diff options
Diffstat (limited to 'fs/nfs/direct.c')
| -rw-r--r-- | fs/nfs/direct.c | 1386 |
1 files changed, 802 insertions, 584 deletions
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index b497c71384e..f11b9eed0de 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -7,11 +7,11 @@ * * There are important applications whose performance or correctness * depends on uncached access to file data. Database clusters - * (multiple copies of the same instance running on separate hosts) + * (multiple copies of the same instance running on separate hosts) * implement their own cache coherency protocol that subsumes file - * system cache protocols. Applications that process datasets - * considerably larger than the client's memory do not always benefit - * from a local cache. A streaming video server, for instance, has no + * system cache protocols. Applications that process datasets + * considerably larger than the client's memory do not always benefit + * from a local cache. A streaming video server, for instance, has no * need to cache the contents of a file. * * When an application requests uncached I/O, all read and write requests @@ -34,674 +34,857 @@ * 08 Jun 2003 Port to 2.5 APIs --cel * 31 Mar 2004 Handle direct I/O without VFS support --cel * 15 Sep 2004 Parallel async reads --cel + * 04 May 2005 support O_DIRECT with aio --cel * */ -#include <linux/config.h> #include <linux/errno.h> #include <linux/sched.h> #include <linux/kernel.h> -#include <linux/smp_lock.h> #include <linux/file.h> #include <linux/pagemap.h> #include <linux/kref.h> +#include <linux/slab.h> +#include <linux/task_io_accounting_ops.h> +#include <linux/module.h> #include <linux/nfs_fs.h> #include <linux/nfs_page.h> #include <linux/sunrpc/clnt.h> -#include <asm/system.h> #include <asm/uaccess.h> -#include <asm/atomic.h> +#include <linux/atomic.h> + +#include "internal.h" +#include "iostat.h" +#include "pnfs.h" #define NFSDBG_FACILITY NFSDBG_VFS -#define MAX_DIRECTIO_SIZE (4096UL << PAGE_SHIFT) -static kmem_cache_t *nfs_direct_cachep; +static struct kmem_cache *nfs_direct_cachep; /* * This represents a set of asynchronous requests that we're waiting on */ struct nfs_direct_req { struct kref kref; /* release manager */ - struct list_head list; /* nfs_read_data structs */ - wait_queue_head_t wait; /* wait for i/o completion */ - struct page ** pages; /* pages in our buffer */ - unsigned int npages; /* count of pages */ - atomic_t complete, /* i/os we're waiting for */ - count, /* bytes actually processed */ + + /* I/O parameters */ + struct nfs_open_context *ctx; /* file open context info */ + struct nfs_lock_context *l_ctx; /* Lock context info */ + struct kiocb * iocb; /* controlling i/o request */ + struct inode * inode; /* target file of i/o */ + + /* completion state */ + atomic_t io_count; /* i/os we're waiting for */ + spinlock_t lock; /* protect completion state */ + ssize_t count, /* bytes actually processed */ + bytes_left, /* bytes left to be sent */ error; /* any reported error */ + struct completion completion; /* wait for i/o completion */ + + /* commit state */ + struct nfs_mds_commit_info mds_cinfo; /* Storage for cinfo */ + struct pnfs_ds_commit_info ds_cinfo; /* Storage for cinfo */ + struct work_struct work; + int flags; +#define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */ +#define NFS_ODIRECT_RESCHED_WRITES (2) /* write verification failed */ + struct nfs_writeverf verf; /* unstable write verifier */ }; +static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops; +static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops; +static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode); +static void nfs_direct_write_schedule_work(struct work_struct *work); -/** - * nfs_get_user_pages - find and set up pages underlying user's buffer - * rw: direction (read or write) - * user_addr: starting address of this segment of user's buffer - * count: size