From cf4ab538f1516606d3ae730dce15d6f33d96b7e1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 8 Mar 2013 12:56:37 -0500 Subject: NFSv4: Fix the string length returned by the idmapper Functions like nfs_map_uid_to_name() and nfs_map_gid_to_group() are expected to return a string without any terminating NUL character. Regression introduced by commit 57e62324e469e092ecc6c94a7a86fe4bd6ac5172 (NFS: Store the legacy idmapper result in the keyring). Reported-by: Dave Chiluk Signed-off-by: Trond Myklebust Cc: Bryan Schumaker Cc: stable@vger.kernel.org [>=3.4] --- fs/nfs/idmap.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c index dc0f98dfa717..c516da5873fd 100644 --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -726,9 +726,9 @@ out1: return ret; } -static int nfs_idmap_instantiate(struct key *key, struct key *authkey, char *data) +static int nfs_idmap_instantiate(struct key *key, struct key *authkey, char *data, size_t datalen) { - return key_instantiate_and_link(key, data, strlen(data) + 1, + return key_instantiate_and_link(key, data, datalen, id_resolver_cache->thread_keyring, authkey); } @@ -738,6 +738,7 @@ static int nfs_idmap_read_and_verify_message(struct idmap_msg *im, struct key *key, struct key *authkey) { char id_str[NFS_UINT_MAXLEN]; + size_t len; int ret = -ENOKEY; /* ret = -ENOKEY */ @@ -747,13 +748,15 @@ static int nfs_idmap_read_and_verify_message(struct idmap_msg *im, case IDMAP_CONV_NAMETOID: if (strcmp(upcall->im_name, im->im_name) != 0) break; - sprintf(id_str, "%d", im->im_id); - ret = nfs_idmap_instantiate(key, authkey, id_str); + /* Note: here we store the NUL terminator too */ + len = sprintf(id_str, "%d", im->im_id) + 1; + ret = nfs_idmap_instantiate(key, authkey, id_str, len); break; case IDMAP_CONV_IDTONAME: if (upcall->im_id != im->im_id) break; - ret = nfs_idmap_instantiate(key, authkey, im->im_name); + len = strlen(im->im_name); + ret = nfs_idmap_instantiate(key, authkey, im->im_name, len); break; default: ret = -EINVAL; -- cgit v1.2.3-59-g8ed1b From 4376c94618c26225e69e17b7c91169c45a90b292 Mon Sep 17 00:00:00 2001 From: fanchaoting Date: Thu, 21 Mar 2013 09:15:30 +0800 Subject: pnfs-block: removing DM device maybe cause oops when call dev_remove when pnfs block using device mapper,if umounting later,it maybe cause oops. we apply "1 + sizeof(bl_umount_request)" memory for msg->data, the memory maybe overflow when we do "memcpy(&dataptr [sizeof(bl_msg)], &bl_umount_request, sizeof(bl_umount_request))", because the size of bl_msg is more than 1 byte. Signed-off-by: fanchaoting Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/blocklayoutdm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/nfs/blocklayout/blocklayoutdm.c b/fs/nfs/blocklayout/blocklayoutdm.c index 737d839bc17b..6fc7b5cae92b 100644 --- a/fs/nfs/blocklayout/blocklayoutdm.c +++ b/fs/nfs/blocklayout/blocklayoutdm.c @@ -55,7 +55,8 @@ static void dev_remove(struct net *net, dev_t dev) bl_pipe_msg.bl_wq = &nn->bl_wq; memset(msg, 0, sizeof(*msg)); - msg->data = kzalloc(1 + sizeof(bl_umount_request), GFP_NOFS); + msg->len = sizeof(bl_msg) + bl_msg.totallen; + msg->data = kzalloc(msg->len, GFP_NOFS); if (!msg->data) goto out; @@ -66,7 +67,6 @@ static void dev_remove(struct net *net, dev_t dev) memcpy(msg->data, &bl_msg, sizeof(bl_msg)); dataptr = (uint8_t *) msg->data; memcpy(&dataptr[sizeof(bl_msg)], &bl_umount_request, sizeof(bl_umount_request)); - msg->len = sizeof(bl_msg) + bl_msg.totallen; add_wait_queue(&nn->bl_wq, &wq); if (rpc_queue_upcall(nn->bl_device_pipe, msg) < 0) { -- cgit v1.2.3-59-g8ed1b From a073dbff359f4741013ae4b8395f5364c5e00b48 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 20 Mar 2013 12:34:32 -0400 Subject: NFSv4.1: Fix a race in pNFS layoutcommit We need to clear the NFS_LSEG_LAYOUTCOMMIT bits atomically with the NFS_INO_LAYOUTCOMMIT