NFS4_WRITE_LT; - case F_UNLCK: - return NFS4_WRITE_LT; - } - BUG(); - return 0; -} - -static inline uint64_t -nfs4_lck_length(struct file_lock *request) -{ - if (request->fl_end == OFFSET_MAX) - return ~(uint64_t)0; - return request->fl_end - request->fl_start + 1; -} - static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request) { struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); struct nfs4_client *clp = server->nfs4_state; - struct nfs_lockargs arg = { + struct nfs_lockt_args arg = { .fh = NFS_FH(inode), - .type = nfs4_lck_type(cmd, request), - .offset = request->fl_start, - .length = nfs4_lck_length(request), + .fl = request, }; - struct nfs_lockres res = { - .server = server, + struct nfs_lockt_res res = { + .denied = request, }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKT], @@ -2979,36 +2953,23 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock .rpc_resp = &res, .rpc_cred = state->owner->so_cred, }; - struct nfs_lowner nlo; struct nfs4_lock_state *lsp; int status; down_read(&clp->cl_sem); - nlo.clientid = clp->cl_clientid; + arg.lock_owner.clientid = clp->cl_clientid; status = nfs4_set_lock_state(state, request); if (status != 0) goto out; lsp = request->fl_u.nfs4_fl.owner; - nlo.id = lsp->ls_id; - arg.u.lockt = &nlo; + arg.lock_owner.id = lsp->ls_id; status = rpc_call_sync(server->client, &msg, 0); - if (!status) { - request->fl_type = F_UNLCK; - } else if (status == -NFS4ERR_DENIED) { - int64_t len, start, end; - start = res.u.denied.offset; - len = res.u.denied.length; - end = start + len - 1; - if (end < 0 || len == 0) - request->fl_end = OFFSET_MAX; - else - request->fl_end = (loff_t)end; - request->fl_start = (loff_t)start; - request->fl_type = F_WRLCK; - if (res.u.denied.type & 1) - request->fl_type = F_RDLCK; - request->fl_pid = 0; - status = 0; + switch (status) { + case 0: + request->fl_type = F_UNLCK; + break; + case -NFS4ERR_DENIED: + status = 0; } out: up_read(&clp->cl_sem); @@ -3048,17 +3009,42 @@ static int do_vfs_lock(struct file *file, struct file_lock *fl) } struct nfs4_unlockdata { - struct nfs_lockargs arg; - struct nfs_locku_opargs luargs; - struct nfs_lockres res; + struct nfs_locku_args arg; + struct nfs_locku_res res; struct nfs4_lock_state *lsp; struct nfs_open_context *ctx; + struct file_lock fl; + const struct nfs_server *server; }; +static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl, + struct nfs_open_context *ctx, + struct nfs4_lock_state *lsp, + struct nfs_seqid *seqid) +{ + struct nfs4_unlockdata *p; + struct inode *inode = lsp->ls_state->inode; + + p = kmalloc(sizeof(*p), GFP_KERNEL); + if (p == NULL) + return NULL; + p->arg.fh = NFS_FH(inode); + p->arg.fl = &p->fl; + p->arg.seqid = seqid; + p->arg.stateid = &lsp->ls_stateid; + p->lsp = lsp; + atomic_inc(&lsp->ls_count); + /* Ensure we don't close file until we're done freeing locks! */ + p->ctx = get_nfs_open_context(ctx); + memcpy(&p->fl, fl, sizeof(p->fl)); + p->server = NFS_SERVER(inode); + return p; +} + static void nfs4_locku_release_calldata(void *data) { struct nfs4_unlockdata *calldata = data; - nfs_free_seqid(calldata->luargs.seqid); + nfs_free_seqid(calldata->arg.seqid); nfs4_put_lock_state(calldata->lsp); put_nfs_open_context(calldata->ctx); kfree(calldata); @@ -3070,19 +3056,19 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) if (RPC_ASSASSINATED(task)) return; - nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid); + nfs_increment_lock_seqid(task->tk_status, calldata->arg.seqid); switch (task->tk_status) { case 0: memcpy(calldata->lsp->ls_stateid.data, - calldata->res.u.stateid.data, + calldata->res.stateid.data, sizeof(calldata->lsp->ls_stateid.data)); break; case -NFS4ERR_STALE_STATEID: case -NFS4ERR_EXPIRED: - nfs4_schedule_state_recovery(calldata->res.server->nfs4_state); + nfs4_schedule_state_recovery(calldata->server->nfs4_state); break; default: - if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) { + if (nfs4_async_handle_error(task, calldata->server) == -EAGAIN) { rpc_restart_call(task); } } @@ -3097,10 +3083,8 @@ static void nfs4_locku_prepare(struct rpc_task *task, void *data) .rpc_resp = &calldata->res, .rpc_cred = calldata->lsp->ls_state->owner->so_cred, }; - int status; - status = nfs_wait_on_sequence(calldata->luargs.seqid, task); - if (status != 0) + if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0) return; if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) { /* Note: exit _without_ running nfs4_locku_done */ @@ -3121,43 +3105,32 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * struct nfs4_unlockdata *calldata; struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); + struct nfs_seqid *seqid; struct nfs4_lock_state *lsp; struct rpc_task *task; int status = 0; /* Is this a delegated lock? */ if (test_bit(NFS_DELEGATED_STATE, &state->flags)) - goto out; + goto out_unlock; + /* Is this open_owner holding any locks on the server? */ + if (test_bit(LK_STATE_IN_USE, &state->flags) == 0) + goto out_unlock; status = nfs4_set_lock_state(state, request); if (status != 0) - goto out; + goto out_unlock; lsp = request->fl_u.nfs4_fl.owner; - /* We might have lost the locks! */ - if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) - goto out; status = -ENOMEM; - calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); + seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (seqid == NULL) + goto out_unlock; + calldata = nfs4_alloc_unlockdata(request, request->fl_file->private_data, + lsp, seqid); if (calldata == NULL) - goto out; - calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); - if (calldata->luargs.seqid == NULL) { - kfree(calldata); - goto out; - } - calldata->luargs.stateid = &lsp->ls_stateid; - calldata->arg.fh = NFS_FH(inode); - calldata->arg.type = nfs4_lck_type(cmd, request); - calldata->arg.offset = request->fl_start; - calldata->arg.length = nfs4_lck_length(request); - calldata->arg.u.locku = &calldata->luargs; - calldata->res.server = server; - calldata->lsp = lsp; - atomic_inc(&lsp->ls_count); - - /* Ensure we don't close file until we're done freeing locks! */ - calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data); - + goto out_free_seqid; + /* Unlock _before_ we do the RPC call */ + do_vfs_lock(request->fl_file, request); task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_locku_ops, calldata); if (!IS_ERR(task)) { status = nfs4_wait_for_completion_rpc_task(task); @@ -3166,7 +3139,10 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * status = PTR_ERR(task); nfs4_locku_release_calldata(calldata); } -out: + return status; +out_free_seqid: + nfs_free_seqid(seqid); +out_unlock: do_vfs_lock(request->fl_file, request); return status; } @@ -3176,27 +3152,19 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; - struct nfs_lock_opargs largs = { + struct nfs_lock_args arg = { + .fh = NFS_FH(inode), + .fl = request, .lock_stateid = &lsp->ls_stateid, .open_stateid = &state->stateid, .lock_owner = { .clientid = server->nfs4_state->cl_clientid, .id = lsp->ls_id, }, + .block = (IS_SETLKW(cmd)) ? 1 : 0, .reclaim = reclaim, }; - struct nfs_lockargs arg = { - .fh = NFS_FH(inode), - .type = nfs4_lck_type(cmd, request), - .offset = request->fl_start, - .length = nfs4_lck_length(request), - .u = { - .lock = &largs, - }, - }; - struct nfs_lockres res = { - .server = server, - }; + struct nfs_lock_res res; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK], .rpc_argp = &arg, @@ -3205,37 +3173,37 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r }; int status = -ENOMEM; - largs.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); - if (largs.lock_seqid == NULL) + arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (arg.lock_seqid == NULL) return -ENOMEM; if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) { struct nfs4_state_owner *owner = state->owner; - largs.open_seqid = nfs_alloc_seqid(&owner->so_seqid); - if (largs.open_seqid == NULL) + arg.open_seqid = nfs_alloc_seqid(&owner->so_seqid); + if (arg.open_seqid == NULL) goto out; - largs.new_lock_owner = 1; + arg.new_lock_owner = 1; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); /* increment open seqid on success, and seqid mutating errors */ - if (largs.new_lock_owner != 0) { - nfs_increment_open_seqid(status, largs.open_seqid); + if (arg.new_lock_owner != 0) { + nfs_increment_open_seqid(status, arg.open_seqid); if (status == 0) nfs_confirm_seqid(&lsp->ls_seqid, 0); } - nfs_free_seqid(largs.open_seqid); + nfs_free_seqid(arg.open_seqid); } else status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); /* increment lock seqid on success, and seqid mutating errors*/ - nfs_increment_lock_seqid(status, largs.lock_seqid); + nfs_increment_lock_seqid(status, arg.lock_seqid); /* save the returned stateid. */ if (status == 0) { - memcpy(lsp->ls_stateid.data, res.u.stateid.data, + memcpy(lsp->ls_stateid.data, res.stateid.data, sizeof(lsp->ls_stateid.data)); lsp->ls_flags |= NFS_LOCK_INITIALIZED; } else if (status == -NFS4ERR_DENIED) status = -EAGAIN; out: - nfs_free_seqid(largs.lock_seqid); + nfs_free_seqid(arg.lock_seqid); return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 3100172822..a7b5de899c 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -742,69 +742,80 @@ static int encode_link(struct xdr_stream *xdr, const struct qstr *name) return 0; } +static inline int nfs4_lock_type(struct file_lock *fl, int block) +{ + if ((fl->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) == F_RDLCK) + return block ? NFS4_READW_LT : NFS4_READ_LT; + return block ? NFS4_WRITEW_LT : NFS4_WRITE_LT; +} + +static inline uint64_t nfs4_lock_length(struct file_lock *fl) +{ + if (fl->fl_end == OFFSET_MAX) + return ~(uint64_t)0; + return fl->fl_end - fl->fl_start + 1; +} + /* * opcode,type,reclaim,offset,length,new_lock_owner = 32 * open_seqid,open_stateid,lock_seqid,lock_owner.clientid, lock_owner.id = 40 */ -static int encode_lock(struct xdr_stream *xdr, const struct nfs_lockargs *arg) +static int encode_lock(struct xdr_stream *xdr, const struct nfs_lock_args *args) { uint32_t *p; - struct nfs_lock_opargs *opargs = arg->u.lock; RESERVE_SPACE(32); WRITE32(OP_LOCK); - WRITE32(arg->type); - WRITE32(opargs->reclaim); - WRITE64(arg->offset); - WRITE64(arg->length); - WRITE32(opargs->new_lock_owner); - if (opargs->new_lock_owner){ + WRITE32(nfs4_lock_type(args->fl, args->block)); + WRITE32(args->reclaim); + WRITE64(args->fl->fl_start); + WRITE64(nfs4_lock_length(args->fl)); + WRITE32(args->new_lock_owner); + if (args->new_lock_owner){ RESERVE_SPACE(40); - WRITE32(opargs->open_seqid->sequence->counter); - WRITEMEM(opargs->open_stateid->data, sizeof(opargs->open_stateid->data)); - WRITE32(opargs->lock_seqid->sequence->counter); - WRITE64(opargs->lock_owner.clientid); + WRITE32(args->open_seqid->sequence->counter); + WRITEMEM(args->open_stateid->data, sizeof(args->open_stateid->data)); + WRITE32(args->lock_seqid->sequence->counter); + WRITE64(args->lock_owner.clientid); WRITE32(4); - WRITE32(opargs->lock_owner.id); + WRITE32(args->lock_owner.id); } else { RESERVE_SPACE(20); - WRITEMEM(opargs->lock_stateid->data, sizeof(opargs->lock_stateid->data)); - WRITE32(opargs->lock_seqid->sequence->counter); + WRITEMEM(args->lock_stateid->data, sizeof(args->lock_stateid->data)); + WRITE32(args->lock_seqid->sequence->counter); } return 0; } -static int encode_lockt(struct xdr_stream *xdr, const struct nfs_lockargs *arg) +static int encode_lockt(struct xdr_stream *xdr, const struct nfs_lockt_args *args) { uint32_t *p; - struct nfs_lowner *opargs = arg->u.lockt; RESERVE_SPACE(40); WRITE32(OP_LOCKT); - WRITE32(arg->type); - WRITE64(arg->offset); - WRITE64(arg->length); - WRITE64(opargs->clientid); + WRITE32(nfs4_lock_type(args->fl, 0)); + WRITE64(args->fl->fl_start); + WRITE64(nfs4_lock_length(args->fl)); + WRITE64(args->lock_owner.clientid); WRITE32(4); - WRITE32(opargs->id); + WRITE32(args->lock_owner.id); return 0; } -static int encode_locku(struct xdr_stream *xdr, const struct nfs_lockargs *arg) +static int encode_locku(struct xdr_stream *xdr, const struct nfs_locku_args *args) { uint32_t *p; - struct nfs_locku_opargs *opargs = arg->u.locku; RESERVE_SPACE(44); WRITE32(OP_LOCKU); - WRITE32(arg->type); - WRITE32(opargs->seqid->sequence->counter); - WRITEMEM(opargs->stateid->data, sizeof(opargs->stateid->data)); - WRITE64(arg->offset); - WRITE64(arg->length); + WRITE32(nfs4_lock_type(args->fl, 0)); + WRITE32(args->seqid->sequence->counter); + WRITEMEM(args->stateid->data, sizeof(args->stateid->data)); + WRITE64(args->fl->fl_start); + WRITE64(nfs4_lock_length(args->fl)); return 0; } @@ -1596,21 +1607,20 @@ out: /* * Encode a LOCK request */ -static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_lockargs *args) +static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_lock_args *args) { struct xdr_stream xdr; struct compound_hdr hdr = { .nops = 2, }; - struct nfs_lock_opargs *opargs = args->u.lock; int status; - status = nfs_wait_on_sequence(opargs->lock_seqid, req->rq_task); + status = nfs_wait_on_sequence(args->lock_seqid, req->rq_task); if (status != 0) goto out; /* Do we need to do an open_to_lock_owner? */ - if (opargs->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED) - opargs->new_lock_owner = 0; + if (args->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED) + args->new_lock_owner = 0; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1624,7 +1634,7 @@ out: /* * Encode a LOCKT request */ -static int nfs4_xdr_enc_lockt(struct rpc_rqst *req, uint32_t *p, struct nfs_lockargs *args) +static int nfs4_xdr_enc_lockt(struct rpc_rqst *req, uint32_t *p, struct nfs_lockt_args *args) { struct xdr_stream xdr; struct compound_hdr hdr = { @@ -1645,7 +1655,7 @@ out: /* * Encode a LOCKU request */ -static int nfs4_xdr_enc_locku(struct rpc_rqst *req, uint32_t *p, struct nfs_lockargs *args) +static int nfs4_xdr_enc_locku(struct rpc_rqst *req, uint32_t *p, struct nfs_locku_args *args) { struct xdr_stream xdr; struct compound_hdr hdr = { @@ -2949,55 +2959,64 @@ static int decode_link(struct xdr_stream *xdr, struct nfs4_change_info *cinfo) /* * We create the owner, so we know a proper owner.id length is 4. */ -static int decode_lock_denied (struct xdr_stream *xdr, struct nfs_lock_denied *denied) +static int decode_lock_denied (struct xdr_stream *xdr, struct file_lock *fl) { + uint64_t offset, length, clientid; uint32_t *p; - uint32_t namelen; + uint32_t