From 88be9f990fe70f0f177ef44a16a477599e91f825 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 5 Jun 2007 10:42:27 -0400 Subject: NFS: Replace vfsmount and dentry in nfs_open_context with struct path Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 0543439a97af..07eea8f64ecf 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -31,6 +31,7 @@ #include #include +#include #include #include #include @@ -70,8 +71,7 @@ struct nfs_access_entry { struct nfs4_state; struct nfs_open_context { atomic_t count; - struct vfsmount *vfsmnt; - struct dentry *dentry; + struct path path; struct rpc_cred *cred; struct nfs4_state *state; fl_owner_t lockowner; -- cgit v1.2.3 From aa53ed541a1fec78a78d02afc8b042d040cc080d Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 5 Jun 2007 14:49:03 -0400 Subject: NFS4: on a O_EXCL OPEN make sure SETATTR sets the fields holding the verifier The Linux NFS4 client simply skips over the bitmask in an O_EXCL open call and so it doesn't bother to reset any fields that may be holding the verifier. This patch has us save the first two words of the bitmask (which is all the current client has #defines for). The client then later checks this bitmask and turns on the appropriate flags in the sattr->ia_verify field for the following SETATTR call. This patch only currently checks to see if the server used the atime and mtime slots for the verifier (which is what the Linux server uses for this). I'm not sure of what other fields the server could reasonably use, but adding checks for others should be trivial. Signed-off-by: Jeff Layton Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 20 ++++++++++++++++++++ fs/nfs/nfs4xdr.c | 9 +++++++-- include/linux/nfs4.h | 1 + include/linux/nfs_xdr.h | 1 + 4 files changed, 29 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3cc75445a68d..fee2d14b158b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -942,6 +942,22 @@ static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, st return res; } +/* + * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-* + * fields corresponding to attributes that were used to store the verifier. + * Make sure we clobber those fields in the later setattr call + */ +static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr) +{ + if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) && + !(sattr->ia_valid & ATTR_ATIME_SET)) + sattr->ia_valid |= ATTR_ATIME; + + if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) && + !(sattr->ia_valid & ATTR_MTIME_SET)) + sattr->ia_valid |= ATTR_MTIME; +} + /* * Returns a referenced nfs4_state */ @@ -973,6 +989,9 @@ static int _nfs4_do_open(struct inode *dir, struct path *path, int flags, struct if (status != 0) goto err_opendata_free; + if (opendata->o_arg.open_flags & O_EXCL) + nfs4_exclusive_attrset(opendata, sattr); + status = -ENOMEM; state = nfs4_opendata_to_nfs4_state(opendata); if (state == NULL) @@ -1784,6 +1803,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, status = nfs4_do_setattr(state->inode, &fattr, sattr, state); if (status == 0) nfs_setattr_update_inode(state->inode, sattr); + nfs_post_op_update_inode(state->inode, &fattr); } if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0) status = nfs4_intent_set_file(nd, &path, state); diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 8003c91ccb9a..1fcca516e6ee 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -3269,7 +3269,7 @@ static int decode_delegation(struct xdr_stream *xdr, struct nfs_openres *res) static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res) { __be32 *p; - uint32_t bmlen; + uint32_t savewords, bmlen, i; int status; status = decode_op_hdr(xdr, OP_OPEN); @@ -3287,7 +3287,12 @@ static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res) goto xdr_error; READ_BUF(bmlen << 2); - p += bmlen; + savewords = min_t(uint32_t, bmlen, NFS4_BITMAP_SIZE); + for (i = 0; i < savewords; ++i) + READ32(res->attrset[i]); + for (; i < NFS4_BITMAP_SIZE; i++) + res->attrset[i] = 0; + return decode_delegation(xdr, res); xdr_error: dprintk("%s: Bitmap too large! Length = %u\n", __FUNCTION__, bmlen); diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index 7e7f33a38fc0..8726491de154 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -15,6 +15,7 @@ #include +#define NFS4_BITMAP_SIZE 2 #define NFS4_VERIFIER_SIZE 8 #define NFS4_STATEID_SIZE 16 #define NFS4_FHSIZE 128 diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 10c26ed0db71..f7100df3a690 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -144,6 +144,7 @@ struct nfs_openres { nfs4_stateid delegation; __u32 do_recall; __u64 maxsize; + __u32 attrset[NFS4_BITMAP_SIZE]; }; /* -- cgit v1.2.3 From c03b40246123b2ced79e2620d1d2c089bb12369a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 13:26:38 -0400 Subject: NFS: Convert struct nfs_page to use krefs Signed-off-by: Trond Myklebust --- fs/nfs/pagelist.c | 14 ++++++++------ fs/nfs/write.c | 6 +++--- include/linux/nfs_page.h | 10 +++++----- 3 files changed, 16 insertions(+), 14 deletions(-) (limited to 'include') diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index f8a4ba533930..257a7f8b2362 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -85,9 +85,8 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode, req->wb_offset = offset; req->wb_pgbase = offset; req->wb_bytes = count; - atomic_set(&req->wb_count, 1); req->wb_context = get_nfs_open_context(ctx); - + kref_init(&req->wb_kref); return req; } @@ -160,11 +159,9 @@ void nfs_clear_request(struct nfs_page *req) * * Note: Should never be called with the spinlock held! */ -void -nfs_release_request(struct nfs_page *req) +static void nfs_free_request(struct kref *kref) { - if (!atomic_dec_and_test(&req->wb_count)) - return; + struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref); /* Release struct file or cached credential */ nfs_clear_request(req); @@ -172,6 +169,11 @@ nfs_release_request(struct nfs_page *req) nfs_page_free(req); } +void nfs_release_request(struct nfs_page *req) +{ + kref_put(&req->wb_kref, nfs_free_request); +} + static int nfs_wait_bit_interruptible(void *word) { int ret = 0; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 9e7c21da864f..e9404328ac02 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -117,7 +117,7 @@ static struct nfs_page *nfs_page_find_request_locked(struct page *page) if (PagePrivate(page)) { req = (struct nfs_page *)page_private(page); if (req != NULL) - atomic_inc(&req->wb_count); + kref_get(&req->wb_kref); } return req; } @@ -398,7 +398,7 @@ static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req) if (PageDirty(req->wb_page)) set_bit(PG_NEED_FLUSH, &req->wb_flags); nfsi->npages++; - atomic_inc(&req->wb_count); + kref_get(&req->wb_kref); return 0; } @@ -531,7 +531,7 @@ static int nfs_wait_on_requests_locked(struct inode *inode, pgoff_t idx_start, u next = req->wb_index + 1; BUG_ON(!NFS_WBACK_BUSY(req)); - atomic_inc(&req->wb_count); + kref_get(&req->wb_kref); spin_unlock(&nfsi->req_lock); error = nfs_wait_on_request(req); nfs_release_request(req); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index bd193af80162..c780e7e39f99 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -16,7 +16,7 @@ #include #include -#include +#include /* * Valid flags for the radix tree @@ -42,7 +42,7 @@ struct nfs_page { unsigned int wb_offset, /* Offset & ~PAGE_CACHE_MASK */ wb_pgbase, /* Start of page data */ wb_bytes; /* Length of request */ - atomic_t wb_count; /* reference count */ + struct kref wb_kref; /* reference count */ unsigned long wb_flags; struct nfs_writeverf wb_verf; /* Commit cookie */ }; @@ -89,7 +89,7 @@ extern void nfs_clear_page_writeback(struct nfs_page *req); /* - * Lock the page of an asynchronous request without incrementing the wb_count + * Lock the page of an asynchronous request without getting a new reference */ static inline int nfs_lock_request_dontget(struct nfs_page *req) @@ -98,14 +98,14 @@ nfs_lock_request_dontget(struct nfs_page *req) } /* - * Lock the page of an asynchronous request + * Lock the page of an asynchronous request and take a reference */ static inline int nfs_lock_request(struct nfs_page *req) { if (test_and_set_bit(PG_BUSY, &req->wb_flags)) return 0; - atomic_inc(&req->wb_count); + kref_get(&req->wb_kref); return 1; } -- cgit v1.2.3 From 9fd367f0f376ccfb2592eed9be0eece70429894f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 15:10:24 -0400 Subject: NFS cleanup: Rename NFS_PAGE_TAG_WRITEBACK to NFS_PAGE_TAG_LOCKED Signed-off-by: Trond Myklebust --- fs/nfs/pagelist.c | 14 +++++++------- fs/nfs/write.c | 16 ++++++++-------- include/linux/nfs_page.h | 5 ++--- 3 files changed, 17 insertions(+), 18 deletions(-) (limited to 'include') diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 257a7f8b2362..23e9dea20902 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -108,29 +108,29 @@ void nfs_unlock_request(struct nfs_page *req) } /** - * nfs_set_page_writeback_locked - Lock a request for writeback + * nfs_set_page_tag_locked - Tag a request as locked * @req: */ -int nfs_set_page_writeback_locked(struct nfs_page *req) +static int nfs_set_page_tag_locked(struct nfs_page *req) { struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode); if (!nfs_lock_request(req)) return 0; - radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK); + radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED); return 1; } /** - * nfs_clear_page_writeback - Unlock request and wake up sleepers + * nfs_clear_page_tag_locked - Clear request tag and wake up sleepers */ -void nfs_clear_page_writeback(struct nfs_page *req) +void nfs_clear_page_tag_locked(struct nfs_page *req) { struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode); if (req->wb_page != NULL) { spin_lock(&nfsi->req_lock); - radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK); + radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED); spin_unlock(&nfsi->req_lock); } nfs_unlock_request(req); @@ -421,7 +421,7 @@ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, idx_start = req->wb_index + 1; if (req->wb_list_head != head) continue; - if (nfs_set_page_writeback_locked(req)) { + if (nfs_set_page_tag_locked(req)) { nfs_list_remove_request(req); nfs_list_add_request(req, dst); res++; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index e9404328ac02..754066cc9146 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -289,7 +289,7 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, BUG(); } radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, - NFS_PAGE_TAG_WRITEBACK); + NFS_PAGE_TAG_LOCKED); ret = test_bit(PG_NEED_FLUSH, &req->wb_flags); spin_unlock(req_lock); nfs_pageio_add_request(pgio, req); @@ -524,7 +524,7 @@ static int nfs_wait_on_requests_locked(struct inode *inode, pgoff_t idx_start, u idx_end = idx_start + npages - 1; next = idx_start; - while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_WRITEBACK)) { + while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_LOCKED)) { if (req->wb_index > idx_end) break; @@ -759,7 +759,7 @@ static void nfs_writepage_release(struct nfs_page *req) nfs_inode_remove_request(req); } else nfs_end_page_writeback(req->wb_page); - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); } static inline int flush_task_priority(int how) @@ -888,7 +888,7 @@ out_bad: } nfs_redirty_request(req); nfs_end_page_writeback(req->wb_page); - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); return -ENOMEM; } @@ -931,7 +931,7 @@ static int nfs_flush_one(struct inode *inode, struct list_head *head, unsigned i nfs_list_remove_request(req); nfs_redirty_request(req); nfs_end_page_writeback(req->wb_page); - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); } return -ENOMEM; } @@ -1049,7 +1049,7 @@ remove_request: nfs_end_page_writeback(page); nfs_inode_remove_request(req); next: - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); } } @@ -1212,7 +1212,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how) nfs_list_remove_request(req); nfs_mark_request_commit(req); dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); } return -ENOMEM; } @@ -1265,7 +1265,7 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata) dprintk(" mismatch\n"); nfs_redirty_request(req); next: - nfs_clear_page_writeback(req); + nfs_clear_page_tag_locked(req); } } diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index c780e7e39f99..042434c39b7e 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -21,7 +21,7 @@ /* * Valid flags for the radix tree */ -#define NFS_PAGE_TAG_WRITEBACK 0 +#define NFS_PAGE_TAG_LOCKED 0 /* * Valid flags for a dirty buffer @@ -84,8 +84,7 @@ extern void nfs_pageio_complete(struct nfs_pageio_descriptor *desc); extern void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *, pgoff_t); extern int nfs_wait_on_request(struct nfs_page *); extern void nfs_unlock_request(struct nfs_page *req); -extern int nfs_set_page_writeback_locked(struct nfs_page *req); -extern void nfs_clear_page_writeback(struct nfs_page *req); +extern void nfs_clear_page_tag_locked(struct nfs_page *req); /* -- cgit v1.2.3 From 5c36968343fcd013a3f7ae93f246c2e75596780b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 15:27:42 -0400 Subject: NFS cleanup: speed up nfs_scan_commit using radix tree tags Add a tag for requests that are waiting for a COMMIT Signed-off-by: Trond Myklebust --- fs/nfs/pagelist.c | 14 +++++++------- fs/nfs/write.c | 6 +++++- include/linux/nfs_page.h | 5 +++-- 3 files changed, 15 insertions(+), 10 deletions(-) (limited to 'include') diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 23e9dea20902..ad90cbe76703 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -381,10 +381,10 @@ void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index) /** * nfs_scan_list - Scan a list for matching requests * @nfsi: NFS inode - * @head: One of the NFS inode request lists * @dst: Destination list * @idx_start: lower bound of page->index to scan * @npages: idx_start + npages sets the upper bound to scan. + * @tag: tag to scan for * * Moves elements from one of the inode request lists. * If the number of requests is set to 0, the entire address_space @@ -392,9 +392,9 @@ void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index) * The requests are *not* checked to ensure that they form a contiguous set. * You must be holding the inode's req_lock when calling this function */ -int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, +int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *dst, pgoff_t idx_start, - unsigned int npages) + unsigned int npages, int tag) { struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES]; struct nfs_page *req; @@ -409,9 +409,9 @@ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, idx_end = idx_start + npages - 1; for (;;) { - found = radix_tree_gang_lookup(&nfsi->nfs_page_tree, + found = radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&pgvec[0], idx_start, - NFS_SCAN_MAXENTRIES); + NFS_SCAN_MAXENTRIES, tag); if (found <= 0) break; for (i = 0; i < found; i++) { @@ -419,10 +419,10 @@ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, if (req->wb_index > idx_end) goto out; idx_start = req->wb_index + 1; - if (req->wb_list_head != head) - continue; if (nfs_set_page_tag_locked(req)) { nfs_list_remove_request(req); + radix_tree_tag_clear(&nfsi->nfs_page_tree, + req->wb_index, tag); nfs_list_add_request(req, dst); res++; } diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 754066cc9146..0f779ca12ec3 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -462,6 +462,9 @@ nfs_mark_request_commit(struct nfs_page *req) nfs_list_add_request(req, &nfsi->commit); nfsi->ncommit++; set_bit(PG_NEED_COMMIT, &(req)->wb_flags); + radix_tree_tag_set(&nfsi->nfs_page_tree, + req->wb_index, + NFS_PAGE_TAG_COMMIT); spin_unlock(&nfsi->req_lock); inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); __mark_inode_dirty(inode, I_DIRTY_DATASYNC); @@ -575,7 +578,8 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, u int res = 0; if (nfsi->ncommit != 0) { - res = nfs_scan_list(nfsi, &nfsi->commit, dst, idx_start, npages); + res = nfs_scan_list(nfsi, dst, idx_start, npages, + NFS_PAGE_TAG_COMMIT); nfsi->ncommit -= res; if ((nfsi->ncommit == 0) != list_empty(&nfsi->commit)) printk(KERN_ERR "NFS: desynchronized value of nfs_i.ncommit.\n"); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 042434c39b7e..481a42105d69 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -22,6 +22,7 @@ * Valid flags for the radix tree */ #define NFS_PAGE_TAG_LOCKED 0 +#define NFS_PAGE_TAG_COMMIT 1 /* * Valid flags for a dirty buffer @@ -71,8 +72,8 @@ extern void nfs_clear_request(struct nfs_page *req); extern void nfs_release_request(struct nfs_page *req); -extern int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, struct list_head *dst, - pgoff_t idx_start, unsigned int npages); +extern int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *dst, + pgoff_t idx_start, unsigned int npages, int tag); extern void nfs_pageio_init(struct nfs_pageio_descriptor *desc, struct inode *inode, int (*doio)(struct inode *, struct list_head *, unsigned int, size_t, int), -- cgit v1.2.3 From 2aefa104313996d1a9582476cee53d1296c834bf Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 15:40:59 -0400 Subject: NFS: Remove the redundant 'dirty' and 'commit' lists from nfs_inode Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 3 --- fs/nfs/write.c | 3 --- include/linux/nfs_fs.h | 5 +---- include/linux/nfs_page.h | 5 +---- 4 files changed, 2 insertions(+), 14 deletions(-) (limited to 'include') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 7bcb3dfa6179..e7d2bba900b5 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1149,14 +1149,11 @@ static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flag inode_init_once(&nfsi->vfs_inode); spin_lock_init(&nfsi->req_lock); - INIT_LIST_HEAD(&nfsi->dirty); - INIT_LIST_HEAD(&nfsi->commit); INIT_LIST_HEAD(&nfsi->open_files); INIT_LIST_HEAD(&nfsi->access_cache_entry_lru); INIT_LIST_HEAD(&nfsi->access_cache_inode_lru); INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC); atomic_set(&nfsi->data_updates, 0); - nfsi->ndirty = 0; nfsi->ncommit = 0; nfsi->npages = 0; nfs4_init_once(nfsi); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 0f779ca12ec3..9ef9ec746bfb 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -459,7 +459,6 @@ nfs_mark_request_commit(struct nfs_page *req) struct nfs_inode *nfsi = NFS_I(inode); spin_lock(&nfsi->req_lock); - nfs_list_add_request(req, &nfsi->commit); nfsi->ncommit++; set_bit(PG_NEED_COMMIT, &(req)->wb_flags); radix_tree_tag_set(&nfsi->nfs_page_tree, @@ -581,8 +580,6 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, u res = nfs_scan_list(nfsi, dst, idx_start, npages, NFS_PAGE_TAG_COMMIT); nfsi->ncommit -= res; - if ((nfsi->ncommit == 0) != list_empty(&nfsi->commit)) - printk(KERN_ERR "NFS: desynchronized value of nfs_i.ncommit.\n"); } return res; } diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 07eea8f64ecf..a94205476736 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -156,12 +156,9 @@ struct nfs_inode { * This is the list of dirty unwritten pages. */ spinlock_t req_lock; - struct list_head dirty; - struct list_head commit; struct radix_tree_root nfs_page_tree; - unsigned int ndirty, - ncommit, + unsigned int ncommit, npages; /* Open contexts for shared mmap writes */ diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 481a42105d69..78e60798d10e 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -34,8 +34,7 @@ struct nfs_inode; struct nfs_page { - struct list_head wb_list, /* Defines state of page: */ - *wb_list_head; /* read/write/commit */ + struct list_head wb_list; /* Defines state of page: */ struct page *wb_page; /* page to read in/write out */ struct nfs_open_context *wb_context; /* File state context info */ atomic_t wb_complete; /* i/os we're waiting for */ @@ -118,7 +117,6 @@ static inline void nfs_list_add_request(struct nfs_page *req, struct list_head *head) { list_add_tail(&req->wb_list, head); - req->wb_list_head = head; } @@ -132,7 +130,6 @@ nfs_list_remove_request(struct nfs_page *req) if (list_empty(&req->wb_list)) return; list_del_init(&req->wb_list); - req->wb_list_head = NULL; } static inline struct nfs_page * -- cgit v1.2.3 From dce34ce298d85b81630401f4feb4bd7ac77fe9c7 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 15:47:53 -0400 Subject: NFS: Prevent integer overflow in nfs_scan_list() Also ensure that nfs_inode ncommit and npages are large enough to represent all possible values for the number of pages. Signed-off-by: Trond Myklebust --- fs/nfs/pagelist.c | 2 ++ include/linux/nfs_fs.h | 2 +- 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index ad90cbe76703..68f6bf122006 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -425,6 +425,8 @@ int nfs_scan_list(struct nfs_inode *nfsi, req->wb_index, tag); nfs_list_add_request(req, dst); res++; + if (res == INT_MAX) + goto out; } } diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index a94205476736..750708ccd708 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -158,7 +158,7 @@ struct nfs_inode { spinlock_t req_lock; struct radix_tree_root nfs_page_tree; - unsigned int ncommit, + unsigned long ncommit, npages; /* Open contexts for shared mmap writes */ -- cgit v1.2.3 From 3bec63db55463365110d00721ed60a31e4614cb6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 17 Jun 2007 16:02:44 -0400 Subject: NFS: Convert struct nfs_open_context to use a kref Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 40 +++++++++++++++++++++++----------------- include/linux/nfs_fs.h | 3 ++- 2 files changed, 25 insertions(+), 18 deletions(-) (limited to 'include') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index e7d2bba900b5..01fc8ab0c562 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -461,7 +461,6 @@ static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, str ctx = kmalloc(sizeof(*ctx), GFP_KERNEL); if (ctx != NULL) { - atomic_set(&ctx->count, 1); ctx->path.dentry = dget(dentry); ctx->path.mnt = mntget(mnt); ctx->cred = get_rpccred(cred); @@ -469,6 +468,7 @@ static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, str ctx->lockowner = current->files; ctx->error = 0; ctx->dir_cookie = 0; + kref_init(&ctx->kref); } return ctx; } @@ -476,27 +476,33 @@ static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, str struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx) { if (ctx != NULL) - atomic_inc(&ctx->count); + kref_get(&ctx->kref); return ctx; } -void put_nfs_open_context(struct nfs_open_context *ctx) +static void nfs_free_open_context(struct kref *kref) { - if (atomic_dec_and_test(&ctx->count)) { - if (!list_empty(&ctx->list)) { - struct inode *inode = ctx->path.dentry->d_inode; - spin_lock(&inode->i_lock); - list_del(&ctx->list); - spin_unlock(&inode->i_lock); - } - if (ctx->state != NULL) - nfs4_close_state(&ctx->path, ctx->state, ctx->mode); - if (ctx->cred != NULL) - put_rpccred(ctx->cred); - dput(ctx->path.dentry); - mntput(ctx->path.mnt); - kfree(ctx); + struct nfs_open_context *ctx = container_of(kref, + struct nfs_open_context, kref); + + if (!list_empty(&ctx->list)) { + struct inode *inode = ctx->path.dentry->d_inode; + spin_lock(&inode->i_lock); + list_del(&ctx->list); + spin_unlock(&inode->i_lock); } + if (ctx->state != NULL) + nfs4_close_state(&ctx->path, ctx->state, ctx->mode); + if (ctx->cred != NULL) + put_rpccred(ctx->cred); + dput(ctx->path.dentry); + mntput(ctx->path.mnt); + kfree(ctx); +} + +void put_nfs_open_context(struct nfs_open_context *ctx) +{ + kref_put(&ctx->kref, nfs_free_open_context); } /* diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 750708ccd708..bf24151d63be 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -30,6 +30,7 @@ #ifdef __KERNEL__ #include +#include #include #include #include @@ -70,7 +71,7 @@ struct nfs_access_entry { struct nfs4_state; struct nfs_open_context { - atomic_t count; + struct kref kref; struct path path; struct rpc_cred *cred; struct nfs4_state *state; -- cgit v1.2.3 From 6529eba08fe7297852391a468d95322913de73fa Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 14 Jun 2007 16:40:14 -0400 Subject: SUNRPC: Move rpc_task->tk_task list into struct rpc_clnt Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 4 ++ include/linux/sunrpc/sched.h | 5 -- net/sunrpc/clnt.c | 5 ++ net/sunrpc/sched.c | 117 +++++++++++++++++++++++++++---------------- 4 files changed, 83 insertions(+), 48 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 66611423c8ee..0801ab5407ce 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -26,6 +26,8 @@ struct rpc_inode; struct rpc_clnt { atomic_t cl_count; /* Number of clones */ atomic_t cl_users; /* number of references */ + struct list_head cl_clients; /* Global list of clients */ + struct list_head cl_tasks; /* List of tasks */ struct rpc_xprt * cl_xprt; /* transport */ struct rpc_procinfo * cl_procinfo; /* procedure info */ u32 cl_prog, /* RPC program number */ @@ -122,6 +124,8 @@ struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); int rpc_shutdown_client(struct rpc_clnt *); int rpc_destroy_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); +void rpc_register_client(struct rpc_clnt *); +void rpc_unregister_client(struct rpc_clnt *); int rpcb_register(u32, u32, int, unsigned short, int *); void rpcb_getport(struct rpc_task *); diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h index 2047fb202a13..3387b008cdfc 100644 --- a/include/linux/sunrpc/sched.h +++ b/include/linux/sunrpc/sched.h @@ -110,11 +110,6 @@ struct rpc_task { if (!list_empty(head) && \ ((task=list_entry((head)->next, struct rpc_task, u.tk_wait.list)),1)) -/* .. and walking list of all tasks */ -#define alltask_for_each(task, pos, head) \ - list_for_each(pos, head) \ - if ((task=list_entry(pos, struct rpc_task, tk_task)),1) - typedef void (*rpc_action)(struct rpc_task *); struct rpc_call_ops { diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index d8fbee40a19c..6631ece14983 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -148,6 +148,7 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s if (clnt->cl_metrics == NULL) goto out_no_stats; clnt->cl_program = program; + INIT_LIST_HEAD(&clnt->cl_tasks); if (!xprt_bound(clnt->cl_xprt)) clnt->cl_autobind = 1; @@ -172,6 +173,7 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s if (clnt->cl_nodelen > UNX_MAXNODENAME) clnt->cl_nodelen = UNX_MAXNODENAME; memcpy(clnt->cl_nodename, utsname()->nodename, clnt->cl_nodelen); + rpc_register_client(clnt); return clnt; out_no_auth: @@ -283,9 +285,11 @@ rpc_clone_client(struct rpc_clnt *clnt) new->cl_autobind = 0; new->cl_oneshot = 0; new->cl_dead = 0; + INIT_LIST_HEAD(&new->cl_tasks); rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval); if (new->cl_auth) atomic_inc(&new->cl_auth->au_count); + rpc_register_client(new); return new; out_no_path: rpc_free_iostats(new->cl_metrics); @@ -357,6 +361,7 @@ rpc_destroy_client(struct rpc_clnt *clnt) if (clnt->cl_server != clnt->cl_inline_name) kfree(clnt->cl_server); out_free: + rpc_unregister_client(clnt); rpc_free_iostats(clnt->cl_metrics); clnt->cl_metrics = NULL; xprt_put(clnt->cl_xprt); diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 944d75396fb3..6309f3b52c53 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -50,9 +50,10 @@ static void rpc_release_task(struct rpc_task *task); static RPC_WAITQ(delay_queue, "delayq"); /* - * All RPC tasks are linked into this list + * All RPC clients are linked into this list */ -static LIST_HEAD(all_tasks); +static LIST_HEAD(all_clients); +static DECLARE_WAIT_QUEUE_HEAD(client_kill_wait); /* * rpciod-related stuff @@ -277,7 +278,8 @@ static void rpc_set_active(struct rpc_task *task) task->tk_pid = rpc_task_id++; #endif /* Add to global list of all tasks */ - list_add_tail(&task->tk_task, &all_tasks); + if (task->tk_client) + list_add_tail(&task->tk_task, &task->tk_client->cl_tasks); spin_unlock(&rpc_sched_lock); } @@ -818,6 +820,7 @@ void rpc_init_task(struct rpc_task *task, struct rpc_clnt *clnt, int flags, cons if (tk_ops->rpc_call_prepare != NULL) task->tk_action = rpc_prepare_task; task->tk_calldata = calldata; + INIT_LIST_HEAD(&task->tk_task); /* Initialize retry counters */ task->tk_garb_retry = 2; @@ -920,11 +923,12 @@ static void rpc_release_task(struct rpc_task *task) #endif dprintk("RPC: %5u release task\n", task->tk_pid); - /* Remove from global task list */ - spin_lock(&rpc_sched_lock); - list_del(&task->tk_task); - spin_unlock(&rpc_sched_lock); - + if (!list_empty(&task->tk_task)) { + /* Remove from client task list */ + spin_lock(&rpc_sched_lock); + list_del(&task->tk_task); + spin_unlock(&rpc_sched_lock); + } BUG_ON (RPC_IS_QUEUED(task)); /* Synchronously delete any running timer */ @@ -966,42 +970,52 @@ EXPORT_SYMBOL(rpc_run_task); * Kill all tasks for the given client. * XXX: kill their descendants as well? */ -void rpc_killall_tasks(struct rpc_clnt *clnt) +static void rpc_killall_tasks_locked(struct list_head *head) { struct rpc_task *rovr; - struct list_head *le; - dprintk("RPC: killing all tasks for client %p\n", clnt); - /* - * Spin lock all_tasks to prevent changes... - */ - spin_lock(&rpc_sched_lock); - alltask_for_each(rovr, le, &all_tasks) { + list_for_each_entry(rovr, head, tk_task) { if (! RPC_IS_ACTIVATED(rovr)) continue; - if (!clnt || rovr->tk_client == clnt) { + if (!