of this segment - * @pages: returned array of page struct pointers underlying user's buffer +static inline void get_dreq(struct nfs_direct_req *dreq) +{ + atomic_inc(&dreq->io_count); +} + +static inline int put_dreq(struct nfs_direct_req *dreq) +{ + return atomic_dec_and_test(&dreq->io_count); +} + +/* + * nfs_direct_select_verf - select the right verifier + * @dreq - direct request possibly spanning multiple servers + * @ds_clp - nfs_client of data server or NULL if MDS / non-pnfs + * @ds_idx - index of data server in data server list, only valid if ds_clp set + * + * returns the correct verifier to use given the role of the server */ -static inline int -nfs_get_user_pages(int rw, unsigned long user_addr, size_t size, - struct page ***pages) -{ - int result = -ENOMEM; - unsigned long page_count; - size_t array_size; - - /* set an arbitrary limit to prevent type overflow */ - /* XXX: this can probably be as large as INT_MAX */ - if (size > MAX_DIRECTIO_SIZE) { - *pages = NULL; - return -EFBIG; +static struct nfs_writeverf * +nfs_direct_select_verf(struct nfs_direct_req *dreq, + struct nfs_client *ds_clp, + int ds_idx) +{ + struct nfs_writeverf *verfp = &dreq->verf; + +#ifdef CONFIG_NFS_V4_1 + if (ds_clp) { + /* pNFS is in use, use the DS verf */ + if (ds_idx >= 0 && ds_idx < dreq->ds_cinfo.nbuckets) + verfp = &dreq->ds_cinfo.buckets[ds_idx].direct_verf; + else + WARN_ON_ONCE(1); } +#endif + return verfp; +} - page_count = (user_addr + size + PAGE_SIZE - 1) >> PAGE_SHIFT; - page_count -= user_addr >> PAGE_SHIFT; - - array_size = (page_count * sizeof(struct page *)); - *pages = kmalloc(array_size, GFP_KERNEL); - if (*pages) { - down_read(¤t->mm->mmap_sem); - result = get_user_pages(current, current->mm, user_addr, - page_count, (rw == READ), 0, - *pages, NULL); - up_read(¤t->mm->mmap_sem); - } - return result; + +/* + * nfs_direct_set_hdr_verf - set the write/commit verifier + * @dreq - direct request possibly spanning multiple servers + * @hdr - pageio header to validate against previously seen verfs + * + * Set the server's (MDS or DS) "seen" verifier + */ +static void nfs_direct_set_hdr_verf(struct nfs_direct_req *dreq, + struct nfs_pgio_header *hdr) +{ + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, hdr->data->ds_clp, + hdr->data->ds_idx); + WARN_ON_ONCE(verfp->committed >= 0); + memcpy(verfp, &hdr->verf, sizeof(struct nfs_writeverf)); + WARN_ON_ONCE(verfp->committed < 0); } -/** - * nfs_free_user_pages - tear down page struct array - * @pages: array of page struct pointers underlying target buffer - * @npages: number of pages in the array - * @do_dirty: dirty the pages as we release them +/* + * nfs_direct_cmp_hdr_verf - compare verifier for pgio header + * @dreq - direct request possibly spanning multiple servers + * @hdr - pageio header to validate against previously seen verf + * + * set the server's "seen" verf if not initialized. + * returns result of comparison between @hdr->verf and the "seen" + * verf of the server used by @hdr (DS or MDS) */ -static void -nfs_free_user_pages(struct page **pages, int npages, int do_dirty) +static int nfs_direct_set_or_cmp_hdr_verf(struct nfs_direct_req *dreq, + struct nfs_pgio_header *hdr) { - int i; - for (i = 0; i < npages; i++) { - if (do_dirty) - set_page_dirty_lock(pages[i]); - page_cache_release(pages[i]); + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, hdr->data->ds_clp, + hdr->data->ds_idx); + if (verfp->committed < 0) { + nfs_direct_set_hdr_verf(dreq, hdr); + return 0; } - kfree(pages); + return memcmp(verfp, &hdr->verf, sizeof(struct nfs_writeverf)); } -/** - * nfs_direct_req_release - release nfs_direct_req structure for direct read - * @kref: kref object embedded in an nfs_direct_req structure +#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4) +/* + * nfs_direct_cmp_commit_data_verf - compare verifier for commit data + * @dreq - direct request possibly spanning multiple