namelen, type; READ_BUF(32); - READ64(denied->offset); - READ64(denied->length); - READ32(denied->type); - READ64(denied->owner.clientid); + READ64(offset); + READ64(length); + READ32(type); + if (fl != NULL) { + fl->fl_start = (loff_t)offset; + fl->fl_end = fl->fl_start + (loff_t)length - 1; + if (length == ~(uint64_t)0) + fl->fl_end = OFFSET_MAX; + fl->fl_type = F_WRLCK; + if (type & 1) + fl->fl_type = F_RDLCK; + fl->fl_pid = 0; + } + READ64(clientid); READ32(namelen); READ_BUF(namelen); - if (namelen == 4) - READ32(denied->owner.id); return -NFS4ERR_DENIED; } -static int decode_lock(struct xdr_stream *xdr, struct nfs_lockres *res) +static int decode_lock(struct xdr_stream *xdr, struct nfs_lock_res *res) { uint32_t *p; int status; status = decode_op_hdr(xdr, OP_LOCK); if (status == 0) { - READ_BUF(sizeof(res->u.stateid.data)); - COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); + READ_BUF(sizeof(res->stateid.data)); + COPYMEM(res->stateid.data, sizeof(res->stateid.data)); } else if (status == -NFS4ERR_DENIED) - return decode_lock_denied(xdr, &res->u.denied); + return decode_lock_denied(xdr, NULL); return status; } -static int decode_lockt(struct xdr_stream *xdr, struct nfs_lockres *res) +static int decode_lockt(struct xdr_stream *xdr, struct nfs_lockt_res *res) { int status; status = decode_op_hdr(xdr, OP_LOCKT); if (status == -NFS4ERR_DENIED) - return decode_lock_denied(xdr, &res->u.denied); + return decode_lock_denied(xdr, res->denied); return status; } -static int decode_locku(struct xdr_stream *xdr, struct nfs_lockres *res) +static int decode_locku(struct xdr_stream *xdr, struct nfs_locku_res *res) { uint32_t *p; int status; status = decode_op_hdr(xdr, OP_LOCKU); if (status == 0) { - READ_BUF(sizeof(res->u.stateid.data)); - COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); + READ_BUF(sizeof(res->stateid.data)); + COPYMEM(res->stateid.data, sizeof(res->stateid.data)); } return status; } @@ -3861,7 +3880,7 @@ out: /* * Decode LOCK response */ -static int nfs4_xdr_dec_lock(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_lockres *res) +static int nfs4_xdr_dec_lock(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_lock_res *res) { struct xdr_stream xdr; struct compound_hdr hdr; @@ -3882,7 +3901,7 @@ out: /* * Decode LOCKT response */ -static int nfs4_xdr_dec_lockt(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_lockres *res) +static int nfs4_xdr_dec_lockt(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_lockt_res *res) { struct xdr_stream xdr; struct compound_hdr hdr; @@ -3903,7 +3922,7 @@ out: /* * Decode LOCKU response */ -static int nfs4_xdr_dec_locku(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_lockres *res) +static int nfs4_xdr_dec_locku(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_locku_res *res) { struct xdr_stream xdr; struct compound_hdr hdr; -- cgit v1.2.2 From a5d16a4d090bd2af86e648ed9bb205903fcf1e86 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:17 +0100 Subject: NFSv4: Convert LOCK rpc call into an asynchronous RPC call In order to allow users to interrupt/cancel it. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 244 +++++++++++++++++++++++++++++++++++++++--------------- fs/nfs/nfs4xdr.c | 6 -- 2 files changed, 175 insertions(+), 75 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 857125705b..718fcc7225 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3100,11 +3100,30 @@ static const struct rpc_call_ops nfs4_locku_ops = { .rpc_release = nfs4_locku_release_calldata, }; +static struct rpc_task *nfs4_do_unlck(struct file_lock *fl, + struct nfs_open_context *ctx, + struct nfs4_lock_state *lsp, + struct nfs_seqid *seqid) +{ + struct nfs4_unlockdata *data; + struct rpc_task *task; + + data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid); + if (data == NULL) { + nfs_free_seqid(seqid); + return ERR_PTR(-ENOMEM); + } + + /* Unlock _before_ we do the RPC call */ + do_vfs_lock(fl->fl_file, fl); + task = rpc_run_task(NFS_CLIENT(lsp->ls_state->inode), RPC_TASK_ASYNC, &nfs4_locku_ops, data); + if (IS_ERR(task)) + nfs4_locku_release_calldata(data); + return task; +} + static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) { - struct nfs4_unlockdata *calldata; - struct inode *inode = state->inode; - struct nfs_server *server = NFS_SERVER(inode); struct nfs_seqid *seqid; struct nfs4_lock_state *lsp; struct rpc_task *task; @@ -3125,86 +3144,173 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock * seqid = nfs_alloc_seqid(&lsp->ls_seqid); if (seqid == NULL) goto out_unlock; - calldata = nfs4_alloc_unlockdata(request, request->fl_file->private_data, - lsp, seqid); - if (calldata == NULL) - goto out_free_seqid; - /* Unlock _before_ we do the RPC call */ - do_vfs_lock(request->fl_file, request); - task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_locku_ops, calldata); - if (!IS_ERR(task)) { - status = nfs4_wait_for_completion_rpc_task(task); - rpc_release_task(task); - } else { - status = PTR_ERR(task); - nfs4_locku_release_calldata(calldata); - } + task = nfs4_do_unlck(request, request->fl_file->private_data, lsp, seqid); + status = PTR_ERR(task); + if (IS_ERR(task)) + goto out_unlock; + status = nfs4_wait_for_completion_rpc_task(task); + rpc_release_task(task); return status; -out_free_seqid: - nfs_free_seqid(seqid); out_unlock: do_vfs_lock(request->fl_file, request); return status; } -static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim) +struct nfs4_lockdata { + struct nfs_lock_args arg; + struct nfs_lock_res res; + struct nfs4_lock_state *lsp; + struct nfs_open_context *ctx; + struct file_lock fl; + int rpc_status; + int cancelled; +}; + +static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl, + struct nfs_open_context *ctx, struct nfs4_lock_state *lsp) { - struct inode *inode = state->inode; + struct nfs4_lockdata *p; + struct inode *inode = lsp->ls_state->inode; struct nfs_server *server = NFS_SERVER(inode); - struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; - struct nfs_lock_args arg = { - .fh = NFS_FH(inode), - .fl = request, - .lock_stateid = &lsp->ls_stateid, - .open_stateid = &state->stateid, - .lock_owner = { - .clientid = server->nfs4_state->cl_clientid, - .id = lsp->ls_id, - }, - .block = (IS_SETLKW(cmd)) ? 1 : 0, - .reclaim = reclaim, - }; - struct nfs_lock_res res; + + p = kzalloc(sizeof(*p), GFP_KERNEL); + if (p == NULL) + return NULL; + + p->arg.fh = NFS_FH(inode); + p->arg.fl = &p->fl; + p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (p->arg.lock_seqid == NULL) + goto out_free; + p->arg.lock_stateid = &lsp->ls_stateid; + p->arg.lock_owner.clientid = server->nfs4_state->cl_clientid; + p->arg.lock_owner.id = lsp->ls_id; + p->lsp = lsp; + atomic_inc(&lsp->ls_count); + p->ctx = get_nfs_open_context(ctx); + memcpy(&p->fl, fl, sizeof(p->fl)); + return p; +out_free: + kfree(p); + return NULL; +} + +static void nfs4_lock_prepare(struct rpc_task *task, void *calldata) +{ + struct nfs4_lockdata *data = calldata; + struct nfs4_state *state = data->lsp->ls_state; + struct nfs4_state_owner *sp = state->owner; struct rpc_message msg = { - .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK], - .rpc_argp = &arg, - .rpc_resp = &res, - .rpc_cred = state->owner->so_cred, + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK], + .rpc_argp = &data->arg, + .rpc_resp = &data->res, + .rpc_cred = sp->so_cred, }; - int status = -ENOMEM; - - arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); - if (arg.lock_seqid == NULL) - return -ENOMEM; - if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) { - struct nfs4_state_owner *owner = state->owner; - arg.open_seqid = nfs_alloc_seqid(&owner->so_seqid); - if (arg.open_seqid == NULL) + if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0) + return; + dprintk("%s: begin!\n", __FUNCTION__); + /* Do we need to do an open_to_lock_owner? */ + if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) { + data->arg.open_seqid = nfs_alloc_seqid(&sp->so_seqid); + if (data->arg.open_seqid == NULL) { + data->rpc_status = -ENOMEM; + task->tk_action = NULL; goto out; - arg.new_lock_owner = 1; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment open seqid on success, and seqid mutating errors */ - if (arg.new_lock_owner != 0) { - nfs_increment_open_seqid(status, arg.open_seqid); - if (status == 0) - nfs_confirm_seqid(&lsp->ls_seqid, 0); } - nfs_free_seqid(arg.open_seqid); - } else - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment lock seqid on success, and seqid mutating errors*/ - nfs_increment_lock_seqid(status, arg.lock_seqid); - /* save the returned stateid. */ - if (status == 0) { - memcpy(lsp->ls_stateid.data, res.stateid.data, - sizeof(lsp->ls_stateid.data)); - lsp->ls_flags |= NFS_LOCK_INITIALIZED; - } else if (status == -NFS4ERR_DENIED) - status = -EAGAIN; + data->arg.open_stateid = &state->stateid; + data->arg.new_lock_owner = 1; + } + rpc_call_setup(task, &msg, 0); out: - nfs_free_seqid(arg.lock_seqid); - return status; + dprintk("%s: done!, ret = %d\n", __FUNCTION__, data->rpc_status); +} + +static void nfs4_lock_done(struct rpc_task *task, void *calldata) +{ + struct nfs4_lockdata *data = calldata; + + dprintk("%s: begin!\n", __FUNCTION__); + + data->rpc_status = task->tk_status; + if (RPC_ASSASSINATED(task)) + goto out; + if (data->arg.new_lock_owner != 0) { + nfs_increment_open_seqid(data->rpc_status, data->arg.open_seqid); + if (data->rpc_status == 0) + nfs_confirm_seqid(&data->lsp->ls_seqid, 0); + else + goto out; + } + if (data->rpc_status == 0) { + memcpy(data->lsp->ls_stateid.data, data->res.stateid.data, + sizeof(data->lsp->ls_stateid.data)); + data->lsp->ls_flags |= NFS_LOCK_INITIALIZED; + } + nfs_increment_lock_seqid(data->rpc_status, data->arg.lock_seqid); +out: + dprintk("%s: done, ret = %d!\n", __FUNCTION__, data->rpc_status); +} + +static void nfs4_lock_release(void *calldata) +{ + struct nfs4_lockdata *data = calldata; + + dprintk("%s: begin!\n", __FUNCTION__); + if (data->arg.open_seqid != NULL) + nfs_free_seqid(data->arg.open_seqid); + if (data->cancelled != 0) { + struct rpc_task *task; + task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp, + data->arg.lock_seqid); + if (!IS_ERR(task)) + rpc_release_task(task); + dprintk("%s: cancelling lock!\n", __FUNCTION__); + } else + nfs_free_seqid(data->arg.lock_seqid); + nfs4_put_lock_state(data->lsp); + put_nfs_open_context(data->ctx); + kfree(data); + dprintk("%s: done!\n", __FUNCTION__); +} + +static const struct rpc_call_ops nfs4_lock_ops = { + .rpc_call_prepare = nfs4_lock_prepare, + .rpc_call_done = nfs4_lock_done, + .rpc_release = nfs4_lock_release, +}; + +static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim) +{ + struct nfs4_lockdata *data; + struct rpc_task *task; + int ret; + + dprintk("%s: begin!\n", __FUNCTION__); + data = nfs4_alloc_lockdata(fl, fl->fl_file->private_data, + fl->fl_u.nfs4_fl.owner); + if (data == NULL) + return -ENOMEM; + if (IS_SETLKW(cmd)) + data->arg.block = 1; + if (reclaim != 0) + data->arg.reclaim = 1; + task = rpc_run_task(NFS_CLIENT(state->inode), RPC_TASK_ASYNC, + &nfs4_lock_ops, data); + if (IS_ERR(task)) { + nfs4_lock_release(data); + return PTR_ERR(task); + } + ret = nfs4_wait_for_completion_rpc_task(task); + if (ret == 0) { + ret = data->rpc_status; + if (ret == -NFS4ERR_DENIED) + ret = -EAGAIN; + } else + data->cancelled = 1; + rpc_release_task(task); + dprintk("%s: done, ret = %d!\n", __FUNCTION__, ret); + return ret; } static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index a7b5de899c..5d6bda43df 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1615,12 +1615,6 @@ static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_lock_ }; int status; - status = nfs_wait_on_sequence(args->lock_seqid, req->rq_task); - if (status != 0) - goto out; - /* Do we need to do an open_to_lock_owner? */ - if (args->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED) - args->new_lock_owner = 0; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); -- cgit v1.2.2 From fe650407a86823bcafbfbee96c7bc6a1b5cd1c76 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:18 +0100 Subject: NFSv4: Make DELEGRETURN an interruptible operation. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 60 insertions(+), 8 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 718fcc7225..f060e6538e 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2885,19 +2885,71 @@ nfs4_proc_setclientid_confirm(struct nfs4_client *clp) return status; } -static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid) +struct nfs4_delegreturndata { + struct nfs4_delegreturnargs args; + struct nfs_fh fh; + nfs4_stateid stateid; + struct rpc_cred *cred; + int rpc_status; +}; + +static void nfs4_delegreturn_prepare(struct rpc_task *task, void *calldata) { - struct nfs4_delegreturnargs args = { - .fhandle = NFS_FH(inode), - .stateid = stateid, - }; + struct nfs4_delegreturndata *data = calldata; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN], - .rpc_argp = &args, - .rpc_cred = cred, + .rpc_argp = &data->args, + .rpc_cred = data->cred, }; + rpc_call_setup(task, &msg, 0); +} - return rpc_call_sync(NFS_CLIENT(inode), &msg, 0); +static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) +{ + struct nfs4_delegreturndata *data = calldata; + data->rpc_status = task->tk_status; +} + +static void nfs4_delegreturn_release(void *calldata) +{ + struct nfs4_delegreturndata *data = calldata; + + put_rpccred(data->cred); + kfree(calldata); +} + +const static struct rpc_call_ops nfs4_delegreturn_ops = { + .rpc_call_prepare = nfs4_delegreturn_prepare, + .rpc_call_done = nfs4_delegreturn_done, + .rpc_release = nfs4_delegreturn_release, +}; + +static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid) +{ + struct nfs4_delegreturndata *data; + struct rpc_task *task; + int status; + + data = kmalloc(sizeof(*data), GFP_KERNEL); + if (data == NULL) + return -ENOMEM; + data->args.fhandle = &data->fh; + data->args.stateid = &data->stateid; + nfs_copy_fh(&data->fh, NFS_FH(inode)); + memcpy(&data->stateid, stateid, sizeof(data->stateid)); + data->cred = get_rpccred(cred); + data->rpc_status = 0; + + task = rpc_run_task(NFS_CLIENT(inode), RPC_TASK_ASYNC, &nfs4_delegreturn_ops, data); + if (IS_ERR(task)) { + nfs4_delegreturn_release(data); + return PTR_ERR(task); + } + status = nfs4_wait_for_completion_rpc_task(task); + if (status == 0) + status = data->rpc_status; + rpc_release_task(task); + return status; } int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid) -- cgit v1.2.2 From 26e976a884be9aa08f8ff906372f25f68df0d948 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:21 +0100 Subject: NFSv4: OPEN/LOCK/LOCKU/CLOSE will automatically renew the NFSv4 lease Cut down on the number of unnecessary RENEW requests on the wire. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 34 +++++++++++++++++++++++++++++++--- 1 file changed, 31 insertions(+), 3 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f060e6538e..76d7d0259b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -174,8 +174,7 @@ static void nfs4_setup_readdir(u64 cookie, u32 *verifier, struct dentry *dentry, kunmap_atomic(start, KM_USER0); } -static void -renew_lease(struct nfs_server *server, unsigned long timestamp) +static void renew_lease(const struct nfs_server *server, unsigned long timestamp) { struct nfs4_client *clp = server->nfs4_state; spin_lock(&clp->cl_lock); @@ -207,6 +206,7 @@ struct nfs4_opendata { struct dentry *dir; struct nfs4_state_owner *owner; struct iattr attrs; + unsigned long timestamp; int rpc_status; int cancelled; }; @@ -548,6 +548,7 @@ static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata) .rpc_resp = &data->c_res, .rpc_cred = data->owner->so_cred, }; + data->timestamp = jiffies; rpc_call_setup(task, &msg, 0); } @@ -558,9 +559,11 @@ static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata) data->rpc_status = task->tk_status; if (RPC_ASSASSINATED(task)) return; - if (data->rpc_status == 0) + if (data->rpc_status == 0) { memcpy(data->o_res.stateid.data, data->c_res.stateid.data, sizeof(data->o_res.stateid.data)); + renew_lease(data->o_res.server, data->timestamp); + } nfs_increment_open_seqid(data->rpc_status, data->c_arg.seqid); nfs_confirm_seqid(&data->owner->so_seqid, data->rpc_status); } @@ -633,6 +636,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) data->o_arg.clientid = sp->so_client->cl_clientid; if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR]; + data->timestamp = jiffies; rpc_call_setup(task, &msg, 0); } @@ -656,6 +660,7 @@ static void nfs4_open_done(struct rpc_task *task, void *calldata) default: data->rpc_status = -ENOTDIR; } + renew_lease(data->o_res.server, data->timestamp); } nfs_increment_open_seqid(data->rpc_status, data->o_arg.seqid); } @@ -1014,6 +1019,7 @@ static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr, .rpc_argp = &arg, .rpc_resp = &res, }; + unsigned long timestamp = jiffies; int status; nfs_fattr_init(fattr); @@ -1025,6 +1031,8 @@ static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr, memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid)); status = rpc_call_sync(server->client, &msg, 0); + if (status == 0 && state != NULL) + renew_lease(server, timestamp); return status; } @@ -1049,6 +1057,7 @@ struct nfs4_closedata { struct nfs_closeargs arg; struct nfs_closeres res; struct nfs_fattr fattr; + unsigned long timestamp; }; static void nfs4_free_closedata(void *data) @@ -1078,6 +1087,7 @@ static void nfs4_close_done(struct rpc_task *task, void *data) case 0: memcpy(&state->stateid, &calldata->res.stateid, sizeof(state->stateid)); + renew_lease(server, calldata->timestamp); break; case -NFS4ERR_STALE_STATEID: case -NFS4ERR_EXPIRED: @@ -1130,6 +1140,7 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) if (mode != 0) msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE]; calldata->arg.open_flags = mode; + calldata->timestamp = jiffies; rpc_call_setup(task, &msg, 0); } @@ -1694,11 +1705,13 @@ static int _nfs4_proc_write(struct nfs_write_data *wdata) wdata->args.bitmask = server->attr_bitmask; wdata->res.server = server; + wdata->timestamp = jiffies; nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, rpcflags); dprintk("NFS reply write: %d\n", status); if (status < 0) return status; + renew_lease(server, wdata->timestamp); nfs_post_op_update_inode(inode, fattr); return wdata->res.count; } @@ -1733,8 +1746,11 @@ static int _nfs4_proc_commit(struct nfs_write_data *cdata) cdata->args.bitmask = server->attr_bitmask; cdata->res.server = server; + cdata->timestamp = jiffies; nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, 0); + if (status >= 0) + renew_lease(server, cdata->timestamp); dprintk("NFS reply commit: %d\n", status); if (status >= 0) nfs_post_op_update_inode(inode, fattr); @@ -2890,6 +2906,8 @@ struct nfs4_delegreturndata { struct nfs_fh fh; nfs4_stateid stateid; struct rpc_cred *cred; + unsigned long timestamp; + const struct nfs_server *server; int rpc_status; }; @@ -2908,6 +2926,8 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) { struct nfs4_delegreturndata *data = calldata; data->rpc_status = task->tk_status; + if (data->rpc_status == 0) + renew_lease(data->server, data->timestamp); } static void nfs4_delegreturn_release(void *calldata) @@ -2938,6 +2958,8 @@ static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, co nfs_copy_fh(&data->fh, NFS_FH(inode)); memcpy(&data->stateid, stateid, sizeof(data->stateid)); data->cred = get_rpccred(cred); + data->timestamp = jiffies; + data->server = NFS_SERVER(inode); data->rpc_status = 0; task = rpc_run_task(NFS_CLIENT(inode), RPC_TASK_ASYNC, &nfs4_delegreturn_ops, data); @@ -3067,6 +3089,7 @@ struct nfs4_unlockdata { struct nfs_open_context *ctx; struct file_lock fl; const struct nfs_server *server; + unsigned long timestamp; }; static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl, @@ -3114,6 +3137,7 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) memcpy(calldata->lsp->ls_stateid.data, calldata->res.stateid.data, sizeof(calldata->lsp->ls_stateid.data)); + renew_lease(calldata->server, calldata->timestamp); break; case -NFS4ERR_STALE_STATEID: case -NFS4ERR_EXPIRED: @@ -3143,6 +3167,7 @@ static void nfs4_locku_prepare(struct rpc_task *task, void *data) task->tk_action = NULL; return; } + calldata->timestamp = jiffies; rpc_call_setup(task, &msg, 0); } @@ -3214,6 +3239,7 @@ struct nfs4_lockdata { struct nfs4_lock_state *lsp; struct nfs_open_context *ctx; struct file_lock fl; + unsigned long timestamp; int rpc_status; int cancelled; }; @@ -3273,6 +3299,7 @@ static void nfs4_lock_prepare(struct rpc_task *task, void *calldata) data->arg.open_stateid = &state->stateid; data->arg.new_lock_owner = 1; } + data->timestamp = jiffies; rpc_call_setup(task, &msg, 0); out: dprintk("%s: done!, ret = %d\n", __FUNCTION__, data->rpc_status); @@ -3298,6 +3325,7 @@ static void nfs4_lock_done(struct rpc_task *task, void *calldata) memcpy(data->lsp->ls_stateid.data, data->res.stateid.data, sizeof(data->lsp->ls_stateid.data)); data->lsp->ls_flags |= NFS_LOCK_INITIALIZED; + renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp); } nfs_increment_lock_seqid(data->rpc_status, data->arg.lock_seqid); out: -- cgit v1.2.2 From 433fbe4c8837e3cc2ba6a6374edf28737d01a2e9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:22 +0100 Subject: NFSv4: State recovery cleanup Use wait_on_bit() when waiting for state recovery to complete. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 3 +-- fs/nfs/nfs4proc.c | 27 ++++++++++++++------------- fs/nfs/nfs4state.c | 23 +++++++++++++---------- 3 files changed, 28 insertions(+), 25 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 4ad5981f93..1d4c5b339b 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -38,7 +38,7 @@ struct idmap; ((err) != NFSERR_NOFILEHANDLE)) enum nfs4_client_state { - NFS4CLNT_OK = 0, + NFS4CLNT_STATE_RECOVER = 0, }; /* @@ -76,7 +76,6 @@ struct nfs4_client { struct work_struct cl_renewd; struct work_struct cl_recoverd; - wait_queue_head_t cl_waitq; struct rpc_wait_queue cl_rpcwaitq; /* used for the setclientid verifier */ diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 76d7d0259b..46623ac3ce 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2736,7 +2736,7 @@ nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server) case -NFS4ERR_EXPIRED: rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL, NULL); nfs4_schedule_state_recovery(clp); - if (test_bit(NFS4CLNT_OK, &clp->cl_state)) + if (test_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0) rpc_wake_up_task(task); task->tk_status = 0; return -EAGAIN; @@ -2753,25 +2753,25 @@ nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server) return 0; } +static int nfs4_wait_bit_interruptible(void *word) +{ + if (signal_pending(current)) + return -ERESTARTSYS; + schedule(); + return 0; +} + static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs4_client *clp) { - DEFINE_WAIT(wait); sigset_t oldset; - int interruptible, res = 0; + int res; might_sleep(); rpc_clnt_sigmask(clnt, &oldset); - interruptible = TASK_UNINTERRUPTIBLE; - if (clnt->cl_intr) - interruptible = TASK_INTERRUPTIBLE; - prepare_to_wait(&clp->cl_waitq, &wait, interruptible); - nfs4_schedule_state_recovery(clp); - if (clnt->cl_intr && signalled()) - res = -ERESTARTSYS; - else if (!test_bit(NFS4CLNT_OK, &clp->cl_state)) - schedule(); - finish_wait(&clp->cl_waitq, &wait); + res = wait_on_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER, + nfs4_wait_bit_interruptible, + TASK_INTERRUPTIBLE); rpc_clnt_sigunmask(clnt, &oldset); return res; } @@ -2814,6 +2814,7 @@ int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct case -NFS4ERR_STALE_CLIENTID: case -NFS4ERR_STALE_STATEID: case -NFS4ERR_EXPIRED: + nfs4_schedule_state_recovery(clp); ret = nfs4_wait_clnt_recover(server->client, clp); if (ret == 0) exception->retry = 1; diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 41a5f1a8a9..4ebf4df7f1 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -106,11 +106,10 @@ nfs4_alloc_client(struct in_addr *addr) INIT_WORK(&clp->cl_recoverd, nfs4_recover_state, clp); INIT_WORK(&clp->cl_renewd, nfs4_renew_state, clp); INIT_LIST_HEAD(&clp->cl_superblocks); - init_waitqueue_head(&clp->cl_waitq); rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS4 client"); clp->cl_rpcclient = ERR_PTR(-EINVAL); clp->cl_boot_time = CURRENT_TIME; - clp->cl_state = 1 << NFS4CLNT_OK; + clp->cl_state = 0; return clp; } @@ -193,7 +192,6 @@ nfs4_put_client(struct nfs4_client *clp) list_del(&clp->cl_servers); spin_unlock(&state_spinlock); BUG_ON(!list_empty(&clp->cl_superblocks)); - wake_up_all(&clp->cl_waitq); rpc_wake_up(&clp->cl_rpcwaitq); nfs4_kill_renewd(clp); nfs4_free_client(clp); @@ -741,6 +739,15 @@ struct reclaimer_args { struct completion complete; }; +static inline void nfs4_clear_recover_bit(struct nfs4_client *clp) +{ + smp_mb__before_clear_bit(); + clear_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state); + smp_mb__after_clear_bit(); + wake_up_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER); + rpc_wake_up(&clp->cl_rpcwaitq); +} + /* * State recovery routine */ @@ -760,9 +767,7 @@ nfs4_recover_state(void *data) wait_for_completion(&args.complete); return; out_failed_clear: - set_bit(NFS4CLNT_OK, &clp->cl_state); - wake_up_all(&clp->cl_waitq); - rpc_wake_up(&clp->cl_rpcwaitq); + nfs4_clear_recover_bit(clp); } /* @@ -773,7 +778,7 @@ nfs4_schedule_state_recovery(struct nfs4_client *clp) { if (!clp) return; - if (test_and_clear_bit(NFS4CLNT_OK, &clp->cl_state)) + if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0) schedule_work(&clp->cl_recoverd); } @@ -943,13 +948,11 @@ restart_loop: } nfs_delegation_reap_unclaimed(clp); out: - set_bit(NFS4CLNT_OK, &clp->cl_state); up_write(&clp->cl_sem); unlock_kernel(); - wake_up_all(&clp->cl_waitq); - rpc_wake_up(&clp->cl_rpcwaitq); if (status == -NFS4ERR_CB_PATH_DOWN) nfs_handle_cb_pathdown(clp); + nfs4_clear_recover_bit(clp); nfs4_put_client(clp); return 0; out_error: -- cgit v1.2.2 From 5043e900f5404c01864fbeb5826aa7de3981bbc1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:23 +0100 Subject: NFS: Convert instances of kernel_thread() to kthread() Convert private implementations in NFSv4 state recovery and delegation code to use kthreads. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 46 ++++++++++++++++------------------------------ 1 file changed, 16 insertions(+), 30 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 4ebf4df7f1..f2148376f1 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -43,6 +43,8 @@ #include #include #include +#include +#include #include #include @@ -57,8 +59,6 @@ const nfs4_stateid zero_stateid; static DEFINE_SPINLOCK(state_spinlock); static LIST_HEAD(nfs4_clientid_list); -static void nfs4_recover_state(void *); - void init_nfsv4_state(struct nfs_server *server) { @@ -103,7 +103,6 @@ nfs4_alloc_client(struct in_addr *addr) INIT_LIST_HEAD(&clp->cl_unused); spin_lock_init(&clp->cl_lock); atomic_set(&clp->cl_count, 1); - INIT_WORK(&clp->cl_recoverd, nfs4_recover_state, clp); INIT_WORK(&clp->cl_renewd, nfs4_renew_state, clp); INIT_LIST_HEAD(&clp->cl_superblocks); rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS4 client"); @@ -734,10 +733,6 @@ out: } static int reclaimer(void *); -struct reclaimer_args { - struct nfs4_client *clp; - struct completion complete; -}; static inline void nfs4_clear_recover_bit(struct nfs4_client *clp) { @@ -751,35 +746,30 @@ static inline void nfs4_clear_recover_bit(struct nfs4_client *clp) /* * State recovery routine */ -void -nfs4_recover_state(void *data) +static void nfs4_recover_state(struct nfs4_client *clp) { - struct nfs4_client *clp = (struct nfs4_client *)data; - struct reclaimer_args args = { - .clp = clp, - }; - might_sleep(); - - init_completion(&args.complete); + struct task_struct *task; - if (kernel_thread(reclaimer, &args, CLONE_KERNEL) < 0) - goto out_failed_clear; - wait_for_completion(&args.complete); - return; -out_failed_clear: + __module_get(THIS_MODULE); + atomic_inc(&clp->cl_count); + task = kthread_run(reclaimer, clp, "%u.%u.%u.%u-reclaim", + NIPQUAD(clp->cl_addr)); + if (!IS_ERR(task)) + return; nfs4_clear_recover_bit(clp); + nfs4_put_client(clp); + module_put(THIS_MODULE); } /* * Schedule a state recovery attempt */ -void -nfs4_schedule_state_recovery(struct nfs4_client *clp) +void nfs4_schedule_state_recovery(struct nfs4_client *clp) { if (!clp) return; if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0) - schedule_work(&clp->cl_recoverd); + nfs4_recover_state(clp); } static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_state *state) @@ -895,18 +885,13 @@ static void nfs4_state_mark_reclaim(struct nfs4_client *clp) static int reclaimer(void *ptr) { - struct reclaimer_args *args = (struct reclaimer_args *)ptr; - struct nfs4_client *clp = args->clp; + struct nfs4_client *clp = ptr; struct nfs4_state_owner *sp; struct nfs4_state_recovery_ops *ops; int status = 0; - daemonize("%u.%u.%u.