(rovr->tk_flags & RPC_TASK_KILLED)) { rovr->tk_flags |= RPC_TASK_KILLED; rpc_exit(rovr, -EIO); rpc_wake_up_task(rovr); } } +} + +void rpc_killall_tasks(struct rpc_clnt *clnt) +{ + dprintk("RPC: killing all tasks for client %p\n", clnt); + /* + * Spin lock all_tasks to prevent changes... + */ + spin_lock(&rpc_sched_lock); + rpc_killall_tasks_locked(&clnt->cl_tasks); spin_unlock(&rpc_sched_lock); } static void rpciod_killall(void) { + struct rpc_clnt *clnt; unsigned long flags; - while (!list_empty(&all_tasks)) { + for(;;) { clear_thread_flag(TIF_SIGPENDING); - rpc_killall_tasks(NULL); + + spin_lock(&rpc_sched_lock); + list_for_each_entry(clnt, &all_clients, cl_clients) + rpc_killall_tasks_locked(&clnt->cl_tasks); + spin_unlock(&rpc_sched_lock); flush_workqueue(rpciod_workqueue); - if (!list_empty(&all_tasks)) { - dprintk("RPC: rpciod_killall: waiting for tasks " + if (!list_empty(&all_clients)) + break; + dprintk("RPC: rpciod_killall: waiting for tasks " "to exit\n"); - yield(); - } + wait_event_timeout(client_kill_wait, + list_empty(&all_clients), 1*HZ); } spin_lock_irqsave(¤t->sighand->siglock, flags); @@ -1009,6 +1023,22 @@ static void rpciod_killall(void) spin_unlock_irqrestore(¤t->sighand->siglock, flags); } +void rpc_register_client(struct rpc_clnt *clnt) +{ + spin_lock(&rpc_sched_lock); + list_add(&clnt->cl_clients, &all_clients); + spin_unlock(&rpc_sched_lock); +} + +void rpc_unregister_client(struct rpc_clnt *clnt) +{ + spin_lock(&rpc_sched_lock); + list_del(&clnt->cl_clients); + if (list_empty(&all_clients)) + wake_up(&client_kill_wait); + spin_unlock(&rpc_sched_lock); +} + /* * Start up the rpciod process if it's not already running. */ @@ -1071,32 +1101,33 @@ rpciod_down(void) #ifdef RPC_DEBUG void rpc_show_tasks(void) { - struct list_head *le; + struct rpc_clnt *clnt; struct rpc_task *t; spin_lock(&rpc_sched_lock); - if (list_empty(&all_tasks)) { - spin_unlock(&rpc_sched_lock); - return; - } + if (list_empty(&all_clients)) + goto out; printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout " "-rpcwait -action- ---ops--\n"); - alltask_for_each(t, le, &all_tasks) { - const char *rpc_waitq = "none"; - - if (RPC_IS_QUEUED(t)) - rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq); - - printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n", - t->tk_pid, - (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1), - t->tk_flags, t->tk_status, - t->tk_client, - (t->tk_client ? t->tk_client->cl_prog : 0), - t->tk_rqstp, t->tk_timeout, - rpc_waitq, - t->tk_action, t->tk_ops); + list_for_each_entry(clnt, &all_clients, cl_clients) { + list_for_each_entry(t, &clnt->cl_tasks, tk_task) { + const char *rpc_waitq = "none"; + + if (RPC_IS_QUEUED(t)) + rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq); + + printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n", + t->tk_pid, + (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1), + t->tk_flags, t->tk_status, + t->tk_client, + (t->tk_client ? t->tk_client->cl_prog : 0), + t->tk_rqstp, t->tk_timeout, + rpc_waitq, + t->tk_action, t->tk_ops); + } } +out: spin_unlock(&rpc_sched_lock); } #endif -- cgit v1.2.3 From 4bef61ff7514396419563ca54fd42ef846485b06 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 16 Jun 2007 14:17:01 -0400 Subject: SUNRPC: Add a per-rpc_clnt spinlock Use that to protect the rpc_clnt->cl_tasks list instead of using a global lock. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 1 + net/sunrpc/clnt.c | 2 ++ net/sunrpc/sched.c | 47 ++++++++++++++++++++++++++------------------- 3 files changed, 30 insertions(+), 20 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 0801ab5407ce..2f4b520a7419 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -28,6 +28,7 @@ struct rpc_clnt { atomic_t cl_users; /* number of references */ struct list_head cl_clients; /* Global list of clients */ struct list_head cl_tasks; /* List of tasks */ + spinlock_t cl_lock; /* spinlock */ struct rpc_xprt * cl_xprt; /* transport */ struct rpc_procinfo * cl_procinfo; /* procedure info */ u32 cl_prog, /* RPC program number */ diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 6631ece14983..424dfdc6862c 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -149,6 +149,7 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s goto out_no_stats; clnt->cl_program = program; INIT_LIST_HEAD(&clnt->cl_tasks); + spin_lock_init(&clnt->cl_lock); if (!xprt_bound(clnt->cl_xprt)) clnt->cl_autobind = 1; @@ -286,6 +287,7 @@ rpc_clone_client(struct rpc_clnt *clnt) new->cl_oneshot = 0; new->cl_dead = 0; INIT_LIST_HEAD(&new->cl_tasks); + spin_lock_init(&new->cl_lock); rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval); if (new->cl_auth) atomic_inc(&new->cl_auth->au_count); diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 6309f3b52c53..f56ebc5a08f7 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -270,17 +270,22 @@ static int rpc_wait_bit_interruptible(void *word) static void rpc_set_active(struct rpc_task *task) { + struct rpc_clnt *clnt; if (test_and_set_bit(RPC_TASK_ACTIVE, &task->tk_runstate) != 0) return; - spin_lock(&rpc_sched_lock); #ifdef RPC_DEBUG task->tk_magic = RPC_TASK_MAGIC_ID; + spin_lock(&rpc_sched_lock); task->tk_pid = rpc_task_id++; + spin_unlock(&rpc_sched_lock); #endif /* Add to global list of all tasks */ - if (task->tk_client) - list_add_tail(&task->tk_task, &task->tk_client->cl_tasks); - spin_unlock(&rpc_sched_lock); + clnt = task->tk_client; + if (clnt != NULL) { + spin_lock(&clnt->cl_lock); + list_add_tail(&task->tk_task, &clnt->cl_tasks); + spin_unlock(&clnt->cl_lock); + } } /* @@ -924,10 +929,11 @@ static void rpc_release_task(struct rpc_task *task) dprintk("RPC: %5u release task\n", task->tk_pid); if (!list_empty(&task->tk_task)) { + struct rpc_clnt *clnt = task->tk_client; /* Remove from client task list */ - spin_lock(&rpc_sched_lock); + spin_lock(&clnt->cl_lock); list_del(&task->tk_task); - spin_unlock(&rpc_sched_lock); + spin_unlock(&clnt->cl_lock); } BUG_ON (RPC_IS_QUEUED(task)); @@ -970,12 +976,19 @@ EXPORT_SYMBOL(rpc_run_task); * Kill all tasks for the given client. * XXX: kill their descendants as well? */ -static void rpc_killall_tasks_locked(struct list_head *head) +void rpc_killall_tasks(struct rpc_clnt *clnt) { struct rpc_task *rovr; - list_for_each_entry(rovr, head, tk_task) { + if (list_empty(&clnt->cl_tasks)) + return; + dprintk("RPC: killing all tasks for client %p\n", clnt); + /* + * Spin lock all_tasks to prevent changes... + */ + spin_lock(&clnt->cl_lock); + list_for_each_entry(rovr, &clnt->cl_tasks, tk_task) { if (! RPC_IS_ACTIVATED(rovr)) continue; if (!(rovr->tk_flags & RPC_TASK_KILLED)) { @@ -984,17 +997,7 @@ static void rpc_killall_tasks_locked(struct list_head *head) rpc_wake_up_task(rovr); } } -} - -void rpc_killall_tasks(struct rpc_clnt *clnt) -{ - dprintk("RPC: killing all tasks for client %p\n", clnt); - /* - * Spin lock all_tasks to prevent changes... - */ - spin_lock(&rpc_sched_lock); - rpc_killall_tasks_locked(&clnt->cl_tasks); - spin_unlock(&rpc_sched_lock); + spin_unlock(&clnt->cl_lock); } static void rpciod_killall(void) @@ -1007,7 +1010,7 @@ static void rpciod_killall(void) spin_lock(&rpc_sched_lock); list_for_each_entry(clnt, &all_clients, cl_clients) - rpc_killall_tasks_locked(&clnt->cl_tasks); + rpc_killall_tasks(clnt); spin_unlock(&rpc_sched_lock); flush_workqueue(rpciod_workqueue); if (!list_empty(&all_clients)) @@ -1110,6 +1113,9 @@ void rpc_show_tasks(void) printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout " "-rpcwait -action- ---ops--\n"); list_for_each_entry(clnt, &all_clients, cl_clients) { + if (list_empty(&clnt->cl_tasks)) + continue; + spin_lock(&clnt->cl_lock); list_for_each_entry(t, &clnt->cl_tasks, tk_task) { const char *rpc_waitq = "none"; @@ -1126,6 +1132,7 @@ void rpc_show_tasks(void) rpc_waitq, t->tk_action, t->tk_ops); } + spin_unlock(&clnt->cl_lock); } out: spin_unlock(&rpc_sched_lock); -- cgit v1.2.3 From 34f52e3591f241b825353ba27def956d8487c400 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 14 Jun 2007 16:40:31 -0400 Subject: SUNRPC: Convert rpc_clnt->cl_users to a kref Signed-off-by: Trond Myklebust --- fs/lockd/host.c | 12 +++------- include/linux/sunrpc/clnt.h | 2 +- net/sunrpc/clnt.c | 57 ++++++++++++++++++++++----------------------- net/sunrpc/rpc_pipe.c | 2 +- net/sunrpc/sched.c | 6 ++--- 5 files changed, 35 insertions(+), 44 deletions(-) (limited to 'include') diff --git a/fs/lockd/host.c b/fs/lockd/host.c index 96070bff93fc..c252a1c95857 100644 --- a/fs/lockd/host.c +++ b/fs/lockd/host.c @@ -161,15 +161,9 @@ nlm_destroy_host(struct nlm_host *host) */ nsm_unmonitor(host); - if ((clnt = host->h_rpcclnt) != NULL) { - if (atomic_read(&clnt->cl_users)) { - printk(KERN_WARNING - "lockd: active RPC handle\n"); - clnt->cl_dead = 1; - } else { - rpc_destroy_client(host->h_rpcclnt); - } - } + clnt = host->h_rpcclnt; + if (clnt != NULL) + rpc_shutdown_client(clnt); kfree(host); } diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 2f4b520a7419..003d8ea70c19 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -24,8 +24,8 @@ struct rpc_inode; * The high-level client handle */ struct rpc_clnt { + struct kref cl_kref; /* Number of references */ atomic_t cl_count; /* Number of clones */ - atomic_t cl_users; /* number of references */ struct list_head cl_clients; /* Global list of clients */ struct list_head cl_tasks; /* List of tasks */ spinlock_t cl_lock; /* spinlock */ diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 424dfdc6862c..254a6e1a5770 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -121,7 +121,6 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s clnt = kzalloc(sizeof(*clnt), GFP_KERNEL); if (!clnt) goto out_err; - atomic_set(&clnt->cl_users, 0); atomic_set(&clnt->cl_count, 1); clnt->cl_parent = clnt; @@ -157,6 +156,8 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s clnt->cl_rtt = &clnt->cl_rtt_default; rpc_init_rtt(&clnt->cl_rtt_default, xprt->timeout.to_initval); + kref_init(&clnt->cl_kref); + err = rpc_setup_pipedir(clnt, program->pipe_dir_name); if (err < 0) goto out_no_path; @@ -272,10 +273,10 @@ rpc_clone_client(struct rpc_clnt *clnt) if (!new) goto out_no_clnt; atomic_set(&new->cl_count, 1); - atomic_set(&new->cl_users, 0); new->cl_metrics = rpc_alloc_iostats(clnt); if (new->cl_metrics == NULL) goto out_no_stats; + kref_init(&new->cl_kref); err = rpc_setup_pipedir(new, clnt->cl_program->pipe_dir_name); if (err != 0) goto out_no_path; @@ -311,40 +312,28 @@ out_no_clnt: int rpc_shutdown_client(struct rpc_clnt *clnt) { - dprintk("RPC: shutting down %s client for %s, tasks=%d\n", - clnt->cl_protname, clnt->cl_server, - atomic_read(&clnt->cl_users)); + dprintk("RPC: shutting down %s client for %s\n", + clnt->cl_protname, clnt->cl_server); - while (atomic_read(&clnt->cl_users) > 0) { + while (!list_empty(&clnt->cl_tasks)) { /* Don't let rpc_release_client destroy us */ clnt->cl_oneshot = 0; clnt->cl_dead = 0; rpc_killall_tasks(clnt); wait_event_timeout(destroy_wait, - !atomic_read(&clnt->cl_users), 1*HZ); - } - - if (atomic_read(&clnt->cl_users) < 0) { - printk(KERN_ERR "RPC: rpc_shutdown_client clnt %p tasks=%d\n", - clnt, atomic_read(&clnt->cl_users)); -#ifdef RPC_DEBUG - rpc_show_tasks(); -#endif - BUG(); + list_empty(&clnt->cl_tasks), 1*HZ); } return rpc_destroy_client(clnt); } /* - * Delete an RPC client + * Free an RPC client */ -int -rpc_destroy_client(struct rpc_clnt *clnt) +static void +rpc_free_client(struct kref *kref) { - if (!atomic_dec_and_test(&clnt->cl_count)) - return 1; - BUG_ON(atomic_read(&clnt->cl_users) != 0); + struct rpc_clnt *clnt = container_of(kref, struct rpc_clnt, cl_kref); dprintk("RPC: destroying %s client for %s\n", clnt->cl_protname, clnt->cl_server); @@ -368,23 +357,33 @@ out_free: clnt->cl_metrics = NULL; xprt_put(clnt->cl_xprt); kfree(clnt); - return 0; } /* - * Release an RPC client + * Release reference to the RPC client */ void rpc_release_client(struct rpc_clnt *clnt) { - dprintk("RPC: rpc_release_client(%p, %d)\n", - clnt, atomic_read(&clnt->cl_users)); + dprintk("RPC: rpc_release_client(%p)\n", clnt); - if (!atomic_dec_and_test(&clnt->cl_users)) - return; - wake_up(&destroy_wait); + if (list_empty(&clnt->cl_tasks)) + wake_up(&destroy_wait); if (clnt->cl_oneshot || clnt->cl_dead) rpc_destroy_client(clnt); + kref_put(&clnt->cl_kref, rpc_free_client); +} + +/* + * Delete an RPC client + */ +int +rpc_destroy_client(struct rpc_clnt *clnt) +{ + if (!atomic_dec_and_test(&clnt->cl_count)) + return 1; + kref_put(&clnt->cl_kref, rpc_free_client); + return 0; } /** diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c index 5887457dc936..826190dacfce 100644 --- a/net/sunrpc/rpc_pipe.c +++ b/net/sunrpc/rpc_pipe.c @@ -344,7 +344,7 @@ rpc_info_open(struct inode *inode, struct file *file) mutex_lock(&inode->i_mutex); clnt = RPC_I(inode)->private; if (clnt) { - atomic_inc(&clnt->cl_users); + kref_get(&clnt->cl_kref); m->private = clnt; } else { single_release(inode, file); diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 0e9fbbd4f987..bb12983580a0 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -846,7 +846,7 @@ void rpc_init_task(struct rpc_task *task, struct rpc_clnt *clnt, int flags, cons task->tk_workqueue = rpciod_workqueue; if (clnt) { - atomic_inc(&clnt->cl_users); + kref_get(&clnt->cl_kref); if (clnt->cl_softrtry) task->tk_flags |= RPC_TASK_SOFT; if (!clnt->cl_intr) @@ -898,9 +898,7 @@ out: cleanup: /* Check whether to release the client */ if (clnt) { - printk("rpc_new_task: failed, users=%d, oneshot=%d\n", - atomic_read(&clnt->cl_users), clnt->cl_oneshot); - atomic_inc(&clnt->cl_users); /* pretend we were used ... */ + kref_get(&clnt->cl_kref); /* pretend we were used ... */ rpc_release_client(clnt); } goto out; -- cgit v1.2.3 From 848f1fe6be2e290691bb6c13cbb8fd92bd0cfaab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 9 Jun 2007 19:39:12 -0400 Subject: SUNRPC: Kill rpc_clnt->cl_dead Its use is at best racy, and there is only one user (lockd), which has additional locking that makes the whole thing redundant. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 3 +-- net/sunrpc/clnt.c | 18 +++--------------- 2 files changed, 4 insertions(+), 17 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 003d8ea70c19..ab3ef6d629a7 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -45,8 +45,7 @@ struct rpc_clnt { cl_intr : 1,/* interruptible */ cl_discrtry : 1,/* disconnect before retry */ cl_autobind : 1,/* use getport() */ - cl_oneshot : 1,/* dispose after use */ - cl_dead : 1;/* abandoned */ + cl_oneshot : 1;/* dispose after use */ struct rpc_rtt * cl_rtt; /* RTO estimator data */ diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 254a6e1a5770..fb65249538d4 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -286,7 +286,6 @@ rpc_clone_client(struct rpc_clnt *clnt) /* Turn off autobind on clones */ new->cl_autobind = 0; new->cl_oneshot = 0; - new->cl_dead = 0; INIT_LIST_HEAD(&new->cl_tasks); spin_lock_init(&new->cl_lock); rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval); @@ -305,9 +304,8 @@ out_no_clnt: /* * Properly shut down an RPC client, terminating all outstanding - * requests. Note that we must be certain that cl_oneshot and - * cl_dead are cleared, or else the client would be destroyed - * when the last task releases it. + * requests. Note that we must be certain that cl_oneshot is cleared, + * or else the client would be destroyed when the last task releases it. */ int rpc_shutdown_client(struct rpc_clnt *clnt) @@ -318,7 +316,6 @@ rpc_shutdown_client(struct rpc_clnt *clnt) while (!list_empty(&clnt->cl_tasks)) { /* Don't let rpc_release_client destroy us */ clnt->cl_oneshot = 0; - clnt->cl_dead = 0; rpc_killall_tasks(clnt); wait_event_timeout(destroy_wait, list_empty(&clnt->cl_tasks), 1*HZ); @@ -369,7 +366,7 @@ rpc_release_client(struct rpc_clnt *clnt) if (list_empty(&clnt->cl_tasks)) wake_up(&destroy_wait); - if (clnt->cl_oneshot || clnt->cl_dead) + if (clnt->cl_oneshot) rpc_destroy_client(clnt); kref_put(&clnt->cl_kref, rpc_free_client); } @@ -483,10 +480,6 @@ int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg, int flags) sigset_t oldset; int status; - /* If this client is slain all further I/O fails */ - if (clnt->cl_dead) - return -EIO; - BUG_ON(flags & RPC_TASK_ASYNC); task = rpc_new_task(clnt, flags, &rpc_default_ops, NULL); @@ -519,11 +512,6 @@ rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, int flags, sigset_t oldset; int status; - /* If this client is slain all further I/O fails */ - status = -EIO; - if (clnt->cl_dead) - goto out_release; - flags |= RPC_TASK_ASYNC; /* Create/initialize a new RPC task */ -- cgit v1.2.3 From 90c5755ff5111ffdcca10a1e8a823dba29f37b6d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 9 Jun 2007 19:49:36 -0400 Subject: SUNRPC: Kill rpc_clnt->cl_oneshot Replace it with explicit calls to rpc_shutdown_client() or rpc_destroy_client() (for the case of asynchronous calls). Signed-off-by: Trond Myklebust --- fs/lockd/mon.c | 2 +- fs/nfs/mount_clnt.c | 4 ++-- include/linux/sunrpc/clnt.h | 10 ++++------ net/sunrpc/clnt.c | 10 +--------- net/sunrpc/rpcb_clnt.c | 6 ++++-- net/sunrpc/sched.c | 14 ++------------ 6 files changed, 14 insertions(+), 32 deletions(-) (limited to 'include') diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c index 2102e2d0134d..3353ed8421a7 100644 --- a/fs/lockd/mon.c +++ b/fs/lockd/mon.c @@ -61,6 +61,7 @@ nsm_mon_unmon(struct nsm_handle *nsm, u32 proc, struct nsm_res *res) status); else status = 0; + rpc_shutdown_client(clnt); out: return status; } @@ -138,7 +139,6 @@ nsm_create(void) .program = &nsm_program, .version = SM_VERSION, .authflavor = RPC_AUTH_NULL, - .flags = (RPC_CLNT_CREATE_ONESHOT), }; return rpc_create(&args); diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c index ca5a266a3140..878d7a5cb6d4 100644 --- a/fs/nfs/mount_clnt.c +++ b/fs/nfs/mount_clnt.c @@ -69,6 +69,7 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh, msg.rpc_proc = &mnt_clnt->cl_procinfo[MNTPROC_MNT]; status = rpc_call_sync(mnt_clnt, &msg, 0); + rpc_shutdown_client(mnt_clnt); return status < 0? status : (result.status? -EACCES : 0); } @@ -84,8 +85,7 @@ mnt_create(char *hostname, struct sockaddr_in *srvaddr, int version, .program = &mnt_program, .version = version, .authflavor = RPC_AUTH_UNIX, - .flags = (RPC_CLNT_CREATE_ONESHOT | - RPC_CLNT_CREATE_INTR), + .flags = RPC_CLNT_CREATE_INTR, }; return rpc_create(&args); diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index ab3ef6d629a7..fe7ea65ed0ae 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -44,8 +44,7 @@ struct rpc_clnt { unsigned int cl_softrtry : 1,/* soft timeouts */ cl_intr : 1,/* interruptible */ cl_discrtry : 1,/* disconnect before retry */ - cl_autobind : 1,/* use getport() */ - cl_oneshot : 1;/* dispose after use */ + cl_autobind : 1;/* use getport() */ struct rpc_rtt * cl_rtt; /* RTO estimator data */ @@ -112,10 +111,9 @@ struct rpc_create_args { #define RPC_CLNT_CREATE_HARDRTRY (1UL << 0) #define RPC_CLNT_CREATE_INTR (1UL << 1) #define RPC_CLNT_CREATE_AUTOBIND (1UL << 2) -#define RPC_CLNT_CREATE_ONESHOT (1UL << 3) -#define RPC_CLNT_CREATE_NONPRIVPORT (1UL << 4) -#define RPC_CLNT_CREATE_NOPING (1UL << 5) -#define RPC_CLNT_CREATE_DISCRTRY (1UL << 6) +#define RPC_CLNT_CREATE_NONPRIVPORT (1UL << 3) +#define RPC_CLNT_CREATE_NOPING (1UL << 4) +#define RPC_CLNT_CREATE_DISCRTRY (1UL << 5) struct rpc_clnt *rpc_create(struct rpc_create_args *args); struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index fb65249538d4..34662dfa9cc0 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -249,8 +249,6 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args) clnt->cl_intr = 1; if (args->flags & RPC_CLNT_CREATE_AUTOBIND) clnt->cl_autobind = 1; - if (args->flags & RPC_CLNT_CREATE_ONESHOT) - clnt->cl_oneshot = 1; if (args->flags & RPC_CLNT_CREATE_DISCRTRY) clnt->cl_discrtry = 1; @@ -285,7 +283,6 @@ rpc_clone_client(struct rpc_clnt *clnt) new->cl_xprt = xprt_get(clnt->cl_xprt); /* Turn off autobind on clones */ new->cl_autobind = 0; - new->cl_oneshot = 0; INIT_LIST_HEAD(&new->cl_tasks); spin_lock_init(&new->cl_lock); rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval); @@ -304,8 +301,7 @@ out_no_clnt: /* * Properly shut down an RPC client, terminating all outstanding - * requests. Note that we must be certain that cl_oneshot is cleared, - * or else the client would be destroyed when the last task releases it. + * requests. */ int rpc_shutdown_client(struct rpc_clnt *clnt) @@ -314,8 +310,6 @@ rpc_shutdown_client(struct rpc_clnt *clnt) clnt->cl_protname, clnt->cl_server); while (!list_empty(&clnt->cl_tasks)) { - /* Don't let rpc_release_client destroy us */ - clnt->cl_oneshot = 0; rpc_killall_tasks(clnt); wait_event_timeout(destroy_wait, list_empty(&clnt->cl_tasks), 1*HZ); @@ -366,8 +360,6 @@ rpc_release_client(struct rpc_clnt *clnt) if (list_empty(&clnt->cl_tasks)) wake_up(&destroy_wait); - if (clnt->cl_oneshot) - rpc_destroy_client(clnt); kref_put(&clnt->cl_kref, rpc_free_client); } diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c index 6c7aa8a1f0c6..00853a326499 100644 --- a/net/sunrpc/rpcb_clnt.c +++ b/net/sunrpc/rpcb_clnt.c @@ -184,8 +184,7 @@ static struct rpc_clnt *rpcb_create(char *hostname, struct sockaddr *srvaddr, .program = &rpcb_program, .version = version, .authflavor = RPC_AUTH_UNIX, - .flags = (RPC_CLNT_CREATE_ONESHOT | - RPC_CLNT_CREATE_NOPING), + .flags = RPC_CLNT_CREATE_NOPING, }; ((struct sockaddr_in *)srvaddr)->sin_port = htons(RPCBIND_PORT); @@ -238,6 +237,7 @@ int rpcb_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay) error = rpc_call_sync(rpcb_clnt, &msg, 0); + rpc_shutdown_client(rpcb_clnt); if (error < 0) printk(KERN_WARNING "RPC: failed to contact local rpcbind " "server (errno %d).\n", -error); @@ -286,6 +286,7 @@ int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog, return PTR_ERR(rpcb_clnt); status = rpc_call_sync(rpcb_clnt, &msg, 0); + rpc_shutdown_client(rpcb_clnt); if (status >= 0) { if (map.r_port != 0) @@ -379,6 +380,7 @@ void rpcb_getport(struct rpc_task *task) } child = rpc_run_task(rpcb_clnt, RPC_TASK_ASYNC, &rpcb_getport_ops, map); + rpc_destroy_client(rpcb_clnt); if (IS_ERR(child)) { status = -EIO; dprintk("RPC: %5u rpcb_getport rpc_run_task failed\n", diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index bb12983580a0..d95fe4e40eb4 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -876,9 +876,7 @@ static void rpc_free_task(struct rcu_head *rcu) } /* - * Create a new task for the specified client. We have to - * clean up after an allocation failure, as the client may - * have specified "oneshot". + * Create a new task for the specified client. */ struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc_call_ops *tk_ops, void *calldata) { @@ -886,7 +884,7 @@ struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc task = rpc_alloc_task(); if (!task) - goto cleanup; + goto out; rpc_init_task(task, clnt, flags, tk_ops, calldata); @@ -894,14 +892,6 @@ struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc task->tk_flags |= RPC_TASK_DYNAMIC; out: return task; - -cleanup: - /* Check whether to release the client */ - if (clnt) { - kref_get(&clnt->cl_kref); /* pretend we were used ... */ - rpc_release_client(clnt); - } - goto out; } -- cgit v1.2.3 From 4c402b40970382ded616eadd544fd63feb76cc79 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 14 Jun 2007 16:40:32 -0400 Subject: SUNRPC: Remove rpc_clnt->cl_count The kref now does most of what cl_count + cl_user used to do. The only remaining role for cl_count is to tell us if we are in a 'shutdown' phase. We can provide that information using a single bit field instead of a full atomic counter. Also rename rpc_destroy_client() to rpc_close_client(), which reflects better what its role is these days. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 4 +--- net/sunrpc/clnt.c | 19 ++----------------- net/sunrpc/rpcb_clnt.c | 2 +- net/sunrpc/sunrpc_syms.c | 1 - 4 files changed, 4 insertions(+), 22 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index fe7ea65ed0ae..cf03494c36e7 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -25,7 +25,6 @@ struct rpc_inode; */ struct rpc_clnt { struct kref cl_kref; /* Number of references */ - atomic_t cl_count; /* Number of clones */ struct list_head cl_clients; /* Global list of clients */ struct list_head cl_tasks; /* List of tasks */ spinlock_t cl_lock; /* spinlock */ @@ -119,8 +118,7 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args); struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, struct rpc_program *, int); struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); -int rpc_shutdown_client(struct rpc_clnt *); -int rpc_destroy_client(struct rpc_clnt *); +void rpc_shutdown_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); void rpc_register_client(struct rpc_clnt *); void rpc_unregister_client(struct rpc_clnt *); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 613c10e4ac31..be5524d20822 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -121,7 +121,6 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s clnt = kzalloc(sizeof(*clnt), GFP_KERNEL); if (!clnt) goto out_err; - atomic_set(&clnt->cl_count, 1); clnt->cl_parent = clnt; clnt->cl_server = clnt->cl_inline_name; @@ -270,7 +269,6 @@ rpc_clone_client(struct rpc_clnt *clnt) new = kmemdup(clnt, sizeof(*new), GFP_KERNEL); if (!new) goto out_no_clnt; - atomic_set(&new->cl_count, 1); new->cl_metrics = rpc_alloc_iostats(clnt); if (new->cl_metrics == NULL) goto out_no_stats; @@ -303,8 +301,7 @@ out_no_clnt: * Properly shut down an RPC client, terminating all outstanding * requests. */ -int -rpc_shutdown_client(struct rpc_clnt *clnt) +void rpc_shutdown_client(struct rpc_clnt *clnt) { dprintk("RPC: shutting down %s client for %s\n", clnt->cl_protname, clnt->cl_server); @@ -315,7 +312,7 @@ rpc_shutdown_client(struct rpc_clnt *clnt) list_empty(&clnt->cl_tasks), 1*HZ); } - return rpc_destroy_client(clnt); + rpc_release_client(clnt); } /* @@ -363,18 +360,6 @@ rpc_release_client(struct rpc_clnt *clnt) kref_put(&clnt->cl_kref, rpc_free_client); } -/* - * Delete an RPC client - */ -int -rpc_destroy_client(struct rpc_clnt *clnt) -{ - if (!atomic_dec_and_test(&clnt->cl_count)) - return 1; - kref_put(&clnt->cl_kref, rpc_free_client); - return 0; -} - /** * rpc_bind_new_program - bind a new RPC program to an existing client * @old - old rpc_client diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c index 00853a326499..cf7db59cdcde 100644 --- a/net/sunrpc/rpcb_clnt.c +++ b/net/sunrpc/rpcb_clnt.c @@ -380,7 +380,7 @@ void rpcb_getport(struct rpc_task *task) } child = rpc_run_task(rpcb_clnt, RPC_TASK_ASYNC, &rpcb_getport_ops, map); - rpc_destroy_client(rpcb_clnt); + rpc_release_client(rpcb_clnt); if (IS_ERR(child)) { status = -EIO; dprintk("RPC: %5u rpcb_getport rpc_run_task failed\n", diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index c46d31ca307b..02e83e15fef5 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -35,7 +35,6 @@ EXPORT_SYMBOL(rpc_wake_up_status); /* RPC client functions */ EXPORT_SYMBOL(rpc_clone_client); EXPORT_SYMBOL(rpc_bind_new_program); -EXPORT_SYMBOL(rpc_destroy_client); EXPORT_SYMBOL(rpc_shutdown_client); EXPORT_SYMBOL(rpc_killall_tasks); EXPORT_SYMBOL(rpc_call_sync); -- cgit v1.2.3 From f61534dfd38f895b203e2aadaba04f21a992ca8c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 14 Jun 2007 17:31:58 -0400 Subject: SUNRPC: Remove redundant calls to rpciod_up()/rpciod_down() Signed-off-by: Trond Myklebust --- fs/lockd/svc.c | 6 ------ fs/nfs/client.c | 15 --------------- fs/nfsd/nfs4callback.c | 12 +++--------- fs/nfsd/nfs4state.c | 1 - include/linux/nfs_fs_sb.h | 1 - net/sunrpc/sunrpc_syms.c | 2 -- 6 files changed, 3 insertions(+), 34 deletions(-) (limited to 'include') diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c index 126b1bf02c0e..26809325469c 100644 --- a/fs/lockd/svc.c +++ b/fs/lockd/svc.c @@ -123,9 +123,6 @@ lockd(struct svc_rqst *rqstp) /* Process request with signals blocked, but allow SIGKILL. */ allow_signal(SIGKILL); - /* kick rpciod */ - rpciod_up(); - dprintk("NFS locking service started (ver " LOCKD_VERSION ").