servers + * @data - commit data to validate against previously seen verf * + * returns result of comparison between @data->verf and the verf of + * the server used by @data (DS or MDS) */ -static void nfs_direct_req_release(struct kref *kref) +static int nfs_direct_cmp_commit_data_verf(struct nfs_direct_req *dreq, + struct nfs_commit_data *data) { - struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref); - kmem_cache_free(nfs_direct_cachep, dreq); + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, data->ds_clp, + data->ds_commit_index); + WARN_ON_ONCE(verfp->committed < 0); + return memcmp(verfp, &data->verf, sizeof(struct nfs_writeverf)); } +#endif /** - * nfs_direct_read_alloc - allocate nfs_read_data structures for direct read - * @count: count of bytes for the read request - * @rsize: local rsize setting + * nfs_direct_IO - NFS address space operation for direct I/O + * @rw: direction (read or write) + * @iocb: target I/O control block + * @iov: array of vectors that define I/O buffer + * @pos: offset in file to begin the operation + * @nr_segs: size of iovec array * - * Note we also set the number of requests we have in the dreq when we are - * done. This prevents races with I/O completion so we will always wait - * until all requests have been dispatched and completed. + * The presence of this routine in the address space ops vector means + * the NFS client supports direct I/O. However, for most direct IO, we + * shunt off direct read and write requests before the VFS gets them, + * so this method is only ever called for swap. */ -static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, unsigned int rsize) +ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, struct iov_iter *iter, loff_t pos) +{ +#ifndef CONFIG_NFS_SWAP + dprintk("NFS: nfs_direct_IO (%pD) off/no(%Ld/%lu) EINVAL\n", + iocb->ki_filp, (long long) pos, iter->nr_segs); + + return -EINVAL; +#else + VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE); + + if (rw == READ || rw == KERNEL_READ) + return nfs_file_direct_read(iocb, iter, pos, + rw == READ ? true : false); + return nfs_file_direct_write(iocb, iter, pos, + rw == WRITE ? true : false); +#endif /* CONFIG_NFS_SWAP */ +} + +static void nfs_direct_release_pages(struct page **pages, unsigned int npages) +{ + unsigned int i; + for (i = 0; i < npages; i++) + page_cache_release(pages[i]); +} + +void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, + struct nfs_direct_req *dreq) +{ + cinfo->lock = &dreq->lock; + cinfo->mds = &dreq->mds_cinfo; + cinfo->ds = &dreq->ds_cinfo; + cinfo->dreq = dreq; + cinfo->completion_ops = &nfs_direct_commit_completion_ops; +} + +static inline struct nfs_direct_req *nfs_direct_req_alloc(void) { - struct list_head *list; struct nfs_direct_req *dreq; - unsigned int reads = 0; - dreq = kmem_cache_alloc(nfs_direct_cachep, SLAB_KERNEL); + dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL); if (!dreq) return NULL; kref_init(&dreq->kref); - init_waitqueue_head(&dreq->wait); - INIT_LIST_HEAD(&dreq->list); - atomic_set(&dreq->count, 0); - atomic_set(&dreq->error, 0); - - list = &dreq->list; - for(;;) { - struct nfs_read_data *data = nfs_readdata_alloc(); - - if (unlikely(!data)) { - while (!list_empty(list)) { - data = list_entry(list->next, - struct nfs_read_data, pages); - list_del(&data->pages); - nfs_readdata_free(data); - } - kref_put(&dreq->kref, nfs_direct_req_release); - return NULL; - } - - INIT_LIST_HEAD(&data->pages); - list_add(&data->pages, list); - - data->req = (struct nfs_page *) dreq; - reads++; - if (nbytes <= rsize) - break; - nbytes -= rsize; - } kref_get(&dreq->kref); - atomic_set(&dreq->complete, reads); + init_completion(&dreq->completion); + INIT_LIST_HEAD(&dreq->mds_cinfo.list); + dreq->verf.committed = NFS_INVALID_STABLE_HOW; /* not set yet */ + INIT_WORK(&dreq->work, nfs_direct_write_schedule_work); + spin_lock_init(&dreq->lock); + return