bit, otherwise we may end up with situations where the two are out of sync. The first half of the problem is to ensure that pnfs_layoutcommit_inode clears the NFS_LSEG_LAYOUTCOMMIT bit through pnfs_list_write_lseg. We still need to keep the reference to those segments until the RPC call is finished, so in order to make it clear _where_ those references come from, we add a helper pnfs_list_write_lseg_done() that cleans up after pnfs_list_write_lseg. Signed-off-by: Trond Myklebust Acked-by: Benny Halevy Cc: stable@vger.kernel.org --- fs/nfs/nfs4proc.c | 14 -------------- fs/nfs/pnfs.c | 19 ++++++++++++++++++- 2 files changed, 18 insertions(+), 15 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index b2671cb0f901..6ccdd4fd9b59 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -6416,22 +6416,8 @@ nfs4_layoutcommit_done(struct rpc_task *task, void *calldata) static void nfs4_layoutcommit_release(void *calldata) { struct nfs4_layoutcommit_data *data = calldata; - struct pnfs_layout_segment *lseg, *tmp; - unsigned long *bitlock = &NFS_I(data->args.inode)->flags; pnfs_cleanup_layoutcommit(data); - /* Matched by references in pnfs_set_layoutcommit */ - list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) { - list_del_init(&lseg->pls_lc_list); - if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT, - &lseg->pls_flags)) - pnfs_put_lseg(lseg); - } - - clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock); - smp_mb__after_clear_bit(); - wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING); - put_rpccred(data->cred); kfree(data); } diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 48ac5aad6258..3d900916fd41 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1746,11 +1746,27 @@ static void pnfs_list_write_lseg(struct inode *inode, struct list_head *listp) list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list) { if (lseg->pls_range.iomode == IOMODE_RW && - test_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) + test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) list_add(&lseg->pls_lc_list, listp); } } +static void pnfs_list_write_lseg_done(struct inode *inode, struct list_head *listp) +{ + struct pnfs_layout_segment *lseg, *tmp; + unsigned long *bitlock = &NFS_I(inode)->flags; + + /* Matched by references in pnfs_set_layoutcommit */ + list_for_each_entry_safe(lseg, tmp, listp, pls_lc_list) { + list_del_init(&lseg->pls_lc_list); + pnfs_put_lseg(lseg); + } + + clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock); + smp_mb__after_clear_bit(); + wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING); +} + void pnfs_set_lo_fail(struct pnfs_layout_segment *lseg) { pnfs_layout_io_set_failed(lseg->pls_layout, lseg->pls_range.iomode); @@ -1795,6 +1811,7 @@ void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data) if (nfss->pnfs_curr_ld->cleanup_layoutcommit) nfss->pnfs_curr_ld->cleanup_layoutcommit(data); + pnfs_list_write_lseg_done(data->args.inode, &data->lseg_list); } /* -- cgit v1.2.3-59-g8ed1b From 24956804349ca0eadcdde032d65e8c00b4214096 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 20 Mar 2013 13:03:00 -0400 Subject: NFSv4.1: Always clear the NFS_INO_LAYOUTCOMMIT in layoutreturn Note that clearing NFS_INO_LAYOUTCOMMIT is tricky, since it requires you to also clear the NFS_LSEG_LAYOUTCOMMIT bits from the layout segments. The only two sites that need to do this are the ones that call pnfs_return_layout() without first doing a layout commit. Signed-off-by: Trond Myklebust Acked-by: Benny Halevy Cc: stable@vger.kernel.org --- fs/nfs/nfs4filelayout.c | 1 - fs/nfs/pnfs.c | 35 +++++++++++++++++++++++++++-------- 2 files changed, 27 insertions(+), 9 deletions(-) diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 49eeb044c109..4fb234d3aefb 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -129,7 +129,6 @@ static void filelayout_fenceme(struct inode *inode, struct pnfs_layout_hdr *lo) { if (!test_and_clear_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) return; - clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(inode)->flags); pnfs_return_layout(inode); } diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 3d900916fd41..5044142c1216 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -417,6 +417,16 @@ should_free_lseg(struct pnfs_layout_range *lseg_range, lo_seg_intersecting(lseg_range, recall_range); } +static bool pnfs_lseg_dec_and_remove_zero(struct pnfs_layout_segment *lseg, + struct list_head *tmp_list) +{ + if (!atomic_dec_and_test(&lseg->pls_refcount)) + return false; + pnfs_layout_remove_lseg(lseg->pls_layout, lseg); + list_add(&lseg->pls_list, tmp_list); + return true; +} + /* Returns 1 if lseg is removed from list, 0 otherwise */ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, struct list_head *tmp_list) @@ -430,11 +440,8 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, */ dprintk("%s: lseg %p ref %d\n", __func__, lseg, atomic_read(&lseg->pls_refcount)); - if (atomic_dec_and_test(&lseg->pls_refcount)) { - pnfs_layout_remove_lseg(lseg->pls_layout, lseg); - list_add(&lseg->pls_list, tmp_list); + if (pnfs_lseg_dec_and_remove_zero(lseg, tmp_list)) rv = 1; - } } return rv; } @@ -777,6 +784,21 @@ send_layoutget(struct pnfs_layout_hdr *lo, return lseg; } +static void pnfs_clear_layoutcommit(struct inode *inode, + struct list_head *head) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct pnfs_layout_segment *lseg, *tmp; + + if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) + return; + list_for_each_entry_safe(lseg, tmp, &nfsi->layout->plh_segs, pls_list) { + if (!test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) + continue; + pnfs_lseg_dec_and_remove_zero(lseg, head); + } +} + /* * Initiates a LAYOUTRETURN(FILE), and removes the pnfs_layout_hdr * when the layout segment list is empty. @@ -808,6 +830,7 @@ _pnfs_return_layout(struct inode *ino) /* Reference matched in nfs4_layoutreturn_release */ pnfs_get_layout_hdr(lo); empty = list_empty(&lo->plh_segs); + pnfs_clear_layoutcommit(ino, &tmp_list); pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL); /* Don't send a LAYOUTRETURN if list was initially empty */ if (empty) { @@ -820,8 +843,6 @@ _pnfs_return_layout(struct inode *ino) spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&tmp_list); - WARN_ON(test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)); - lrp = kzalloc(sizeof(*lrp), GFP_KERNEL); if (unlikely(lrp == NULL)) { status = -ENOMEM; @@ -1458,7 +1479,6 @@ static void pnfs_ld_handle_write_error(struct nfs_write_data *data) dprintk("pnfs write error = %d\n", hdr->pnfs_error); if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & PNFS_LAYOUTRET_ON_ERROR) { - clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); pnfs_return_layout(hdr->inode); } if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) @@ -1613,7 +1633,6 @@ static void pnfs_ld_handle_read_error(struct nfs_read_data *data) dprintk("pnfs read error = %d\n", hdr->pnfs_error); if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & PNFS_LAYOUTRET_ON_ERROR) { - clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); pnfs_return_layout(hdr->inode); } if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) -- cgit v1.2.3-59-g8ed1b From 240286725d854331422cb15957f8d9bf2741d4e3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 20 Mar 2013 13:23:33 -0400 Subject: NFSv4.1: Add a helper pnfs_commit_and_return_layout In order to be able to safely return the layout in nfs4_proc_setattr, we need to block new uses of the layout, wait for all outstanding users of the layout to complete, commit the layout and then return it. This patch adds a helper in order to do