\n"); if (!nlm_timeout) @@ -202,9 +199,6 @@ lockd(struct svc_rqst *rqstp) /* Exit the RPC thread */ svc_exit_thread(rqstp); - /* release rpciod */ - rpciod_down(); - /* Release module */ unlock_kernel(); module_put_and_exit(0); diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 881fa4900923..71d4c4cdac52 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -102,19 +102,10 @@ static struct nfs_client *nfs_alloc_client(const char *hostname, int nfsversion) { struct nfs_client *clp; - int error; if ((clp = kzalloc(sizeof(*clp), GFP_KERNEL)) == NULL) goto error_0; - error = rpciod_up(); - if (error < 0) { - dprintk("%s: couldn't start rpciod! Error = %d\n", - __FUNCTION__, error); - goto error_1; - } - __set_bit(NFS_CS_RPCIOD, &clp->cl_res_state); - if (nfsversion == 4) { if (nfs_callback_up() < 0) goto error_2; @@ -154,9 +145,6 @@ error_3: if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state)) nfs_callback_down(); error_2: - rpciod_down(); - __clear_bit(NFS_CS_RPCIOD, &clp->cl_res_state); -error_1: kfree(clp); error_0: return NULL; @@ -198,9 +186,6 @@ static void nfs_free_client(struct nfs_client *clp) if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state)) nfs_callback_down(); - if (__test_and_clear_bit(NFS_CS_RPCIOD, &clp->cl_res_state)) - rpciod_down(); - kfree(clp->cl_hostname); kfree(clp); diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c index 864090edc28b..6b1b487db1ec 100644 --- a/fs/nfsd/nfs4callback.c +++ b/fs/nfsd/nfs4callback.c @@ -429,29 +429,23 @@ nfsd4_probe_callback(struct nfs4_client *clp) goto out_err; } - /* Kick rpciod, put the call on the wire. */ - if (rpciod_up() != 0) - goto out_clnt; - /* the task holds a reference to the nfs4_client struct */ atomic_inc(&clp->cl_count); msg.rpc_cred = nfsd4_lookupcred(clp,0); if (IS_ERR(msg.rpc_cred)) - goto out_rpciod; + goto out_release_clp; status = rpc_call_async(cb->cb_client, &msg, RPC_TASK_ASYNC, &nfs4_cb_null_ops, NULL); put_rpccred(msg.rpc_cred); if (status != 0) { dprintk("NFSD: asynchronous NFSPROC4_CB_NULL failed!\n"); - goto out_rpciod; + goto out_release_clp; } return; -out_rpciod: +out_release_clp: atomic_dec(&clp->cl_count); - rpciod_down(); -out_clnt: rpc_shutdown_client(cb->cb_client); out_err: cb->cb_client = NULL; diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index 3cc8ce422ab1..8c52913d7cb6 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -378,7 +378,6 @@ shutdown_callback_client(struct nfs4_client *clp) if (clnt) { clp->cl_callback.cb_client = NULL; rpc_shutdown_client(clnt); - rpciod_down(); } } diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 52b4378311c8..144d955dc46a 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -16,7 +16,6 @@ struct nfs_client { #define NFS_CS_INITING 1 /* busy initialising */ int cl_nfsversion; /* NFS protocol version */ unsigned long cl_res_state; /* NFS resources state */ -#define NFS_CS_RPCIOD 0 /* - rpciod started */ #define NFS_CS_CALLBACK 1 /* - callback started */ #define NFS_CS_IDMAP 2 /* - idmap started */ #define NFS_CS_RENEWD 3 /* - renewd started */ diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index 02e83e15fef5..b99b11b11461 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -28,8 +28,6 @@ EXPORT_SYMBOL(rpc_init_task); EXPORT_SYMBOL(rpc_sleep_on); EXPORT_SYMBOL(rpc_wake_up_next); EXPORT_SYMBOL(rpc_wake_up_task); -EXPORT_SYMBOL(rpciod_down); -EXPORT_SYMBOL(rpciod_up); EXPORT_SYMBOL(rpc_wake_up_status); /* RPC client functions */ -- cgit v1.2.3 From 188fef11db219f13f32d055ba59985e7d1a349fe Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 16 Jun 2007 14:18:40 -0400 Subject: SUNRPC: Move rpc_register_client and friends into net/sunrpc/clnt.c Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 2 -- net/sunrpc/clnt.c | 57 +++++++++++++++++++++++++++++++++++++++++ net/sunrpc/sched.c | 62 --------------------------------------------- 3 files changed, 57 insertions(+), 64 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index cf03494c36e7..a451351c7eff 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -120,8 +120,6 @@ struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); void rpc_shutdown_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); -void rpc_register_client(struct rpc_clnt *); -void rpc_unregister_client(struct rpc_clnt *); int rpcb_register(u32, u32, int, unsigned short, int *); void rpcb_getport(struct rpc_task *); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index fe838e996ee3..4f39ab1b04d6 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -44,6 +44,12 @@ dprintk("RPC: %5u %s (status %d)\n", t->tk_pid, \ __FUNCTION__, t->tk_status) +/* + * All RPC clients are linked into this list + */ +static LIST_HEAD(all_clients); +static DEFINE_SPINLOCK(rpc_client_lock); + static DECLARE_WAIT_QUEUE_HEAD(destroy_wait); @@ -66,6 +72,19 @@ static void call_connect_status(struct rpc_task *task); static __be32 * call_header(struct rpc_task *task); static __be32 * call_verify(struct rpc_task *task); +static void rpc_register_client(struct rpc_clnt *clnt) +{ + spin_lock(&rpc_client_lock); + list_add(&clnt->cl_clients, &all_clients); + spin_unlock(&rpc_client_lock); +} + +static void rpc_unregister_client(struct rpc_clnt *clnt) +{ + spin_lock(&rpc_client_lock); + list_del(&clnt->cl_clients); + spin_unlock(&rpc_client_lock); +} static int rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name) @@ -1410,3 +1429,41 @@ int rpc_ping(struct rpc_clnt *clnt, int flags) put_rpccred(msg.rpc_cred); return err; } + +#ifdef RPC_DEBUG +void rpc_show_tasks(void) +{ + struct rpc_clnt *clnt; + struct rpc_task *t; + + spin_lock(&rpc_client_lock); + if (list_empty(&all_clients)) + goto out; + printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout " + "-rpcwait -action- ---ops--\n"); + list_for_each_entry(clnt, &all_clients, cl_clients) { + if (list_empty(&clnt->cl_tasks)) + continue; + spin_lock(&clnt->cl_lock); + list_for_each_entry(t, &clnt->cl_tasks, tk_task) { + const char *rpc_waitq = "none"; + + if (RPC_IS_QUEUED(t)) + rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq); + + printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n", + t->tk_pid, + (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1), + t->tk_flags, t->tk_status, + t->tk_client, + (t->tk_client ? t->tk_client->cl_prog : 0), + t->tk_rqstp, t->tk_timeout, + rpc_waitq, + t->tk_action, t->tk_ops); + } + spin_unlock(&clnt->cl_lock); + } +out: + spin_unlock(&rpc_client_lock); +} +#endif diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 05825154ddd9..c0f8d25caf57 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -47,11 +47,6 @@ static void rpc_release_task(struct rpc_task *task); */ static RPC_WAITQ(delay_queue, "delayq"); -/* - * All RPC clients are linked into this list - */ -static LIST_HEAD(all_clients); - /* * rpciod-related stuff */ @@ -59,11 +54,6 @@ static DEFINE_MUTEX(rpciod_mutex); static atomic_t rpciod_users = ATOMIC_INIT(0); struct workqueue_struct *rpciod_workqueue; -/* - * Spinlock for other critical sections of code. - */ -static DEFINE_SPINLOCK(rpc_sched_lock); - /* * Disable the timer for a given RPC task. Should be called with * queue->lock and bh_disabled in order to avoid races within @@ -994,20 +984,6 @@ void rpc_killall_tasks(struct rpc_clnt *clnt) spin_unlock(&clnt->cl_lock); } -void rpc_register_client(struct rpc_clnt *clnt) -{ - spin_lock(&rpc_sched_lock); - list_add(&clnt->cl_clients, &all_clients); - spin_unlock(&rpc_sched_lock); -} - -void rpc_unregister_client(struct rpc_clnt *clnt) -{ - spin_lock(&rpc_sched_lock); - list_del(&clnt->cl_clients); - spin_unlock(&rpc_sched_lock); -} - /* * Start up the rpciod process if it's not already running. */ @@ -1059,44 +1035,6 @@ rpciod_down(void) mutex_unlock(&rpciod_mutex); } -#ifdef RPC_DEBUG -void rpc_show_tasks(void) -{ - struct rpc_clnt *clnt; - struct rpc_task *t; - - spin_lock(&rpc_sched_lock); - if (list_empty(&all_clients)) - goto out; - printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout " - "-rpcwait -action- ---ops--\n"); - list_for_each_entry(clnt, &all_clients, cl_clients) { - if (list_empty(&clnt->cl_tasks)) - continue; - spin_lock(&clnt->cl_lock); - list_for_each_entry(t, &clnt->cl_tasks, tk_task) { - const char *rpc_waitq = "none"; - - if (RPC_IS_QUEUED(t)) - rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq); - - printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n", - t->tk_pid, - (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1), - t->tk_flags, t->tk_status, - t->tk_client, - (t->tk_client ? t->tk_client->cl_prog : 0), - t->tk_rqstp, t->tk_timeout, - rpc_waitq, - t->tk_action, t->tk_ops); - } - spin_unlock(&clnt->cl_lock); - } -out: - spin_unlock(&rpc_sched_lock); -} -#endif - void rpc_destroy_mempool(void) { -- cgit v1.2.3 From 4a8c1344dccb848dbcf0edabc8b5c51a8ecf2808 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 7 Jun 2007 10:14:14 -0400 Subject: SUNRPC: Add a backpointer from the struct rpc_cred to the rpc_auth Cleans up an issue whereby rpcsec_gss uses the rpc_clnt->cl_auth. If we want to be able to add several rpc_auths to a single rpc_clnt, then this abuse must go. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 4 ++++ net/sunrpc/auth_gss/auth_gss.c | 3 ++- net/sunrpc/auth_null.c | 1 + net/sunrpc/auth_unix.c | 1 + 4 files changed, 8 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 534cdc7be58d..8ef27afeea73 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -30,8 +30,11 @@ struct auth_cred { /* * Client user credentials */ +struct rpc_auth; +struct rpc_credops; struct rpc_cred { struct hlist_node cr_hash; /* hash chain */ + struct rpc_auth * cr_auth; struct rpc_credops * cr_ops; unsigned long cr_expire; /* when to gc */ atomic_t cr_count; /* ref count */ @@ -60,6 +63,7 @@ struct rpc_cred_cache { unsigned long expire; /* cache expiry interval */ }; +struct rpc_authops; struct rpc_auth { unsigned int au_cslack; /* call cred size estimate */ /* guess at number of u32's auth adds before diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 4e4ccc5b6fea..e894e2fc360d 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -386,7 +386,7 @@ static inline int gss_refresh_upcall(struct rpc_task *task) { struct rpc_cred *cred = task->tk_msg.rpc_cred; - struct gss_auth *gss_auth = container_of(task->tk_client->cl_auth, + struct gss_auth *gss_auth = container_of(cred->cr_auth, struct gss_auth, rpc_auth); struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); @@ -741,6 +741,7 @@ gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) * fail to flag the credential as RPCAUTH_CRED_UPTODATE. */ cred->gc_flags = 0; + cred->gc_base.cr_auth = auth; cred->gc_base.cr_ops = &gss_credops; cred->gc_base.cr_flags = RPCAUTH_CRED_NEW; cred->gc_service = gss_auth->service; diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c index 3df9fccab2f8..890bd9b3794b 100644 --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -133,6 +133,7 @@ struct rpc_credops null_credops = { static struct rpc_cred null_cred = { + .cr_auth = &null_auth, .cr_ops = &null_credops, .cr_count = ATOMIC_INIT(1), .cr_flags = RPCAUTH_CRED_UPTODATE, diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 4e7733aee36e..82300b83045e 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -92,6 +92,7 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) if (i < NFS_NGROUPS) cred->uc_gids[i] = NOGROUP; } + cred->uc_base.cr_auth = &unix_auth; cred->uc_base.cr_ops = &unix_credops; return (struct rpc_cred *) cred; -- cgit v1.2.3 From 6e84c7b66a0aa0be16a7728d1e687c57978dac2c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 7 Jun 2007 15:31:36 -0400 Subject: SUNRPC: Add a downcall queue to struct rpc_inode Currently, the downcall queue is tied to the struct gss_auth, which means that different RPCSEC_GSS pseudoflavours must use different upcall pipes. Add a list to struct rpc_inode that can be used instead. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/rpc_pipe_fs.h | 1 + net/sunrpc/auth_gss/auth_gss.c | 29 ++++++++++------------------- net/sunrpc/rpc_pipe.c | 1 + 3 files changed, 12 insertions(+), 19 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/rpc_pipe_fs.h b/include/linux/sunrpc/rpc_pipe_fs.h index ad293760f6eb..430cea104817 100644 --- a/include/linux/sunrpc/rpc_pipe_fs.h +++ b/include/linux/sunrpc/rpc_pipe_fs.h @@ -23,6 +23,7 @@ struct rpc_inode { void *private; struct list_head pipe; struct list_head in_upcall; + struct list_head in_downcall; int pipelen; int nreaders; int nwriters; diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index e407a352440f..50809086fa1b 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -85,7 +85,6 @@ struct gss_auth { struct rpc_auth rpc_auth; struct gss_api_mech *mech; enum rpc_gss_svc service; - struct list_head upcalls; struct rpc_clnt *client; struct dentry *dentry; }; @@ -268,10 +267,10 @@ gss_release_msg(struct gss_upcall_msg *gss_msg) } static struct gss_upcall_msg * -__gss_find_upcall(struct gss_auth *gss_auth, uid_t uid) +__gss_find_upcall(struct rpc_inode *rpci, uid_t uid) { struct gss_upcall_msg *pos; - list_for_each_entry(pos, &gss_auth->upcalls, list) { + list_for_each_entry(pos, &rpci->in_downcall, list) { if (pos->uid != uid) continue; atomic_inc(&pos->count); @@ -290,13 +289,14 @@ static inline struct gss_upcall_msg * gss_add_msg(struct gss_auth *gss_auth, struct gss_upcall_msg *gss_msg) { struct inode *inode = gss_auth->dentry->d_inode; + struct rpc_inode *rpci = RPC_I(inode); struct gss_upcall_msg *old; spin_lock(&inode->i_lock); - old = __gss_find_upcall(gss_auth, gss_msg->uid); + old = __gss_find_upcall(rpci, gss_msg->uid); if (old == NULL) { atomic_inc(&gss_msg->count); - list_add(&gss_msg->list, &gss_auth->upcalls); + list_add(&gss_msg->list, &rpci->in_downcall); } else gss_msg = old; spin_unlock(&inode->i_lock); @@ -493,7 +493,6 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) const void *p, *end; void *buf; struct rpc_clnt *clnt; - struct gss_auth *gss_auth; struct gss_upcall_msg *gss_msg; struct inode *inode = filp->f_path.dentry->d_inode; struct gss_cl_ctx *ctx; @@ -526,9 +525,8 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) err = -ENOENT; /* Find a matching upcall */ - gss_auth = container_of(clnt->cl_auth, struct gss_auth, rpc_auth); spin_lock(&inode->i_lock); - gss_msg = __gss_find_upcall(gss_auth, uid); + gss_msg = __gss_find_upcall(RPC_I(inode), uid); if (gss_msg == NULL) { spin_unlock(&inode->i_lock); goto err_put_ctx; @@ -536,7 +534,7 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) list_del_init(&gss_msg->list); spin_unlock(&inode->i_lock); - p = gss_fill_context(p, end, ctx, gss_auth->mech); + p = gss_fill_context(p, end, ctx, gss_msg->auth->mech); if (IS_ERR(p)) { err = PTR_ERR(p); gss_msg->msg.errno = (err == -EACCES) ? -EACCES : -EAGAIN; @@ -563,18 +561,12 @@ static void gss_pipe_release(struct inode *inode) { struct rpc_inode *rpci = RPC_I(inode); - struct rpc_clnt *clnt; - struct rpc_auth *auth; - struct gss_auth *gss_auth; + struct gss_upcall_msg *gss_msg; - clnt = rpci->private; - auth = clnt->cl_auth; - gss_auth = container_of(auth, struct gss_auth, rpc_auth); spin_lock(&inode->i_lock); - while (!list_empty(&gss_auth->upcalls)) { - struct gss_upcall_msg *gss_msg; + while (!list_empty(&rpci->in_downcall)) { - gss_msg = list_entry(gss_auth->upcalls.next, + gss_msg = list_entry(rpci->in_downcall.next, struct gss_upcall_msg, list); gss_msg->msg.errno = -EPIPE; atomic_inc(&gss_msg->count); @@ -637,7 +629,6 @@ gss_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor) gss_auth->service = gss_pseudoflavor_to_service(gss_auth->mech, flavor); if (gss_auth->service == 0) goto err_put_mech; - INIT_LIST_HEAD(&gss_auth->upcalls); auth = &gss_auth->rpc_auth; auth->au_cslack = GSS_CRED_SLACK >> 2; auth->au_rslack = GSS_VERF_SLACK >> 2; diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c index 826190dacfce..2320f1e42da4 100644 --- a/net/sunrpc/rpc_pipe.c +++ b/net/sunrpc/rpc_pipe.c @@ -833,6 +833,7 @@ init_once(void * foo, struct kmem_cache * cachep, unsigned long flags) rpci->nreaders = 0; rpci->nwriters = 0; INIT_LIST_HEAD(&rpci->in_upcall); + INIT_LIST_HEAD(&rpci->in_downcall); INIT_LIST_HEAD(&rpci->pipe); rpci->pipelen = 0; init_waitqueue_head(&rpci->waitq); -- cgit v1.2.3 From 03a1256f06cf1f58e33971fb4a524479e75c200e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 8 Jun 2007 14:14:53 -0400 Subject: SUNRPC: Add a field to track the number of kernel users of an rpc_pipe This allows us to correctly deduce when we need to remove the pipe. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/rpc_pipe_fs.h | 1 + net/sunrpc/rpc_pipe.c | 12 ++++++++---- 2 files changed, 9 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/rpc_pipe_fs.h b/include/linux/sunrpc/rpc_pipe_fs.h index 430cea104817..51b977a4ca20 100644 --- a/include/linux/sunrpc/rpc_pipe_fs.h +++ b/include/linux/sunrpc/rpc_pipe_fs.h @@ -27,6 +27,7 @@ struct rpc_inode { int pipelen; int nreaders; int nwriters; + int nkern_readwriters; wait_queue_head_t waitq; #define RPC_PIPE_WAIT_FOR_OPEN 1 int flags; diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c index e5fd796e897e..e787b6a43eee 100644 --- a/net/sunrpc/rpc_pipe.c +++ b/net/sunrpc/rpc_pipe.c @@ -737,6 +737,7 @@ rpc_mkpipe(struct dentry *parent, const char *name, void *private, struct rpc_pi dput (dentry); dentry = ERR_PTR(-EBUSY); } + rpci->nkern_readwriters++; goto out; } inode = rpc_get_inode(dir->i_sb, S_IFIFO | S_IRUSR | S_IWUSR); @@ -749,6 +750,7 @@ rpc_mkpipe(struct dentry *parent, const char *name, void *private, struct rpc_pi rpci->private = private; rpci->flags = flags; rpci->ops = ops; + rpci->nkern_readwriters = 1; inode_dir_notify(dir, DN_CREATE); dget(dentry); out: @@ -773,10 +775,12 @@ rpc_unlink(struct dentry *dentry) parent = dget_parent(dentry); dir = parent->d_inode; mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); - rpc_close_pipes(dentry->d_inode); - error = simple_unlink(dir, dentry); - if (!error) - d_delete(dentry); + if (--RPC_I(dentry->d_inode)->nkern_readwriters == 0) { + rpc_close_pipes(dentry->d_inode); + error = simple_unlink(dir, dentry); + if (!error) + d_delete(dentry); + } dput(dentry); mutex_unlock(&dir->i_mutex); dput(parent); -- cgit v1.2.3 From 3ab9bb7243489f9db3abf3d05521ddfc6b184c0a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 9 Jun 2007 15:41:42 -0400 Subject: SUNRPC: Fix a memory leak in the auth credcache code The leak only affects the RPCSEC_GSS caches, since they are the only ones that are dynamically allocated... Rename the existing rpcauth_free_credcache() to rpcauth_clear_credcache() in order to better describe its role, then add a new function rpcauth_destroy_credcache() that actually frees the cache in addition to clearing it out. Also move the call to destroy the credcache in gss_destroy() to come before the rpc upcall pipe is unlinked. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 3 ++- net/sunrpc/auth.c | 18 ++++++++++++++++-- net/sunrpc/auth_gss/auth_gss.c | 3 ++- net/sunrpc/auth_unix.c | 2 +- net/sunrpc/sunrpc_syms.c | 2 +- 5 files changed, 22 insertions(+), 6 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 8ef27afeea73..3972b8414c88 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -143,7 +143,8 @@ int rpcauth_refreshcred(struct rpc_task *); void rpcauth_invalcred(struct rpc_task *); int rpcauth_uptodatecred(struct rpc_task *); int rpcauth_init_credcache(struct rpc_auth *, unsigned long); -void rpcauth_free_credcache(struct rpc_auth *); +void rpcauth_destroy_credcache(struct rpc_auth *); +void rpcauth_clear_credcache(struct rpc_cred_cache *); static inline struct rpc_cred * get_rpccred(struct rpc_cred *cred) diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 9527f2bb1744..f6b6c81cbc3e 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -137,9 +137,8 @@ void rpcauth_destroy_credlist(struct hlist_head *head) * that are not referenced. */ void -rpcauth_free_credcache(struct rpc_auth *auth) +rpcauth_clear_credcache(struct rpc_cred_cache *cache) { - struct rpc_cred_cache *cache = auth->au_credcache; HLIST_HEAD(free); struct hlist_node *pos, *next; struct rpc_cred *cred; @@ -157,6 +156,21 @@ rpcauth_free_credcache(struct rpc_auth *auth) rpcauth_destroy_credlist(&free); } +/* + * Destroy the RPC credential cache + */ +void +rpcauth_destroy_credcache(struct rpc_auth *auth) +{ + struct rpc_cred_cache *cache = auth->au_credcache; + + if (cache) { + auth->au_credcache = NULL; + rpcauth_clear_credcache(cache); + kfree(cache); + } +} + static void rpcauth_prune_expired(struct rpc_auth *auth, struct rpc_cred *cred, struct hlist_head *free) { diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 50809086fa1b..8b4c02f8befb 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -665,12 +665,13 @@ gss_destroy(struct rpc_auth *auth) dprintk("RPC: destroying GSS authenticator %p flavor %d\n", auth, auth->au_flavor); + rpcauth_destroy_credcache(auth); + gss_auth = container_of(auth, struct gss_auth, rpc_auth); rpc_unlink(gss_auth->dentry); gss_auth->dentry = NULL; gss_mech_put(gss_auth->mech); - rpcauth_free_credcache(auth); kfree(gss_auth); module_put(THIS_MODULE); } diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 82300b83045e..5622783011a4 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -50,7 +50,7 @@ static void unx_destroy(struct rpc_auth *auth) { dprintk("RPC: destroying UNIX authenticator %p\n", auth); - rpcauth_free_credcache(auth); + rpcauth_clear_credcache(auth->au_credcache); } /* diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index b99b11b11461..3e19e7af6799 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -57,7 +57,7 @@ EXPORT_SYMBOL(rpcauth_unregister); EXPORT_SYMBOL(rpcauth_create); EXPORT_SYMBOL(rpcauth_lookupcred); EXPORT_SYMBOL(rpcauth_lookup_credcache); -EXPORT_SYMBOL(rpcauth_free_credcache); +EXPORT_SYMBOL(rpcauth_destroy_credcache); EXPORT_SYMBOL(rpcauth_init_credcache); EXPORT_SYMBOL(put_rpccred); -- cgit v1.2.3 From 64c91a1f1c8bc4295fd6b90df8adf911a7dd64f4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 10:17:16 -0400 Subject: SUNRPC: Make rpc_ping() static Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 1 - net/sunrpc/clnt.c | 4 +++- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index a451351c7eff..a0e51e193284 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -136,7 +136,6 @@ void rpc_clnt_sigunmask(struct rpc_clnt *clnt, sigset_t *oldset); void rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int); size_t rpc_max_payload(struct rpc_clnt *); void rpc_force_rebind(struct rpc_clnt *); -int rpc_ping(struct rpc_clnt *clnt, int flags); size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t); char * rpc_peeraddr2str(struct rpc_clnt *, enum rpc_display_format_t); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 76eef19cf995..4e91f3110938 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -72,6 +72,8 @@ static void call_connect_status(struct rpc_task *task); static __be32 * call_header(struct rpc_task *task); static __be32 * call_verify(struct rpc_task *task); +static int rpc_ping(struct rpc_clnt *clnt, int flags); + static void rpc_register_client(struct rpc_clnt *clnt) { spin_lock(&rpc_client_lock); @@ -1441,7 +1443,7 @@ static struct rpc_procinfo rpcproc_null = { .p_decode = rpcproc_decode_null, }; -int rpc_ping(struct rpc_clnt *clnt, int flags) +static int rpc_ping(struct rpc_clnt *clnt, int flags) { struct rpc_message msg = { .rpc_proc = &rpcproc_null, -- cgit v1.2.3 From 5e1550d6a2c2dd33ff0ca5febefd8e9c65c6ca1e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 10:17:16 -0400 Subject: SUNRPC: Add the helper function 'rpc_call_null()' Does a NULL RPC call and returns a pointer to the resulting rpc_task. The call may be either synchronous or asynchronous. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 2 ++ net/sunrpc/clnt.c | 10 ++++++++++ 2 files changed, 12 insertions(+) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index a0e51e193284..097984b03857 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -130,6 +130,8 @@ int rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, void *calldata); int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg, int flags); +struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred, + int flags); void rpc_restart_call(struct rpc_task *); void rpc_clnt_sigmask(struct rpc_clnt *clnt, sigset_t *oldset); void rpc_clnt_sigunmask(struct rpc_clnt *clnt, sigset_t *oldset); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 4e91f3110938..5a28ffac99ea 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -1455,6 +1455,16 @@ static int rpc_ping(struct rpc_clnt *clnt, int flags) return err; } +struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred, int flags) +{ + struct rpc_message msg = { + .rpc_proc = &rpcproc_null, + .rpc_cred = cred, + }; + return rpc_do_run_task(clnt, &msg, flags, &rpc_default_ops, NULL); +} +EXPORT_SYMBOL(rpc_call_null); + #ifdef RPC_DEBUG void rpc_show_tasks(void) { -- cgit v1.2.3 From de7a8ce38aea529876db3890b61947bc4bc004da Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 10:46:47 -0400 Subject: SUNRPC: Rename rpcauth_destroy() to rpcauth_release() Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 2 +- net/sunrpc/auth.c | 4 ++-- net/sunrpc/clnt.c | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 3972b8414c88..bc77c730325c 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -128,7 +128,7 @@ extern struct rpc_authops authdes_ops; int rpcauth_register(struct rpc_authops *); int rpcauth_unregister(struct rpc_authops *); struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *); -void rpcauth_destroy(struct rpc_auth *); +void rpcauth_release(struct rpc_auth *); struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int); struct rpc_cred * rpcauth_bindcred(struct rpc_task *); diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 584f24311a80..1686dc74c6a9 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -93,7 +93,7 @@ rpcauth_create(rpc_authflavor_t pseudoflavor, struct rpc_clnt *clnt) if (IS_ERR(auth)) return auth; if (clnt->cl_auth) - rpcauth_destroy(clnt->cl_auth); + rpcauth_release(clnt->cl_auth); clnt->cl_auth = auth; out: @@ -101,7 +101,7 @@ out: } void -rpcauth_destroy(struct rpc_auth *auth) +rpcauth_release(struct rpc_auth *auth) { if (!atomic_dec_and_test(&auth->au_count)) return; diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 5a28ffac99ea..98df44e453fe 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -353,7 +353,7 @@ rpc_free_client(struct kref *kref) dprintk("RPC: destroying %s client for %s\n", clnt->cl_protname, clnt->cl_server); if (clnt->cl_auth) { - rpcauth_destroy(clnt->cl_auth); + rpcauth_release(clnt->cl_auth); clnt->cl_auth = NULL; } if (!IS_ERR(clnt->cl_dentry)) { -- cgit v1.2.3 From f1c0a8615090359d57e096157feb9f900cbb233c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 20:17:58 -0400 Subject: SUNRPC: Mark auth and cred operation tables as constant. Also do the same for gss_api operation tables. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 15 ++++++--------- include/linux/sunrpc/gss_api.h | 2 +- net/sunrpc/auth.c | 8 ++++---- net/sunrpc/auth_gss/auth_gss.c | 8 ++++---- net/sunrpc/auth_gss/gss_krb5_mech.c | 2 +- net/sunrpc/auth_gss/gss_spkm3_mech.c | 2 +- net/sunrpc/auth_null.c | 4 ++-- net/sunrpc/auth_unix.c | 6 +++--- 8 files changed, 22 insertions(+), 25 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index bc77c730325c..e606c2804685 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -35,7 +35,7 @@ struct rpc_credops; struct rpc_cred { struct hlist_node cr_hash; /* hash chain */ struct rpc_auth * cr_auth; - struct rpc_credops * cr_ops; + const struct rpc_credops *cr_ops; unsigned long cr_expire; /* when to gc */ atomic_t cr_count; /* ref count */ unsigned short cr_flags; /* various flags */ @@ -73,7 +73,7 @@ struct rpc_auth { unsigned int au_verfsize; unsigned int au_flags; /* various flags */ - struct rpc_authops * au_ops; /* operations */ + const struct rpc_authops *au_ops; /* operations */ rpc_authflavor_t au_flavor; /* pseudoflavor (note may * differ from the flavor in * au_ops->au_flavor in gss @@ -119,14 +119,11 @@ struct rpc_credops { void *, __be32 *, void *); }; -extern struct rpc_authops authunix_ops; -extern struct rpc_authops authnull_ops; -#ifdef CONFIG_SUNRPC_SECURE -extern struct rpc_authops authdes_ops; -#endif +extern const struct rpc_authops authunix_ops; +extern const struct rpc_authops authnull_ops; -int rpcauth_register(struct rpc_authops *); -int rpcauth_unregister(struct rpc_authops *); +int rpcauth_register(const struct rpc_authops *); +int rpcauth_unregister(const struct rpc_authops *); struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *); void rpcauth_release(struct rpc_auth *); struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); diff --git a/include/linux/sunrpc/gss_api.h b/include/linux/sunrpc/gss_api.h index 5eca9e442051..bbac101ac372 100644 --- a/include/linux/sunrpc/gss_api.h +++ b/include/linux/sunrpc/gss_api.h @@ -77,7 +77,7 @@ struct gss_api_mech { struct module *gm_owner; struct xdr_netobj gm_oid; char *gm_name; - struct gss_api_ops *gm_ops; + const struct gss_api_ops *gm_ops; /* pseudoflavors supported by this mechanism: */ int gm_pf_num; struct pf_desc * gm_pfs; diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 1686dc74c6a9..d3f0f944c0b5 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -19,7 +19,7 @@ #endif static DEFINE_SPINLOCK(rpc_authflavor_lock); -static struct rpc_authops * auth_flavors[RPC_AUTH_MAXFLAVOR] = { +static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = { &authnull_ops, /* AUTH_NULL */ &authunix_ops, /* AUTH_UNIX */ NULL, /* others can be loadable modules */ @@ -33,7 +33,7 @@ pseudoflavor_to_flavor(u32 flavor) { } int -rpcauth_register(struct rpc_authops *ops) +rpcauth_register(const struct rpc_authops *ops) { rpc_authflavor_t flavor; int ret = -EPERM; @@ -50,7 +50,7 @@ rpcauth_register(struct rpc_authops *ops) } int -rpcauth_unregister(struct rpc_authops *ops) +rpcauth_unregister(const struct rpc_authops *ops) { rpc_authflavor_t flavor; int ret = -EPERM; @@ -70,7 +70,7 @@ struct rpc_auth * rpcauth_create(rpc_authflavor_t pseudoflavor, struct rpc_clnt *clnt) { struct rpc_auth *auth; - struct rpc_authops *ops; + const struct rpc_authops *ops; u32 flavor = pseudoflavor_to_flavor(pseudoflavor); auth = ERR_PTR(-EINVAL); diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 459dc9b1d1ad..177a9e413c0a 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -54,9 +54,9 @@ #include #include -static struct rpc_authops authgss_ops; +static const struct rpc_authops authgss_ops; -static struct rpc_credops gss_credops; +static const struct rpc_credops gss_credops; #ifdef RPC_DEBUG # define RPCDBG_FACILITY RPCDBG_AUTH @@ -1193,7 +1193,7 @@ out: return status; } -static struct rpc_authops authgss_ops = { +static const struct rpc_authops authgss_ops = { .owner = THIS_MODULE, .au_flavor = RPC_AUTH_GSS, #ifdef RPC_DEBUG @@ -1205,7 +1205,7 @@ static struct rpc_authops authgss_ops = { .crcreate = gss_create_cred }; -static struct rpc_credops gss_credops = { +static const struct rpc_credops gss_credops = { .cr_name = "AUTH_GSS", .crdestroy = gss_destroy_cred, .cr_init = gss_cred_init, diff --git a/net/sunrpc/auth_gss/gss_krb5_mech.c b/net/sunrpc/auth_gss/gss_krb5_mech.c index 7b1943217053..71b9daefdff3 100644 --- a/net/sunrpc/auth_gss/gss_krb5_mech.c +++ b/net/sunrpc/auth_gss/gss_krb5_mech.c @@ -201,7 +201,7 @@ gss_delete_sec_context_kerberos(void *internal_ctx) { kfree(kctx); } -static struct gss_api_ops gss_kerberos_ops = { +static const struct gss_api_ops gss_kerberos_ops = { .gss_import_sec_context = gss_import_sec_context_kerberos, .gss_get_mic = gss_get_mic_kerberos, .gss_verify_mic = gss_verify_mic_kerberos, diff --git a/net/sunrpc/auth_gss/gss_spkm3_mech.c b/net/sunrpc/auth_gss/gss_spkm3_mech.c index 7e15aa68ae64..577d590e755f 100644 --- a/net/sunrpc/auth_gss/gss_spkm3_mech.c +++ b/net/sunrpc/auth_gss/gss_spkm3_mech.c @@ -202,7 +202,7 @@ gss_get_mic_spkm3(struct gss_ctx *ctx, return err; } -static struct gss_api_ops gss_spkm3_ops = { +static const struct gss_api_ops gss_spkm3_ops = { .gss_import_sec_context = gss_import_sec_context_spkm3, .gss_get_mic = gss_get_mic_spkm3, .gss_verify_mic = gss_verify_mic_spkm3, diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c index 890bd9b3794b..fe9b6aaf91eb 100644 --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -101,7 +101,7 @@ nul_validate(struct rpc_task *task, __be32 *p) return p; } -struct rpc_authops authnull_ops = { +const struct rpc_authops authnull_ops = { .owner = THIS_MODULE, .au_flavor = RPC_AUTH_NULL, #ifdef RPC_DEBUG @@ -122,7 +122,7 @@ struct rpc_auth null_auth = { }; static -struct rpc_credops null_credops = { +const struct rpc_credops null_credops = { .cr_name = "AUTH_NULL", .crdestroy = nul_destroy_cred, .crmatch = nul_match, diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index e54782e75d59..6600c7ad72a9 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -34,7 +34,7 @@ struct unx_cred { static struct rpc_auth unix_auth; static struct rpc_cred_cache unix_cred_cache; -static struct rpc_credops unix_credops; +static const struct rpc_credops unix_credops; static struct rpc_auth * unx_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor) @@ -205,7 +205,7 @@ unx_validate(struct rpc_task *task, __be32 *p) return p; } -struct rpc_authops authunix_ops = { +const struct rpc_authops authunix_ops = { .owner = THIS_MODULE, .au_flavor = RPC_AUTH_UNIX, #ifdef RPC_DEBUG @@ -233,7 +233,7 @@ struct rpc_auth unix_auth = { }; static -struct rpc_credops unix_credops = { +const struct rpc_credops unix_credops = { .cr_name = "AUTH_UNIX", .crdestroy = unx_destroy_cred, .crmatch = unx_match, -- cgit v1.2.3 From 5fe4755e2526a2aa82b7ed8daeb3aed74a236925 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 19:55:31 -0400 Subject: SUNRPC: Clean up rpc credential initialisation Add a helper rpc_cred_init() Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 1 + include/linux/sunrpc/auth_gss.h | 5 ----- net/sunrpc/auth.c | 24 ++++++++++++++++++------ net/sunrpc/auth_gss/auth_gss.c | 6 +----- net/sunrpc/auth_unix.c | 10 ++-------- 5 files changed, 22 insertions(+), 24 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index e606c2804685..d5bfc67461fc 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -127,6 +127,7 @@ int rpcauth_unregister(const struct rpc_authops *); struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *); void rpcauth_release(struct rpc_auth *); struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); +void rpcauth_init_cred(struct rpc_cred *, const struct auth_cred *, struct rpc_auth *, const struct rpc_credops *); struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int); struct rpc_cred * rpcauth_bindcred(struct rpc_task *); void rpcauth_holdcred(struct rpc_task *); diff --git a/include/linux/sunrpc/auth_gss.h b/include/linux/sunrpc/auth_gss.h index 2db2fbf34947..0bd1d06777b9 100644 --- a/include/linux/sunrpc/auth_gss.h +++ b/include/linux/sunrpc/auth_gss.h @@ -85,11 +85,6 @@ struct gss_cred { struct gss_upcall_msg *gc_upcall; }; -#define gc_uid gc_base.cr_uid -#define gc_count gc_base.cr_count -#define gc_flags gc_base.cr_flags -#define gc_expire gc_base.cr_expire - #endif /* __KERNEL__ */ #endif /* _LINUX_SUNRPC_AUTH_GSS_H */ diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index d3f0f944c0b5..2156327da45b 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -264,13 +264,9 @@ retry: if (!cred) { new = auth->au_ops->crcreate(auth, acred, flags); - if (!IS_ERR(new)) { -#ifdef RPC_DEBUG - new->cr_magic = RPCAUTH_CRED_MAGIC; -#endif + if (!IS_ERR(new)) goto retry; - } else - cred = new; + cred = new; } else if ((cred->cr_flags & RPCAUTH_CRED_NEW) && cred->cr_ops->cr_init != NULL && !(flags & RPCAUTH_LOOKUP_NEW)) { @@ -302,6 +298,22 @@ rpcauth_lookupcred(struct rpc_auth *auth, int flags) return ret; } +void +rpcauth_init_cred(struct rpc_cred *cred, const struct auth_cred *acred, + struct rpc_auth *auth, const struct rpc_credops *ops) +{ + INIT_HLIST_NODE(&cred->cr_hash); + atomic_set(&cred->cr_count, 1); + cred->cr_auth = auth; + cred->cr_ops = ops; + cred->cr_expire = jiffies; +#ifdef RPC_DEBUG + cred->cr_magic = RPCAUTH_CRED_MAGIC; +#endif + cred->cr_uid = acred->uid; +} +EXPORT_SYMBOL(rpcauth_init_cred); + struct rpc_cred * rpcauth_bindcred(struct rpc_task *task) { diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 177a9e413c0a..766de0a41b22 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -727,15 +727,11 @@ gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) if (!(cred = kzalloc(sizeof(*cred), GFP_KERNEL))) goto out_err; - atomic_set(&cred->gc_count, 1); - cred->gc_uid = acred->uid; + rpcauth_init_cred(&cred->gc_base, acred, auth, &gss_credops); /* * Note: in order to force a call to call_refresh(), we deliberately * fail to flag the credential as RPCAUTH_CRED_UPTODATE. */ - cred->gc_flags = 0; - cred->gc_base.cr_auth = auth; - cred->gc_base.cr_ops = &gss_credops; cred->gc_base.cr_flags = RPCAUTH_CRED_NEW; cred->gc_service = gss_auth->service; return &cred->gc_base; diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 6600c7ad72a9..2f1bdb5c86b3 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -20,9 +20,6 @@ struct unx_cred { gid_t uc_gids[NFS_NGROUPS]; }; #define uc_uid uc_base.cr_uid -#define uc_count uc_base.cr_count -#define uc_flags uc_base.cr_flags -#define uc_expire uc_base.cr_expire #define UNX_CRED_EXPIRE (60 * HZ) @@ -74,8 +71,8 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) if (!(cred = kmalloc(sizeof(*cred), GFP_KERNEL))) return ERR_PTR(-ENOMEM); - atomic_set(&cred->uc_count, 1); - cred->uc_flags = RPCAUTH_CRED_UPTODATE; + rpcauth_init_cred(&cred->uc_base, acred, auth, &unix_credops); + cred->uc_base.cr_flags = RPCAUTH_CRED_UPTODATE; if (flags & RPCAUTH_LOOKUP_ROOTCREDS) { cred->uc_uid = 0; cred->uc_gid = 0; @@ -85,15 +82,12 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) if (groups > NFS_NGROUPS) groups = NFS_NGROUPS; - cred->uc_uid = acred->uid; cred->uc_gid = acred->gid; for (i = 0; i < groups; i++) cred->uc_gids[i] = GROUP_AT(acred->group_info, i); if (i < NFS_NGROUPS) cred->uc_gids[i] = NOGROUP; } - cred->uc_base.cr_auth = &unix_auth; - cred->uc_base.cr_ops = &unix_credops; return (struct rpc_cred *) cred; } -- cgit v1.2.3 From fc432dd90760a629c57026e57f65ff80a1a31d2f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 25 Jun 2007 10:15:15 -0400 Subject: SUNRPC: Enforce atomic updates of rpc_cred->cr_flags Convert to the use of atomic bitops... Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 10 +++++----- net/sunrpc/auth.c | 22 ++++++++++++---------- net/sunrpc/auth_gss/auth_gss.c | 22 +++++++++++----------- net/sunrpc/auth_null.c | 4 ++-- net/sunrpc/auth_unix.c | 4 ++-- 5 files changed, 32 insertions(+), 30 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index d5bfc67461fc..8586503d5ebd 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -36,19 +36,19 @@ struct rpc_cred { struct hlist_node cr_hash; /* hash chain */ struct rpc_auth * cr_auth; const struct rpc_credops *cr_ops; - unsigned long cr_expire; /* when to gc */ - atomic_t cr_count; /* ref count */ - unsigned short cr_flags; /* various flags */ #ifdef RPC_DEBUG unsigned long cr_magic; /* 0x0f4aa4f0 */ #endif + unsigned long cr_expire; /* when to gc */ + unsigned long cr_flags; /* various flags */ + atomic_t cr_count; /* ref count */ uid_t cr_uid; /* per-flavor data */ }; -#define RPCAUTH_CRED_NEW 0x0001 -#define RPCAUTH_CRED_UPTODATE 0x0002 +#define RPCAUTH_CRED_NEW 0 +#define RPCAUTH_CRED_UPTODATE 1 #define RPCAUTH_CRED_MAGIC 0x0f4aa4f0 diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 2156327da45b..4d7c78b05d1e 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -190,8 +190,8 @@ rpcauth_prune_expired(struct rpc_auth *auth, struct rpc_cred *cred, struct hlist if (atomic_read(&cred->cr_count) != 1) return; if (time_after(jiffies, cred->cr_expire + auth->au_credcache->expire)) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; - if (!(cred->cr_flags & RPCAUTH_CRED_UPTODATE)) { + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); + if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) { __hlist_del(&cred->cr_hash); hlist_add_head(&cred->cr_hash, free); } @@ -267,7 +267,7 @@ retry: if (!IS_ERR(new)) goto retry; cred = new; - } else if ((cred->cr_flags & RPCAUTH_CRED_NEW) + } else if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags) && cred->cr_ops->cr_init != NULL && !(flags & RPCAUTH_LOOKUP_NEW)) { int res = cred->cr_ops->cr_init(auth, cred); @@ -440,17 +440,19 @@ rpcauth_refreshcred(struct rpc_task *task) void rpcauth_invalcred(struct rpc_task *task) { + struct rpc_cred *cred = task->tk_msg.rpc_cred; + dprintk("RPC: %5u invalidating %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, task->tk_msg.rpc_cred); - spin_lock(&rpc_credcache_lock); - if (task->tk_msg.rpc_cred) - task->tk_msg.rpc_cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; - spin_unlock(&rpc_credcache_lock); + task->tk_pid, task->tk_auth->au_ops->au_name, cred); + if (cred) + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); } int rpcauth_uptodatecred(struct rpc_task *task) { - return !(task->tk_msg.rpc_cred) || - (task->tk_msg.rpc_cred->cr_flags & RPCAUTH_CRED_UPTODATE); + struct rpc_cred *cred = task->tk_msg.rpc_cred; + + return cred == NULL || + test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) != 0; } diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 766de0a41b22..55c47ae0a258 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -114,8 +114,8 @@ gss_cred_set_ctx(struct rpc_cred *cred, struct gss_cl_ctx *ctx) write_lock(&gss_ctx_lock); old = gss_cred->gc_ctx; gss_cred->gc_ctx = ctx; - cred->cr_flags |= RPCAUTH_CRED_UPTODATE; - cred->cr_flags &= ~RPCAUTH_CRED_NEW; + set_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); + clear_bit(RPCAUTH_CRED_NEW, &cred->cr_flags); write_unlock(&gss_ctx_lock); if (old) gss_put_ctx(old); @@ -128,7 +128,7 @@ gss_cred_is_uptodate_ctx(struct rpc_cred *cred) int res = 0; read_lock(&gss_ctx_lock); - if ((cred->cr_flags & RPCAUTH_CRED_UPTODATE) && gss_cred->gc_ctx) + if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) && gss_cred->gc_ctx) res = 1; read_unlock(&gss_ctx_lock); return res; @@ -732,7 +732,7 @@ gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) * Note: in order to force a call to call_refresh(), we deliberately * fail to flag the credential as RPCAUTH_CRED_UPTODATE. */ - cred->gc_base.cr_flags = RPCAUTH_CRED_NEW; + cred->gc_base.cr_flags = 1UL << RPCAUTH_CRED_NEW; cred->gc_service = gss_auth->service; return &cred->gc_base; @@ -764,7 +764,7 @@ gss_match(struct auth_cred *acred, struct rpc_cred *rc, int flags) * we don't really care if the credential has expired or not, * since the caller should be prepared to reinitialise it. */ - if ((flags & RPCAUTH_LOOKUP_NEW) && (rc->cr_flags & RPCAUTH_CRED_NEW)) + if ((flags & RPCAUTH_LOOKUP_NEW) && test_bit(RPCAUTH_CRED_NEW, &rc->cr_flags)) goto out; /* Don't match with creds that have expired. */ if (gss_cred->gc_ctx && time_after(jiffies, gss_cred->gc_ctx->gc_expiry)) @@ -820,7 +820,7 @@ gss_marshal(struct rpc_task *task, __be32 *p) mic.data = (u8 *)(p + 1); maj_stat = gss_get_mic(ctx->gc_gss_ctx, &verf_buf, &mic); if (maj_stat == GSS_S_CONTEXT_EXPIRED) { - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); } else if (maj_stat != 0) { printk("gss_marshal: gss_get_mic FAILED (%d)\n", maj_stat); goto out_put_ctx; @@ -873,7 +873,7 @@ gss_validate(struct rpc_task *task, __be32 *p) maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &verf_buf, &mic); if (maj_stat == GSS_S_CONTEXT_EXPIRED) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); if (maj_stat) goto out_bad; /* We leave it to unwrap to calculate au_rslack. For now we just @@ -927,7 +927,7 @@ gss_wrap_req_integ(struct rpc_cred *cred, struct gss_cl_ctx *ctx, maj_stat = gss_get_mic(ctx->gc_gss_ctx, &integ_buf, &mic); status = -EIO; /* XXX? */ if (maj_stat == GSS_S_CONTEXT_EXPIRED) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); else if (maj_stat) return status; q = xdr_encode_opaque(p, NULL, mic.len); @@ -1026,7 +1026,7 @@ gss_wrap_req_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx, /* We're assuming that when GSS_S_CONTEXT_EXPIRED, the encryption was * done anyway, so it's safe to put the request on the wire: */ if (maj_stat == GSS_S_CONTEXT_EXPIRED) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); else if (maj_stat) return status; @@ -1113,7 +1113,7 @@ gss_unwrap_resp_integ(struct rpc_cred *cred, struct gss_cl_ctx *ctx, maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &integ_buf, &mic); if (maj_stat == GSS_S_CONTEXT_EXPIRED) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); if (maj_stat != GSS_S_COMPLETE) return status; return 0; @@ -1138,7 +1138,7 @@ gss_unwrap_resp_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx, maj_stat = gss_unwrap(ctx->gc_gss_ctx, offset, rcv_buf); if (maj_stat == GSS_S_CONTEXT_EXPIRED) - cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; + clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); if (maj_stat != GSS_S_COMPLETE) return status; if (ntohl(*(*p)++) != rqstp->rq_seqno) diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c index fe9b6aaf91eb..6c905fb11c5d 100644 --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -76,7 +76,7 @@ nul_marshal(struct rpc_task *task, __be32 *p) static int nul_refresh(struct rpc_task *task) { - task->tk_msg.rpc_cred->cr_flags |= RPCAUTH_CRED_UPTODATE; + set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags); return 0; } @@ -136,7 +136,7 @@ struct rpc_cred null_cred = { .cr_auth = &null_auth, .cr_ops = &null_credops, .cr_count = ATOMIC_INIT(1), - .cr_flags = RPCAUTH_CRED_UPTODATE, + .cr_flags = 1UL << RPCAUTH_CRED_UPTODATE, #ifdef RPC_DEBUG .cr_magic = RPCAUTH_CRED_MAGIC, #endif diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index f17dabbab1c7..29d50ffa69d6 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -72,7 +72,7 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) return ERR_PTR(-ENOMEM); rpcauth_init_cred(&cred->uc_base, acred, auth, &unix_credops); - cred->uc_base.cr_flags = RPCAUTH_CRED_UPTODATE; + cred->uc_base.cr_flags = 1UL << RPCAUTH_CRED_UPTODATE; if (flags & RPCAUTH_LOOKUP_ROOTCREDS) { cred->uc_uid = 0; cred->uc_gid = 0; @@ -172,7 +172,7 @@ unx_marshal(struct rpc_task *task, __be32 *p) static int unx_refresh(struct rpc_task *task) { - task->tk_msg.rpc_cred->cr_flags |= RPCAUTH_CRED_UPTODATE; + set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags); return 0; } -- cgit v1.2.3 From e092bdcd939416ef911090890096fe07d0281a5e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2007 19:45:36 -0400 Subject: SUNRPC: cleanup rpc credential cache garbage collection Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 1 + net/sunrpc/auth.c | 121 ++++++++++++++++++++++++++------------------ net/sunrpc/auth_null.c | 1 + 3 files changed, 74 insertions(+), 49 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 8586503d5ebd..4e78f0c5f014 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -34,6 +34,7 @@ struct rpc_auth; struct rpc_credops; struct rpc_cred { struct hlist_node cr_hash; /* hash chain */ + struct list_head cr_lru; /* lru garbage collection */ struct rpc_auth * cr_auth; const struct rpc_credops *cr_ops; #ifdef RPC_DEBUG diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 4d7c78b05d1e..00f9649b0901 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -25,6 +25,8 @@ static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = { NULL, /* others can be loadable modules */ }; +static LIST_HEAD(cred_unused); + static u32 pseudoflavor_to_flavor(u32 flavor) { if (flavor >= RPC_AUTH_MAXFLAVOR) @@ -134,13 +136,13 @@ rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire) * Destroy a list of credentials */ static inline -void rpcauth_destroy_credlist(struct hlist_head *head) +void rpcauth_destroy_credlist(struct list_head *head) { struct rpc_cred *cred; - while (!hlist_empty(head)) { - cred = hlist_entry(head->first, struct rpc_cred, cr_hash); - hlist_del_init(&cred->cr_hash); + while (!list_empty(head)) { + cred = list_entry(head->next, struct rpc_cred, cr_lru); + list_del_init(&cred->cr_lru); put_rpccred(cred); } } @@ -152,17 +154,20 @@ void rpcauth_destroy_credlist(struct hlist_head *head) void rpcauth_clear_credcache(struct rpc_cred_cache *cache) { - HLIST_HEAD(free); - struct hlist_node *pos, *next; + LIST_HEAD(free); + struct hlist_head *head; struct rpc_cred *cred; int i; spin_lock(&rpc_credcache_lock); for (i = 0; i < RPC_CREDCACHE_NR; i++) { - hlist_for_each_safe(pos, next, &cache->hashtable[i]) { - cred = hlist_entry(pos, struct rpc_cred, cr_hash); - __hlist_del(&cred->cr_hash); - hlist_add_head(&cred->cr_hash, &free); + head = &cache->hashtable[i]; + while (!hlist_empty(head)) { + cred = hlist_entry(head->first, struct rpc_cred, cr_hash); + get_rpccred(cred); + list_move_tail(&cred->cr_lru, &free); + smp_wmb(); + hlist_del_init(&cred->cr_hash); } } spin_unlock(&rpc_credcache_lock); @@ -184,38 +189,39 @@ rpcauth_destroy_credcache(struct rpc_auth *auth) } } +/* + * Remove stale credentials. Avoid sleeping inside the loop. + */ static void -rpcauth_prune_expired(struct rpc_auth *auth, struct rpc_cred *cred, struct hlist_head *free) +rpcauth_prune_expired(struct list_head *free) { - if (atomic_read(&cred->cr_count) != 1) - return; - if (time_after(jiffies, cred->cr_expire + auth->au_credcache->expire)) - clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); - if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) { - __hlist_del(&cred->cr_hash); - hlist_add_head(&cred->cr_hash, free); + struct rpc_cred *cred; + + while (!list_empty(&cred_unused)) { + cred = list_entry(cred_unused.next, struct rpc_cred, cr_lru); + if (time_after(jiffies, cred->cr_expire + + cred->cr_auth->au_credcache->expire)) + break; + list_del_init(&cred->cr_lru); + if (atomic_read(&cred->cr_count) != 0) + continue; + get_rpccred(cred); + list_add_tail(&cred->cr_lru, free); + smp_wmb(); + hlist_del_init(&cred->cr_hash); } } /* - * Remove stale credentials. Avoid sleeping inside the loop. + * Run garbage collector. */ static void -rpcauth_gc_credcache(struct rpc_auth *auth, struct hlist_head *free) +rpcauth_gc_credcache(struct rpc_cred_cache *cache, struct list_head *free) { - struct rpc_cred_cache *cache = auth->au_credcache; - struct hlist_node *pos, *next; - struct rpc_cred *cred; - int i; - - dprintk("RPC: gc'ing RPC credentials for auth %p\n", auth); - for (i = 0; i < RPC_CREDCACHE_NR; i++) { - hlist_for_each_safe(pos, next, &cache->hashtable[i]) { - cred = hlist_entry(pos, struct rpc_cred, cr_hash); - rpcauth_prune_expired(auth, cred, free); - } - } + if (time_before(jiffies, cache->nextgc)) + return; cache->nextgc = jiffies + cache->expire; + rpcauth_prune_expired(free); } /* @@ -225,39 +231,35 @@ struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, int flags) { + LIST_HEAD(free); struct rpc_cred_cache *cache = auth->au_credcache; - HLIST_HEAD(free); - struct hlist_node *pos, *next; + struct hlist_node *pos; struct rpc_cred *new = NULL, - *cred = NULL; + *cred = NULL, + *entry; int nr = 0; if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS)) nr = acred->uid & RPC_CREDCACHE_MASK; retry: spin_lock(&rpc_credcache_lock); - if (time_before(cache->nextgc, jiffies)) - rpcauth_gc_credcache(auth, &free); - hlist_for_each_safe(pos, next, &cache->hashtable[nr]) { - struct rpc_cred *entry; - entry = hlist_entry(pos, struct rpc_cred, cr_hash); - if (entry->cr_ops->crmatch(acred, entry, flags)) { - hlist_del(&entry->cr_hash); - cred = entry; - break; - } - rpcauth_prune_expired(auth, entry, &free); + hlist_for_each_entry(entry, pos, &cache->hashtable[nr], cr_hash) { + if (!entry->cr_ops->crmatch(acred, entry, flags)) + continue; + cred = get_rpccred(entry); + hlist_del(&entry->cr_hash); + break; } if (new) { if (cred) - hlist_add_head(&new->cr_hash, &free); + list_add_tail(&new->cr_lru, &free); else cred = new; } if (cred) { hlist_add_head(&cred->cr_hash, &cache->hashtable[nr]); - get_rpccred(cred); } + rpcauth_gc_credcache(cache, &free); spin_unlock(&rpc_credcache_lock); rpcauth_destroy_credlist(&free); @@ -303,6 +305,7 @@ rpcauth_init_cred(struct rpc_cred *cred, const struct auth_cred *acred, struct rpc_auth *auth, const struct rpc_credops *ops) { INIT_HLIST_NODE(&cred->cr_hash); + INIT_LIST_HEAD(&cred->cr_lru); atomic_set(&cred->cr_count, 1); cred->cr_auth = auth; cred->cr_ops = ops; @@ -353,9 +356,29 @@ rpcauth_holdcred(struct rpc_task *task) void put_rpccred(struct rpc_cred *cred) { - cred->cr_expire = jiffies; + /* Fast path for unhashed credentials */ + if (!hlist_unhashed(&cred->cr_hash)) + goto need_lock; + if (!atomic_dec_and_test(&cred->cr_count)) return; + goto out_destroy; + +need_lock: + if (!atomic_dec_and_lock(&cred->cr_count, &rpc_credcache_lock)) + return; + if (!list_empty(&cred->cr_lru)) + list_del_init(&cred->cr_lru); + if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) + hlist_del(&cred->cr_hash); + else if (!hlist_unhashed(&cred->cr_hash)) { + cred->cr_expire = jiffies; + list_add_tail(&cred->cr_lru, &cred_unused); + spin_unlock(&rpc_credcache_lock); + return; + } + spin_unlock(&rpc_credcache_lock); +out_destroy: cred->cr_ops->crdestroy(cred); } diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c index 6c905fb11c5d..537d0e8589dd 100644 --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -133,6 +133,7 @@ const struct rpc_credops null_credops = { static struct rpc_cred null_cred = { + .cr_lru = LIST_HEAD_INIT(null_cred.cr_lru), .cr_auth = &null_auth, .cr_ops = &null_credops, .cr_count = ATOMIC_INIT(1), -- cgit v1.2.3 From 31be5bf15f3dafffce110eb1afadccbf2e3067b4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 24 Jun 2007 15:55:26 -0400 Subject: SUNRPC: Convert the credcache lookup code to use RCU Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 3 ++ net/sunrpc/auth.c | 91 ++++++++++++++++++++++++++---------------- net/sunrpc/auth_gss/auth_gss.c | 22 +++++++--- net/sunrpc/auth_unix.c | 18 +++++++-- 4 files changed, 91 insertions(+), 43 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 4e78f0c5f014..5974e8a493c4 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -16,6 +16,7 @@ #include #include +#include /* size of the nodename buffer */ #define UNX_MAXNODENAME 32 @@ -35,6 +36,7 @@ struct rpc_credops; struct rpc_cred { struct hlist_node cr_hash; /* hash chain */ struct list_head cr_lru; /* lru garbage collection */ + struct rcu_head cr_rcu; struct rpc_auth * cr_auth; const struct rpc_credops *cr_ops; #ifdef RPC_DEBUG @@ -50,6 +52,7 @@ struct rpc_cred { }; #define RPCAUTH_CRED_NEW 0 #define RPCAUTH_CRED_UPTODATE 1 +#define RPCAUTH_CRED_HASHED 2 #define RPCAUTH_CRED_MAGIC 0x0f4aa4f0 diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 00f9649b0901..ad7bde2c437e 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -112,6 +112,14 @@ rpcauth_release(struct rpc_auth *auth) static DEFINE_SPINLOCK(rpc_credcache_lock); +static void +rpcauth_unhash_cred_locked(struct rpc_cred *cred) +{ + hlist_del_rcu(&cred->cr_hash); + smp_mb__before_clear_bit(); + clear_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags); +} + /* * Initialize RPC credential cache */ @@ -166,8 +174,7 @@ rpcauth_clear_credcache(struct rpc_cred_cache *cache) cred = hlist_entry(head->first, struct rpc_cred, cr_hash); get_rpccred(cred); list_move_tail(&cred->cr_lru, &free); - smp_wmb(); - hlist_del_init(&cred->cr_hash); + rpcauth_unhash_cred_locked(cred); } } spin_unlock(&rpc_credcache_lock); @@ -207,8 +214,7 @@ rpcauth_prune_expired(struct list_head *free) continue; get_rpccred(cred); list_add_tail(&cred->cr_lru, free); - smp_wmb(); - hlist_del_init(&cred->cr_hash); + rpcauth_unhash_cred_locked(cred); } } @@ -218,10 +224,12 @@ rpcauth_prune_expired(struct list_head *free) static void rpcauth_gc_credcache(struct rpc_cred_cache *cache, struct list_head *free) { - if (time_before(jiffies, cache->nextgc)) + if (list_empty(&cred_unused) || time_before(jiffies, cache->nextgc)) return; + spin_lock(&rpc_credcache_lock); cache->nextgc = jiffies + cache->expire; rpcauth_prune_expired(free); + spin_unlock(&rpc_credcache_lock); } /* @@ -234,42 +242,57 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, LIST_HEAD(free); struct rpc_cred_cache *cache = auth->au_credcache; struct hlist_node *pos; - struct rpc_cred *new = NULL, - *cred = NULL, - *entry; + struct rpc_cred *cred = NULL, + *entry, *new; int nr = 0; if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS)) nr = acred->uid & RPC_CREDCACHE_MASK; -retry: - spin_lock(&rpc_credcache_lock); - hlist_for_each_entry(entry, pos, &cache->hashtable[nr], cr_hash) { + + rcu_read_lock(); + hlist_for_each_entry_rcu(entry, pos, &cache->hashtable[nr], cr_hash) { if (!entry->cr_ops->crmatch(acred, entry, flags)) continue; + spin_lock(&rpc_credcache_lock); + if (test_bit(RPCAUTH_CRED_HASHED, &entry->cr_flags) == 0) { + spin_unlock(&rpc_credcache_lock); + continue; + } cred = get_rpccred(entry); - hlist_del(&entry->cr_hash); + spin_unlock(&rpc_credcache_lock); break; } - if (new) { - if (cred) - list_add_tail(&new->cr_lru, &free); - else - cred = new; - } - if (cred) { - hlist_add_head(&cred->cr_hash, &cache->hashtable[nr]); + rcu_read_unlock(); + + if (cred != NULL) { + rpcauth_gc_credcache(cache, &free); + goto found; } - rpcauth_gc_credcache(cache, &free); - spin_unlock(&rpc_credcache_lock); - rpcauth_destroy_credlist(&free); + new = auth->au_ops->crcreate(auth, acred, flags); + if (IS_ERR(new)) { + cred = new; + goto out; + } - if (!cred) { - new = auth->au_ops->crcreate(auth, acred, flags); - if (!IS_ERR(new)) - goto retry; + spin_lock(&rpc_credcache_lock); + hlist_for_each_entry(entry, pos, &cache->hashtable[nr], cr_hash) { + if (!entry->cr_ops->crmatch(acred, entry, flags)) + continue; + cred = get_rpccred(entry); + break; + } + if (cred == NULL) { cred = new; - } else if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags) + set_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags); + hlist_add_head_rcu(&cred->cr_hash, &cache->hashtable[nr]); + } else + list_add_tail(&new->cr_lru, &free); + rpcauth_prune_expired(&free); + cache->nextgc = jiffies + cache->expire; + spin_unlock(&rpc_credcache_lock); +found: + if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags) && cred->cr_ops->cr_init != NULL && !(flags & RPCAUTH_LOOKUP_NEW)) { int res = cred->cr_ops->cr_init(auth, cred); @@ -278,8 +301,9 @@ retry: cred = ERR_PTR(res); } } - - return (struct rpc_cred *) cred; + rpcauth_destroy_credlist(&free); +out: + return cred; } struct rpc_cred * @@ -357,21 +381,20 @@ void put_rpccred(struct rpc_cred *cred) { /* Fast path for unhashed credentials */ - if (!hlist_unhashed(&cred->cr_hash)) + if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0) goto need_lock; if (!atomic_dec_and_test(&cred->cr_count)) return; goto out_destroy; - need_lock: if (!atomic_dec_and_lock(&cred->cr_count, &rpc_credcache_lock)) return; if (!list_empty(&cred->cr_lru)) list_del_init(&cred->cr_lru); if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) - hlist_del(&cred->cr_hash); - else if (!hlist_unhashed(&cred->cr_hash)) { + rpcauth_unhash_cred_locked(cred); + else if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0) { cred->cr_expire = jiffies; list_add_tail(&cred->cr_lru, &cred_unused); spin_unlock(&rpc_credcache_lock); diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 55c47ae0a258..068fa6dfb64e 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -694,15 +694,25 @@ gss_destroy_ctx(struct gss_cl_ctx *ctx) } static void -gss_destroy_cred(struct rpc_cred *rc) +gss_free_cred(struct gss_cred *gss_cred) { - struct gss_cred *cred = container_of(rc, struct gss_cred, gc_base); + dprintk("RPC: gss_free_cred %p\n", gss_cred); + if (gss_cred->gc_ctx) + gss_put_ctx(gss_cred->gc_ctx); + kfree(gss_cred); +} - dprintk("RPC: gss_destroy_cred \n"); +static void +gss_free_cred_callback(struct rcu_head *head) +{ + struct gss_cred *gss_cred = container_of(head, struct gss_cred, gc_base.cr_rcu); + gss_free_cred(gss_cred); +} - if (cred->gc_ctx) - gss_put_ctx(cred->gc_ctx); - kfree(cred); +static void +gss_destroy_cred(struct rpc_cred *cred) +{ + call_rcu(&cred->cr_rcu, gss_free_cred_callback); } /* diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 29d50ffa69d6..f7ff6ad3259e 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -93,11 +93,23 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags) } static void -unx_destroy_cred(struct rpc_cred *rcred) +unx_free_cred(struct unx_cred *unx_cred) { - struct unx_cred *cred = container_of(rcred, struct unx_cred, uc_base); + dprintk("RPC: unx_free_cred %p\n", unx_cred); + kfree(unx_cred); +} + +static void +unx_free_cred_callback(struct rcu_head *head) +{ + struct unx_cred *unx_cred = container_of(head, struct unx_cred, uc_base.cr_rcu); + unx_free_cred(unx_cred); +} - kfree(cred); +static void +unx_destroy_cred(struct rpc_cred *cred) +{ + call_rcu(&cred->cr_rcu, unx_free_cred_callback); } /* -- cgit v1.2.3 From 9499b4341b56935f61af9e7e354e7d11e70f5258 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 24 Jun 2007 15:57:57 -0400 Subject: SUNRPC: Give credential cache a local spinlock Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 3 +++ net/sunrpc/auth.c | 46 +++++++++++++++++++++++++++++++-------------- net/sunrpc/auth_unix.c | 5 +++++ net/sunrpc/sunrpc_syms.c | 1 + 4 files changed, 41 insertions(+), 14 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 5974e8a493c4..e5a3b5141ed2 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -63,6 +63,7 @@ struct rpc_cred { #define RPC_CREDCACHE_MASK (RPC_CREDCACHE_NR - 1) struct rpc_cred_cache { struct hlist_head hashtable[RPC_CREDCACHE_NR]; + spinlock_t lock; unsigned long nextgc; /* next garbage collection */ unsigned long expire; /* cache expiry interval */ }; @@ -126,6 +127,8 @@ struct rpc_credops { extern const struct rpc_authops authunix_ops; extern const struct rpc_authops authnull_ops; +void __init rpc_init_authunix(void); + int rpcauth_register(const struct rpc_authops *); int rpcauth_unregister(const struct rpc_authops *); struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *); diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index ad7bde2c437e..cf1198d10ee9 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -120,6 +120,18 @@ rpcauth_unhash_cred_locked(struct rpc_cred *cred) clear_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags); } +static void +rpcauth_unhash_cred(struct rpc_cred *cred) +{ + spinlock_t *cache_lock; + + cache_lock = &cred->cr_auth->au_credcache->lock; + spin_lock(cache_lock); + if (atomic_read(&cred->cr_count) == 0) + rpcauth_unhash_cred_locked(cred); + spin_unlock(cache_lock); +} + /* * Initialize RPC credential cache */ @@ -134,6 +146,7 @@ rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire) return -ENOMEM; for (i = 0; i < RPC_CREDCACHE_NR; i++) INIT_HLIST_HEAD(&new->hashtable[i]); + spin_lock_init(&new->lock); new->expire = expire; new->nextgc = jiffies + (expire >> 1); auth->au_credcache = new; @@ -168,6 +181,7 @@ rpcauth_clear_credcache(struct rpc_cred_cache *cache) int i; spin_lock(&rpc_credcache_lock); + spin_lock(&cache->lock); for (i = 0; i < RPC_CREDCACHE_NR; i++) { head = &cache->hashtable[i]; while (!hlist_empty(head)) { @@ -177,6 +191,7 @@ rpcauth_clear_credcache(struct rpc_cred_cache *cache) rpcauth_unhash_cred_locked(cred); } } + spin_unlock(&cache->lock); spin_unlock(&rpc_credcache_lock); rpcauth_destroy_credlist(&free); } @@ -202,6 +217,7 @@ rpcauth_destroy_credcache(struct rpc_auth *auth) static void rpcauth_prune_expired(struct list_head *free) { + spinlock_t *cache_lock; struct rpc_cred *cred; while (!list_empty(&cred_unused)) { @@ -212,9 +228,14 @@ rpcauth_prune_expired(struct list_head *free) list_del_init(&cred->cr_lru); if (atomic_read(&cred->cr_count) != 0) continue; - get_rpccred(cred); - list_add_tail(&cred->cr_lru, free); - rpcauth_unhash_cred_locked(cred); + cache_lock = &cred->cr_auth->au_credcache->lock; + spin_lock(cache_lock); + if (atomic_read(&cred->cr_count) == 0) { + get_rpccred(cred); + list_add_tail(&cred->cr_lru, free); + rpcauth_unhash_cred_locked(cred); + } + spin_unlock(cache_lock); } } @@ -253,21 +274,19 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, hlist_for_each_entry_rcu(entry, pos, &cache->hashtable[nr], cr_hash) { if (!entry->cr_ops->crmatch(acred, entry, flags)) continue; - spin_lock(&rpc_credcache_lock); + spin_lock(&cache->lock); if (test_bit(RPCAUTH_CRED_HASHED, &entry->cr_flags) == 0) { - spin_unlock(&rpc_credcache_lock); + spin_unlock(&cache->lock); continue; } cred = get_rpccred(entry); - spin_unlock(&rpc_credcache_lock); + spin_unlock(&cache->lock); break; } rcu_read_unlock(); - if (cred != NULL) { - rpcauth_gc_credcache(cache, &free); + if (cred != NULL) goto found; - } new = auth->au_ops->crcreate(auth, acred, flags); if (IS_ERR(new)) { @@ -275,7 +294,7 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, goto out; } - spin_lock(&rpc_credcache_lock); + spin_lock(&cache->lock); hlist_for_each_entry(entry, pos, &cache->hashtable[nr], cr_hash) { if (!entry->cr_ops->crmatch(acred, entry, flags)) continue; @@ -288,9 +307,7 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, hlist_add_head_rcu(&cred->cr_hash, &cache->hashtable[nr]); } else list_add_tail(&new->cr_lru, &free); - rpcauth_prune_expired(&free); - cache->nextgc = jiffies + cache->expire; - spin_unlock(&rpc_credcache_lock); + spin_unlock(&cache->lock); found: if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags) && cred->cr_ops->cr_init != NULL @@ -301,6 +318,7 @@ found: cred = ERR_PTR(res); } } + rpcauth_gc_credcache(cache, &free); rpcauth_destroy_credlist(&free); out: return cred; @@ -393,7 +411,7 @@ need_lock: if (!list_empty(&cred->cr_lru)) list_del_init(&cred->cr_lru); if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) - rpcauth_unhash_cred_locked(cred); + rpcauth_unhash_cred(cred); else if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0) { cred->cr_expire = jiffies; list_add_tail(&cred->cr_lru, &cred_unused); diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index f7ff6ad3259e..205878a3caa5 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -213,6 +213,11 @@ unx_validate(struct rpc_task *task, __be32 *p) return p; } +void __init rpc_init_authunix(void) +{ + spin_lock_init(&unix_cred_cache.lock); +} + const struct rpc_authops authunix_ops = { .owner = THIS_MODULE, .au_flavor = RPC_AUTH_UNIX, diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index 3e19e7af6799..018065fca84b 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -152,6 +152,7 @@ init_sunrpc(void) cache_register(&ip_map_cache); cache_register(&unix_gid_cache); init_socket_xprt(); + rpc_init_authunix(); out: return err; } -- cgit v1.2.3 From f5c2187cfef628784d8a09b6d0f77888246d0c0f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 25 Jun 2007 17:11:20 -0400 Subject: SUNRPC: Convert the credential garbage collector into a shrinker callback Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 6 ++-- net/sunrpc/auth.c | 63 ++++++++++++++++++++++++++++++------------ net/sunrpc/auth_gss/auth_gss.c | 3 +- net/sunrpc/auth_unix.c | 6 +--- net/sunrpc/sunrpc_syms.c | 3 +- 5 files changed, 52 insertions(+), 29 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index e5a3b5141ed2..7a69ca3bebaf 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -64,8 +64,6 @@ struct rpc_cred { struct rpc_cred_cache { struct hlist_head hashtable[RPC_CREDCACHE_NR]; spinlock_t lock; - unsigned long nextgc; /* next garbage collection */ - unsigned long expire; /* cache expiry interval */ }; struct rpc_authops; @@ -128,6 +126,8 @@ extern const struct rpc_authops authunix_ops; extern const struct rpc_authops authnull_ops; void __init rpc_init_authunix(void); +void __init rpcauth_init_module(void); +void __exit rpcauth_remove_module(void); int rpcauth_register(const struct rpc_authops *); int rpcauth_unregister(const struct rpc_authops *); @@ -147,7 +147,7 @@ int rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp, int rpcauth_refreshcred(struct rpc_task *); void rpcauth_invalcred(struct rpc_task *); int rpcauth_uptodatecred(struct rpc_task *); -int rpcauth_init_credcache(struct rpc_auth *, unsigned long); +int rpcauth_init_credcache(struct rpc_auth *); void rpcauth_destroy_credcache(struct rpc_auth *); void rpcauth_clear_credcache(struct rpc_cred_cache *); diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index cf1198d10ee9..81f4c776c558 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -26,6 +26,7 @@ static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = { }; static LIST_HEAD(cred_unused); +static unsigned long number_cred_unused; static u32 pseudoflavor_to_flavor(u32 flavor) { @@ -136,7 +137,7 @@ rpcauth_unhash_cred(struct rpc_cred *cred) * Initialize RPC credential cache */ int -rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire) +rpcauth_init_credcache(struct rpc_auth *auth) { struct rpc_cred_cache *new; int i; @@ -147,8 +148,6 @@ rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire) for (i = 0; i < RPC_CREDCACHE_NR; i++) INIT_HLIST_HEAD(&new->hashtable[i]); spin_lock_init(&new->lock); - new->expire = expire; - new->nextgc = jiffies + (expire >> 1); auth->au_credcache = new; return 0; } @@ -187,7 +186,11 @@ rpcauth_clear_credcache(struct rpc_cred_cache *cache) while (!hlist_empty(head)) { cred = hlist_entry(head->first, struct rpc_cred, cr_hash); get_rpccred(cred); - list_move_tail(&cred->cr_lru, &free); + if (!list_empty(&cred->cr_lru)) { + list_del(&cred->cr_lru); + number_cred_unused--; + } + list_add_tail(&cred->cr_lru, &free); rpcauth_unhash_cred_locked(cred); } } @@ -214,18 +217,16 @@ rpcauth_destroy_credcache(struct rpc_auth *auth) /* * Remove stale credentials. Avoid sleeping inside the loop. */ -static void -rpcauth_prune_expired(struct list_head *free) +static int +rpcauth_prune_expired(struct list_head *free, int nr_to_scan) { spinlock_t *cache_lock; struct rpc_cred *cred; while (!list_empty(&cred_unused)) { cred = list_entry(cred_unused.next, struct rpc_cred, cr_lru); - if (time_after(jiffies, cred->cr_expire + - cred->cr_auth->au_credcache->expire)) - break; list_del_init(&cred->cr_lru); + number_cred_unused--; if (atomic_read(&cred->cr_count) != 0) continue; cache_lock = &cred->cr_auth->au_credcache->lock; @@ -234,23 +235,32 @@ rpcauth_prune_expired(struct list_head *free) get_rpccred(cred); list_add_tail(&cred->cr_lru, free); rpcauth_unhash_cred_locked(cred); + nr_to_scan--; } spin_unlock(cache_lock); + if (nr_to_scan == 0) + break; } + return nr_to_scan; } /* - * Run garbage collector. + * Run memory cache shrinker. */ -static void -rpcauth_gc_credcache(struct rpc_cred_cache *cache, struct list_head *free) +static int +rpcauth_cache_shrinker(int nr_to_scan, gfp_t gfp_mask) { - if (list_empty(&cred_unused) || time_before(jiffies, cache->nextgc)) - return; + LIST_HEAD(free); + int res; + + if (list_empty(&cred_unused)) + return 0; spin_lock(&rpc_credcache_lock); - cache->nextgc = jiffies + cache->expire; - rpcauth_prune_expired(free); + nr_to_scan = rpcauth_prune_expired(&free, nr_to_scan); + res = (number_cred_unused / 100) * sysctl_vfs_cache_pressure; spin_unlock(&rpc_credcache_lock); + rpcauth_destroy_credlist(&free); + return res; } /* @@ -318,7 +328,6 @@ found: cred = ERR_PTR(res); } } - rpcauth_gc_credcache(cache, &free); rpcauth_destroy_credlist(&free); out: return cred; @@ -408,13 +417,16 @@ put_rpccred(struct rpc_cred *cred) need_lock: if (!atomic_dec_and_lock(&cred->cr_count, &rpc_credcache_lock)) return; - if (!list_empty(&cred->cr_lru)) + if (!list_empty(&cred->cr_lru)) { + number_cred_unused--; list_del_init(&cred->cr_lru); + } if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0) rpcauth_unhash_cred(cred); else if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0) { cred->cr_expire = jiffies; list_add_tail(&cred->cr_lru, &cred_unused); + number_cred_unused++; spin_unlock(&rpc_credcache_lock); return; } @@ -520,3 +532,18 @@ rpcauth_uptodatecred(struct rpc_task *task) return cred == NULL || test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) != 0; } + + +static struct shrinker *rpc_cred_shrinker; + +void __init rpcauth_init_module(void) +{ + rpc_init_authunix(); + rpc_cred_shrinker = set_shrinker(DEFAULT_SEEKS, rpcauth_cache_shrinker); +} + +void __exit rpcauth_remove_module(void) +{ + if (rpc_cred_shrinker != NULL) + remove_shrinker(rpc_cred_shrinker); +} diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 068fa6dfb64e..8653a92144ae 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -64,7 +64,6 @@ static const struct rpc_credops gss_credops; #define NFS_NGROUPS 16 -#define GSS_CRED_EXPIRE (60 * HZ) /* XXX: reasonable? */ #define GSS_CRED_SLACK 1024 /* XXX: unused */ /* length of a krb5 verifier (48), plus data added before arguments when * using integrity (two 4-byte integers): */ @@ -643,7 +642,7 @@ gss_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor) goto err_put_mech; } - err = rpcauth_init_credcache(auth, GSS_CRED_EXPIRE); + err = rpcauth_init_credcache(auth); if (err) goto err_unlink_pipe; diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 205878a3caa5..d9c50d810d15 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -21,8 +21,6 @@ struct unx_cred { }; #define uc_uid uc_base.cr_uid -#define UNX_CRED_EXPIRE (60 * HZ) - #define UNX_WRITESLACK (21 + (UNX_MAXNODENAME >> 2)) #ifdef RPC_DEBUG @@ -38,8 +36,7 @@ unx_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor) { dprintk("RPC: creating UNIX authenticator for client %p\n", clnt); - if (atomic_inc_return(&unix_auth.au_count) == 1) - unix_cred_cache.nextgc = jiffies + (unix_cred_cache.expire >> 1); + atomic_inc(&unix_auth.au_count); return &unix_auth; } @@ -232,7 +229,6 @@ const struct rpc_authops authunix_ops = { static struct rpc_cred_cache unix_cred_cache = { - .expire = UNX_CRED_EXPIRE, }; static diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index 018065fca84b..384c4ad5ab86 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -152,7 +152,7 @@ init_sunrpc(void) cache_register(&ip_map_cache); cache_register(&unix_gid_cache); init_socket_xprt(); - rpc_init_authunix(); + rpcauth_init_module(); out: return err; } @@ -160,6 +160,7 @@ out: static void __exit cleanup_sunrpc(void) { + rpcauth_remove_module(); cleanup_socket_xprt(); unregister_rpc_pipefs(); rpc_destroy_mempool(); -- cgit v1.2.3 From 5d28dc82074f1e64b22c9424b161abc1f5d6bcdb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 26 Jun 2007 19:18:38 -0400 Subject: SUNRPC: Convert gss_ctx_lock to an RCU lock Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth_gss.h | 1 + net/sunrpc/auth_gss/auth_gss.c | 53 ++++++++++++++++++++++++++++------------- 2 files changed, 37 insertions(+), 17 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/auth_gss.h b/include/linux/sunrpc/auth_gss.h index 0bd1d06777b9..67658e17a375 100644 --- a/include/linux/sunrpc/auth_gss.h +++ b/include/linux/sunrpc/auth_gss.h @@ -75,6 +75,7 @@ struct gss_cl_ctx { struct xdr_netobj gc_wire_ctx; u32 gc_win; unsigned long gc_expiry; + struct rcu_head gc_rcu; }; struct gss_upcall_msg; diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 8653a92144ae..15da6f82db36 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -78,8 +78,6 @@ static const struct rpc_credops gss_credops; /* dump the buffer in `emacs-hexl' style */ #define isprint(c) ((c > 0x1f) && (c < 0x7f)) -static DEFINE_RWLOCK(gss_ctx_lock); - struct gss_auth { struct rpc_auth rpc_auth; struct gss_api_mech *mech; @@ -88,7 +86,7 @@ struct gss_auth { struct dentry *dentry; }; -static void gss_destroy_ctx(struct gss_cl_ctx *); +static void gss_free_ctx(struct gss_cl_ctx *); static struct rpc_pipe_ops gss_upcall_ops; static inline struct gss_cl_ctx * @@ -102,20 +100,24 @@ static inline void gss_put_ctx(struct gss_cl_ctx *ctx) { if (atomic_dec_and_test(&ctx->count)) - gss_destroy_ctx(ctx); + gss_free_ctx(ctx); } +/* gss_cred_set_ctx: + * called by gss_upcall_callback and gss_create_upcall in order + * to set the gss context. The actual exchange of an old context + * and a new one is protected by the inode->i_lock. + */ static void gss_cred_set_ctx(struct rpc_cred *cred, struct gss_cl_ctx *ctx) { struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); struct gss_cl_ctx *old; - write_lock(&gss_ctx_lock); + old = gss_cred->gc_ctx; - gss_cred->gc_ctx = ctx; + rcu_assign_pointer(gss_cred->gc_ctx, ctx); set_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); clear_bit(RPCAUTH_CRED_NEW, &cred->cr_flags); - write_unlock(&gss_ctx_lock); if (old) gss_put_ctx(old); } @@ -126,10 +128,10 @@ gss_cred_is_uptodate_ctx(struct rpc_cred *cred) struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); int res = 0; - read_lock(&gss_ctx_lock); + rcu_read_lock(); if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) && gss_cred->gc_ctx) res = 1; - read_unlock(&gss_ctx_lock); + rcu_read_unlock(); return res; } @@ -168,10 +170,10 @@ gss_cred_get_ctx(struct rpc_cred *cred) struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); struct gss_cl_ctx *ctx = NULL; - read_lock(&gss_ctx_lock); + rcu_read_lock(); if (gss_cred->gc_ctx) ctx = gss_get_ctx(gss_cred->gc_ctx); - read_unlock(&gss_ctx_lock); + rcu_read_unlock(); return ctx; } @@ -333,11 +335,11 @@ gss_upcall_callback(struct rpc_task *task) struct gss_upcall_msg *gss_msg = gss_cred->gc_upcall; struct inode *inode = gss_msg->auth->dentry->d_inode; + spin_lock(&inode->i_lock); if (gss_msg->ctx) gss_cred_set_ctx(task->tk_msg.rpc_cred, gss_get_ctx(gss_msg->ctx)); else task->tk_status = gss_msg->msg.errno; - spin_lock(&inode->i_lock); gss_cred->gc_upcall = NULL; rpc_wake_up_status(&gss_msg->rpc_waitqueue, gss_msg->msg.errno); spin_unlock(&inode->i_lock); @@ -440,7 +442,6 @@ gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred) prepare_to_wait(&gss_msg->waitqueue, &wait, TASK_INTERRUPTIBLE); spin_lock(&inode->i_lock); if (gss_msg->ctx != NULL || gss_msg->msg.errno < 0) { - spin_unlock(&inode->i_lock); break; } spin_unlock(&inode->i_lock); @@ -454,6 +455,7 @@ gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred) gss_cred_set_ctx(cred, gss_get_ctx(gss_msg->ctx)); else err = gss_msg->msg.errno; + spin_unlock(&inode->i_lock); out_intr: finish_wait(&gss_msg->waitqueue, &wait); gss_release_msg(gss_msg); @@ -681,9 +683,9 @@ gss_destroy(struct rpc_auth *auth) * to create a new cred or context, so they check that things have been * allocated before freeing them. */ static void -gss_destroy_ctx(struct gss_cl_ctx *ctx) +gss_do_free_ctx(struct gss_cl_ctx *ctx) { - dprintk("RPC: gss_destroy_ctx\n"); + dprintk("RPC: gss_free_ctx\n"); if (ctx->gc_gss_ctx) gss_delete_sec_context(&ctx->gc_gss_ctx); @@ -692,12 +694,23 @@ gss_destroy_ctx(struct gss_cl_ctx *ctx) kfree(ctx); } +static void +gss_free_ctx_callback(struct rcu_head *head) +{ + struct gss_cl_ctx *ctx = container_of(head, struct gss_cl_ctx, gc_rcu); + gss_do_free_ctx(ctx); +} + +static void +gss_free_ctx(struct gss_cl_ctx *ctx) +{ + call_rcu(&ctx->gc_rcu, gss_free_ctx_callback); +} + static void gss_free_cred(struct gss_cred *gss_cred) { dprintk("RPC: gss_free_cred %p\n", gss_cred); - if (gss_cred->gc_ctx) - gss_put_ctx(gss_cred->gc_ctx); kfree(gss_cred); } @@ -711,7 +724,13 @@ gss_free_cred_callback(struct rcu_head *head) static void gss_destroy_cred(struct rpc_cred *cred) { + struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); + struct gss_cl_ctx *ctx = gss_cred->gc_ctx; + + rcu_assign_pointer(gss_cred->gc_ctx, NULL); call_rcu(&cred->cr_rcu, gss_free_cred_callback); + if (ctx) + gss_put_ctx(ctx); } /* -- cgit v1.2.3 From 1be27f36601973815171db684c711d30557cf50c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 27 Jun 2007 14:29:04 -0400 Subject: SUNRPC: Remove the tk_auth macro... We should almost always be deferencing the rpc_auth struct by means of the credential's cr_auth field instead of the rpc_clnt->cl_auth anyway. Fix up that historical mistake, and remove the macro that propagated it. Signed-off-by: Trond Myklebust --- fs/nfs/nfs2xdr.c | 6 +++--- fs/nfs/nfs3xdr.c | 8 ++++---- fs/nfs/nfs4xdr.c | 10 +++++----- include/linux/sunrpc/sched.h | 1 - net/sunrpc/auth.c | 25 +++++++++++++------------ net/sunrpc/auth_gss/auth_gss.c | 4 ++-- net/sunrpc/auth_unix.c | 2 +- net/sunrpc/clnt.c | 2 +- 8 files changed, 29 insertions(+), 29 deletions(-) (limited to 'include') diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c index cd3ca7b5d3db..7fcc78f2aa71 100644 --- a/fs/nfs/nfs2xdr.c +++ b/fs/nfs/nfs2xdr.c @@ -223,7 +223,7 @@ nfs_xdr_diropargs(struct rpc_rqst *req, __be32 *p, struct nfs_diropargs *args) static int nfs_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; u32 offset = (u32)args->offset; u32 count = args->count; @@ -380,7 +380,7 @@ static int nfs_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs_readdirargs *args) { struct rpc_task *task = req->rq_task; - struct rpc_auth *auth = task->tk_auth; + struct rpc_auth *auth = task->tk_msg.rpc_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -541,7 +541,7 @@ nfs_xdr_diropres(struct rpc_rqst *req, __be32 *p, struct nfs_diropok *res) static int nfs_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs_readlinkargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index b51df8eb9f01..b4647a22f349 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -319,7 +319,7 @@ nfs3_xdr_accessargs(struct rpc_rqst *req, __be32 *p, struct nfs3_accessargs *arg static int nfs3_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -458,7 +458,7 @@ nfs3_xdr_linkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_linkargs *args) static int nfs3_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readdirargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -643,7 +643,7 @@ static int nfs3_xdr_getaclargs(struct rpc_rqst *req, __be32 *p, struct nfs3_getaclargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); @@ -773,7 +773,7 @@ nfs3_xdr_accessres(struct rpc_rqst *req, __be32 *p, struct nfs3_accessres *res) static int nfs3_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readlinkargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 1fcca516e6ee..859b13633258 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1071,7 +1071,7 @@ static int encode_read(struct xdr_stream *xdr, const struct nfs_readargs *args) static int encode_readdir(struct xdr_stream *xdr, const struct nfs4_readdir_arg *readdir, struct rpc_rqst *req) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; uint32_t attrs[2] = { FATTR4_WORD0_RDATTR_ERROR|FATTR4_WORD0_FILEID, FATTR4_WORD1_MOUNTED_ON_FILEID, @@ -1117,7 +1117,7 @@ static int encode_readdir(struct xdr_stream *xdr, const struct nfs4_readdir_arg static int encode_readlink(struct xdr_stream *xdr, const struct nfs4_readlink *readlink, struct rpc_rqst *req) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; unsigned int replen; __be32 *p; @@ -1735,7 +1735,7 @@ out: */ static int nfs4_xdr_enc_read(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) { - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; struct xdr_stream xdr; struct compound_hdr hdr = { .nops = 2, @@ -1795,7 +1795,7 @@ nfs4_xdr_enc_getacl(struct rpc_rqst *req, __be32 *p, struct nfs_getaclargs *args) { struct xdr_stream xdr; - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; struct compound_hdr hdr = { .nops = 2, }; @@ -2030,7 +2030,7 @@ static int nfs4_xdr_enc_fs_locations(struct rpc_rqst *req, __be32 *p, struct nfs struct compound_hdr hdr = { .nops = 3, }; - struct rpc_auth *auth = req->rq_task->tk_auth; + struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; int replen; int status; diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h index 3387b008cdfc..8ea077db0099 100644 --- a/include/linux/sunrpc/sched.h +++ b/include/linux/sunrpc/sched.h @@ -98,7 +98,6 @@ struct rpc_task { unsigned short tk_pid; /* debugging aid */ #endif }; -#define tk_auth tk_client->cl_auth #define tk_xprt tk_client->cl_xprt /* support walking a list of tasks on a wait queue */ diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 81f4c776c558..74baf87ccff9 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -371,7 +371,7 @@ EXPORT_SYMBOL(rpcauth_init_cred); struct rpc_cred * rpcauth_bindcred(struct rpc_task *task) { - struct rpc_auth *auth = task->tk_auth; + struct rpc_auth *auth = task->tk_client->cl_auth; struct auth_cred acred = { .uid = current->fsuid, .gid = current->fsgid, @@ -381,7 +381,7 @@ rpcauth_bindcred(struct rpc_task *task) int flags = 0; dprintk("RPC: %5u looking up %s cred\n", - task->tk_pid, task->tk_auth->au_ops->au_name); + task->tk_pid, task->tk_client->cl_auth->au_ops->au_name); get_group_info(acred.group_info); if (task->tk_flags & RPC_TASK_ROOTCREDS) flags |= RPCAUTH_LOOKUP_ROOTCREDS; @@ -397,11 +397,12 @@ rpcauth_bindcred(struct rpc_task *task) void rpcauth_holdcred(struct rpc_task *task) { - dprintk("RPC: %5u holding %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, - task->tk_msg.rpc_cred); - if (task->tk_msg.rpc_cred) - get_rpccred(task->tk_msg.rpc_cred); + struct rpc_cred *cred = task->tk_msg.rpc_cred; + if (cred != NULL) { + get_rpccred(cred); + dprintk("RPC: %5u holding %s cred %p\n", task->tk_pid, + cred->cr_auth->au_ops->au_name, cred); + } } void @@ -441,7 +442,7 @@ rpcauth_unbindcred(struct rpc_task *task) struct rpc_cred *cred = task->tk_msg.rpc_cred; dprintk("RPC: %5u releasing %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, cred); + task->tk_pid, cred->cr_auth->au_ops->au_name, cred); put_rpccred(cred); task->tk_msg.rpc_cred = NULL; @@ -453,7 +454,7 @@ rpcauth_marshcred(struct rpc_task *task, __be32 *p) struct rpc_cred *cred = task->tk_msg.rpc_cred; dprintk("RPC: %5u marshaling %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, cred); + task->tk_pid, cred->cr_auth->au_ops->au_name, cred); return cred->cr_ops->crmarshal(task, p); } @@ -464,7 +465,7 @@ rpcauth_checkverf(struct rpc_task *task, __be32 *p) struct rpc_cred *cred = task->tk_msg.rpc_cred; dprintk("RPC: %5u validating %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, cred); + task->tk_pid, cred->cr_auth->au_ops->au_name, cred); return cred->cr_ops->crvalidate(task, p); } @@ -505,7 +506,7 @@ rpcauth_refreshcred(struct rpc_task *task) int err; dprintk("RPC: %5u refreshing %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, cred); + task->tk_pid, cred->cr_auth->au_ops->au_name, cred); err = cred->cr_ops->crrefresh(task); if (err < 0) @@ -519,7 +520,7 @@ rpcauth_invalcred(struct rpc_task *task) struct rpc_cred *cred = task->tk_msg.rpc_cred; dprintk("RPC: %5u invalidating %s cred %p\n", - task->tk_pid, task->tk_auth->au_ops->au_name, cred); + task->tk_pid, cred->cr_auth->au_ops->au_name, cred); if (cred) clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags); } diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 15da6f82db36..debcda86467c 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -906,7 +906,7 @@ gss_validate(struct rpc_task *task, __be32 *p) goto out_bad; /* We leave it to unwrap to calculate au_rslack. For now we just * calculate the length of the verifier: */ - task->tk_auth->au_verfsize = XDR_QUADLEN(len) + 2; + cred->cr_auth->au_verfsize = XDR_QUADLEN(len) + 2; gss_put_ctx(ctx); dprintk("RPC: %5u gss_validate: gss_verify_mic succeeded.\n", task->tk_pid); @@ -1206,7 +1206,7 @@ gss_unwrap_resp(struct rpc_task *task, break; } /* take into account extra slack for integrity and privacy cases: */ - task->tk_auth->au_rslack = task->tk_auth->au_verfsize + (p - savedp) + cred->cr_auth->au_rslack = cred->cr_auth->au_verfsize + (p - savedp) + (savedlen - head->iov_len); out_decode: status = decode(rqstp, p, obj); diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index d9c50d810d15..5ed91e5bcee4 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -204,7 +204,7 @@ unx_validate(struct rpc_task *task, __be32 *p) printk("RPC: giant verf size: %u\n", size); return NULL; } - task->tk_auth->au_rslack = (size >> 2) + 2; + task->tk_msg.rpc_cred->cr_auth->au_rslack = (size >> 2) + 2; p += (size >> 2); return p; diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 28a789419f64..50af8bbe7f20 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -787,7 +787,7 @@ call_reserveresult(struct rpc_task *task) static void call_allocate(struct rpc_task *task) { - unsigned int slack = task->tk_auth->au_cslack; + unsigned int slack = task->tk_msg.rpc_cred->cr_auth->au_cslack; struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = task->tk_xprt; struct rpc_procinfo *proc = task->tk_msg.rpc_proc; -- cgit v1.2.3 From 587142f85f796cf0b823dd3080e815f02ff6b952 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 2 Jul 2007 09:57:54 -0400 Subject: NFS: Replace NFS_I(inode)->req_lock with inode->i_lock There is no justification for keeping a special spinlock for the exclusive use of the NFS writeback code. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 1 - fs/nfs/pagelist.c | 11 ++++--- fs/nfs/write.c | 84 ++++++++++++++++++++++++-------------------------- include/linux/nfs_fs.h | 1 - 4 files changed, 46 insertions(+), 51 deletions(-) (limited to 'include') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 01fc8ab0c562..9d5124166d20 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1154,7 +1154,6 @@ static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flag struct nfs_inode *nfsi = (struct nfs_inode *) foo; inode_init_once(&nfsi->vfs_inode); - spin_lock_init(&nfsi->req_lock); INIT_LIST_HEAD(&nfsi->open_files); INIT_LIST_HEAD(&nfsi->access_cache_entry_lru); INIT_LIST_HEAD(&nfsi->access_cache_inode_lru); diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 8d2642f24b8f..f56dae5216f4 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -126,12 +126,13 @@ static int nfs_set_page_tag_locked(struct nfs_page *req) */ void nfs_clear_page_tag_locked(struct nfs_page *req) { - struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode); + struct inode *inode = req->wb_context->path.dentry->d_inode; + struct nfs_inode *nfsi = NFS_I(inode); if (req->wb_page != NULL) { - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); } nfs_unlock_request(req); } @@ -390,7 +391,7 @@ void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index) * If the number of requests is set to 0, the entire address_space * starting at index idx_start, is scanned. * The requests are *not* checked to ensure that they form a contiguous set. - * You must be holding the inode's req_lock when calling this function + * You must be holding the inode's i_lock when calling this function */ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *dst, pgoff_t idx_start, @@ -430,7 +431,7 @@ int nfs_scan_list(struct nfs_inode *nfsi, } } /* for latency reduction */ - cond_resched_lock(&nfsi->req_lock); + cond_resched_lock(&nfsi->vfs_inode.i_lock); } out: return res; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 9ef9ec746bfb..73ac992ece85 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -124,12 +124,12 @@ static struct nfs_page *nfs_page_find_request_locked(struct page *page) static struct nfs_page *nfs_page_find_request(struct page *page) { + struct inode *inode = page->mapping->host; struct nfs_page *req = NULL; - spinlock_t *req_lock = &NFS_I(page->mapping->host)->req_lock; - spin_lock(req_lock); + spin_lock(&inode->i_lock); req = nfs_page_find_request_locked(page); - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); return req; } @@ -251,16 +251,16 @@ static void nfs_end_page_writeback(struct page *page) static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, struct page *page) { + struct inode *inode = page->mapping->host; + struct nfs_inode *nfsi = NFS_I(inode); struct nfs_page *req; - struct nfs_inode *nfsi = NFS_I(page->mapping->host); - spinlock_t *req_lock = &nfsi->req_lock; int ret; - spin_lock(req_lock); + spin_lock(&inode->i_lock); for(;;) { req = nfs_page_find_request_locked(page); if (req == NULL) { - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); return 1; } if (nfs_lock_request_dontget(req)) @@ -270,28 +270,28 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, * succeed provided that someone hasn't already marked the * request as dirty (in which case we don't care). */ - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); ret = nfs_wait_on_request(req); nfs_release_request(req); if (ret != 0) return ret; - spin_lock(req_lock); + spin_lock(&inode->i_lock); } if (test_bit(PG_NEED_COMMIT, &req->wb_flags)) { /* This request is marked for commit */ - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); nfs_unlock_request(req); nfs_pageio_complete(pgio); return 1; } if (nfs_set_page_writeback(page) != 0) { - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); BUG(); } radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED); ret = test_bit(PG_NEED_FLUSH, &req->wb_flags); - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); nfs_pageio_add_request(pgio, req); return ret; } @@ -412,7 +412,7 @@ static void nfs_inode_remove_request(struct nfs_page *req) BUG_ON (!NFS_WBACK_BUSY(req)); - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); set_page_private(req->wb_page, 0); ClearPagePrivate(req->wb_page); radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index); @@ -420,11 +420,11 @@ static void nfs_inode_remove_request(struct nfs_page *req) __set_page_dirty_nobuffers(req->wb_page); nfsi->npages--; if (!nfsi->npages) { - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); nfs_end_data_update(inode); iput(inode); } else - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); nfs_clear_request(req); nfs_release_request(req); } @@ -458,13 +458,13 @@ nfs_mark_request_commit(struct nfs_page *req) struct inode *inode = req->wb_context->path.dentry->d_inode; struct nfs_inode *nfsi = NFS_I(inode); - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); nfsi->ncommit++; set_bit(PG_NEED_COMMIT, &(req)->wb_flags); radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_COMMIT); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); __mark_inode_dirty(inode, I_DIRTY_DATASYNC); } @@ -534,10 +534,10 @@ static int nfs_wait_on_requests_locked(struct inode *inode, pgoff_t idx_start, u BUG_ON(!NFS_WBACK_BUSY(req)); kref_get(&req->wb_kref); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); error = nfs_wait_on_request(req); nfs_release_request(req); - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); if (error < 0) return error; res++; @@ -602,7 +602,6 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx, { struct address_space *mapping = page->mapping; struct inode *inode = mapping->host; - struct nfs_inode *nfsi = NFS_I(inode); struct nfs_page *req, *new = NULL; pgoff_t rqend, end; @@ -612,13 +611,13 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx, /* Loop over all inode entries and see if we find * A request for the page we wish to update */ - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); req = nfs_page_find_request_locked(page); if (req) { if (!nfs_lock_request_dontget(req)) { int error; - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); error = nfs_wait_on_request(req); nfs_release_request(req); if (error < 0) { @@ -628,7 +627,7 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx, } continue; } - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); if (new) nfs_release_request(new); break; @@ -639,14 +638,14 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx, nfs_lock_request_dontget(new); error = nfs_inode_add_request(inode, new); if (error) { - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); nfs_unlock_request(new); return ERR_PTR(error); } - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); return new; } - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); new = nfs_create_request(ctx, inode, page, offset, bytes); if (IS_ERR(new)) @@ -974,9 +973,9 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata) } if (nfs_write_need_commit(data)) { - spinlock_t *req_lock = &NFS_I(page->mapping->host)->req_lock; + struct inode *inode = page->mapping->host; - spin_lock(req_lock); + spin_lock(&inode->i_lock); if (test_bit(PG_NEED_RESCHED, &req->wb_flags)) { /* Do nothing we need to resend the writes */ } else if (!test_and_set_bit(PG_NEED_COMMIT, &req->wb_flags)) { @@ -987,7 +986,7 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata) clear_bit(PG_NEED_COMMIT, &req->wb_flags); dprintk(" server reboot detected\n"); } - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); } else dprintk(" OK\n"); @@ -1277,13 +1276,12 @@ static const struct rpc_call_ops nfs_commit_ops = { int nfs_commit_inode(struct inode *inode, int how) { - struct nfs_inode *nfsi = NFS_I(inode); LIST_HEAD(head); int res; - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); res = nfs_scan_commit(inode, &head, 0, 0); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); if (res) { int error = nfs_commit_list(inode, &head, how); if (error < 0) @@ -1301,7 +1299,6 @@ static inline int nfs_commit_list(struct inode *inode, struct list_head *head, i long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_control *wbc, int how) { struct inode *inode = mapping->host; - struct nfs_inode *nfsi = NFS_I(inode); pgoff_t idx_start, idx_end; unsigned int npages = 0; LIST_HEAD(head); @@ -1323,7 +1320,7 @@ long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_contr } } how &= ~FLUSH_NOCOMMIT; - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); do { ret = nfs_wait_on_requests_locked(inode, idx_start, npages); if (ret != 0) @@ -1334,18 +1331,19 @@ long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_contr if (pages == 0) break; if (how & FLUSH_INVALIDATE) { - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); nfs_cancel_commit_list(&head); ret = pages; - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); continue; } pages += nfs_scan_commit(inode, &head, 0, 0); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); ret = nfs_commit_list(inode, &head, how); - spin_lock(&nfsi->req_lock); + spin_lock(&inode->i_lock); + } while (ret >= 0); - spin_unlock(&nfsi->req_lock); + spin_unlock(&inode->i_lock); return ret; } @@ -1439,7 +1437,6 @@ int nfs_set_page_dirty(struct page *page) { struct address_space *mapping = page->mapping; struct inode *inode; - spinlock_t *req_lock; struct nfs_page *req; int ret; @@ -1448,18 +1445,17 @@ int nfs_set_page_dirty(struct page *page) inode = mapping->host; if (!inode) goto out_raced; - req_lock = &NFS_I(inode)->req_lock; - spin_lock(req_lock); + spin_lock(&inode->i_lock); req = nfs_page_find_request_locked(page); if (req != NULL) { /* Mark any existing write requests for flushing */ ret = !test_and_set_bit(PG_NEED_FLUSH, &req->wb_flags); - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); nfs_release_request(req); return ret; } ret = __set_page_dirty_nobuffers(page); - spin_unlock(req_lock); + spin_unlock(&inode->i_lock); return ret; out_raced: return !TestSetPageDirty(page); diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index bf24151d63be..cf395351cdd4 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -156,7 +156,6 @@ struct nfs_inode { /* * This is the list of dirty unwritten pages. */ - spinlock_t req_lock; struct radix_tree_root nfs_page_tree; unsigned long ncommit, -- cgit v1.2.3 From 7af654f8d1b7460415af5d1d326233478dd0f563 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 2 Jul 2007 12:49:23 -0400 Subject: NFSv4: Don't reuse expired nfs4_state_owner structs That just confuses certain NFSv4 servers. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 10 ---------- fs/nfs/nfs4state.c | 28 ---------------------------- include/linux/nfs_fs_sb.h | 2 -- 3 files changed, 40 deletions(-) (limited to 'include') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 71d4c4cdac52..6b424407d631 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -131,7 +131,6 @@ static struct nfs_client *nfs_alloc_client(const char *hostname, init_rwsem(&clp->cl_sem); INIT_LIST_HEAD(&clp->cl_delegations); INIT_LIST_HEAD(&clp->cl_state_owners); - INIT_LIST_HEAD(&clp->cl_unused); spin_lock_init(&clp->cl_lock); INIT_DELAYED_WORK(&clp->cl_renewd, nfs4_renew_state); rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS client"); @@ -155,15 +154,6 @@ static void nfs4_shutdown_client(struct nfs_client *clp) #ifdef CONFIG_NFS_V4 if (__test_and_clear_bit(NFS_CS_RENEWD, &clp->cl_res_state)) nfs4_kill_renewd(clp); - while (!list_empty(&clp->cl_unused)) { - struct nfs4_state_owner *sp; - - sp = list_entry(clp->cl_unused.next, - struct nfs4_state_owner, - so_list); - list_del(&sp->so_list); - kfree(sp); - } BUG_ON(!list_empty(&clp->cl_state_owners)); if (__test_and_clear_bit(NFS_CS_IDMAP, &clp->cl_res_state)) nfs_idmap_delete(clp); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 0030248d63ec..2b00c45aebea 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -75,21 +75,6 @@ nfs4_alloc_lockowner_id(struct nfs_client *clp) return clp->cl_lockowner_id ++; } -static struct nfs4_state_owner * -nfs4_client_grab_unused(struct nfs_client *clp, struct rpc_cred *cred) -{ - struct nfs4_state_owner *sp = NULL; - - if (!list_empty(&clp->cl_unused)) { - sp = list_entry(clp->cl_unused.next, struct nfs4_state_owner, so_list); - atomic_inc(&sp->so_count); - sp->so_cred = get_rpccred(cred); - list_move(&sp->so_list, &clp->cl_state_owners); - clp->cl_nunused--; - } - return sp; -} - struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) { struct nfs4_state_owner *sp; @@ -178,8 +163,6 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct new = nfs4_alloc_state_owner(); spin_lock(&clp->cl_lock); sp = nfs4_find_state_owner(clp, cred); - if (sp == NULL) - sp = nfs4_client_grab_unused(clp, cred); if (sp == NULL && new != NULL) { list_add(&new->so_list, &clp->cl_state_owners); new->so_client = clp; @@ -206,17 +189,6 @@ void nfs4_put_state_owner(struct nfs4_state_owner *sp) if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) return; - if (clp->cl_nunused >= OPENOWNER_POOL_SIZE) - goto out_free; - if (list_empty(&sp->so_list)) - goto out_free; - list_move(&sp->so_list, &clp->cl_unused); - clp->cl_nunused++; - spin_unlock(&clp->cl_lock); - put_rpccred(cred); - cred = NULL; - return; -out_free: list_del(&sp->so_list); spin_unlock(&clp->cl_lock); put_rpccred(cred); diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 144d955dc46a..2cef0a68aa77 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -44,8 +44,6 @@ struct nfs_client { struct list_head cl_delegations; struct list_head cl_state_owners; - struct list_head cl_unused; - int cl_nunused; spinlock_t cl_lock; unsigned long cl_lease_time; -- cgit v1.2.3 From 9f958ab8858c75df800e0121b1920182820cbc39 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 2 Jul 2007 13:58:33 -0400 Subject: NFSv4: Reduce the chances of an open_owner identifier collision Currently we just use a 32-bit counter. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 3 +- fs/nfs/nfs4_fs.h | 17 ++-- fs/nfs/nfs4proc.c | 8 +- fs/nfs/nfs4state.c | 193 ++++++++++++++++++++++++++++++++++++---------- fs/nfs/nfs4xdr.c | 27 ++++--- include/linux/nfs_fs_sb.h | 5 +- include/linux/nfs_xdr.h | 4 +- 7 files changed, 190 insertions(+), 67 deletions(-) (limited to 'include') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 6b424407d631..ccb455053ee4 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -130,7 +130,6 @@ static struct nfs_client *nfs_alloc_client(const char *hostname, #ifdef CONFIG_NFS_V4 init_rwsem(&clp->cl_sem); INIT_LIST_HEAD(&clp->cl_delegations); - INIT_LIST_HEAD(&clp->cl_state_owners); spin_lock_init(&clp->cl_lock); INIT_DELAYED_WORK(&clp->cl_renewd, nfs4_renew_state); rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS client"); @@ -154,7 +153,7 @@ static void nfs4_shutdown_client(struct nfs_client *clp) #ifdef CONFIG_NFS_V4 if (__test_and_clear_bit(NFS_CS_RENEWD, &clp->cl_res_state)) nfs4_kill_renewd(clp); - BUG_ON(!list_empty(&clp->cl_state_owners)); + BUG_ON(!RB_EMPTY_ROOT(&clp->cl_state_owners)); if (__test_and_clear_bit(NFS_CS_IDMAP, &clp->cl_res_state)) nfs_idmap_delete(clp); #endif diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index c97a0ad8430e..44b56c915f72 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -70,19 +70,25 @@ static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status seqid->flags |= NFS_SEQID_CONFIRMED; } +struct nfs_unique_id { + struct rb_node rb_node; + __u64 id; +}; + /* * NFS4 state_owners and lock_owners are simply labels for ordered * sequences of RPC calls. Their sole purpose is to provide once-only * semantics by allowing the server to identify replayed requests. */ struct nfs4_state_owner { - spinlock_t so_lock; - struct list_head so_list; /* per-clientid list of state_owners */ + struct nfs_unique_id so_owner_id; struct nfs_client *so_client; - u32 so_id; /* 32-bit identifier, unique */ - atomic_t so_count; + struct rb_node so_client_node; struct rpc_cred *so_cred; /* Associated cred */ + + spinlock_t so_lock; + atomic_t so_count; struct list_head so_states; struct list_head so_delegations; struct nfs_seqid_counter so_seqid; @@ -108,7 +114,7 @@ struct nfs4_lock_state { #define NFS_LOCK_INITIALIZED 1 int ls_flags; struct nfs_seqid_counter ls_seqid; - u32 ls_id; + struct nfs_unique_id ls_id; nfs4_stateid ls_stateid; atomic_t ls_count; }; @@ -189,7 +195,6 @@ extern void nfs4_renew_state(struct work_struct *); /* nfs4state.c */ struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp); -extern u32 nfs4_alloc_lockowner_id(struct nfs_client *); extern struct nfs4_state_owner * nfs4_get_state_owner(struct nfs_server *, struct rpc_cred *); extern void nfs4_put_state_owner(struct nfs4_state_owner *); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 84d0b7e0dd67..1840ebc78fd3 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -253,7 +253,7 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path, p->o_arg.fh = NFS_FH(dir); p->o_arg.open_flags = flags, p->o_arg.clientid = server->nfs_client->cl_clientid; - p->o_arg.id = sp->so_id; + p->o_arg.id = sp->so_owner_id.id; p->o_arg.name = &p->path.dentry->d_name; p->o_arg.server = server; p->o_arg.bitmask = server->attr_bitmask; @@ -651,7 +651,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) 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.id = sp->so_owner_id.