%u-reclaim", NIPQUAD(clp->cl_addr)); allow_signal(SIGKILL); - atomic_inc(&clp->cl_count); - complete(&args->complete); - /* Ensure exclusive access to NFSv4 state */ lock_kernel(); down_write(&clp->cl_sem); @@ -954,6 +939,7 @@ out: nfs_handle_cb_pathdown(clp); nfs4_clear_recover_bit(clp); nfs4_put_client(clp); + module_put_and_exit(0); return 0; out_error: printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n", -- cgit v1.2.2 From 58d9714a44a79bba9b414da3ffbf3c753dc5915f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:24 +0100 Subject: NFSv4: Send RENEW requests to the server only when we're holding state Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ fs/nfs/delegation.h | 1 + fs/nfs/nfs4_fs.h | 1 + fs/nfs/nfs4proc.c | 20 +++++++++++++++++++- fs/nfs/nfs4renewd.c | 9 ++++++++- 5 files changed, 75 insertions(+), 2 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 618a327027..75dfb1c717 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -8,6 +8,7 @@ */ #include #include +#include #include #include #include @@ -231,6 +232,51 @@ restart: spin_unlock(&clp->cl_lock); } +int nfs_do_expire_all_delegations(void *ptr) +{ + struct nfs4_client *clp = ptr; + struct nfs_delegation *delegation; + struct inode *inode; + int err = 0; + + allow_signal(SIGKILL); +restart: + spin_lock(&clp->cl_lock); + if (test_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) != 0) + goto out; + if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) + goto out; + list_for_each_entry(delegation, &clp->cl_delegations, super_list) { + inode = igrab(delegation->inode); + if (inode == NULL) + continue; + spin_unlock(&clp->cl_lock); + err = nfs_inode_return_delegation(inode); + iput(inode); + if (!err) + goto restart; + } +out: + spin_unlock(&clp->cl_lock); + nfs4_put_client(clp); + module_put_and_exit(0); +} + +void nfs_expire_all_delegations(struct nfs4_client *clp) +{ + struct task_struct *task; + + __module_get(THIS_MODULE); + atomic_inc(&clp->cl_count); + task = kthread_run(nfs_do_expire_all_delegations, clp, + "%u.%u.%u.%u-delegreturn", + NIPQUAD(clp->cl_addr)); + if (!IS_ERR(task)) + return; + nfs4_put_client(clp); + module_put(THIS_MODULE); +} + /* * Return all delegations following an NFS4ERR_CB_PATH_DOWN error. */ diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 2fcc30de92..fbc50ec271 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -30,6 +30,7 @@ int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *s struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle); void nfs_return_all_delegations(struct super_block *sb); +void nfs_expire_all_delegations(struct nfs4_client *clp); void nfs_handle_cb_pathdown(struct nfs4_client *clp); void nfs_delegation_mark_reclaim(struct nfs4_client *clp); diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 1d4c5b339b..75fe646c28 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -39,6 +39,7 @@ struct idmap; enum nfs4_client_state { NFS4CLNT_STATE_RECOVER = 0, + NFS4CLNT_LEASE_EXPIRED, }; /* diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 46623ac3ce..cc33a1c32c 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -63,6 +63,7 @@ static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinf static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception); +static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs4_client *clp); extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus); extern struct rpc_procinfo nfs4_procedures[]; @@ -765,6 +766,15 @@ out: return -EACCES; } +int nfs4_recover_expired_lease(struct nfs_server *server) +{ + struct nfs4_client *clp = server->nfs4_state; + + if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) + nfs4_schedule_state_recovery(clp); + return nfs4_wait_clnt_recover(server->client, clp); +} + /* * OPEN_EXPIRED: * reclaim state on the server after a network partition. @@ -840,6 +850,9 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred int open_flags = flags & (FMODE_READ|FMODE_WRITE); int err; + err = nfs4_recover_expired_lease(server); + if (err != 0) + return err; /* Protect against reboot recovery - NOTE ORDER! */ down_read(&clp->cl_sem); /* Protect against delegation recall */ @@ -921,12 +934,16 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st int status; /* Protect against reboot recovery conflicts */ - down_read(&clp->cl_sem); status = -ENOMEM; if (!(sp = nfs4_get_state_owner(server, cred))) { dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n"); goto out_err; } + status = nfs4_recover_expired_lease(server); + if (status != 0) + goto out_err; + down_read(&clp->cl_sem); + status = -ENOMEM; opendata = nfs4_opendata_alloc(dentry, sp, flags, sattr); if (opendata == NULL) goto err_put_state_owner; @@ -2897,6 +2914,7 @@ nfs4_proc_setclientid_confirm(struct nfs4_client *clp) spin_lock(&clp->cl_lock); clp->cl_lease_time = fsinfo.lease_time * HZ; clp->cl_last_renewal = now; + clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state); spin_unlock(&clp->cl_lock); } return status; diff --git a/fs/nfs/nfs4renewd.c b/fs/nfs/nfs4renewd.c index a3001628ad..f62c2f7a4f 100644 --- a/fs/nfs/nfs4renewd.c +++ b/fs/nfs/nfs4renewd.c @@ -54,6 +54,7 @@ #include #include #include "nfs4_fs.h" +#include "delegation.h" #define NFSDBG_FACILITY NFSDBG_PROC @@ -68,7 +69,7 @@ nfs4_renew_state(void *data) dprintk("%s: start\n", __FUNCTION__); /* Are there any active superblocks? */ if (list_empty(&clp->cl_superblocks)) - goto out; + goto out; spin_lock(&clp->cl_lock); lease = clp->cl_lease_time; last = clp->cl_last_renewal; @@ -76,6 +77,12 @@ nfs4_renew_state(void *data) timeout = (2 * lease) / 3 + (long)last - (long)now; /* Are we close to a lease timeout? */ if (time_after(now, last + lease/3)) { + if (list_empty(&clp->cl_state_owners)) { + set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state); + spin_unlock(&clp->cl_lock); + nfs_expire_all_delegations(clp); + goto out; + } spin_unlock(&clp->cl_lock); /* Queue an asynchronous RENEW. */ nfs4_proc_async_renew(clp); -- cgit v1.2.2 From b4454fe1a7cb76a248d0641c9d68a44a1b8d9a1f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:25 +0100 Subject: NFSv4: Remove requirement for machine creds for the "renew" operation In RFC3530, the RENEW operation is allowed to use either the same principal, RPC security flavour and (if RPCSEC_GSS), the same mechanism and service that was used for SETCLIENTID_CONFIRM OR Any principal, RPC security flavour and service combination that currently has an OPEN file on the server. Choose the latter since that doesn't require us to keep credentials for the same principal for the entire duration of the mount. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 5 +++-- fs/nfs/nfs4proc.c | 38 ++++++++++++++++++-------------------- fs/nfs/nfs4renewd.c | 7 +++++-- fs/nfs/nfs4state.c | 16 +++++++++++++++- 4 files changed, 41 insertions(+), 25 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 75fe646c28..4a9c53e079 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -213,8 +213,8 @@ extern ssize_t nfs4_listxattr(struct dentry *, char *, size_t); extern int nfs4_map_errors(int err); extern int nfs4_proc_setclientid(struct nfs4_client *, u32, unsigned short); extern int nfs4_proc_setclientid_confirm(struct nfs4_client *); -extern int nfs4_proc_async_renew(struct nfs4_client *); -extern int nfs4_proc_renew(struct nfs4_client *); +extern int nfs4_proc_async_renew(struct nfs4_client *, struct rpc_cred *); +extern int nfs4_proc_renew(struct nfs4_client *, struct rpc_cred *); extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state); extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *); @@ -240,6 +240,7 @@ extern struct nfs4_client *nfs4_get_client(struct in_addr *); extern void nfs4_put_client(struct nfs4_client *clp); extern int nfs4_init_client(struct nfs4_client *clp); extern struct nfs4_client *nfs4_find_client(struct in_addr *); +struct rpc_cred *nfs4_get_renew_cred(struct nfs4_client *clp); extern u32 nfs4_alloc_lockowner_id(struct nfs4_client *); extern struct nfs4_state_owner * nfs4_get_state_owner(struct nfs_server *, struct rpc_cred *); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index cc33a1c32c..e384019312 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -850,9 +850,14 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred int open_flags = flags & (FMODE_READ|FMODE_WRITE); int err; + err = -ENOMEM; + if (!(sp = nfs4_get_state_owner(server, cred))) { + dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__); + return err; + } err = nfs4_recover_expired_lease(server); if (err != 0) - return err; + goto out_put_state_owner; /* Protect against reboot recovery - NOTE ORDER! */ down_read(&clp->cl_sem); /* Protect against delegation recall */ @@ -862,10 +867,6 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred if (delegation == NULL || (delegation->type & open_flags) != open_flags) goto out_err; err = -ENOMEM; - if (!(sp = nfs4_get_state_owner(server, cred))) { - dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__); - goto out_err; - } state = nfs4_get_open_state(inode, sp); if (state == NULL) goto out_err; @@ -877,13 +878,13 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred spin_unlock(&inode->i_lock); goto out_ok; } else if (state->state != 0) - goto out_err; + goto out_put_open_state; lock_kernel(); err = _nfs4_do_access(inode, cred, open_flags); unlock_kernel(); if (err != 0) - goto out_err; + goto out_put_open_state; set_bit(NFS_DELEGATED_STATE, &state->flags); update_open_stateid(state, &delegation->stateid, open_flags); out_ok: @@ -891,17 +892,16 @@ out_ok: up_read(&nfsi->rwsem); up_read(&clp->cl_sem); *res = state; - return 0; + return 0; +out_put_open_state: + nfs4_put_open_state(state); out_err: - if (sp != NULL) { - if (state != NULL) - nfs4_put_open_state(state); - nfs4_put_state_owner(sp); - } up_read(&nfsi->rwsem); up_read(&clp->cl_sem); if (err != -EACCES) nfs_inode_return_delegation(inode); +out_put_state_owner: + nfs4_put_state_owner(sp); return err; } @@ -941,7 +941,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st } status = nfs4_recover_expired_lease(server); if (status != 0) - goto out_err; + goto err_put_state_owner; down_read(&clp->cl_sem); status = -ENOMEM; opendata = nfs4_opendata_alloc(dentry, sp, flags, sattr); @@ -2518,26 +2518,24 @@ static const struct rpc_call_ops nfs4_renew_ops = { .rpc_call_done = nfs4_renew_done, }; -int -nfs4_proc_async_renew(struct nfs4_client *clp) +int nfs4_proc_async_renew(struct nfs4_client *clp, struct rpc_cred *cred) { struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENEW], .rpc_argp = clp, - .rpc_cred = clp->cl_cred, + .rpc_cred = cred, }; return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT, &nfs4_renew_ops, (void *)jiffies); } -int -nfs4_proc_renew(struct nfs4_client *clp) +int nfs4_proc_renew(struct nfs4_client *clp, struct rpc_cred *cred) { struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENEW], .rpc_argp = clp, - .rpc_cred = clp->cl_cred, + .rpc_cred = cred, }; unsigned long now = jiffies; int status; diff --git a/fs/nfs/nfs4renewd.c b/fs/nfs/nfs4renewd.c index f62c2f7a4f..5d764d8e6d 100644 --- a/fs/nfs/nfs4renewd.c +++ b/fs/nfs/nfs4renewd.c @@ -62,6 +62,7 @@ void nfs4_renew_state(void *data) { struct nfs4_client *clp = (struct nfs4_client *)data; + struct rpc_cred *cred; long lease, timeout; unsigned long last, now; @@ -77,7 +78,8 @@ nfs4_renew_state(void *data) timeout = (2 * lease) / 3 + (long)last - (long)now; /* Are we close to a lease timeout? */ if (time_after(now, last + lease/3)) { - if (list_empty(&clp->cl_state_owners)) { + cred = nfs4_get_renew_cred(clp); + if (cred == NULL) { set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state); spin_unlock(&clp->cl_lock); nfs_expire_all_delegations(clp); @@ -85,7 +87,8 @@ nfs4_renew_state(void *data) } spin_unlock(&clp->cl_lock); /* Queue an asynchronous RENEW. */ - nfs4_proc_async_renew(clp); + nfs4_proc_async_renew(clp, cred); + put_rpccred(cred); timeout = (2 * lease) / 3; spin_lock(&clp->cl_lock); } else diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index f2148376f1..18f6ed1a0b 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -232,6 +232,20 @@ nfs4_client_grab_unused(struct nfs4_client *clp, struct rpc_cred *cred) return sp; } +struct rpc_cred *nfs4_get_renew_cred(struct nfs4_client *clp) +{ + struct nfs4_state_owner *sp; + struct rpc_cred *cred = NULL; + + list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + if (list_empty(&sp->so_states)) + continue; + cred = get_rpccred(sp->so_cred); + break; + } + return cred; +} + static struct nfs4_state_owner * nfs4_find_state_owner(struct nfs4_client *clp, struct rpc_cred *cred) { @@ -899,7 +913,7 @@ static int reclaimer(void *ptr) if (list_empty(&clp->cl_superblocks)) goto out; restart_loop: - status = nfs4_proc_renew(clp); + status = nfs4_proc_renew(clp, clp->cl_cred); switch (status) { case 0: case -NFS4ERR_CB_PATH_DOWN: -- cgit v1.2.2 From 286d7d6a0cb38d3d4316a1dfea9b0c0fc5a6455b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:26 +0100 Subject: NFSv4: Remove requirement for machine creds for the "setclientid" operation Use a cred from the nfs4_client->cl_state_owners list. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 14 ----------- fs/nfs/nfs4_fs.h | 6 ++--- fs/nfs/nfs4proc.c | 10 ++++---- fs/nfs/nfs4state.c | 71 ++++++++++++++++++++++++++++++++++-------------------- 4 files changed, 52 insertions(+), 49 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 432f41cd75..740d3cd402 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1822,23 +1822,9 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, clnt->cl_softrtry = 1; clnt->cl_chatty = 1; clp->cl_rpcclient = clnt; - clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0); - if (IS_ERR(clp->cl_cred)) { - up_write(&clp->cl_sem); - err = PTR_ERR(clp->cl_cred); - clp->cl_cred = NULL; - goto out_fail; - } memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr)); nfs_idmap_new(clp); } - if (list_empty(&clp->cl_superblocks)) { - err = nfs4_init_client(clp); - if (err != 0) { - up_write(&clp->cl_sem); - goto out_fail; - } - } list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks); clnt = rpc_clone_client(clp->cl_rpcclient); if (!IS_ERR(clnt)) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 4a9c53e079..0f5e4e7cdd 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -68,7 +68,6 @@ struct nfs4_client { atomic_t cl_count; struct rpc_clnt * cl_rpcclient; - struct rpc_cred * cl_cred; struct list_head cl_superblocks; /* List of nfs_server structs */ @@ -211,8 +210,8 @@ extern ssize_t nfs4_listxattr(struct dentry *, char *, size_t); /* nfs4proc.c */ extern int nfs4_map_errors(int err); -extern int nfs4_proc_setclientid(struct nfs4_client *, u32, unsigned short); -extern int nfs4_proc_setclientid_confirm(struct nfs4_client *); +extern int nfs4_proc_setclientid(struct nfs4_client *, u32, unsigned short, struct rpc_cred *); +extern int nfs4_proc_setclientid_confirm(struct nfs4_client *, struct rpc_cred *); extern int nfs4_proc_async_renew(struct nfs4_client *, struct rpc_cred *); extern int nfs4_proc_renew(struct nfs4_client *, struct rpc_cred *); extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state); @@ -238,7 +237,6 @@ extern void init_nfsv4_state(struct nfs_server *); extern void destroy_nfsv4_state(struct nfs_server *); extern struct nfs4_client *nfs4_get_client(struct in_addr *); extern void nfs4_put_client(struct nfs4_client *clp); -extern int nfs4_init_client(struct nfs4_client *clp); extern struct nfs4_client *nfs4_find_client(struct in_addr *); struct rpc_cred *nfs4_get_renew_cred(struct nfs4_client *clp); extern u32 nfs4_alloc_lockowner_id(struct nfs4_client *); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index e384019312..b334915499 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2846,7 +2846,7 @@ int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct return nfs4_map_errors(ret); } -int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short port) +int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short port, struct rpc_cred *cred) { nfs4_verifier sc_verifier; struct nfs4_setclientid setclientid = { @@ -2857,7 +2857,7 @@ int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short p .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID], .rpc_argp = &setclientid, .rpc_resp = clp, - .rpc_cred = clp->cl_cred, + .rpc_cred = cred, }; u32 *p; int loop = 0; @@ -2871,7 +2871,7 @@ int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short p setclientid.sc_name_len = scnprintf(setclientid.sc_name, sizeof(setclientid.sc_name), "%s/%u.%u.%u.%u %s %u", clp->cl_ipaddr, NIPQUAD(clp->cl_addr.s_addr), - clp->cl_cred->cr_ops->cr_name, + cred->cr_ops->cr_name, clp->cl_id_uniquifier); setclientid.sc_netid_len = scnprintf(setclientid.sc_netid, sizeof(setclientid.sc_netid), "tcp"); @@ -2894,14 +2894,14 @@ int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short p } int -nfs4_proc_setclientid_confirm(struct nfs4_client *clp) +nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred) { struct nfs_fsinfo fsinfo; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM], .rpc_argp = clp, .rpc_resp = &fsinfo, - .rpc_cred = clp->cl_cred, + .rpc_cred = cred, }; unsigned long now; int status; diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 18f6ed1a0b..afad0255e7 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -91,11 +91,10 @@ nfs4_alloc_client(struct in_addr *addr) if (nfs_callback_up() < 0) return NULL; - if ((clp = kmalloc(sizeof(*clp), GFP_KERNEL)) == NULL) { + if ((clp = kzalloc(sizeof(*clp), GFP_KERNEL)) == NULL) { nfs_callback_down(); return NULL; } - memset(clp, 0, sizeof(*clp)); memcpy(&clp->cl_addr, addr, sizeof(clp->cl_addr)); init_rwsem(&clp->cl_sem); INIT_LIST_HEAD(&clp->cl_delegations); @@ -108,7 +107,7 @@ nfs4_alloc_client(struct in_addr *addr) rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS4 client"); clp->cl_rpcclient = ERR_PTR(-EINVAL); clp->cl_boot_time = CURRENT_TIME; - clp->cl_state = 0; + clp->cl_state = 1 << NFS4CLNT_LEASE_EXPIRED; return clp; } @@ -125,8 +124,6 @@ nfs4_free_client(struct nfs4_client *clp) kfree(sp); } BUG_ON(!list_empty(&clp->cl_state_owners)); - if (clp->cl_cred) - put_rpccred(clp->cl_cred); nfs_idmap_delete(clp); if (!IS_ERR(clp->cl_rpcclient)) rpc_shutdown_client(clp->cl_rpcclient); @@ -196,21 +193,17 @@ nfs4_put_client(struct nfs4_client *clp) nfs4_free_client(clp); } -static int __nfs4_init_client(struct nfs4_client *clp) +static int nfs4_init_client(struct nfs4_client *clp, struct rpc_cred *cred) { - int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, nfs_callback_tcpport); + int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, + nfs_callback_tcpport, cred); if (status == 0) - status = nfs4_proc_setclientid_confirm(clp); + status = nfs4_proc_setclientid_confirm(clp, cred); if (status == 0) nfs4_schedule_state_renewal(clp); return status; } -int nfs4_init_client(struct nfs4_client *clp) -{ - return nfs4_map_errors(__nfs4_init_client(clp)); -} - u32 nfs4_alloc_lockowner_id(struct nfs4_client *clp) { @@ -246,6 +239,18 @@ struct rpc_cred *nfs4_get_renew_cred(struct nfs4_client *clp) return cred; } +struct rpc_cred *nfs4_get_setclientid_cred(struct nfs4_client *clp) +{ + struct nfs4_state_owner *sp; + + if (!list_empty(&clp->cl_state_owners)) { + sp = list_entry(clp->cl_state_owners.next, + struct nfs4_state_owner, so_list); + return get_rpccred(sp->so_cred); + } + return NULL; +} + static struct nfs4_state_owner * nfs4_find_state_owner(struct nfs4_client *clp, struct rpc_cred *cred) { @@ -902,6 +907,7 @@ static int reclaimer(void *ptr) struct nfs4_client *clp = ptr; struct nfs4_state_owner *sp; struct nfs4_state_recovery_ops *ops; + struct rpc_cred *cred; int status = 0; allow_signal(SIGKILL); @@ -913,20 +919,33 @@ static int reclaimer(void *ptr) if (list_empty(&clp->cl_superblocks)) goto out; restart_loop: - status = nfs4_proc_renew(clp, clp->cl_cred); - switch (status) { - case 0: - case -NFS4ERR_CB_PATH_DOWN: - goto out; - case -NFS4ERR_STALE_CLIENTID: - case -NFS4ERR_LEASE_MOVED: - ops = &nfs4_reboot_recovery_ops; - break; - default: - ops = &nfs4_network_partition_recovery_ops; - }; + ops = &nfs4_network_partition_recovery_ops; + /* Are there any open files on this volume? */ + cred = nfs4_get_renew_cred(clp); + if (cred != NULL) { + /* Yes there are: try to renew the old lease */ + status = nfs4_proc_renew(clp, cred); + switch (status) { + case 0: + case -NFS4ERR_CB_PATH_DOWN: + put_rpccred(cred); + goto out; + case -NFS4ERR_STALE_CLIENTID: + case -NFS4ERR_LEASE_MOVED: + ops = &nfs4_reboot_recovery_ops; + } + } else { + /* "reboot" to ensure we clear all state on the server */ + clp->cl_boot_time = CURRENT_TIME; + cred = nfs4_get_setclientid_cred(clp); + } + /* We're going to have to re-establish a clientid */ nfs4_state_mark_reclaim(clp); - status = __nfs4_init_client(clp); + status = -ENOENT; + if (cred != NULL) { + status = nfs4_init_client(clp, cred); + put_rpccred(cred); + } if (status) goto out_error; /* Mark all delegations for reclaim */ -- cgit v1.2.2 From ce1a8e6796150233f5098100f70217521dc7c08f Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Wed, 30 Nov 2005 18:08:17 -0500 Subject: NFS: use generic_write_checks() to sanity check direct writes Replace ad hoc write parameter sanity checking in nfs_file_direct_write() with a call to generic_write_checks(). This should make the proper checks modulo the O_LARGEFILE flag, and should catch NFSv2-specific limitations by virtue of i_sb->s_maxbytes. Test plan: Posix compliance testing with both NFSv2 and NFSv3. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 43 +++++++++++++++++++------------------------ 1 file changed, 19 insertions(+), 24 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index ae2be07441..f69d95aa78 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -660,10 +660,10 @@ nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t .iov_len = count, }; - dprintk("nfs: direct read(%s/%s, %lu@%lu)\n", + dprintk("nfs: direct read(%s/%s, %lu@%Ld)\n", file->f_dentry->d_parent->d_name.name, file->f_dentry->d_name.name, - (unsigned long) count, (unsigned long) pos); + (unsigned long) count, (long long) pos); if (!is_sync_kiocb(iocb)) goto out; @@ -716,9 +716,7 @@ out: ssize_t nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t pos) { - ssize_t retval = -EINVAL; - loff_t *ppos = &iocb->ki_pos; - unsigned long limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur; + ssize_t retval; struct file *file = iocb->ki_filp; struct nfs_open_context *ctx = (struct nfs_open_context *) file->private_data; @@ -726,35 +724,32 @@ nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, struct inode *inode = mapping->host; struct iovec iov = { .iov_base = (char __user *)buf, - .iov_len = count, }; - dfprintk(VFS, "nfs: direct write(%s/%s(%ld), %lu@%lu)\n", + dfprintk(VFS, "nfs: direct write(%s/%s, %lu@%Ld)\n", file->f_dentry->d_parent->d_name.name, - file->f_dentry->d_name.name, inode->i_ino, - (unsigned long) count, (unsigned long) pos); + file->f_dentry->d_name.name, + (unsigned long) count, (long long) pos); + retval = -EINVAL; if (!is_sync_kiocb(iocb)) goto out; - if (count < 0) - goto out; - if (pos < 0) + + retval = generic_write_checks(file, &pos, &count, 0); + if (retval) goto out; - retval = -EFAULT; - if (!access_ok(VERIFY_READ, iov.iov_base, iov.iov_len)) + + retval = -EINVAL; + if ((ssize_t) count < 0) goto out; - retval = -EFBIG; - if (limit != RLIM_INFINITY) { - if (pos >= limit) { - send_sig(SIGXFSZ, current, 0); - goto out; - } - if (count > limit - (unsigned long) pos) - count = limit - (unsigned long) pos; - } retval = 0; if (!count) goto out; + iov.iov_len = count, + + retval = -EFAULT; + if (!access_ok(VERIFY_READ, iov.iov_base, iov.iov_len)) + goto out; retval = nfs_sync_mapping(mapping); if (retval) @@ -764,7 +759,7 @@ nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, if (mapping->nrpages) invalidate_inode_pages2(mapping); if (retval > 0) - *ppos = pos + retval; + iocb->ki_pos = pos + retval; out: return retval; -- cgit v1.2.2 From 6b59a75460eb9527145d7bd55068e5d32bee8a44 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Wed, 30 Nov 2005 18:08:19 -0500 Subject: NFS: Fix error recovery code in fs/nfs/inode.c:__init_nfs() Red Hat found a problem in the error recovery logic in __init_nfs. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 740d3cd402..da58207b36 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -2152,11 +2152,11 @@ out: #ifdef CONFIG_PROC_FS rpc_proc_unregister("nfs"); #endif - nfs_destroy_writepagecache(); #ifdef CONFIG_NFS_DIRECTIO -out0: nfs_destroy_directcache(); +out0: #endif + nfs_destroy_writepagecache(); out1: nfs_destroy_readpagecache(); out2: -- cgit v1.2.2 From dc20f803904dbf30f834dcc43c14701dfce32491 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Wed, 30 Nov 2005 18:08:57 -0500 Subject: NFS: get rid of useless kernel log message nfs_statfs() generates a log message when GETATTR returns an error. This is usually a useless message. Make it a dprintk. Test plan: None Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index da58207b36..cc753928e4 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -575,11 +575,10 @@ nfs_statfs(struct super_block *sb, struct kstatfs *buf) buf->f_namelen = server->namelen; out: unlock_kernel(); - return 0; out_err: - printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error); + dprintk("%s: statfs error = %d\n", __FUNCTION__, -error); buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; goto out; -- cgit v1.2.2 From 325cfed9ae901320e9234b18c21434b783dbe342 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Wed, 30 Nov 2005 18:08:55 -0500 Subject: NFS: make "inode number mismatch" message more useful To help NFS users and server developers, make the "inode number mismatch" message display more useful information. Test-plan: None. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index cc753928e4..4e6558df54 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1409,14 +1409,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) if ((fattr->valid & NFS_ATTR_FATTR) == 0) return 0; - if (nfsi->fileid != fattr->fileid) { - printk(KERN_ERR "%s: inode number mismatch\n" - "expected (%s/0x%Lx), got (%s/0x%Lx)\n", - __FUNCTION__, - inode->i_sb->s_id, (long long)nfsi->fileid, - inode->i_sb->s_id, (long long)fattr->fileid); - goto out_err; - } + if (nfsi->fileid != fattr->fileid) + goto out_fileid; /* * Make sure the inode's type hasn't changed. @@ -1538,6 +1532,13 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) */ nfs_invalidate_inode(inode); return -ESTALE; + + out_fileid: + printk(KERN_ERR "NFS: server %s error: fileid changed\n" + "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n", + NFS_SERVER(inode)->hostname, inode->i_sb->s_id, + (long long)nfsi->fileid, (long long)fattr->fileid); + goto out_err; } /* -- cgit v1.2.2 From 40859d7ee64ed6bfad8a4e93f9bb5c1074afadff Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Wed, 30 Nov 2005 18:09:02 -0500 Subject: NFS: support large reads and writes on the wire Most NFS server implementations allow up to 64KB reads and writes on the wire. The Solaris NFS server allows up to a megabyte, for instance. Now the Linux NFS client supports transfer sizes up to 1MB, too. This will help reduce protocol and context switch overhead on read/write intensive NFS workloads, and support larger atomic read and write operations on servers that support them. Test-plan: Connectathon and iozone on mount point with wsize=rsize>32768 over TCP. Tests with NFS over UDP to verify the maximum RPC payload size cap. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 5 +++-- fs/nfs/inode.c | 25 ++++++++++--------------- fs/nfs/nfsroot.c | 4 ++-- fs/nfs/read.c | 6 +++--- fs/nfs/write.c | 29 ++++++++++++++++++++++------- 5 files changed, 40 insertions(+), 29 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index f69d95aa78..fd7ac5e841 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -154,6 +154,7 @@ static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, unsigned int struct list_head *list; struct nfs_direct_req *dreq; unsigned int reads = 0; + unsigned int rpages = (rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; dreq = kmem_cache_alloc(nfs_direct_cachep, SLAB_KERNEL); if (!dreq) @@ -167,7 +168,7 @@ static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, unsigned int list = &dreq->list; for(;;) { - struct nfs_read_data *data = nfs_readdata_alloc(); + struct nfs_read_data *data = nfs_readdata_alloc(rpages); if (unlikely(!data)) { while (!list_empty(list)) { @@ -431,7 +432,7 @@ static ssize_t nfs_direct_write_seg(struct inode *inode, struct nfs_writeverf first_verf; struct nfs_write_data *wdata; - wdata = nfs_writedata_alloc(); + wdata = nfs_writedata_alloc(NFS_SERVER(inode)->wpages); if (!wdata) return -ENOMEM; diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 4e6558df54..acde2c5725 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -221,10 +221,10 @@ nfs_calc_block_size(u64 tsize) static inline unsigned long nfs_block_size(unsigned long bsize, unsigned char *nrbitsp) { - if (bsize < 1024) - bsize = NFS_DEF_FILE_IO_BUFFER_SIZE; - else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE) - bsize = NFS_MAX_FILE_IO_BUFFER_SIZE; + if (bsize < NFS_MIN_FILE_IO_SIZE) + bsize = NFS_DEF_FILE_IO_SIZE; + else if (bsize >= NFS_MAX_FILE_IO_SIZE) + bsize = NFS_MAX_FILE_IO_SIZE; return nfs_block_bits(bsize, nrbitsp); } @@ -307,20 +307,15 @@ nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor) max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL); if (server->rsize > max_rpc_payload) server->rsize = max_rpc_payload; - if (server->wsize > max_rpc_payload) - server->wsize = max_rpc_payload; - + if (server->rsize > NFS_MAX_FILE_IO_SIZE) + server->rsize = NFS_MAX_FILE_IO_SIZE; server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; - if (server->rpages > NFS_READ_MAXIOV) { - server->rpages = NFS_READ_MAXIOV; - server->rsize = server->rpages << PAGE_CACHE_SHIFT; - } + if (server->wsize > max_rpc_payload) + server->wsize = max_rpc_payload; + if (server->wsize > NFS_MAX_FILE_IO_SIZE) + server->wsize = NFS_MAX_FILE_IO_SIZE; server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; - if (server->wpages > NFS_WRITE_MAXIOV) { - server->wpages = NFS_WRITE_MAXIOV; - server->wsize = server->wpages << PAGE_CACHE_SHIFT; - } if (sb->s_blocksize == 0) sb->s_blocksize = nfs_block_bits(server->wsize, diff --git a/fs/nfs/nfsroot.c b/fs/nfs/nfsroot.c index 1b272a135a..985cc53b8d 100644 --- a/fs/nfs/nfsroot.c +++ b/fs/nfs/nfsroot.c @@ -296,8 +296,8 @@ static int __init root_nfs_name(char *name) nfs_port = -1; nfs_data.version = NFS_MOUNT_VERSION; nfs_data.flags = NFS_MOUNT_NONLM; /* No lockd in nfs root yet */ - nfs_data.rsize = NFS_DEF_FILE_IO_BUFFER_SIZE; - nfs_data.wsize = NFS_DEF_FILE_IO_BUFFER_SIZE; + nfs_data.rsize = NFS_DEF_FILE_IO_SIZE; + nfs_data.wsize = NFS_DEF_FILE_IO_SIZE; nfs_data.acregmin = 3; nfs_data.acregmax = 60; nfs_data.acdirmin = 30; diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 21486242c3..05eb43fadf 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -83,7 +83,7 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode, int result; struct nfs_read_data *rdata; - rdata = nfs_readdata_alloc(); + rdata = nfs_readdata_alloc(1); if (!rdata) return -ENOMEM; @@ -283,7 +283,7 @@ static int nfs_pagein_multi(struct list_head *head, struct inode *inode) nbytes = req->wb_bytes; for(;;) { - data = nfs_readdata_alloc(); + data = nfs_readdata_alloc(1); if (!data) goto out_bad; INIT_LIST_HEAD(&data->pages); @@ -339,7 +339,7 @@ static int nfs_pagein_one(struct list_head *head, struct inode *inode) if (NFS_SERVER(inode)->rsize < PAGE_CACHE_SIZE) return nfs_pagein_multi(head, inode); - data = nfs_readdata_alloc(); + data = nfs_readdata_alloc(NFS_SERVER(inode)->rpages); if (!data) goto out_bad; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 80bc4ea1b8..1ce0c200df 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -89,18 +89,33 @@ static mempool_t *nfs_commit_mempool; static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion); -static inline struct nfs_write_data *nfs_commit_alloc(void) +static inline struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount) { struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS); + if (p) { memset(p, 0, sizeof(*p)); INIT_LIST_HEAD(&p->pages); + if (pagecount < NFS_PAGEVEC_SIZE) + p->pagevec = &p->page_array[0]; + else { + size_t size = ++pagecount * sizeof(struct page *); + p->pagevec = kmalloc(size, GFP_NOFS); + if (p->pagevec) { + memset(p->pagevec, 0, size); + } else { + mempool_free(p, nfs_commit_mempool); + p = NULL; + } + } } return p; } static inline void nfs_commit_free(struct nfs_write_data *p) { + if (p && (p->pagevec != &p->page_array[0])) + kfree(p->pagevec); mempool_free(p, nfs_commit_mempool); } @@ -167,7 +182,7 @@ static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode, int result, written = 0; struct nfs_write_data *wdata; - wdata = nfs_writedata_alloc(); + wdata = nfs_writedata_alloc(1); if (!wdata) return -ENOMEM; @@ -909,7 +924,7 @@ static int nfs_flush_multi(struct list_head *head, struct inode *inode, int how) nbytes = req->wb_bytes; for (;;) { - data = nfs_writedata_alloc(); + data = nfs_writedata_alloc(1); if (!data) goto out_bad; list_add(&data->pages, &list); @@ -973,7 +988,7 @@ static int nfs_flush_one(struct list_head *head, struct inode *inode, int how) if (NFS_SERVER(inode)->wsize < PAGE_CACHE_SIZE) return nfs_flush_multi(head, inode, how); - data = nfs_writedata_alloc(); + data = nfs_writedata_alloc(NFS_SERVER(inode)->wpages); if (!data) goto out_bad; @@ -1241,12 +1256,12 @@ static void nfs_commit_rpcsetup(struct list_head *head, * Commit dirty pages */ static int -nfs_commit_list(struct list_head *head, int how) +nfs_commit_list(struct inode *inode, struct list_head *head, int how) { struct nfs_write_data *data; struct nfs_page *req; - data = nfs_commit_alloc(); + data = nfs_commit_alloc(NFS_SERVER(inode)->wpages); if (!data) goto out_bad; @@ -1351,7 +1366,7 @@ int nfs_commit_inode(struct inode *inode, int how) res = nfs_scan_commit(inode, &head, 0, 0); spin_unlock(&nfsi->req_lock); if (res) { - error = nfs_commit_list(&head, how); + error = nfs_commit_list(inode, &head, how); if (error < 0) return error; } -- cgit v1.2.2 From 24174119c73983d5217da8f56a12c79a9b57e056 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:33 +0100 Subject: NFSv4: Ensure that we return the delegation on the target of a rename too. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index c0d1a21457..e9255198f7 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1550,8 +1550,10 @@ go_ahead: } nfs_inode_return_delegation(old_inode); - if (new_inode) + if (new_inode != NULL) { + nfs_inode_return_delegation(new_inode); d_delete(new_dentry); + } nfs_begin_data_update(old_dir); nfs_begin_data_update(new_dir); -- cgit v1.2.2 From 70b9ecbdb9c5fdc731f8780bffd45d9519020c4a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:34 +0100 Subject: NFS: Make stat() return updated mtimes after a write() The SuS states that a call to write() will cause mtime to be updated on the file. In order to satisfy that requirement, we need to flush out any cached writes in nfs_getattr(). Speed things up slightly by not committing the writes. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 2 ++ fs/nfs/write.c | 23 ++++++++++++----------- 2 files changed, 14 insertions(+), 11 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index acde2c5725..2c7f8aac1d 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -952,6 +952,8 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME; int err; + /* Flush out writes to the server in order to update c/mtime */ + nfs_sync_inode(inode, 0, 0, FLUSH_WAIT|FLUSH_NOCOMMIT); if (__IS_FLG(inode, MS_NOATIME)) need_atime = 0; else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode)) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 1ce0c200df..9449b68355 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1377,22 +1377,23 @@ int nfs_commit_inode(struct inode *inode, int how) int nfs_sync_inode(struct inode *inode, unsigned long idx_start, unsigned int npages, int how) { - int error, - wait; + int nocommit = how & FLUSH_NOCOMMIT; + int wait = how & FLUSH_WAIT; + int error; - wait = how & FLUSH_WAIT; - how &= ~FLUSH_WAIT; + how &= ~(FLUSH_WAIT|FLUSH_NOCOMMIT); do { - error = 0; - if (wait) + if (wait) { error = nfs_wait_on_requests(inode, idx_start, npages); - if (error == 0) - error = nfs_flush_inode(inode, idx_start, npages, how); -#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) - if (error == 0) + if (error != 0) + continue; + } + error = nfs_flush_inode(inode, idx_start, npages, how); + if (error != 0) + continue; + if (!nocommit) error = nfs_commit_inode(inode, how); -#endif } while (error > 0); return error; } -- cgit v1.2.2 From 566dd6064e89b15ff2dec666a421bebf0f98f26c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:35 +0100 Subject: NFS: Make directIO aware of compound pages... ...and avoid calling set_page_dirty on them Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index fd7ac5e841..10ae377e68 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -122,9 +122,10 @@ nfs_free_user_pages(struct page **pages, int npages, int do_dirty) { int i; for (i = 0; i < npages; i++) { - if (do_dirty) - set_page_dirty_lock(pages[i]); - page_cache_release(pages[i]); + struct page *page = pages[i]; + if (do_dirty && !PageCompound(page)) + set_page_dirty_lock(page); + page_cache_release(page); } kfree(pages); } -- cgit v1.2.2 From beb2a5ec386e5ce6891ebd1c06b913da04354b40 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:37 +0100 Subject: NFSv4: Ensure change attribute returned by GETATTR callback conforms to spec According to RFC3530 we're supposed to cache the change attribute at the time the client receives a write delegation. If the inode is clean, a CB_GETATTR callback by the server to the client is supposed to return the cached change attribute. If, OTOH, the inode is dirty, the client should bump the cached change attribute by 1. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 4 +++- fs/nfs/delegation.c | 1 + fs/nfs/delegation.h | 1 + 3 files changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 65f1e19e4d..462cfceb50 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -35,7 +35,9 @@ unsigned nfs4_callback_getattr(struct cb_getattrargs *args, struct cb_getattrres if (delegation == NULL || (delegation->type & FMODE_WRITE) == 0) goto out_iput; res->size = i_size_read(inode); - res->change_attr = NFS_CHANGE_ATTR(inode); + res->change_attr = delegation->change_attr; + if (nfsi->npages != 0) + res->change_attr++; res->ctime = inode->i_ctime; res->mtime = inode->i_mtime; res->bitmap[0] = (FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE) & diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 75dfb1c717..d2ee09b38c 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -131,6 +131,7 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct sizeof(delegation->stateid.data)); delegation->type = res->delegation_type; delegation->maxsize = res->maxsize; + delegation->change_attr = nfsi->change_attr; delegation->cred = get_rpccred(cred); delegation->inode = inode; diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index fbc50ec271..7a0b2bfce7 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -21,6 +21,7 @@ struct nfs_delegation { #define NFS_DELEGATION_NEED_RECLAIM 1 long flags; loff_t maxsize; + __u64 change_attr; }; int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); -- cgit v1.2.2 From fa178f29c0f8a0dce748181a5351f4a92fd4f455 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:38 +0100 Subject: NFSv4: Ensure DELEGRETURN returns attributes Upon return of a write delegation, the server will almost always bump the change attribute. Ensure that we pick up that change so that we don't invalidate our data cache unnecessarily. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 2 -- fs/nfs/nfs4proc.c | 17 +++++++++++++---- fs/nfs/nfs4xdr.c | 33 ++++++++++++++++++++++----------- 3 files changed, 35 insertions(+), 17 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index d2ee09b38c..66cc720e39 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -159,8 +159,6 @@ static int nfs_do_return_delegation(struct inode *inode, struct nfs_delegation * { int res = 0; - __nfs_revalidate_inode(NFS_SERVER(inode), inode); - res = nfs4_proc_delegreturn(inode, delegation->cred, &delegation->stateid); nfs_free_delegation(delegation); return res; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index b334915499..984ca3454d 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2920,11 +2920,12 @@ nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred) struct nfs4_delegreturndata { struct nfs4_delegreturnargs args; + struct nfs4_delegreturnres res; struct nfs_fh fh; nfs4_stateid stateid; struct rpc_cred *cred; unsigned long timestamp; - const struct nfs_server *server; + struct nfs_fattr fattr; int rpc_status; }; @@ -2934,8 +2935,10 @@ static void nfs4_delegreturn_prepare(struct rpc_task *task, void *calldata) struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN], .rpc_argp = &data->args, + .rpc_resp = &data->res, .rpc_cred = data->cred, }; + nfs_fattr_init(data->res.fattr); rpc_call_setup(task, &msg, 0); } @@ -2944,7 +2947,7 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) struct nfs4_delegreturndata *data = calldata; data->rpc_status = task->tk_status; if (data->rpc_status == 0) - renew_lease(data->server, data->timestamp); + renew_lease(data->res.server, data->timestamp); } static void nfs4_delegreturn_release(void *calldata) @@ -2964,6 +2967,7 @@ const static struct rpc_call_ops nfs4_delegreturn_ops = { static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid) { struct nfs4_delegreturndata *data; + struct nfs_server *server = NFS_SERVER(inode); struct rpc_task *task; int status; @@ -2972,11 +2976,13 @@ static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, co return -ENOMEM; data->args.fhandle = &data->fh; data->args.stateid = &data->stateid; + data->args.bitmask = server->attr_bitmask; nfs_copy_fh(&data->fh, NFS_FH(inode)); memcpy(&data->stateid, stateid, sizeof(data->stateid)); + data->res.fattr = &data->fattr; + data->res.server = server; data->cred = get_rpccred(cred); data->timestamp = jiffies; - data->server = NFS_SERVER(inode); data->rpc_status = 0; task = rpc_run_task(NFS_CLIENT(inode), RPC_TASK_ASYNC, &nfs4_delegreturn_ops, data); @@ -2985,8 +2991,11 @@ static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, co return PTR_ERR(task); } status = nfs4_wait_for_completion_rpc_task(task); - if (status == 0) + if (status == 0) { status = data->rpc_status; + if (status == 0) + nfs_post_op_update_inode(inode, &data->fattr); + } rpc_release_task(task); return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 5d6bda43df..12be1d6821 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -392,9 +392,11 @@ static int nfs_stat_to_errno(int); decode_getattr_maxsz) #define NFS4_enc_delegreturn_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - encode_delegreturn_maxsz) + encode_delegreturn_maxsz + \ + encode_getattr_maxsz) #define NFS4_dec_delegreturn_sz (compound_decode_hdr_maxsz + \ - decode_delegreturn_maxsz) + decode_delegreturn_maxsz + \ + decode_getattr_maxsz) #define NFS4_enc_getacl_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_getattr_maxsz) @@ -1983,14 +1985,20 @@ static int nfs4_xdr_enc_delegreturn(struct rpc_rqst *req, uint32_t *p, const str { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); - if ((status = encode_putfh(&xdr, args->fhandle)) == 0) - status = encode_delegreturn(&xdr, args->stateid); + status = encode_putfh(&xdr, args->fhandle); + if (status != 0) + goto out; + status = encode_delegreturn(&xdr, args->stateid); + if (status != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); +out: return status; } @@ -4184,7 +4192,7 @@ static int nfs4_xdr_dec_setclientid_confirm(struct rpc_rqst *req, uint32_t *p, s /* * DELEGRETURN request */ -static int nfs4_xdr_dec_delegreturn(struct rpc_rqst *rqstp, uint32_t *p, void *dummy) +static int nfs4_xdr_dec_delegreturn(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_delegreturnres *res) { struct xdr_stream xdr; struct compound_hdr hdr; @@ -4192,11 +4200,14 @@ static int nfs4_xdr_dec_delegreturn(struct rpc_rqst *rqstp, uint32_t *p, void *d xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p); status = decode_compound_hdr(&xdr, &hdr); - if (status == 0) { - status = decode_putfh(&xdr); - if (status == 0) - status = decode_delegreturn(&xdr); - } + if (status != 0) + goto out; + status = decode_putfh(&xdr); + if (status != 0) + goto out; + status = decode_delegreturn(&xdr); + decode_getfattr(&xdr, res->fattr, res->server); +out: return status; } -- cgit v1.2.2 From a895b4a198dd06f8353328867e4f6cfd28b63081 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:40 +0100 Subject: NFS: Clean up weak cache consistency code ...and ensure that nfs_update_inode() respects wcc Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 60 ++++++++++++++++++++++++++++++++++++++-------------------- 1 file changed, 40 insertions(+), 20 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 2c7f8aac1d..7270b1d73d 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1248,6 +1248,33 @@ void nfs_end_data_update(struct inode *inode) atomic_dec(&nfsi->data_updates); } +static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr) +{ + struct nfs_inode *nfsi = NFS_I(inode); + + if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0 + && nfsi->change_attr == fattr->pre_change_attr) { + nfsi->change_attr = fattr->change_attr; + nfsi->cache_change_attribute = jiffies; + } + + /* If we have atomic WCC data, we may update some attributes */ + if ((fattr->valid & NFS_ATTR_WCC) != 0) { + if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) { + memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime)); + nfsi->cache_change_attribute = jiffies; + } + if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) { + memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime)); + nfsi->cache_change_attribute = jiffies; + } + if (inode->i_size == fattr->pre_size && nfsi->npages == 0) { + inode->i_size = fattr->size; + nfsi->cache_change_attribute = jiffies; + } + } +} + /** * nfs_check_inode_attributes - verify consistency of the inode attribute cache * @inode - pointer to inode @@ -1264,22 +1291,20 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat int data_unstable; + if ((fattr->valid & NFS_ATTR_FATTR) == 0) + return 0; + /* Are we in the process of updating data on the server? */ data_unstable = nfs_caches_unstable(inode); - if (fattr->valid & NFS_ATTR_FATTR_V4) { - if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0 - && nfsi->change_attr == fattr->pre_change_attr) - nfsi->change_attr = fattr->change_attr; - if (nfsi->change_attr != fattr->change_attr) { - nfsi->cache_validity |= NFS_INO_INVALID_ATTR; - if (!data_unstable) - nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE; - } - } + /* Do atomic weak cache consistency updates */ + nfs_wcc_update_inode(inode, fattr); - if ((fattr->valid & NFS_ATTR_FATTR) == 0) { - return 0; + if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 && + nfsi->change_attr != fattr->change_attr) { + nfsi->cache_validity |= NFS_INO_INVALID_ATTR; + if (!data_unstable) + nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE; } /* Has the inode gone and changed behind our back? */ @@ -1291,14 +1316,6 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat cur_size = i_size_read(inode); new_isize = nfs_size_to_loff_t(fattr->size); - /* If we have atomic WCC data, we may update some attributes */ - if ((fattr->valid & NFS_ATTR_WCC) != 0) { - if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) - memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime)); - if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) - memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime)); - } - /* Verify a few of the more important attributes */ if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) { nfsi->cache_validity |= NFS_INO_INVALID_ATTR; @@ -1426,6 +1443,9 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) if (data_stable) nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); + /* Do atomic weak cache consistency updates */ + nfs_wcc_update_inode(inode, fattr); + /* Check if our cached file size is stale */ new_isize = nfs_size_to_loff_t(fattr->size); cur_isize = i_size_read(inode); -- cgit v1.2.2 From a72b44222d222749d54b3e370d825094352e389f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:41 +0100 Subject: NFSv4: Allow user to set the port used by the NFSv4 callback channel Signed-off-by: Trond Myklebust --- fs/nfs/Makefile | 1 + fs/nfs/callback.c | 3 ++- fs/nfs/callback.h | 1 + fs/nfs/inode.c | 37 ++++++++++++++++++++++++++-- fs/nfs/sysctl.c | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 113 insertions(+), 3 deletions(-) create mode 100644 fs/nfs/sysctl.c (limited to 'fs/nfs') diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile index 8b3bb715d1..ec61fd56a1 100644 --- a/fs/nfs/Makefile +++ b/fs/nfs/Makefile @@ -13,4 +13,5 @@ nfs-$(CONFIG_NFS_V4) += nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \ delegation.o idmap.o \ callback.o callback_xdr.o callback_proc.o nfs-$(CONFIG_NFS_DIRECTIO) += direct.o +nfs-$(CONFIG_SYSCTL) += sysctl.o nfs-objs := $(nfs-y) diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c index 30cae36028..fcd97406a7 100644 --- a/fs/nfs/callback.c +++ b/fs/nfs/callback.c @@ -34,6 +34,7 @@ static struct nfs_callback_data nfs_callback_info; static DECLARE_MUTEX(nfs_callback_sema); static struct svc_program nfs4_callback_program; +unsigned int nfs_callback_set_tcpport; unsigned short nfs_callback_tcpport; /* @@ -98,7 +99,7 @@ int nfs_callback_up(void) if (!serv) goto out_err; /* FIXME: We don't want to register this socket with the portmapper */ - ret = svc_makesock(serv, IPPROTO_TCP, 0); + ret = svc_makesock(serv, IPPROTO_TCP, nfs_callback_set_tcpport); if (ret < 0) goto out_destroy; if (!list_empty(&serv->sv_permsocks)) { diff --git a/fs/nfs/callback.h b/fs/nfs/callback.h index a0db2d4f94..b252e7fe53 100644 --- a/fs/nfs/callback.h +++ b/fs/nfs/callback.h @@ -65,6 +65,7 @@ extern unsigned nfs4_callback_recall(struct cb_recallargs *args, void *dummy); extern int nfs_callback_up(void); extern int nfs_callback_down(void); +extern unsigned int nfs_callback_set_tcpport; extern unsigned short nfs_callback_tcpport; #endif /* __LINUX_FS_NFS_CALLBACK_H */ diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 7270b1d73d..648cb1aef3 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -40,6 +40,7 @@ #include #include "nfs4_fs.h" +#include "callback.h" #include "delegation.h" #define NFSDBG_FACILITY NFSDBG_VFS @@ -2036,6 +2037,21 @@ static struct file_system_type nfs4_fs_type = { .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA, }; +static const int nfs_set_port_min = 0; +static const int nfs_set_port_max = 65535; +static int param_set_port(const char *val, struct kernel_param *kp) +{ + char *endp; + int num = simple_strtol(val, &endp, 0); + if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max) + return -EINVAL; + *((int *)kp->arg) = num; + return 0; +} + +module_param_call(callback_tcpport, param_set_port, param_get_int, + &nfs_callback_set_tcpport, 0644); + #define nfs4_init_once(nfsi) \ do { \ INIT_LIST_HEAD(&(nfsi)->open_states); \ @@ -2043,8 +2059,25 @@ static struct file_system_type nfs4_fs_type = { nfsi->delegation_state = 0; \ init_rwsem(&nfsi->rwsem); \ } while(0) -#define register_nfs4fs() register_filesystem(&nfs4_fs_type) -#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type) + +static inline int register_nfs4fs(void) +{ + int ret; + + ret = nfs_register_sysctl(); + if (ret != 0) + return ret; + ret = register_filesystem(&nfs4_fs_type); + if (ret != 0) + nfs_unregister_sysctl(); + return ret; +} + +static inline void unregister_nfs4fs(void) +{ + unregister_filesystem(&nfs4_fs_type); + nfs_unregister_sysctl(); +} #else #define nfs4_init_once(nfsi) \ do { } while (0) diff --git a/fs/nfs/sysctl.c b/fs/nfs/sysctl.c new file mode 100644 index 0000000000..fdc64b59a4 --- /dev/null +++ b/fs/nfs/sysctl.c @@ -0,0 +1,74 @@ +/* + * linux/fs/nfs/sysctl.c + * + * Sysctl interface to NFS parameters + */ +#include +#include +#include +#include +#include +#include +#include +#include + +#include "callback.h" + +static const int nfs_set_port_min = 0; +static const int nfs_set_port_max = 65535; +static struct ctl_table_header *nfs_callback_sysctl_table; +/* + * Something that isn't CTL_ANY, CTL_NONE or a value that may clash. + * Use the same values as fs/lockd/svc.c + */ +#define CTL_UNNUMBERED -2 + +static ctl_table nfs_cb_sysctls[] = { +#ifdef CONFIG_NFS_V4 + { + .ctl_name = CTL_UNNUMBERED, + .procname = "nfs_callback_tcpport", + .data = &nfs_callback_set_tcpport, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .extra1 = (int *)&nfs_set_port_min, + .extra2 = (int *)&nfs_set_port_max, + }, +#endif + { .ctl_name = 0 } +}; + +static ctl_table nfs_cb_sysctl_dir[] = { + { + .ctl_name = CTL_UNNUMBERED, + .procname = "nfs", + .mode = 0555, + .child = nfs_cb_sysctls, + }, + { .ctl_name = 0 } +}; + +static ctl_table nfs_cb_sysctl_root[] = { + { + .ctl_name = CTL_FS, + .procname = "fs", + .mode = 0555, + .child = nfs_cb_sysctl_dir, + }, + { .ctl_name = 0 } +}; + +int nfs_register_sysctl(void) +{ + nfs_callback_sysctl_table = register_sysctl_table(nfs_cb_sysctl_root, 0); + if (nfs_callback_sysctl_table == NULL) + return -ENOMEM; + return 0; +} + +void nfs_unregister_sysctl(void) +{ + unregister_sysctl_table(nfs_callback_sysctl_table); + nfs_callback_sysctl_table = NULL; +} -- cgit v1.2.2 From 03c21733938aad0758f5f88e1cc7ede69fc3c910 Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Tue, 3 Jan 2006 09:55:48 +0100 Subject: NFSv3: try get_root user-supplied security_flavor Thanks to Ed Keizer for bug and root cause. He says: "... we could only mount the top-level Solaris share. We could not mount deeper into the tree. Investigation showed that Solaris allows UNIX authenticated FSINFO only on the top level of the share. This is a problem because we share/export our home directories one level higher than we mount them. I.e. we share the partition and not the individual home directories. This prevented access to home directories." We still may need to try auth_sys for the case where the client doesn't have appropriate credentials. Signed-off-by: J. Bruce Fields Signed-off-by: Trond Myklebust --- fs/nfs/nfs3proc.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index c172a75846..ed67567f05 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -68,26 +68,38 @@ nfs3_async_handle_jukebox(struct rpc_task *task) return 1; } -/* - * Bare-bones access to getattr: this is for nfs_read_super. - */ static int -nfs3_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, - struct nfs_fsinfo *info) +do_proc_get_root(struct rpc_clnt *client, struct nfs_fh *fhandle, + struct nfs_fsinfo *info) { int status; dprintk("%s: call fsinfo\n", __FUNCTION__); nfs_fattr_init(info->fattr); - status = rpc_call(server->client_sys, NFS3PROC_FSINFO, fhandle, info, 0); + status = rpc_call(client, NFS3PROC_FSINFO, fhandle, info, 0); dprintk("%s: reply fsinfo: %d\n", __FUNCTION__, status); if (!(info->fattr->valid & NFS_ATTR_FATTR)) { - status = rpc_call(server->client_sys, NFS3PROC_GETATTR, fhandle, info->fattr, 0); + status = rpc_call(client, NFS3PROC_GETATTR, fhandle, info->fattr, 0); dprintk("%s: reply getattr: %d\n", __FUNCTION__, status); } return status; } +/* + * Bare-bones access to getattr: this is for nfs_read_super. + */ +static int +nfs3_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, + struct nfs_fsinfo *info) +{ + int status; + + status = do_proc_get_root(server->client, fhandle, info); + if (status && server->client_sys != server->client) + status = do_proc_get_root(server->client_sys, fhandle, info); + return status; +} + /* * One function for each procedure in the NFS protocol. */ -- cgit v1.2.2 From f518e35aec984036903c1003e867f833747a9d79 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Tue, 3 Jan 2006 09:55:52 +0100 Subject: SUNRPC: get rid of cl_chatty Clean up: Every ULP that uses the in-kernel RPC client, except the NLM client, sets cl_chatty. There's no reason why NLM shouldn't set it, so just get rid of cl_chatty and always be verbose. Test-plan: Compile with CONFIG_NFS enabled. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 2 -- fs/nfs/mount_clnt.c | 1 - 2 files changed, 3 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 648cb1aef3..4625479a6b 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -413,7 +413,6 @@ nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data) clnt->cl_intr = 1; clnt->cl_softrtry = 1; - clnt->cl_chatty = 1; return clnt; @@ -1838,7 +1837,6 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, } clnt->cl_intr = 1; clnt->cl_softrtry = 1; - clnt->cl_chatty = 1; clp->cl_rpcclient = clnt; memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr)); nfs_idmap_new(clp); diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c index 0e82617f2d..db99b8f678 100644 --- a/fs/nfs/mount_clnt.c +++ b/fs/nfs/mount_clnt.c @@ -82,7 +82,6 @@ mnt_create(char *hostname, struct sockaddr_in *srvaddr, int version, RPC_AUTH_UNIX); if (!IS_ERR(clnt)) { clnt->cl_softrtry = 1; - clnt->cl_chatty = 1; clnt->cl_oneshot = 1; clnt->cl_intr = 1; } -- cgit v1.2.2 From cf3fff54a46e1f8fa4cc1deb783172a392077eb0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:53 +0100 Subject: NFS: Send valid mode bits to the server inode->i_mode contains a lot more than just the mode bits. Make sure that we mask away this extra stuff in SETATTR calls to the server. Signed-off-by: Trond Myklebust --- fs/nfs/nfs3xdr.c | 2 +- fs/nfs/nfs4xdr.c | 2 +- fs/nfs/proc.c | 3 +++ 3 files changed, 5 insertions(+), 2 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index 0498bd3660..b6c0b5012b 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -182,7 +182,7 @@ xdr_encode_sattr(u32 *p, struct iattr *attr) { if (attr->ia_valid & ATTR_MODE) { *p++ = xdr_one; - *p++ = htonl(attr->ia_mode); + *p++ = htonl(attr->ia_mode & S_IALLUGO); } else { *p++ = xdr_zero; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 12be1d6821..4bbf5ef577 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -566,7 +566,7 @@ static int encode_attrs(struct xdr_stream *xdr, const struct iattr *iap, const s } if (iap->ia_valid & ATTR_MODE) { bmval1 |= FATTR4_WORD1_MODE; - WRITE32(iap->ia_mode); + WRITE32(iap->ia_mode & S_IALLUGO); } if (iap->ia_valid & ATTR_UID) { bmval1 |= FATTR4_WORD1_OWNER; diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index 6145e82b45..f5150d71c0 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -111,6 +111,9 @@ nfs_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr, }; int status; + /* Mask out the non-modebit related stuff from attr->ia_mode */ + sattr->ia_mode &= S_IALLUGO; + dprintk("NFS call setattr\n"); nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(inode), NFSPROC_SETATTR, &arg, fattr, 0); -- cgit v1.2.2 From eadb8c147154bff982f02accf31b847a1f142ace Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:54 +0100 Subject: NFS: get rid of some needless code obfuscation in xdr_encode_sattr(). Signed-off-by: Trond Myklebust --- fs/nfs/nfs2xdr.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c index 59049e864c..7fc0560c89 100644 --- a/fs/nfs/nfs2xdr.c +++ b/fs/nfs/nfs2xdr.c @@ -146,23 +146,23 @@ xdr_decode_fattr(u32 *p, struct nfs_fattr *fattr) return p; } -#define SATTR(p, attr, flag, field) \ - *p++ = (attr->ia_valid & flag) ? htonl(attr->field) : ~(u32) 0 static inline u32 * xdr_encode_sattr(u32 *p, struct iattr *attr) { - SATTR(p, attr, ATTR_MODE, ia_mode); - SATTR(p, attr, ATTR_UID, ia_uid); - SATTR(p, attr, ATTR_GID, ia_gid); - SATTR(p, attr, ATTR_SIZE, ia_size); + const u32 not_set = __constant_htonl(0xFFFFFFFF); + + *p++ = (attr->ia_valid & ATTR_MODE) ? htonl(attr->ia_mode) : not_set; + *p++ = (attr->ia_valid & ATTR_UID) ? htonl(attr->ia_uid) : not_set; + *p++ = (attr->ia_valid & ATTR_GID) ? htonl(attr->ia_gid) : not_set; + *p++ = (attr->ia_valid & ATTR_SIZE) ? htonl(attr->ia_size) : not_set; if (attr->ia_valid & ATTR_ATIME_SET) { p = xdr_encode_time(p, &attr->ia_atime); } else if (attr->ia_valid & ATTR_ATIME) { p = xdr_encode_current_server_time(p, &attr->ia_atime); } else { - *p++ = ~(u32) 0; - *p++ = ~(u32) 0; + *p++ = not_set; + *p++ = not_set; } if (attr->ia_valid & ATTR_MTIME_SET) { @@ -170,12 +170,11 @@ xdr_encode_sattr(u32 *p, struct iattr *attr) } else if (attr->ia_valid & ATTR_MTIME) { p = xdr_encode_current_server_time(p, &attr->ia_mtime); } else { - *p++ = ~(u32) 0; - *p++ = ~(u32) 0; + *p++ = not_set; + *p++ = not_set; } return p; } -#undef SATTR /* * NFS encode functions -- cgit v1.2.2 From 58df095b732529ade8f4051b41d7c29731afecd6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:57 +0100 Subject: NFSv4: Allow entries in the idmap cache to expire If someone changes the uid/gid mapping in userland, then we do eventually want those changes to be propagated to the kernel. Currently the kernel assumes that it may cache entries forever. Add an expiration time + garbage collector for idmap entries. Signed-off-by: Trond Myklebust --- fs/nfs/idmap.c | 9 +++++++++ fs/nfs/inode.c | 14 ++++++++++++++ fs/nfs/sysctl.c | 10 ++++++++++ 3 files changed, 33 insertions(+) (limited to 'fs/nfs') diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c index ffb8df91dc..821edd3033 100644 --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -54,7 +54,11 @@ #define IDMAP_HASH_SZ 128 +/* Default cache timeout is 10 minutes */ +unsigned int nfs_idmap_cache_timeout = 600 * HZ; + struct idmap_hashent { + unsigned long ih_expires; __u32 ih_id; int ih_namelen; char ih_name[IDMAP_NAMESZ]; @@ -149,6 +153,8 @@ idmap_lookup_name(struct idmap_hashtable *h, const char *name, size_t len) if (he->ih_namelen != len || memcmp(he->ih_name, name, len) != 0) return NULL; + if (time_after(jiffies, he->ih_expires)) + return NULL; return he; } @@ -164,6 +170,8 @@ idmap_lookup_id(struct idmap_hashtable *h, __u32 id) struct idmap_hashent *he = idmap_id_hash(h, id); if (he->ih_id != id || he->ih_namelen == 0) return NULL; + if (time_after(jiffies, he->ih_expires)) + return NULL; return he; } @@ -192,6 +200,7 @@ idmap_update_entry(struct idmap_hashent *he, const char *name, memcpy(he->ih_name, name, namelen); he->ih_name[namelen] = '\0'; he->ih_namelen = namelen; + he->ih_expires = jiffies + nfs_idmap_cache_timeout; } /* diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 4625479a6b..e7bd0d9260 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -2050,6 +2050,20 @@ static int param_set_port(const char *val, struct kernel_param *kp) module_param_call(callback_tcpport, param_set_port, param_get_int, &nfs_callback_set_tcpport, 0644); +static int param_set_idmap_timeout(const char *val, struct kernel_param *kp) +{ + char *endp; + int num = simple_strtol(val, &endp, 0); + int jif = num * HZ; + if (endp == val || *endp || num < 0 || jif < num) + return -EINVAL; + *((int *)kp->arg) = jif; + return 0; +} + +module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int, + &nfs_idmap_cache_timeout, 0644); + #define nfs4_init_once(nfsi) \ do { \ INIT_LIST_HEAD(&(nfsi)->open_states); \ diff --git a/fs/nfs/sysctl.c b/fs/nfs/sysctl.c index fdc64b59a4..4c486eb867 100644 --- a/fs/nfs/sysctl.c +++ b/fs/nfs/sysctl.c @@ -11,6 +11,7 @@ #include #include #include +#include #include "callback.h" @@ -35,6 +36,15 @@ static ctl_table nfs_cb_sysctls[] = { .extra1 = (int *)&nfs_set_port_min, .extra2 = (int *)&nfs_set_port_max, }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "idmap_cache_timeout", + .data = &nfs_idmap_cache_timeout, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec_jiffies, + .strategy = &sysctl_jiffies, + }, #endif { .ctl_name = 0 } }; -- cgit v1.2.2 From 26c78e156b1d1b2387ec33b5f2fb62d6e0a186a3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jan 2006 09:55:58 +0100 Subject: NFSv4: Fix an Oops in nfs_do_expire_all_delegations If the loop errors, we need to exit. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 66cc720e39..c6f07c1c71 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -236,7 +236,6 @@ int nfs_do_expire_all_delegations(void *ptr) struct nfs4_client *clp = ptr; struct nfs_delegation *delegation; struct inode *inode; - int err = 0; allow_signal(SIGKILL); restart: @@ -250,10 +249,9 @@ restart: if (inode == NULL) continue; spin_unlock(&clp->cl_lock); - err = nfs_inode_return_delegation(inode); + nfs_inode_return_delegation(inode); iput(inode); - if (!err) - goto restart; + goto restart; } out: spin_unlock(&clp->cl_lock); -- cgit v1.2.2 From 28fd129827b00e12829d48a5290f46277600619b Mon Sep 17 00:00:00 2001 From: OGAWA Hirofumi Date: Sun, 8 Jan 2006 01:02:14 -0800 Subject: [PATCH] Fix and add EXPORT_SYMBOL(filemap_write_and_wait) This patch add EXPORT_SYMBOL(filemap_write_and_wait) and use it. See mm/filemap.c: And changes the filemap_write_and_wait() and filemap_write_and_wait_range(). Current filemap_write_and_wait() doesn't wait if filemap_fdatawrite() returns error. However, even if filemap_fdatawrite() returned an error, it may have submitted the partially data pages to the device. (e.g. in the case of -ENOSPC) Andrew Morton writes, If filemap_fdatawrite() returns an error, this might be due to some I/O problem: dead disk, unplugged cable, etc. Given the generally crappy quality of the kernel's handling of such exceptions, there's a good chance that the filemap_fdatawait() will get stuck in D state forever. So, this patch doesn't wait if filemap_fdatawrite() returns the -EIO. Trond, could you please review the nfs part? Especially I'm not sure, nfs must use the "filemap_fdatawrite(inode->i_mapping) == 0", or not. Acked-by: Trond Myklebust Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/nfs/inode.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index e7bd0d9260..3e4ba9cb7f 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -644,10 +644,7 @@ int nfs_sync_mapping(struct address_space *mapping) if (mapping->nrpages == 0) return 0; unmap_mapping_range(mapping, 0, 0, 0); - ret = filemap_fdatawrite(mapping); - if (ret != 0) - goto out; - ret = filemap_fdatawait(mapping); + ret = filemap_write_and_wait(mapping); if (ret != 0) goto out; ret = nfs_wb_all(mapping->host); @@ -864,8 +861,7 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) nfs_begin_data_update(inode); /* Write all dirty data if we're changing file permissions or size */ if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) { - if (filemap_fdatawrite(inode->i_mapping) == 0) - filemap_fdatawait(inode->i_mapping); + filemap_write_and_wait(inode->i_mapping); nfs_wb_all(inode); } /* -- cgit v1.2.2 From 21b6bf143d05d77c350d9c6764ae090a877b66ea Mon Sep 17 00:00:00 2001 From: Jorn Dreyer Date: Sun, 8 Jan 2006 01:03:30 -0800 Subject: [PATCH] nfsroot: do not silently stop parsing on an unknown option It would be helpful if the kernel did not silently stop parsing nfs options, but instead warned about any he does not recognize. The attached patch adds one printk to do just that. It took me a couple of hours to find my configuration mistake. Cc: "David S. Miller" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/nfs/nfsroot.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/nfs') diff --git a/fs/nfs/nfsroot.c b/fs/nfs/nfsroot.c index 985cc53b8d..e897e00c2c 100644 --- a/fs/nfs/nfsroot.c +++ b/fs/nfs/nfsroot.c @@ -275,7 +275,9 @@ static int __init root_nfs_parse(char *name, char *buf) case Opt_noacl: nfs_data.flags |= NFS_MOUNT_NOACL; break; - default : + default: + printk(KERN_WARNING "Root-NFS: unknown " + "option: %s\n", p); return 0; } } -- cgit v1.2.2 From 1b1dcc1b57a49136f118a0f16367256ff9994a69 Mon Sep 17 00:00:00 2001 From: Jes Sorensen Date: Mon, 9 Jan 2006 15:59:24 -0800 Subject: [PATCH] mutex subsystem, semaphore to mutex: VFS, ->i_sem This patch converts the inode semaphore to a mutex. I have tested it on XFS and compiled as much as one can consider on an ia64. Anyway your luck with it might be different. Modified-by: Ingo Molnar (finished the conversion) Signed-off-by: Jes Sorensen Signed-off-by: Ingo Molnar --- fs/nfs/dir.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index e9255198f7..a1554bead6 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -194,7 +194,7 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page *page) spin_unlock(&inode->i_lock); /* Ensure consistent page alignment of the data. * Note: assumes we have exclusive access to this mapping either - * through inode->i_sem or some other mechanism. + * through inode->i_mutex or some other mechanism. */ if (page->index == 0) invalidate_inode_pages2_range(inode->i_mapping, PAGE_CACHE_SIZE, -1); @@ -573,7 +573,7 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin) { - down(&filp->f_dentry->d_inode->i_sem); + mutex_lock(&filp->f_dentry->d_inode->i_mutex); switch (origin) { case 1: offset += filp->f_pos; @@ -589,7 +589,7 @@ loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin) ((struct nfs_open_context *)filp->private_data)->dir_cookie = 0; } out: - up(&filp->f_dentry->d_inode->i_sem); + mutex_unlock(&filp->f_dentry->d_inode->i_mutex); return offset; } @@ -1001,7 +1001,7 @@ static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd) openflags &= ~(O_CREAT|O_TRUNC); /* - * Note: we're not holding inode->i_sem and so may be racing with + * Note: we're not holding inode->i_mutex and so may be racing with * operations that change the directory. We therefore save the * change attribute *before* we do the RPC call. */ @@ -1051,7 +1051,7 @@ static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc) return dentry; if (!desc->plus || !(entry->fattr->valid & NFS_ATTR_FATTR)) return NULL; - /* Note: caller is already holding the dir->i_sem! */ + /* Note: caller is already holding the dir->i_mutex! */ dentry = d_alloc(parent, &name); if (dentry == NULL) return NULL; -- cgit v1.2.2 From fc33a7bb9c6dd8f6e4a014976200f8fdabb3a45c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 9 Jan 2006 20:52:17 -0800 Subject: [PATCH] per-mountpoint noatime/nodiratime Turn noatime and nodiratime into per-mount instead of per-sb flags. After all the preparations this is a rather trivial patch. The mount code needs to treat the two options as per-mount instead of per-superblock, and touch_atime needs to be changed to check the new MNT_ flags in addition to the MS_ flags that are kept for filesystems that are always noatime/nodiratime but not user settable anymore. Besides that core code only nfs needed an update because it's leaving atime updates to the server and thus sets the S_NOATIME flag on every inode, but needs to know whether it's a real noatime mount for an getattr optimization. While we're at it I've killed the IS_NOATIME/IS_NODIRATIME macros that were only used by touch_atime. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/nfs/inode.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'fs/nfs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 3e4ba9cb7f..a77ee95b7e 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -950,11 +950,20 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) /* Flush out writes to the server in order to update c/mtime */ nfs_sync_inode(inode, 0, 0, FLUSH_WAIT|FLUSH_NOCOMMIT); - if (__IS_FLG(inode, MS_NOATIME)) - need_atime = 0; - else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode)) + + /* + * We may force a getattr if the user cares about atime. + * + * Note that we only have to check the vfsmount flags here: + * - NFS always sets S_NOATIME by so checking it would give a + * bogus result + * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is + * no point in checking those. + */ + if ((mnt->mnt_flags & MNT_NOATIME) || + ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))) need_atime = 0; - /* We may force a getattr if the user cares about atime */ + if (need_atime) err = __nfs_revalidate_inode(NFS_SERVER(inode), inode); else -- cgit v1.2.2