all this safely. Signed-off-by: Trond Myklebust Cc: Boaz Harrosh --- fs/nfs/nfs4proc.c | 2 +- fs/nfs/pnfs.c | 27 +++++++++++++++++++++++++++ fs/nfs/pnfs.h | 6 ++++++ 3 files changed, 34 insertions(+), 1 deletion(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 6ccdd4fd9b59..26431cf62ddb 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2632,7 +2632,7 @@ nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr, int status; if (pnfs_ld_layoutret_on_setattr(inode)) - pnfs_return_layout(inode); + pnfs_commit_and_return_layout(inode); nfs_fattr_init(fattr); diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 5044142c1216..4bdffe0ba025 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -866,6 +866,33 @@ out: } EXPORT_SYMBOL_GPL(_pnfs_return_layout); +int +pnfs_commit_and_return_layout(struct inode *inode) +{ + struct pnfs_layout_hdr *lo; + int ret; + + spin_lock(&inode->i_lock); + lo = NFS_I(inode)->layout; + if (lo == NULL) { + spin_unlock(&inode->i_lock); + return 0; + } + pnfs_get_layout_hdr(lo); + /* Block new layoutgets and read/write to ds */ + lo->plh_block_lgets++; + spin_unlock(&inode->i_lock); + filemap_fdatawait(inode->i_mapping); + ret = pnfs_layoutcommit_inode(inode, true); + if (ret == 0) + ret = _pnfs_return_layout(inode); + spin_lock(&inode->i_lock); + lo->plh_block_lgets--; + spin_unlock(&inode->i_lock); + pnfs_put_layout_hdr(lo); + return ret; +} + bool pnfs_roc(struct inode *ino) { struct pnfs_layout_hdr *lo; diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 94ba80417748..f5f8a470a647 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -219,6 +219,7 @@ void pnfs_set_layoutcommit(struct nfs_write_data *wdata); void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data); int pnfs_layoutcommit_inode(struct inode *inode, bool sync); int _pnfs_return_layout(struct inode *); +int pnfs_commit_and_return_layout(struct inode *); void pnfs_ld_write_done(struct nfs_write_data *); void pnfs_ld_read_done(struct nfs_read_data *); struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino, @@ -407,6 +408,11 @@ static inline int pnfs_return_layout(struct inode *ino) return 0; } +static inline int pnfs_commit_and_return_layout(struct inode *inode) +{ + return 0; +} + static inline bool pnfs_ld_layoutret_on_setattr(struct inode *inode) { -- cgit v1.2.3-59-g8ed1b From 1166fde6a923c30f4351515b6a9a1efc513e7d00 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 25 Mar 2013 11:23:40 -0400 Subject: SUNRPC: Add barriers to ensure read ordering in rpc_wake_up_task_queue_locked We need to be careful when testing task->tk_waitqueue in rpc_wake_up_task_queue_locked, because it can be changed while we are holding the queue->lock. By adding appropriate memory barriers, we can ensure that it is safe to test task->tk_waitqueue for equality if the RPC_TASK_QUEUED bit is set. Signed-off-by: Trond Myklebust Cc: stable@vger.kernel.org --- net/sunrpc/sched.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index fb20f25ddec9..f8529fc8e542 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -180,6 +180,8 @@ static void __rpc_add_wait_queue(struct rpc_wait_queue *queue, list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]); task->tk_waitqueue = queue; queue->qlen++; + /* barrier matches the read in rpc_wake_up_task_queue_locked() */ + smp_wmb(); rpc_set_queued(task); dprintk("RPC: %5u added to queue %p \"%s\"\n", @@ -430,8 +432,11 @@ static void __rpc_do_wake_up_task(struct rpc_wait_queue *queue, struct rpc_task */ static void rpc_wake_up_task_queue_locked(struct rpc_wait_queue *queue, struct rpc_task *task) { - if (RPC_IS_QUEUED(task) && task->tk_waitqueue == queue) - __rpc_do_wake_up_task(queue, task); + if (RPC_IS_QUEUED(task)) { + smp_rmb(); + if (task->tk_waitqueue == queue) + __rpc_do_wake_up_task(queue, task); + } } /* -- cgit v1.2.3-59-g8ed1b