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]; @@ -3029,7 +3029,7 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock if (status != 0) goto out; lsp = request->fl_u.nfs4_fl.owner; - arg.lock_owner.id = lsp->ls_id; + arg.lock_owner.id = lsp->ls_id.id; status = rpc_call_sync(server->client, &msg, 0); switch (status) { case 0: @@ -3243,7 +3243,7 @@ static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl, goto out_free; p->arg.lock_stateid = &lsp->ls_stateid; p->arg.lock_owner.clientid = server->nfs_client->cl_clientid; - p->arg.lock_owner.id = lsp->ls_id; + p->arg.lock_owner.id = lsp->ls_id.id; p->lsp = lsp; atomic_inc(&lsp->ls_count); p->ctx = get_nfs_open_context(ctx); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 0f79d56e97f0..ab0b5ab60e60 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -44,6 +44,7 @@ #include #include #include +#include #include #include @@ -69,18 +70,14 @@ static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred) return status; } -u32 -nfs4_alloc_lockowner_id(struct nfs_client *clp) -{ - return clp->cl_lockowner_id ++; -} - struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) { struct nfs4_state_owner *sp; + struct rb_node *pos; struct rpc_cred *cred = NULL; - list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { + sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); if (list_empty(&sp->so_states)) continue; cred = get_rpccred(sp->so_cred); @@ -92,32 +89,129 @@ struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp) { struct nfs4_state_owner *sp; + struct rb_node *pos; - if (!list_empty(&clp->cl_state_owners)) { - sp = list_entry(clp->cl_state_owners.next, - struct nfs4_state_owner, so_list); + pos = rb_first(&clp->cl_state_owners); + if (pos != NULL) { + sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); return get_rpccred(sp->so_cred); } return NULL; } +static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new, + __u64 minval, int maxbits) +{ + struct rb_node **p, *parent; + struct nfs_unique_id *pos; + __u64 mask = ~0ULL; + + if (maxbits < 64) + mask = (1ULL << maxbits) - 1ULL; + + /* Ensure distribution is more or less flat */ + get_random_bytes(&new->id, sizeof(new->id)); + new->id &= mask; + if (new->id < minval) + new->id += minval; +retry: + p = &root->rb_node; + parent = NULL; + + while (*p != NULL) { + parent = *p; + pos = rb_entry(parent, struct nfs_unique_id, rb_node); + + if (new->id < pos->id) + p = &(*p)->rb_left; + else if (new->id > pos->id) + p = &(*p)->rb_right; + else + goto id_exists; + } + rb_link_node(&new->rb_node, parent, p); + rb_insert_color(&new->rb_node, root); + return; +id_exists: + for (;;) { + new->id++; + if (new->id < minval || (new->id & mask) != new->id) { + new->id = minval; + break; + } + parent = rb_next(parent); + if (parent == NULL) + break; + pos = rb_entry(parent, struct nfs_unique_id, rb_node); + if (new->id < pos->id) + break; + } + goto retry; +} + +static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id) +{ + rb_erase(&id->rb_node, root); +} + static struct nfs4_state_owner * nfs4_find_state_owner(struct nfs_client *clp, struct rpc_cred *cred) { + struct rb_node **p = &clp->cl_state_owners.rb_node, + *parent = NULL; struct nfs4_state_owner *sp, *res = NULL; - list_for_each_entry(sp, &clp->cl_state_owners, so_list) { - if (sp->so_cred != cred) - continue; - atomic_inc(&sp->so_count); - /* Move to the head of the list */ - list_move(&sp->so_list, &clp->cl_state_owners); - res = sp; - break; + while (*p != NULL) { + parent = *p; + sp = rb_entry(parent, struct nfs4_state_owner, so_client_node); + + if (cred < sp->so_cred) + p = &parent->rb_left; + else if (cred > sp->so_cred) + p = &parent->rb_right; + else { + atomic_inc(&sp->so_count); + res = sp; + break; + } } return res; } +static struct nfs4_state_owner * +nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new) +{ + struct rb_node **p = &clp->cl_state_owners.rb_node, + *parent = NULL; + struct nfs4_state_owner *sp; + + while (*p != NULL) { + parent = *p; + sp = rb_entry(parent, struct nfs4_state_owner, so_client_node); + + if (new->so_cred < sp->so_cred) + p = &parent->rb_left; + else if (new->so_cred > sp->so_cred) + p = &parent->rb_right; + else { + atomic_inc(&sp->so_count); + return sp; + } + } + nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64); + rb_link_node(&new->so_client_node, parent, p); + rb_insert_color(&new->so_client_node, &clp->cl_state_owners); + return new; +} + +static void +nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp) +{ + if (!RB_EMPTY_NODE(&sp->so_client_node)) + rb_erase(&sp->so_client_node, &clp->cl_state_owners); + nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id); +} + /* * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to * create a new state_owner. @@ -145,10 +239,14 @@ nfs4_alloc_state_owner(void) void nfs4_drop_state_owner(struct nfs4_state_owner *sp) { - struct nfs_client *clp = sp->so_client; - spin_lock(&clp->cl_lock); - list_del_init(&sp->so_list); - spin_unlock(&clp->cl_lock); + if (!RB_EMPTY_NODE(&sp->so_client_node)) { + struct nfs_client *clp = sp->so_client; + + spin_lock(&clp->cl_lock); + rb_erase(&sp->so_client_node, &clp->cl_state_owners); + RB_CLEAR_NODE(&sp->so_client_node); + spin_unlock(&clp->cl_lock); + } } /* @@ -160,22 +258,24 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct struct nfs_client *clp = server->nfs_client; struct nfs4_state_owner *sp, *new; - new = nfs4_alloc_state_owner(); spin_lock(&clp->cl_lock); sp = nfs4_find_state_owner(clp, cred); - if (sp == NULL && new != NULL) { - list_add(&new->so_list, &clp->cl_state_owners); - new->so_client = clp; - new->so_id = nfs4_alloc_lockowner_id(clp); - new->so_cred = get_rpccred(cred); - sp = new; - new = NULL; - } spin_unlock(&clp->cl_lock); - kfree(new); if (sp != NULL) return sp; - return NULL; + new = nfs4_alloc_state_owner(); + if (new == NULL) + return NULL; + new->so_client = clp; + new->so_cred = cred; + spin_lock(&clp->cl_lock); + sp = nfs4_insert_state_owner(clp, new); + spin_unlock(&clp->cl_lock); + if (sp == new) + get_rpccred(cred); + else + kfree(new); + return sp; } /* @@ -189,7 +289,7 @@ void nfs4_put_state_owner(struct nfs4_state_owner *sp) if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) return; - list_del(&sp->so_list); + nfs4_remove_state_owner(clp, sp); spin_unlock(&clp->cl_lock); put_rpccred(cred); kfree(sp); @@ -386,12 +486,22 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f atomic_set(&lsp->ls_count, 1); lsp->ls_owner = fl_owner; spin_lock(&clp->cl_lock); - lsp->ls_id = nfs4_alloc_lockowner_id(clp); + nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64); spin_unlock(&clp->cl_lock); INIT_LIST_HEAD(&lsp->ls_locks); return lsp; } +static void nfs4_free_lock_state(struct nfs4_lock_state *lsp) +{ + struct nfs_client *clp = lsp->ls_state->owner->so_client; + + spin_lock(&clp->cl_lock); + nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id); + spin_unlock(&clp->cl_lock); + kfree(lsp); +} + /* * Return a compatible lock_state. If no initialized lock_state structure * exists, return an uninitialized one. @@ -421,7 +531,8 @@ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_ return NULL; } spin_unlock(&state->state_lock); - kfree(new); + if (new != NULL) + nfs4_free_lock_state(new); return lsp; } @@ -442,7 +553,7 @@ void nfs4_put_lock_state(struct nfs4_lock_state *lsp) if (list_empty(&state->lock_states)) clear_bit(LK_STATE_IN_USE, &state->flags); spin_unlock(&state->state_lock); - kfree(lsp); + nfs4_free_lock_state(lsp); } static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src) @@ -719,11 +830,13 @@ out_err: static void nfs4_state_mark_reclaim(struct nfs_client *clp) { struct nfs4_state_owner *sp; + struct rb_node *pos; struct nfs4_state *state; struct nfs4_lock_state *lock; /* Reset all sequence ids to zero */ - list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { + sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); sp->so_seqid.counter = 0; sp->so_seqid.flags = 0; spin_lock(&sp->so_lock); @@ -742,6 +855,7 @@ static int reclaimer(void *ptr) { struct nfs_client *clp = ptr; struct nfs4_state_owner *sp; + struct rb_node *pos; struct nfs4_state_recovery_ops *ops; struct rpc_cred *cred; int status = 0; @@ -787,7 +901,8 @@ restart_loop: /* Mark all delegations for reclaim */ nfs_delegation_mark_reclaim(clp); /* Note: list is protected by exclusive lock on cl->cl_sem */ - list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { + sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); status = nfs4_reclaim_open_state(ops, sp); if (status < 0) { if (status == -NFS4ERR_NO_GRACE) { diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 4c8f67d47523..c08738441f73 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -68,10 +68,10 @@ static int nfs4_stat_to_errno(int); #endif /* lock,open owner id: - * we currently use size 1 (u32) out of (NFS4_OPAQUE_LIMIT >> 2) + * we currently use size 2 (u64) out of (NFS4_OPAQUE_LIMIT >> 2) */ -#define open_owner_id_maxsz (1 + 1) -#define lock_owner_id_maxsz (1 + 1) +#define open_owner_id_maxsz (1 + 4) +#define lock_owner_id_maxsz (1 + 4) #define compound_encode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2)) #define compound_decode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2)) #define op_encode_hdr_maxsz (1) @@ -827,13 +827,14 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lock_args *args) WRITE64(nfs4_lock_length(args->fl)); WRITE32(args->new_lock_owner); if (args->new_lock_owner){ - RESERVE_SPACE(4+NFS4_STATEID_SIZE+20); + RESERVE_SPACE(4+NFS4_STATEID_SIZE+32); WRITE32(args->open_seqid->sequence->counter); WRITEMEM(args->open_stateid->data, NFS4_STATEID_SIZE); WRITE32(args->lock_seqid->sequence->counter); WRITE64(args->lock_owner.clientid); - WRITE32(4); - WRITE32(args->lock_owner.id); + WRITE32(16); + WRITEMEM("lock id:", 8); + WRITE64(args->lock_owner.id); } else { RESERVE_SPACE(NFS4_STATEID_SIZE+4); @@ -848,14 +849,15 @@ static int encode_lockt(struct xdr_stream *xdr, const struct nfs_lockt_args *arg { __be32 *p; - RESERVE_SPACE(40); + RESERVE_SPACE(52); WRITE32(OP_LOCKT); 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(args->lock_owner.id); + WRITE32(16); + WRITEMEM("lock id:", 8); + WRITE64(args->lock_owner.id); return 0; } @@ -920,10 +922,11 @@ static inline void encode_openhdr(struct xdr_stream *xdr, const struct nfs_opena WRITE32(OP_OPEN); WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); - RESERVE_SPACE(16); + RESERVE_SPACE(28); WRITE64(arg->clientid); - WRITE32(4); - WRITE32(arg->id); + WRITE32(16); + WRITEMEM("open id:", 8); + WRITE64(arg->id); } static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_openargs *arg) diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 2cef0a68aa77..0cac49bc0955 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -34,7 +34,8 @@ struct nfs_client { nfs4_verifier cl_confirm; unsigned long cl_state; - u32 cl_lockowner_id; + struct rb_root cl_openowner_id; + struct rb_root cl_lockowner_id; /* * The following rwsem ensures exclusive access to the server @@ -43,7 +44,7 @@ struct nfs_client { struct rw_semaphore cl_sem; struct list_head cl_delegations; - struct list_head cl_state_owners; + struct rb_root cl_state_owners; spinlock_t cl_lock; unsigned long cl_lease_time; diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index f7100df3a690..38d77681cf27 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -119,7 +119,7 @@ struct nfs_openargs { struct nfs_seqid * seqid; int open_flags; __u64 clientid; - __u32 id; + __u64 id; union { struct iattr * attrs; /* UNCHECKED, GUARDED */ nfs4_verifier verifier; /* EXCLUSIVE */ @@ -181,7 +181,7 @@ struct nfs_closeres { * */ struct nfs_lowner { __u64 clientid; - u32 id; + __u64 id; }; struct nfs_lock_args { -- cgit v1.2.3 From 412c77cee6d6e73fbe1dc3d67f52163efed33fc4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 3 Jul 2007 16:10:55 -0400 Subject: NFSv4: Defer inode revalidation when setting up a delegation Currently we force a synchronous call to __nfs_revalidate_inode() in nfs_inode_set_delegation(). This not only ensures that we cannot call nfs_inode_set_delegation from an asynchronous context, but it also slows down any call to open(). Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 10 ++++++---- fs/nfs/inode.c | 4 +++- include/linux/nfs_fs.h | 1 + 3 files changed, 10 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 56f4f6a99d4e..20ac403469a0 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -124,10 +124,6 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct struct nfs_delegation *delegation; int status = 0; - /* Ensure we first revalidate the attributes and page cache! */ - if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR))) - __nfs_revalidate_inode(NFS_SERVER(inode), inode); - delegation = kmalloc(sizeof(*delegation), GFP_KERNEL); if (delegation == NULL) return -ENOMEM; @@ -154,6 +150,12 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct status = -EIO; } } + + /* Ensure we revalidate the attributes and page cache! */ + spin_lock(&inode->i_lock); + nfsi->cache_validity |= NFS_INO_REVAL_FORCED; + spin_unlock(&inode->i_lock); + spin_unlock(&clp->cl_lock); kfree(delegation); return status; diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 9d5124166d20..3d9fccf4ef93 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1072,8 +1072,10 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) 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)) + if (!nfs_have_delegation(inode, FMODE_READ) || + (nfsi->cache_validity & NFS_INO_REVAL_FORCED)) nfsi->cache_validity |= invalid; + nfsi->cache_validity &= ~NFS_INO_REVAL_FORCED; return 0; out_changed: diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index cf395351cdd4..e94971040de9 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -184,6 +184,7 @@ struct nfs_inode { #define NFS_INO_INVALID_ACCESS 0x0008 /* cached access cred invalid */ #define NFS_INO_INVALID_ACL 0x0010 /* cached acls are invalid */ #define NFS_INO_REVAL_PAGECACHE 0x0020 /* must revalidate pagecache */ +#define NFS_INO_REVAL_FORCED 0x0040 /* force revalidation ignoring a delegation */ /* * Bit offsets in flags field -- cgit v1.2.3 From 433c92379d9c2c59c2ebc7628fe4fb02cfc2daf8 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:12:14 -0400 Subject: NFS: Clean up nfs_size_to_loff_t() Use the same file size limit that lockd uses. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index e94971040de9..7deb5b0347f7 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -503,12 +503,10 @@ extern int nfsroot_mount(struct sockaddr_in *, char *, struct nfs_fh *, * inline functions */ -static inline loff_t -nfs_size_to_loff_t(__u64 size) +static inline loff_t nfs_size_to_loff_t(__u64 size) { - loff_t maxsz = (((loff_t) ULONG_MAX) << PAGE_CACHE_SHIFT) + PAGE_CACHE_SIZE - 1; - if (size > maxsz) - return maxsz; + if (size > (__u64) OFFSET_MAX - 1) + return OFFSET_MAX - 1; return (loff_t) size; } -- cgit v1.2.3 From 5680d48be88d12cd987e5579a6072a4ca34ca6ea Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:12:24 -0400 Subject: NFS: Clean-up: Define macros for maximum host and export path name lengths Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/super.c | 4 ++-- include/linux/nfs_mount.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/fs/nfs/super.c b/fs/nfs/super.c index 14c7923697d2..e7d197085834 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -867,12 +867,12 @@ static int nfs4_get_sb(struct file_system_type *fs_type, } } - p = nfs_copy_user_string(NULL, &data->hostname, 256); + p = nfs_copy_user_string(NULL, &data->hostname, NFS4_MAXNAMLEN); if (IS_ERR(p)) goto out_err; hostname = p; - p = nfs_copy_user_string(NULL, &data->mnt_path, 1024); + p = nfs_copy_user_string(NULL, &data->mnt_path, NFS4_MAXPATHLEN); if (IS_ERR(p)) goto out_err; mntpath = p; diff --git a/include/linux/nfs_mount.h b/include/linux/nfs_mount.h index cc8b9c59acb8..0b82a17c705b 100644 --- a/include/linux/nfs_mount.h +++ b/include/linux/nfs_mount.h @@ -37,7 +37,7 @@ struct nfs_mount_data { int acdirmin; /* 1 */ int acdirmax; /* 1 */ struct sockaddr_in addr; /* 1 */ - char hostname[256]; /* 1 */ + char hostname[NFS_MAXNAMLEN + 1]; /* 1 */ int namlen; /* 2 */ unsigned int bsize; /* 3 */ struct nfs3_fh root; /* 4 */ -- cgit v1.2.3 From f18289931d705f9c4634b361341a1677bea97aca Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:12:51 -0400 Subject: NFS: Add a new NFS debugging flag just for mount processing Note to self: fix up /usr/sbin/rpcdebug too Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 7deb5b0347f7..04f659f1e560 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -553,6 +553,7 @@ extern void * nfs_root_data(void); #define NFSDBG_ROOT 0x0080 #define NFSDBG_CALLBACK 0x0100 #define NFSDBG_CLIENT 0x0200 +#define NFSDBG_MOUNT 0x0400 #define NFSDBG_ALL 0xFFFF #ifdef __KERNEL__ -- cgit v1.2.3 From cce63cd6374e6f1b4ea897ece1454feb13993d7c Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:13:12 -0400 Subject: SUNRPC: Rename rpcb_getport_external routine In preparation for handling NFS mount option parsing in the kernel, rename rpcb_getport_external as rpcb_get_port_sync, and make it available always (instead of only when CONFIG_ROOT_NFS is enabled). Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/nfsroot.c | 2 +- include/linux/sunrpc/clnt.h | 7 ++----- net/sunrpc/rpcb_clnt.c | 21 +++++++++++---------- 3 files changed, 14 insertions(+), 16 deletions(-) (limited to 'include') diff --git a/fs/nfs/nfsroot.c b/fs/nfs/nfsroot.c index 49d1008ce1d7..f0db4703b1c9 100644 --- a/fs/nfs/nfsroot.c +++ b/fs/nfs/nfsroot.c @@ -428,7 +428,7 @@ static int __init root_nfs_getport(int program, int version, int proto) printk(KERN_NOTICE "Looking up port of RPC %d/%d on %u.%u.%u.%u\n", program, version, NIPQUAD(servaddr)); set_sockaddr(&sin, servaddr, 0); - return rpcb_getport_external(&sin, program, version, proto); + return rpcb_getport_sync(&sin, program, version, proto); } diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 097984b03857..b28d919c7758 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -120,8 +120,10 @@ struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); void rpc_shutdown_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); + int rpcb_register(u32, u32, int, unsigned short, int *); void rpcb_getport(struct rpc_task *); +int rpcb_getport_sync(struct sockaddr_in *, __u32, __u32, int); void rpc_call_setup(struct rpc_task *, struct rpc_message *, int); @@ -141,10 +143,5 @@ void rpc_force_rebind(struct rpc_clnt *); size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t); char * rpc_peeraddr2str(struct rpc_clnt *, enum rpc_display_format_t); -/* - * Helper function for NFSroot support - */ -int rpcb_getport_external(struct sockaddr_in *, __u32, __u32, int); - #endif /* __KERNEL__ */ #endif /* _LINUX_SUNRPC_CLNT_H */ diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c index 9a20f380ab09..fc881a675eb9 100644 --- a/net/sunrpc/rpcb_clnt.c +++ b/net/sunrpc/rpcb_clnt.c @@ -12,6 +12,8 @@ * Copyright (C) 1996, Olaf Kirch */ +#include + #include #include #include @@ -247,21 +249,20 @@ int rpcb_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay) return error; } -#ifdef CONFIG_ROOT_NFS /** - * rpcb_getport_external - obtain the port for an RPC service on a given host + * rpcb_getport_sync - obtain the port for an RPC service on a given host * @sin: address of remote peer * @prog: RPC program number to bind * @vers: RPC version number to bind * @prot: transport protocol to use to make this request * * Called from outside the RPC client in a synchronous task context. + * Uses default timeout parameters specified by underlying transport. * - * For now, this supports only version 2 queries, but is used only by - * mount_clnt for NFS_ROOT. + * XXX: Needs to support IPv6, and rpcbind versions 3 and 4 */ -int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog, - __u32 vers, int prot) +int rpcb_getport_sync(struct sockaddr_in *sin, __u32 prog, + __u32 vers, int prot) { struct rpcbind_args map = { .r_prog = prog, @@ -278,10 +279,10 @@ int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog, char hostname[40]; int status; - dprintk("RPC: rpcb_getport_external(%u.%u.%u.%u, %u, %u, %d)\n", - NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot); + dprintk("RPC: %s(" NIPQUAD_FMT ", %u, %u, %d)\n", + __FUNCTION__, NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot); - sprintf(hostname, "%u.%u.%u.%u", NIPQUAD(sin->sin_addr.s_addr)); + sprintf(hostname, NIPQUAD_FMT, NIPQUAD(sin->sin_addr.s_addr)); rpcb_clnt = rpcb_create(hostname, (struct sockaddr *)sin, prot, 2, 0); if (IS_ERR(rpcb_clnt)) return PTR_ERR(rpcb_clnt); @@ -296,7 +297,7 @@ int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog, } return status; } -#endif +EXPORT_SYMBOL_GPL(rpcb_getport_sync); /** * rpcb_getport - obtain the port for a given RPC service on a given host -- cgit v1.2.3 From 45160d6275814e0c86206e6981f0b92c61a50a21 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:13:17 -0400 Subject: SUNRPC: Rename rpcb_getport to be consistent with new rpcb_getport_sync name Clean up, for consistency. Rename rpcb_getport as rpcb_getport_async, to match the naming scheme of rpcb_getport_sync. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 2 +- net/sunrpc/rpcb_clnt.c | 37 +++++++++++++++++++------------------ net/sunrpc/xprtsock.c | 4 ++-- 3 files changed, 22 insertions(+), 21 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index b28d919c7758..c1b37972b0d5 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -122,8 +122,8 @@ void rpc_shutdown_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); int rpcb_register(u32, u32, int, unsigned short, int *); -void rpcb_getport(struct rpc_task *); int rpcb_getport_sync(struct sockaddr_in *, __u32, __u32, int); +void rpcb_getport_async(struct rpc_task *); void rpc_call_setup(struct rpc_task *, struct rpc_message *, int); diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c index fc881a675eb9..d1740dbab991 100644 --- a/net/sunrpc/rpcb_clnt.c +++ b/net/sunrpc/rpcb_clnt.c @@ -300,13 +300,13 @@ int rpcb_getport_sync(struct sockaddr_in *sin, __u32 prog, EXPORT_SYMBOL_GPL(rpcb_getport_sync); /** - * rpcb_getport - obtain the port for a given RPC service on a given host + * rpcb_getport_async - obtain the port for a given RPC service on a given host * @task: task that is waiting for portmapper request * * This one can be called for an ongoing RPC request, and can be used in * an async (rpciod) context. */ -void rpcb_getport(struct rpc_task *task) +void rpcb_getport_async(struct rpc_task *task) { struct rpc_clnt *clnt = task->tk_client; int bind_version; @@ -317,17 +317,17 @@ void rpcb_getport(struct rpc_task *task) struct sockaddr addr; int status; - dprintk("RPC: %5u rpcb_getport(%s, %u, %u, %d)\n", - task->tk_pid, clnt->cl_server, - clnt->cl_prog, clnt->cl_vers, xprt->prot); + dprintk("RPC: %5u %s(%s, %u, %u, %d)\n", + task->tk_pid, __FUNCTION__, + clnt->cl_server, clnt->cl_prog, clnt->cl_vers, xprt->prot); /* Autobind on cloned rpc clients is discouraged */ BUG_ON(clnt->cl_parent != clnt); if (xprt_test_and_set_binding(xprt)) { status = -EACCES; /* tell caller to check again */ - dprintk("RPC: %5u rpcb_getport waiting for another binder\n", - task->tk_pid); + dprintk("RPC: %5u %s: waiting for another binder\n", + task->tk_pid, __FUNCTION__); goto bailout_nowake; } @@ -338,27 +338,28 @@ void rpcb_getport(struct rpc_task *task) /* Someone else may have bound if we slept */ if (xprt_bound(xprt)) { status = 0; - dprintk("RPC: %5u rpcb_getport already bound\n", task->tk_pid); + dprintk("RPC: %5u %s: already bound\n", + task->tk_pid, __FUNCTION__); goto bailout_nofree; } if (rpcb_next_version[xprt->bind_index].rpc_proc == NULL) { xprt->bind_index = 0; status = -EACCES; /* tell caller to try again later */ - dprintk("RPC: %5u rpcb_getport no more getport versions " - "available\n", task->tk_pid); + dprintk("RPC: %5u %s: no more getport versions available\n", + task->tk_pid, __FUNCTION__); goto bailout_nofree; } bind_version = rpcb_next_version[xprt->bind_index].rpc_vers; - dprintk("RPC: %5u rpcb_getport trying rpcbind version %u\n", - task->tk_pid, bind_version); + dprintk("RPC: %5u %s: trying rpcbind version %u\n", + task->tk_pid, __FUNCTION__, bind_version); map = kzalloc(sizeof(struct rpcbind_args), GFP_ATOMIC); if (!map) { status = -ENOMEM; - dprintk("RPC: %5u rpcb_getport no memory available\n", - task->tk_pid); + dprintk("RPC: %5u %s: no memory available\n", + task->tk_pid, __FUNCTION__); goto bailout_nofree; } map->r_prog = clnt->cl_prog; @@ -376,8 +377,8 @@ void rpcb_getport(struct rpc_task *task) rpcb_clnt = rpcb_create(clnt->cl_server, &addr, xprt->prot, bind_version, 0); if (IS_ERR(rpcb_clnt)) { status = PTR_ERR(rpcb_clnt); - dprintk("RPC: %5u rpcb_getport rpcb_create failed, error %ld\n", - task->tk_pid, PTR_ERR(rpcb_clnt)); + dprintk("RPC: %5u %s: rpcb_create failed, error %ld\n", + task->tk_pid, __FUNCTION__, PTR_ERR(rpcb_clnt)); goto bailout; } @@ -385,8 +386,8 @@ void rpcb_getport(struct rpc_task *task) rpc_release_client(rpcb_clnt); if (IS_ERR(child)) { status = -EIO; - dprintk("RPC: %5u rpcb_getport rpc_run_task failed\n", - task->tk_pid); + dprintk("RPC: %5u %s: rpc_run_task failed\n", + task->tk_pid, __FUNCTION__); goto bailout_nofree; } rpc_put_task(child); diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index ee6ad3baf680..a8f7c5fd752c 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1473,7 +1473,7 @@ static struct rpc_xprt_ops xs_udp_ops = { .set_buffer_size = xs_udp_set_buffer_size, .reserve_xprt = xprt_reserve_xprt_cong, .release_xprt = xprt_release_xprt_cong, - .rpcbind = rpcb_getport, + .rpcbind = rpcb_getport_async, .set_port = xs_set_port, .connect = xs_connect, .buf_alloc = rpc_malloc, @@ -1490,7 +1490,7 @@ static struct rpc_xprt_ops xs_udp_ops = { static struct rpc_xprt_ops xs_tcp_ops = { .reserve_xprt = xprt_reserve_xprt, .release_xprt = xs_tcp_release_xprt, - .rpcbind = rpcb_getport, + .rpcbind = rpcb_getport_async, .set_port = xs_set_port, .connect = xs_connect, .buf_alloc = rpc_malloc, -- cgit v1.2.3 From 3ea97309e6b18bce200211b3f9188e8023321adc Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:13:27 -0400 Subject: NFS: Remake nfsroot_mount as a permanent part of NFS client In preparation for supporting NFSv2 and NFSv3 mount option handling in the kernel NFS client, convert mount_clnt.c to be a permanent part of the NFS client, instead of built only when CONFIG_ROOT_NFS is enabled. In addition, we also replace the "struct sockaddr_in *" argument with something more generic, to help support IPv6 at some later point. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/Makefile | 4 ++-- fs/nfs/mount_clnt.c | 55 +++++++++++++++++++++++++------------------------- fs/nfs/nfsroot.c | 3 ++- include/linux/nfs_fs.h | 5 ++--- 4 files changed, 34 insertions(+), 33 deletions(-) (limited to 'include') diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile index f4580b44eef4..b55cb236cf74 100644 --- a/fs/nfs/Makefile +++ b/fs/nfs/Makefile @@ -6,8 +6,8 @@ obj-$(CONFIG_NFS_FS) += nfs.o nfs-y := client.o dir.o file.o getroot.o inode.o super.o nfs2xdr.o \ pagelist.o proc.o read.o symlink.o unlink.o \ - write.o namespace.o -nfs-$(CONFIG_ROOT_NFS) += nfsroot.o mount_clnt.o + write.o namespace.o mount_clnt.o +nfs-$(CONFIG_ROOT_NFS) += nfsroot.o nfs-$(CONFIG_NFS_V3) += nfs3proc.o nfs3xdr.o nfs-$(CONFIG_NFS_V3_ACL) += nfs3acl.o nfs-$(CONFIG_NFS_V4) += nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \ diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c index 2892ec843066..ee4899d96629 100644 --- a/fs/nfs/mount_clnt.c +++ b/fs/nfs/mount_clnt.c @@ -18,7 +18,7 @@ #include #ifdef RPC_DEBUG -# define NFSDBG_FACILITY NFSDBG_ROOT +# define NFSDBG_FACILITY NFSDBG_MOUNT #endif /* @@ -28,7 +28,6 @@ #define MOUNT_UMNT 3 */ -static struct rpc_clnt * mnt_create(struct sockaddr_in *, int, int); static struct rpc_program mnt_program; struct mnt_fhstatus { @@ -36,14 +35,21 @@ struct mnt_fhstatus { struct nfs_fh * fh; }; -/* - * Obtain an NFS file handle for the given host and path +/** + * nfs_mount - Obtain an NFS file handle for the given host and path + * @addr: pointer to server's address + * @len: size of server's address + * @hostname: name of server host, or NULL + * @path: pointer to string containing export path to mount + * @version: mount version to use for this request + * @protocol: transport protocol to use for thie request + * @fh: pointer to location to place returned file handle + * + * Uses default timeout parameters specified by underlying transport. */ -int -nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh, - int version, int protocol) +int nfs_mount(struct sockaddr *addr, size_t len, char *hostname, char *path, + int version, int protocol, struct nfs_fh *fh) { - struct rpc_clnt *mnt_clnt; struct mnt_fhstatus result = { .fh = fh }; @@ -51,12 +57,23 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh, .rpc_argp = path, .rpc_resp = &result, }; + struct rpc_create_args args = { + .protocol = protocol, + .address = addr, + .addrsize = len, + .servername = hostname, + .program = &mnt_program, + .version = version, + .authflavor = RPC_AUTH_UNIX, + .flags = RPC_CLNT_CREATE_INTR, + }; + struct rpc_clnt *mnt_clnt; int status; - dprintk("NFS: nfs_mount(%08x:%s)\n", - (unsigned)ntohl(addr->sin_addr.s_addr), path); + dprintk("NFS: sending MNT request for %s:%s\n", + (hostname ? hostname : "server"), path); - mnt_clnt = mnt_create(addr, version, protocol); + mnt_clnt = rpc_create(&args); if (IS_ERR(mnt_clnt)) return PTR_ERR(mnt_clnt); @@ -70,22 +87,6 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh, return status < 0? status : (result.status? -EACCES : 0); } -static struct rpc_clnt *mnt_create(struct sockaddr_in *srvaddr, int version, - int protocol) -{ - struct rpc_create_args args = { - .protocol = protocol, - .address = (struct sockaddr *)srvaddr, - .addrsize = sizeof(*srvaddr), - .program = &mnt_program, - .version = version, - .authflavor = RPC_AUTH_UNIX, - .flags = RPC_CLNT_CREATE_INTR, - }; - - return rpc_create(&args); -} - /* * XDR encode/decode functions for MOUNT */ diff --git a/fs/nfs/nfsroot.c b/fs/nfs/nfsroot.c index f0db4703b1c9..3490322d1145 100644 --- a/fs/nfs/nfsroot.c +++ b/fs/nfs/nfsroot.c @@ -496,7 +496,8 @@ static int __init root_nfs_get_handle(void) NFS_MNT3_VERSION : NFS_MNT_VERSION; set_sockaddr(&sin, servaddr, htons(mount_port)); - status = nfsroot_mount(&sin, nfs_path, &fh, version, protocol); + status = nfs_mount((struct sockaddr *) &sin, sizeof(sin), NULL, + nfs_path, version, protocol, &fh); if (status < 0) printk(KERN_ERR "Root-NFS: Server returned error %d " "while mounting %s\n", status, nfs_path); diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 04f659f1e560..c098ae194f79 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -494,10 +494,9 @@ static inline void nfs3_forget_cached_acls(struct inode *inode) /* * linux/fs/mount_clnt.c - * (Used only by nfsroot module) */ -extern int nfsroot_mount(struct sockaddr_in *, char *, struct nfs_fh *, - int, int); +extern int nfs_mount(struct sockaddr *, size_t, char *, char *, + int, int, struct nfs_fh *); /* * inline functions -- cgit v1.2.3 From 8007122520f0a3599bdc4df47358a5d83b2574aa Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sun, 1 Jul 2007 12:13:59 -0400 Subject: NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/super.c | 91 ++++++++++++++++++++++++++++++++++++++++++---- include/linux/nfs4_mount.h | 2 +- 2 files changed, 85 insertions(+), 8 deletions(-) (limited to 'include') diff --git a/fs/nfs/super.c b/fs/nfs/super.c index 757aa3b7e64b..064e69d2fdde 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -1541,8 +1541,90 @@ static int nfs4_validate_mount_data(struct nfs4_mount_data **options, *ip_addr = c; break; - default: - goto out_bad_version; + default: { + unsigned int len; + struct nfs_parsed_mount_data args = { + .rsize = NFS_MAX_FILE_IO_SIZE, + .wsize = NFS_MAX_FILE_IO_SIZE, + .timeo = 600, + .retrans = 2, + .acregmin = 3, + .acregmax = 60, + .acdirmin = 30, + .acdirmax = 60, + .nfs_server.protocol = IPPROTO_TCP, + }; + + if (nfs_parse_mount_options((char *) *options, &args) == 0) + return -EINVAL; + + if (!nfs_verify_server_address((struct sockaddr *) + &args.nfs_server.address)) + return -EINVAL; + *addr = args.nfs_server.address; + + switch (args.auth_flavor_len) { + case 0: + *authflavour = RPC_AUTH_UNIX; + break; + case 1: + *authflavour = (rpc_authflavor_t) args.auth_flavors[0]; + break; + default: + goto out_inval_auth; + } + + /* + * Translate to nfs4_mount_data, which nfs4_fill_super + * can deal with. + */ + data = kzalloc(sizeof(*data), GFP_KERNEL); + if (data == NULL) + return -ENOMEM; + *options = data; + + data->version = 1; + data->flags = args.flags & NFS4_MOUNT_FLAGMASK; + data->rsize = args.rsize; + data->wsize = args.wsize; + data->timeo = args.timeo; + data->retrans = args.retrans; + data->acregmin = args.acregmin; + data->acregmax = args.acregmax; + data->acdirmin = args.acdirmin; + data->acdirmax = args.acdirmax; + data->proto = args.nfs_server.protocol; + + /* + * Split "dev_name" into "hostname:mntpath". + */ + c = strchr(dev_name, ':'); + if (c == NULL) + return -EINVAL; + /* while calculating len, pretend ':' is '\0' */ + len = c - dev_name; + if (len > NFS4_MAXNAMLEN) + return -EINVAL; + *hostname = kzalloc(len, GFP_KERNEL); + if (*hostname == NULL) + return -ENOMEM; + strncpy(*hostname, dev_name, len - 1); + + c++; /* step over the ':' */ + len = strlen(c); + if (len > NFS4_MAXPATHLEN) + return -EINVAL; + *mntpath = kzalloc(len + 1, GFP_KERNEL); + if (*mntpath == NULL) + return -ENOMEM; + strncpy(*mntpath, c, len); + + dprintk("MNTPATH: %s\n", *mntpath); + + *ip_addr = args.client_address; + + break; + } } return 0; @@ -1559,11 +1641,6 @@ out_inval_auth: out_no_address: dfprintk(MOUNT, "NFS4: mount program didn't pass remote address\n"); return -EINVAL; - -out_bad_version: - dfprintk(MOUNT, "NFS4: bad nfs_mount_data version %d\n", - data->version); - return -EINVAL; } /* diff --git a/include/linux/nfs4_mount.h b/include/linux/nfs4_mount.h index 26b4c83f831d..d8d7480e5a47 100644 --- a/include/linux/nfs4_mount.h +++ b/include/linux/nfs4_mount.h @@ -65,6 +65,6 @@ struct nfs4_mount_data { #define NFS4_MOUNT_NOCTO 0x0010 /* 1 */ #define NFS4_MOUNT_NOAC 0x0020 /* 1 */ #define NFS4_MOUNT_STRICTLOCK 0x1000 /* 1 */ -#define NFS4_MOUNT_FLAGMASK 0xFFFF +#define NFS4_MOUNT_FLAGMASK 0x1033 #endif -- cgit v1.2.3 From 75180df2ed467866ada839fe73cf7cc7d75c0a22 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 May 2007 16:53:28 -0400 Subject: NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust --- fs/nfs/super.c | 43 ++++++++++++++++++++++++++++++++++++++----- include/linux/nfs4_mount.h | 3 ++- include/linux/nfs_mount.h | 1 + 3 files changed, 41 insertions(+), 6 deletions(-) (limited to 'include') diff --git a/fs/nfs/super.c b/fs/nfs/super.c index 064e69d2fdde..1b555cd41e3b 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -100,6 +100,7 @@ enum { Opt_udp, Opt_tcp, Opt_acl, Opt_noacl, Opt_rdirplus, Opt_nordirplus, + Opt_sharecache, Opt_nosharecache, /* Mount options that take integer arguments */ Opt_port, @@ -146,6 +147,8 @@ static match_table_t nfs_mount_option_tokens = { { Opt_noacl, "noacl" }, { Opt_rdirplus, "rdirplus" }, { Opt_nordirplus, "nordirplus" }, + { Opt_sharecache, "sharecache" }, + { Opt_nosharecache, "nosharecache" }, { Opt_port, "port=%u" }, { Opt_rsize, "rsize=%u" }, @@ -450,6 +453,7 @@ static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, { NFS_MOUNT_NONLM, ",nolock", "" }, { NFS_MOUNT_NOACL, ",noacl", "" }, { NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" }, + { NFS_MOUNT_UNSHARED, ",nosharecache", ""}, { 0, NULL, NULL } }; const struct proc_nfs_info *nfs_infop; @@ -714,6 +718,12 @@ static int nfs_parse_mount_options(char *raw, case Opt_nordirplus: mnt->flags |= NFS_MOUNT_NORDIRPLUS; break; + case Opt_sharecache: + mnt->flags &= ~NFS_MOUNT_UNSHARED; + break; + case Opt_nosharecache: + mnt->flags |= NFS_MOUNT_UNSHARED; + break; case Opt_port: if (match_int(args, &option)) @@ -1309,6 +1319,9 @@ static int nfs_compare_super(struct super_block *sb, void *data) if (old->nfs_client != server->nfs_client) return 0; + /* Note: NFS_MOUNT_UNSHARED == NFS4_MOUNT_UNSHARED */ + if (old->flags & NFS_MOUNT_UNSHARED) + return 0; if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0) return 0; return 1; @@ -1322,6 +1335,7 @@ static int nfs_get_sb(struct file_system_type *fs_type, struct nfs_fh mntfh; struct nfs_mount_data *data = raw_data; struct dentry *mntroot; + int (*compare_super)(struct super_block *, void *) = nfs_compare_super; int error; /* Validate the mount data */ @@ -1336,8 +1350,11 @@ static int nfs_get_sb(struct file_system_type *fs_type, goto out; } + if (server->flags & NFS_MOUNT_UNSHARED) + compare_super = NULL; + /* Get a superblock - note that we may end up sharing one that already exists */ - s = sget(fs_type, nfs_compare_super, nfs_set_super, server); + s = sget(fs_type, compare_super, nfs_set_super, server); if (IS_ERR(s)) { error = PTR_ERR(s); goto out_err_nosb; @@ -1402,6 +1419,7 @@ static int nfs_xdev_get_sb(struct file_system_type *fs_type, int flags, struct super_block *s; struct nfs_server *server; struct dentry *mntroot; + int (*compare_super)(struct super_block *, void *) = nfs_compare_super; int error; dprintk("--> nfs_xdev_get_sb()\n"); @@ -1413,8 +1431,11 @@ static int nfs_xdev_get_sb(struct file_system_type *fs_type, int flags, goto out_err_noserver; } + if (server->flags & NFS_MOUNT_UNSHARED) + compare_super = NULL; + /* Get a superblock - note that we may end up sharing one that already exists */ - s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); + s = sget(&nfs_fs_type, compare_super, nfs_set_super, server); if (IS_ERR(s)) { error = PTR_ERR(s); goto out_err_nosb; @@ -1657,6 +1678,7 @@ static int nfs4_get_sb(struct file_system_type *fs_type, struct nfs_fh mntfh; struct dentry *mntroot; char *mntpath = NULL, *hostname = NULL, *ip_addr = NULL; + int (*compare_super)(struct super_block *, void *) = nfs_compare_super; int error; /* Validate the mount data */ @@ -1673,8 +1695,11 @@ static int nfs4_get_sb(struct file_system_type *fs_type, goto out; } + if (server->flags & NFS4_MOUNT_UNSHARED) + compare_super = NULL; + /* Get a superblock - note that we may end up sharing one that already exists */ - s = sget(fs_type, nfs_compare_super, nfs_set_super, server); + s = sget(fs_type, compare_super, nfs_set_super, server); if (IS_ERR(s)) { error = PTR_ERR(s); goto out_free; @@ -1740,6 +1765,7 @@ static int nfs4_xdev_get_sb(struct file_system_type *fs_type, int flags, struct super_block *s; struct nfs_server *server; struct dentry *mntroot; + int (*compare_super)(struct super_block *, void *) = nfs_compare_super; int error; dprintk("--> nfs4_xdev_get_sb()\n"); @@ -1751,8 +1777,11 @@ static int nfs4_xdev_get_sb(struct file_system_type *fs_type, int flags, goto out_err_noserver; } + if (server->flags & NFS4_MOUNT_UNSHARED) + compare_super = NULL; + /* Get a superblock - note that we may end up sharing one that already exists */ - s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); + s = sget(&nfs_fs_type, compare_super, nfs_set_super, server); if (IS_ERR(s)) { error = PTR_ERR(s); goto out_err_nosb; @@ -1807,6 +1836,7 @@ static int nfs4_referral_get_sb(struct file_system_type *fs_type, int flags, struct nfs_server *server; struct dentry *mntroot; struct nfs_fh mntfh; + int (*compare_super)(struct super_block *, void *) = nfs_compare_super; int error; dprintk("--> nfs4_referral_get_sb()\n"); @@ -1818,8 +1848,11 @@ static int nfs4_referral_get_sb(struct file_system_type *fs_type, int flags, goto out_err_noserver; } + if (server->flags & NFS4_MOUNT_UNSHARED) + compare_super = NULL; + /* Get a superblock - note that we may end up sharing one that already exists */ - s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); + s = sget(&nfs_fs_type, compare_super, nfs_set_super, server); if (IS_ERR(s)) { error = PTR_ERR(s); goto out_err_nosb; diff --git a/include/linux/nfs4_mount.h b/include/linux/nfs4_mount.h index d8d7480e5a47..a0dcf6655657 100644 --- a/include/linux/nfs4_mount.h +++ b/include/linux/nfs4_mount.h @@ -65,6 +65,7 @@ struct nfs4_mount_data { #define NFS4_MOUNT_NOCTO 0x0010 /* 1 */ #define NFS4_MOUNT_NOAC 0x0020 /* 1 */ #define NFS4_MOUNT_STRICTLOCK 0x1000 /* 1 */ -#define NFS4_MOUNT_FLAGMASK 0x1033 +#define NFS4_MOUNT_UNSHARED 0x8000 /* 1 */ +#define NFS4_MOUNT_FLAGMASK 0x9033 #endif diff --git a/include/linux/nfs_mount.h b/include/linux/nfs_mount.h index 0b82a17c705b..a3ade89a64d2 100644 --- a/include/linux/nfs_mount.h +++ b/include/linux/nfs_mount.h @@ -62,6 +62,7 @@ struct nfs_mount_data { #define NFS_MOUNT_STRICTLOCK 0x1000 /* reserved for NFSv4 */ #define NFS_MOUNT_SECFLAVOUR 0x2000 /* 5 */ #define NFS_MOUNT_NORDIRPLUS 0x4000 /* 5 */ +#define NFS_MOUNT_UNSHARED 0x8000 /* 5 */ #define NFS_MOUNT_FLAGMASK 0xFFFF #endif -- cgit v1.2.3 From 96802a095171f5b35cf0e1e0d4be943e6696a253 Mon Sep 17 00:00:00 2001 From: Frank van Maarseveen Date: Sun, 8 Jul 2007 13:08:54 +0200 Subject: SUNRPC: cleanup transport creation argument passing Cleanup argument passing to functions for creating an RPC transport. Signed-off-by: Frank van Maarseveen Signed-off-by: Trond Myklebust --- include/linux/sunrpc/xprt.h | 15 +++++++++++---- net/sunrpc/clnt.c | 9 +++++++-- net/sunrpc/xprt.c | 15 ++++++--------- net/sunrpc/xprtsock.c | 36 ++++++++++++++++-------------------- 4 files changed, 40 insertions(+), 35 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h index 34f7590506fa..ea828b09e4ad 100644 --- a/include/linux/sunrpc/xprt.h +++ b/include/linux/sunrpc/xprt.h @@ -17,6 +17,8 @@ #include #include +#ifdef __KERNEL__ + extern unsigned int xprt_udp_slot_table_entries; extern unsigned int xprt_tcp_slot_table_entries; @@ -194,7 +196,12 @@ struct rpc_xprt { char * address_strings[RPC_DISPLAY_MAX]; }; -#ifdef __KERNEL__ +struct rpc_xprtsock_create { + int proto; /* IPPROTO_UDP or IPPROTO_TCP */ + struct sockaddr * dstaddr; /* remote peer address */ + size_t addrlen; + struct rpc_timeout * timeout; /* optional timeout parameters */ +}; /* * Transport operations used by ULPs @@ -204,7 +211,7 @@ void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long /* * Generic internal transport functions */ -struct rpc_xprt * xprt_create_transport(int proto, struct sockaddr *addr, size_t size, struct rpc_timeout *toparms); +struct rpc_xprt * xprt_create_transport(struct rpc_xprtsock_create *args); void xprt_connect(struct rpc_task *task); void xprt_reserve(struct rpc_task *task); int xprt_reserve_xprt(struct rpc_task *task); @@ -242,8 +249,8 @@ void xprt_disconnect(struct rpc_xprt *xprt); /* * Socket transport setup operations */ -struct rpc_xprt * xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to); -struct rpc_xprt * xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to); +struct rpc_xprt * xs_setup_udp(struct rpc_xprtsock_create *args); +struct rpc_xprt * xs_setup_tcp(struct rpc_xprtsock_create *args); int init_socket_xprt(void); void cleanup_socket_xprt(void); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 0d9b5275fac3..25bbf2d7f603 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -234,10 +234,15 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args) { struct rpc_xprt *xprt; struct rpc_clnt *clnt; + struct rpc_xprtsock_create xprtargs = { + .proto = args->protocol, + .dstaddr = args->address, + .addrlen = args->addrsize, + .timeout = args->timeout + }; char servername[20]; - xprt = xprt_create_transport(args->protocol, args->address, - args->addrsize, args->timeout); + xprt = xprt_create_transport(&xprtargs); if (IS_ERR(xprt)) return (struct rpc_clnt *)xprt; diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c index 518acb74a5bb..c8c2edccad7e 100644 --- a/net/sunrpc/xprt.c +++ b/net/sunrpc/xprt.c @@ -886,27 +886,24 @@ void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long i /** * xprt_create_transport - create an RPC transport - * @proto: requested transport protocol - * @ap: remote peer address - * @size: length of address - * @to: timeout parameters + * @args: rpc transport creation arguments * */ -struct rpc_xprt *xprt_create_transport(int proto, struct sockaddr *ap, size_t size, struct rpc_timeout *to) +struct rpc_xprt *xprt_create_transport(struct rpc_xprtsock_create *args) { struct rpc_xprt *xprt; struct rpc_rqst *req; - switch (proto) { + switch (args->proto) { case IPPROTO_UDP: - xprt = xs_setup_udp(ap, size, to); + xprt = xs_setup_udp(args); break; case IPPROTO_TCP: - xprt = xs_setup_tcp(ap, size, to); + xprt = xs_setup_tcp(args); break; default: printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n", - proto); + args->proto); return ERR_PTR(-EIO); } if (IS_ERR(xprt)) { diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index a8f7c5fd752c..6b7cea57651c 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1502,12 +1502,12 @@ static struct rpc_xprt_ops xs_tcp_ops = { .print_stats = xs_tcp_print_stats, }; -static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, unsigned int slot_table_size) +static struct rpc_xprt *xs_setup_xprt(struct rpc_xprtsock_create *args, unsigned int slot_table_size) { struct rpc_xprt *xprt; struct sock_xprt *new; - if (addrlen > sizeof(xprt->addr)) { + if (args->addrlen > sizeof(xprt->addr)) { dprintk("RPC: xs_setup_xprt: address too large\n"); return ERR_PTR(-EBADF); } @@ -1529,8 +1529,8 @@ static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, uns return ERR_PTR(-ENOMEM); } - memcpy(&xprt->addr, addr, addrlen); - xprt->addrlen = addrlen; + memcpy(&xprt->addr, args->dstaddr, args->addrlen); + xprt->addrlen = args->addrlen; new->port = xs_get_random_port(); return xprt; @@ -1538,22 +1538,20 @@ static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, uns /** * xs_setup_udp - Set up transport to use a UDP socket - * @addr: address of remote server - * @addrlen: length of address in bytes - * @to: timeout parameters + * @args: rpc transport creation arguments * */ -struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to) +struct rpc_xprt *xs_setup_udp(struct rpc_xprtsock_create *args) { struct rpc_xprt *xprt; struct sock_xprt *transport; - xprt = xs_setup_xprt(addr, addrlen, xprt_udp_slot_table_entries); + xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries); if (IS_ERR(xprt)) return xprt; transport = container_of(xprt, struct sock_xprt, xprt); - if (ntohs(((struct sockaddr_in *)addr)->sin_port) != 0) + if (ntohs(((struct sockaddr_in *)args->dstaddr)->sin_port) != 0) xprt_set_bound(xprt); xprt->prot = IPPROTO_UDP; @@ -1569,8 +1567,8 @@ struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_ xprt->ops = &xs_udp_ops; - if (to) - xprt->timeout = *to; + if (args->timeout) + xprt->timeout = *args->timeout; else xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); @@ -1583,22 +1581,20 @@ struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_ /** * xs_setup_tcp - Set up transport to use a TCP socket - * @addr: address of remote server - * @addrlen: length of address in bytes - * @to: timeout parameters + * @args: rpc transport creation arguments * */ -struct rpc_xprt *xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to) +struct rpc_xprt *xs_setup_tcp(struct rpc_xprtsock_create *args) { struct rpc_xprt *xprt; struct sock_xprt *transport; - xprt = xs_setup_xprt(addr, addrlen, xprt_tcp_slot_table_entries); + xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries); if (IS_ERR(xprt)) return xprt; transport = container_of(xprt, struct sock_xprt, xprt); - if (ntohs(((struct sockaddr_in *)addr)->sin_port) != 0) + if (ntohs(((struct sockaddr_in *)args->dstaddr)->sin_port) != 0) xprt_set_bound(xprt); xprt->prot = IPPROTO_TCP; @@ -1613,8 +1609,8 @@ struct rpc_xprt *xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_ xprt->ops = &xs_tcp_ops; - if (to) - xprt->timeout = *to; + if (args->timeout) + xprt->timeout = *args->timeout; else xprt_set_timeout(&xprt->timeout, 2, 60 * HZ); -- cgit v1.2.3 From a97476926ec061f90b77da478620ea6dc71a3237 Mon Sep 17 00:00:00 2001 From: Frank van Maarseveen Date: Mon, 9 Jul 2007 22:21:39 +0200 Subject: SUNRPC server: record the destination address of a request Save the destination address of an incoming request over TCP like is done already for UDP. It is necessary later for callbacks by the server. Signed-off-by: Frank van Maarseveen Signed-off-by: Trond Myklebust --- include/linux/sunrpc/svcsock.h | 1 + net/sunrpc/svcsock.c | 20 ++++++++++++++++++++ 2 files changed, 21 insertions(+) (limited to 'include') diff --git a/include/linux/sunrpc/svcsock.h b/include/linux/sunrpc/svcsock.h index e21dd93ac4b7..a53e0fa855d2 100644 --- a/include/linux/sunrpc/svcsock.h +++ b/include/linux/sunrpc/svcsock.h @@ -59,6 +59,7 @@ struct svc_sock { /* cache of various info for TCP sockets */ void *sk_info_authunix; + struct sockaddr_storage sk_local; /* local address */ struct sockaddr_storage sk_remote; /* remote peer's address */ int sk_remotelen; /* length of address */ }; diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c index 5baf48de2558..64b9b8c743c4 100644 --- a/net/sunrpc/svcsock.c +++ b/net/sunrpc/svcsock.c @@ -644,6 +644,7 @@ svc_recvfrom(struct svc_rqst *rqstp, struct kvec *iov, int nr, int buflen) struct msghdr msg = { .msg_flags = MSG_DONTWAIT, }; + struct sockaddr *sin; int len; len = kernel_recvmsg(svsk->sk_sock, &msg, iov, nr, buflen, @@ -654,6 +655,19 @@ svc_recvfrom(struct svc_rqst *rqstp, struct kvec *iov, int nr, int buflen) memcpy(&rqstp->rq_addr, &svsk->sk_remote, svsk->sk_remotelen); rqstp->rq_addrlen = svsk->sk_remotelen; + /* Destination address in request is needed for binding the + * source address in RPC callbacks later. + */ + sin = (struct sockaddr *)&svsk->sk_local; + switch (sin->sa_family) { + case AF_INET: + rqstp->rq_daddr.addr = ((struct sockaddr_in *)sin)->sin_addr; + break; + case AF_INET6: + rqstp->rq_daddr.addr6 = ((struct sockaddr_in6 *)sin)->sin6_addr; + break; + } + dprintk("svc: socket %p recvfrom(%p, %Zu) = %d\n", svsk, iov[0].iov_base, iov[0].iov_len, len); @@ -1064,6 +1078,12 @@ svc_tcp_accept(struct svc_sock *svsk) goto failed; memcpy(&newsvsk->sk_remote, sin, slen); newsvsk->sk_remotelen = slen; + err = kernel_getsockname(newsock, sin, &slen); + if (unlikely(err < 0)) { + dprintk("svc_tcp_accept: kernel_getsockname error %d\n", -err); + slen = offsetof(struct sockaddr, sa_data); + } + memcpy(&newsvsk->sk_local, sin, slen); svc_sock_received(newsvsk); -- cgit v1.2.3 From d3bc9a1deb8964d774af8535814cb91bf8f6def0 Mon Sep 17 00:00:00 2001 From: Frank van Maarseveen Date: Mon, 9 Jul 2007 22:23:35 +0200 Subject: SUNRPC client: add interface for binding to a local address In addition to binding to a local privileged port the NFS client should allow binding to a specific local address. This is used by the server for callbacks. The patch adds the necessary interface. Signed-off-by: Frank van Maarseveen Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 1 + include/linux/sunrpc/xprt.h | 1 + net/sunrpc/clnt.c | 1 + net/sunrpc/xprtsock.c | 24 ++++++++++++++++-------- 4 files changed, 19 insertions(+), 8 deletions(-) (limited to 'include') diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index c1b37972b0d5..c0d9d14983b3 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -98,6 +98,7 @@ struct rpc_create_args { int protocol; struct sockaddr *address; size_t addrsize; + struct sockaddr *saddress; struct rpc_timeout *timeout; char *servername; struct rpc_program *program; diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h index ea828b09e4ad..d11cedd14f0f 100644 --- a/include/linux/sunrpc/xprt.h +++ b/include/linux/sunrpc/xprt.h @@ -198,6 +198,7 @@ struct rpc_xprt { struct rpc_xprtsock_create { int proto; /* IPPROTO_UDP or IPPROTO_TCP */ + struct sockaddr * srcaddr; /* optional local address */ struct sockaddr * dstaddr; /* remote peer address */ size_t addrlen; struct rpc_timeout * timeout; /* optional timeout parameters */ diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 25bbf2d7f603..5d3fe7b22488 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -236,6 +236,7 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args) struct rpc_clnt *clnt; struct rpc_xprtsock_create xprtargs = { .proto = args->protocol, + .srcaddr = args->saddress, .dstaddr = args->address, .addrlen = args->addrsize, .timeout = args->timeout diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 6b7cea57651c..4ae7eed7f617 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -235,6 +235,7 @@ struct sock_xprt { * Connection of transports */ struct delayed_work connect_worker; + struct sockaddr_storage addr; unsigned short port; /* @@ -1145,31 +1146,36 @@ static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) sap->sin_port = htons(port); } -static int xs_bindresvport(struct sock_xprt *transport, struct socket *sock) +static int xs_bind(struct sock_xprt *transport, struct socket *sock) { struct sockaddr_in myaddr = { .sin_family = AF_INET, }; + struct sockaddr_in *sa; int err; unsigned short port = transport->port; + if (!transport->xprt.resvport) + port = 0; + sa = (struct sockaddr_in *)&transport->addr; + myaddr.sin_addr = sa->sin_addr; do { myaddr.sin_port = htons(port); err = kernel_bind(sock, (struct sockaddr *) &myaddr, sizeof(myaddr)); + if (!transport->xprt.resvport) + break; if (err == 0) { transport->port = port; - dprintk("RPC: xs_bindresvport bound to port %u\n", - port); - return 0; + break; } if (port <= xprt_min_resvport) port = xprt_max_resvport; else port--; } while (err == -EADDRINUSE && port != transport->port); - - dprintk("RPC: can't bind to reserved port (%d).\n", -err); + dprintk("RPC: xs_bind "NIPQUAD_FMT":%u: %s (%d)\n", + NIPQUAD(myaddr.sin_addr), port, err ? "failed" : "ok", err); return err; } @@ -1228,7 +1234,7 @@ static void xs_udp_connect_worker(struct work_struct *work) } xs_reclassify_socket(sock); - if (xprt->resvport && xs_bindresvport(transport, sock) < 0) { + if (xs_bind(transport, sock)) { sock_release(sock); goto out; } @@ -1315,7 +1321,7 @@ static void xs_tcp_connect_worker(struct work_struct *work) } xs_reclassify_socket(sock); - if (xprt->resvport && xs_bindresvport(transport, sock) < 0) { + if (xs_bind(transport, sock)) { sock_release(sock); goto out; } @@ -1531,6 +1537,8 @@ static struct rpc_xprt *xs_setup_xprt(struct rpc_xprtsock_create *args, unsigned memcpy(&xprt->addr, args->dstaddr, args->addrlen); xprt->addrlen = args->addrlen; + if (args->srcaddr) + memcpy(&new->addr, args->srcaddr, args->addrlen); new->port = xs_get_random_port(); return xprt; -- cgit v1.2.3 From c98451bdb2f3e6d6cc1e03adad641e9497512b49 Mon Sep 17 00:00:00 2001 From: Frank van Maarseveen Date: Mon, 9 Jul 2007 22:25:29 +0200 Subject: NLM: fix source address of callback to client Use the destination address of the original NLM request as the source address in callbacks to the client. Signed-off-by: Frank van Maarseveen Signed-off-by: Trond Myklebust --- fs/lockd/host.c | 27 +++++++++++++++++++-------- include/linux/lockd/lockd.h | 1 + 2 files changed, 20 insertions(+), 8 deletions(-) (limited to 'include') diff --git a/fs/lockd/host.c b/fs/lockd/host.c index c252a1c95857..572601e98dcd 100644 --- a/fs/lockd/host.c +++ b/fs/lockd/host.c @@ -44,9 +44,8 @@ static struct nsm_handle * nsm_find(const struct sockaddr_in *sin, */ static struct nlm_host * nlm_lookup_host(int server, const struct sockaddr_in *sin, - int proto, int version, - const char *hostname, - int hostname_len) + int proto, int version, const char *hostname, + int hostname_len, const struct sockaddr_in *ssin) { struct hlist_head *chain; struct hlist_node *pos; @@ -54,7 +53,9 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin, struct nsm_handle *nsm = NULL; int hash; - dprintk("lockd: nlm_lookup_host(%u.%u.%u.%u, p=%d, v=%d, my role=%s, name=%.*s)\n", + dprintk("lockd: nlm_lookup_host("NIPQUAD_FMT"->"NIPQUAD_FMT + ", p=%d, v=%d, my role=%s, name=%.*s)\n", + NIPQUAD(ssin->sin_addr.s_addr), NIPQUAD(sin->sin_addr.s_addr), proto, version, server? "server" : "client", hostname_len, @@ -91,6 +92,8 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin, continue; if (host->h_server != server) continue; + if (!nlm_cmp_addr(&host->h_saddr, ssin)) + continue; /* Move to head of hash chain. */ hlist_del(&host->h_hash); @@ -118,6 +121,7 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin, host->h_name = nsm->sm_name; host->h_addr = *sin; host->h_addr.sin_port = 0; /* ouch! */ + host->h_saddr = *ssin; host->h_version = version; host->h_proto = proto; host->h_rpcclnt = NULL; @@ -174,8 +178,10 @@ struct nlm_host * nlmclnt_lookup_host(const struct sockaddr_in *sin, int proto, int version, const char *hostname, int hostname_len) { + struct sockaddr_in ssin = {0}; + return nlm_lookup_host(0, sin, proto, version, - hostname, hostname_len); + hostname, hostname_len, &ssin); } /* @@ -185,9 +191,12 @@ struct nlm_host * nlmsvc_lookup_host(struct svc_rqst *rqstp, const char *hostname, int hostname_len) { + struct sockaddr_in ssin = {0}; + + ssin.sin_addr = rqstp->rq_daddr.addr; return nlm_lookup_host(1, svc_addr_in(rqstp), rqstp->rq_prot, rqstp->rq_vers, - hostname, hostname_len); + hostname, hostname_len, &ssin); } /* @@ -198,8 +207,9 @@ nlm_bind_host(struct nlm_host *host) { struct rpc_clnt *clnt; - dprintk("lockd: nlm_bind_host(%08x)\n", - (unsigned)ntohl(host->h_addr.sin_addr.s_addr)); + dprintk("lockd: nlm_bind_host("NIPQUAD_FMT"->"NIPQUAD_FMT")\n", + NIPQUAD(host->h_saddr.sin_addr), + NIPQUAD(host->h_addr.sin_addr)); /* Lock host handle */ mutex_lock(&host->h_mutex); @@ -226,6 +236,7 @@ nlm_bind_host(struct nlm_host *host) .protocol = host->h_proto, .address = (struct sockaddr *)&host->h_addr, .addrsize = sizeof(host->h_addr), + .saddress = (struct sockaddr *)&host->h_saddr, .timeout = &timeparms, .servername = host->h_name, .program = &nlm_program, diff --git a/include/linux/lockd/lockd.h b/include/linux/lockd/lockd.h index 05707e2fccae..e2d1ce36b367 100644 --- a/include/linux/lockd/lockd.h +++ b/include/linux/lockd/lockd.h @@ -39,6 +39,7 @@ struct nlm_host { struct hlist_node h_hash; /* doubly linked list */ struct sockaddr_in h_addr; /* peer address */ + struct sockaddr_in h_saddr; /* our address (optional) */ struct rpc_clnt * h_rpcclnt; /* RPC client to talk to peer */ char * h_name; /* remote hostname */ u32 h_version; /* interface version */ -- cgit v1.2.3