dreq; } -/** - * nfs_direct_read_result - handle a read reply for a direct read request - * @data: address of NFS READ operation control block - * @status: status of this NFS READ operation - * - * We must hold a reference to all the pages in this direct read request - * until the RPCs complete. This could be long *after* we are woken up in - * nfs_direct_read_wait (for instance, if someone hits ^C on a slow server). - */ -static void nfs_direct_read_result(struct nfs_read_data *data, int status) +static void nfs_direct_req_free(struct kref *kref) { - struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req; + struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref); - if (likely(status >= 0)) - atomic_add(data->res.count, &dreq->count); - else - atomic_set(&dreq->error, status); + if (dreq->l_ctx != NULL) + nfs_put_lock_context(dreq->l_ctx); + if (dreq->ctx != NULL) + put_nfs_open_context(dreq->ctx); + kmem_cache_free(nfs_direct_cachep, dreq); +} - if (unlikely(atomic_dec_and_test(&dreq->complete))) { - nfs_free_user_pages(dreq->pages, dreq->npages, 1); - wake_up(&dreq->wait); - kref_put(&dreq->kref, nfs_direct_req_release); - } +static void nfs_direct_req_release(struct nfs_direct_req *dreq) +{ + kref_put(&dreq->kref, nfs_direct_req_free); } -/** - * nfs_direct_read_schedule - dispatch NFS READ operations for a direct read - * @dreq: address of nfs_direct_req struct for this request - * @inode: target inode - * @ctx: target file open context - * @user_addr: starting address of this segment of user's buffer - * @count: size of this segment - * @file_offset: offset in file to begin the operation - * - * For each nfs_read_data struct that was allocated on the list, dispatch - * an NFS READ operation - */ -static void nfs_direct_read_schedule(struct nfs_direct_req *dreq, - struct inode *inode, struct nfs_open_context *ctx, - unsigned long user_addr, size_t count, loff_t file_offset) -{ - struct list_head *list = &dreq->list; - struct page **pages = dreq->pages; - unsigned int curpage, pgbase; - unsigned int rsize = NFS_SERVER(inode)->rsize; - - curpage = 0; - pgbase = user_addr & ~PAGE_MASK; - do { - struct nfs_read_data *data; - unsigned int bytes; - - bytes = rsize; - if (count < rsize) - bytes = count; - - data = list_entry(list->next, struct nfs_read_data, pages); - list_del_init(&data->pages); - - data->inode = inode; - data->cred = ctx->cred; - data->args.fh = NFS_FH(inode); - data->args.context = ctx; - data->args.offset = file_offset; - data->args.pgbase = pgbase; - data->args.pages = &pages[curpage]; - data->args.count = bytes; - data->res.fattr = &data->fattr; - data->res.eof = 0; - data->res.count = bytes; - - NFS_PROTO(inode)->read_setup(data); - - data->task.tk_cookie = (unsigned long) inode; - data->task.tk_calldata = data; - data->task.tk_release = nfs_readdata_release; - data->complete = nfs_direct_read_result; - - lock_kernel(); - rpc_execute(&data->task); - unlock_kernel(); - - dfprintk(VFS, "NFS: %4d initiated direct read call (req %s/%Ld, %u bytes @ offset %Lu)\n", - data->task.tk_pid, - inode->i_sb->s_id, - (long long)NFS_FILEID(inode), - bytes, - (unsigned long long)data->args.offset); - - file_offset += bytes; - pgbase += bytes; - curpage += pgbase >> PAGE_SHIFT; - pgbase &= ~PAGE_MASK; - - count -= bytes; - } while (count != 0); +ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq) +{ + return dreq->bytes_left; } +EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left); -/** - * nfs_direct_read_wait - wait for I/O completion for direct reads - * @dreq: request on which we are to wait - * @intr: whether or not this wait can be interrupted - * +/* * Collects and returns the final error value/byte-count. */ -static ssize_t nfs_direct_read_wait(struct nfs_direct_req *dreq, int intr) +static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq) { - int result = 0; + ssize_t result = -EIOCBQUEUED; - if (intr) { - result = wait_event_interruptible(dreq->wait, - (atomic_read(&dreq->complete) == 0)); - } else { - wait_event(dreq->wait, (atomic_read(&dreq->complete) == 0)); - } + /* Async requests don't wait here */ + if (dreq->iocb) + goto out; + + result = wait_for_completion_killable(&dreq->completion); if (!result) - result = atomic_read(&dreq->error); + result = dreq->error; if (!result) - result = atomic_read(&dreq->count); + result = dreq->count; - kref_put(&dreq->kref, nfs_direct_req_release); +out: return (ssize_t) result; } -/** - * nfs_direct_read_seg - Read in one iov segment. Generate separate - * read RPCs for each "rsize" bytes. - * @inode: target inode - * @ctx: target file open context - * @user_addr: starting address of this segment of user's buffer - * @count: size of this segment - * @file_offset: offset in file to begin the operation - * @pages: array of addresses of page structs defining user's buffer - * @nr_pages: number of pages in the array - * +/* + * Synchronous I/O uses a stack-allocated iocb. Thus we can't trust + * the iocb is still valid here if this is a synchronous request. */ -static ssize_t nfs_direct_read_seg(struct inode *inode, - struct nfs_open_context *ctx, unsigned long user_addr, - size_t count, loff_t file_offset, struct page **pages, - unsigned int nr_pages) -{ - ssize_t result; - sigset_t oldset; - struct rpc_clnt *clnt = NFS_CLIENT(inode); - struct nfs_direct_req *dreq; - - dreq = nfs_direct_read_alloc(count, NFS_SERVER(inode)->rsize); - if (!dreq) - return -ENOMEM; +static void nfs_direct_complete(struct nfs_direct_req *dreq, bool write) +{ + struct inode *inode = dreq->inode; - dreq->pages = pages; - dreq->npages = nr_pages; + if (dreq->iocb && write) { + loff_t pos = dreq->iocb->ki_pos + dreq->count; - rpc_clnt_sigmask(clnt, &oldset); - nfs_direct_read_schedule(dreq, inode, ctx, user_addr, count, - file_offset); - result = nfs_direct_read_wait(dreq, clnt->cl_intr); - rpc_clnt_sigunmask(clnt, &oldset); + spin_lock(&inode->i_lock); + if (i_size_read(inode) < pos) + i_size_write(inode, pos); + spin_unlock(&inode->i_lock); + } - return result; -} + if (write) + nfs_zap_mapping(inode, inode->i_mapping); -/** - * nfs_direct_read - For each iov segment, map the user's buffer - * then generate read RPCs. - * @inode: target inode - * @ctx: target file open context - * @iov: array of vectors that define I/O buffer - * file_offset: offset in file to begin the operation - * nr_segs: size of iovec array - * - * We've already pushed out any non-direct writes so that this read - * will see them when we read from the server. - */ -static ssize_t -nfs_direct_read(struct inode *inode, struct nfs_open_context *ctx, - const struct iovec *iov, loff_t file_offset, - unsigned long nr_segs) -{ - ssize_t tot_bytes = 0; - unsigned long seg = 0; - - while ((seg < nr_segs) && (tot_bytes >= 0)) { - ssize_t result; - int page_count; - struct page **pages; - const struct iovec *vec = &iov[seg++]; - unsigned long user_addr = (unsigned long) vec->iov_base; - size_t size = vec->iov_len; - - page_count = nfs_get_user_pages(READ, user_addr, size, &pages); - if (page_count < 0) { - nfs_free_user_pages(pages, 0, 0); - if (tot_bytes > 0) - break; - return page_count; - } + inode_dio_done(inode); - result = nfs_direct_read_seg(inode, ctx, user_addr, size, - file_offset, pages, page_count); - - if (result <= 0) { - if (tot_bytes > 0) - break; - return result; - } - tot_bytes += result; - file_offset += result; - if (result < size) - break; + if (dreq->iocb) { + long res = (long) dreq->error; + if (!res) + res = (long) dreq->count; + aio_complete(dreq->iocb, res, 0); } - return tot_bytes; -} + complete_all(&dreq->completion); -/** - * nfs_direct_write_seg - Write out one iov segment. Generate separate - * write RPCs for each "wsize" bytes, then commit. - * @inode: target inode - * @ctx: target file open context - * user_addr: starting address of this segment of user's buffer - * count: size of this segment - * file_offset: offset in file to begin the operation - * @pages: array of addresses of page structs defining user's buffer - * nr_pages: size of pages array - */ -static ssize_t nfs_direct_write_seg(struct inode *inode, - struct nfs_open_context *ctx, unsigned long user_addr, - size_t count, loff_t file_offset, struct page **pages, - int nr_pages) -{ - const unsigned int wsize = NFS_SERVER(inode)->wsize; - size_t request; - int curpage, need_commit; - ssize_t result, tot_bytes; - struct nfs_writeverf first_verf; - struct nfs_write_data *wdata; - - wdata = nfs_writedata_alloc(); - if (!wdata) - return -ENOMEM; + nfs_direct_req_release(dreq); +} - wdata->inode = inode; - wdata->cred = ctx->cred; - wdata->args.fh = NFS_FH(inode); - wdata->args.context = ctx; - wdata->args.stable = NFS_UNSTABLE; - if (IS_SYNC(inode) || NFS_PROTO(inode)->version == 2 || count <= wsize) - wdata->args.stable = NFS_FILE_SYNC; - wdata->res.fattr = &wdata->fattr; - wdata->res.verf = &wdata->verf; - - nfs_begin_data_update(inode); -retry: - need_commit = 0; - tot_bytes = 0; - curpage = 0; - request = count; - wdata->args.pgbase = user_addr & ~PAGE_MASK; - wdata->args.offset = file_offset; - do { - wdata->args.count = request; - if (wdata->args.count > wsize) - wdata->args.count = wsize; - wdata->args.pages = &pages[curpage]; - - dprintk("NFS: direct write: c=%u o=%Ld ua=%lu, pb=%u, cp=%u\n", - wdata->args.count, (long long) wdata->args.offset, - user_addr + tot_bytes, wdata->args.pgbase, curpage); - - lock_kernel(); - result = NFS_PROTO(inode)->write(wdata); - unlock_kernel(); - - if (result <= 0) { - if (tot_bytes > 0) - break; - goto out; - } +static void nfs_direct_readpage_release(struct nfs_page *req) +{ + dprintk("NFS: direct read done (%s/%llu %d@%lld)\n", + req->wb_context->dentry->d_inode->i_sb->s_id, + (unsigned long long)NFS_FILEID(req->wb_context->dentry->d_inode), + req->wb_bytes, + (long long)req_offset(req)); + nfs_release_request(req); +} - if (tot_bytes == 0) - memcpy(&first_verf.verifier, &wdata->verf.verifier, - sizeof(first_verf.verifier)); - if (wdata->verf.committed != NFS_FILE_SYNC) { - need_commit = 1; - if (memcmp(&first_verf.verifier, &wdata->verf.verifier, - sizeof(first_verf.verifier))); - goto sync_retry; - } +static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) +{ + unsigned long bytes = 0; + struct nfs_direct_req *dreq = hdr->dreq; - tot_bytes += result; + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out_put; - /* in case of a short write: stop now, let the app recover */ - if (result < wdata->args.count) - break; + spin_lock(&dreq->lock); + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0)) + dreq->error = hdr->error; + else + dreq->count += hdr->good_bytes; + spin_unlock(&dreq->lock); + + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (!PageCompound(page) && bytes < hdr->good_bytes) + set_page_dirty(page); + bytes += req->wb_bytes; + nfs_list_remove_request(req); + nfs_direct_readpage_release(req); + } +out_put: + if (put_dreq(dreq)) + nfs_direct_complete(dreq, false); + hdr->release(hdr); +} - wdata->args.offset += result; - wdata->args.pgbase += result; - curpage += wdata->args.pgbase >> PAGE_SHIFT; - wdata->args.pgbase &= ~PAGE_MASK; - request -= result; - } while (request != 0); +static void nfs_read_sync_pgio_error(struct list_head *head) +{ + struct nfs_page *req; - /* - * Commit data written so far, even in the event of an error - */ - if (need_commit) { - wdata->args.count = tot_bytes; - wdata->args.offset = file_offset; - - lock_kernel(); - result = NFS_PROTO(inode)->commit(wdata); - unlock_kernel(); - - if (result < 0 || memcmp(&first_verf.verifier, - &wdata->verf.verifier, - sizeof(first_verf.verifier)) != 0) - goto sync_retry; + while (!list_empty(head)) { + req = nfs_list_entry(head->next); + nfs_list_remove_request(req); + nfs_release_request(req); } - result = tot_bytes; - -out: - nfs_end_data_update(inode); - nfs_writedata_free(wdata); - return result; +} -sync_retry: - wdata->args.stable = NFS_FILE_SYNC; - goto retry; +static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr) +{ + get_dreq(hdr->dreq); } -/** - * nfs_direct_write - For each iov segment, map the user's buffer - * then generate write and commit RPCs. - * @inode: target inode - * @ctx: target file open context - * @iov: array of vectors that define I/O buffer - * file_offset: offset in file to begin the operation - * nr_segs: size of iovec array - * - * Upon return, generic_file_direct_IO invalidates any cached pages - * that non-direct readers might access, so they will pick up these - * writes immediately. - */ -static ssize_t nfs_direct_write(struct inode *inode, - struct nfs_open_context *ctx, const struct iovec *iov, - loff_t file_offset, unsigned long nr_segs) -{ - ssize_t tot_bytes = 0; - unsigned long seg = 0; - - while ((seg < nr_segs) && (tot_bytes >= 0)) { - ssize_t result; - int page_count; - struct page **pages; - const struct iovec *vec = &iov[seg++]; - unsigned long user_addr = (unsigned long) vec->iov_base; - size_t size = vec->iov_len; - - page_count = nfs_get_user_pages(WRITE, user_addr, size, &pages); - if (page_count < 0) { - nfs_free_user_pages(pages, 0, 0); - if (tot_bytes > 0) - break; - return page_count; - } +static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = { + .error_cleanup = nfs_read_sync_pgio_error, + .init_hdr = nfs_direct_pgio_init, + .completion = nfs_direct_read_completion, +}; - result = nfs_direct_write_seg(inode, ctx, user_addr, size, - file_offset, pages, page_count); - nfs_free_user_pages(pages, page_count, 0); +/* + * For each rsize'd chunk of the user's buffer, dispatch an NFS READ + * operation. If nfs_readdata_alloc() or get_user_pages() fails, + * bail and stop sending more reads. Read length accounting is + * handled automatically by nfs_direct_read_result(). Otherwise, if + * no requests have been sent, just return an error. + */ - if (result <= 0) { - if (tot_bytes > 0) +static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, + struct iov_iter *iter, + loff_t pos) +{ + struct nfs_pageio_descriptor desc; + struct inode *inode = dreq->inode; + ssize_t result = -EINVAL; + size_t requested_bytes = 0; + size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE); + + nfs_pageio_init_read(&desc, dreq->inode, false, + &nfs_direct_read_completion_ops); + get_dreq(dreq); + desc.pg_dreq = dreq; + atomic_inc(&inode->i_dio_count); + + while (iov_iter_count(iter)) { + struct page **pagevec; + size_t bytes; + size_t pgbase; + unsigned npages, i; + + result = iov_iter_get_pages_alloc(iter, &pagevec, + rsize, &pgbase); + if (result < 0) + break; + + bytes = result; + iov_iter_advance(iter, bytes); + npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; + for (i = 0; i < npages; i++) { + struct nfs_page *req; + unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); + /* XXX do we need to do the eof zeroing found in async_filler? */ + req = nfs_create_request(dreq->ctx, pagevec[i], NULL, + pgbase, req_len); + if (IS_ERR(req)) { + result = PTR_ERR(req); break; - return result; + } + req->wb_index = pos >> PAGE_SHIFT; + req->wb_offset = pos & ~PAGE_MASK; + if (!nfs_pageio_add_request(&desc, req)) { + result = desc.pg_error; + nfs_release_request(req); + break; + } + pgbase = 0; + bytes -= req_len; + requested_bytes += req_len; + pos += req_len; + dreq->bytes_left -= req_len; } - tot_bytes += result; - file_offset += result; - if (result < size) + nfs_direct_release_pages(pagevec, npages); + kvfree(pagevec); + if (result < 0) break; } - return tot_bytes; -} -/** - * nfs_direct_IO - NFS address space operation for direct I/O - * rw: direction (read or write) - * @iocb: target I/O control block - * @iov: array of vectors that define I/O buffer - * file_offset: offset in file to begin the operation - * nr_segs: size of iovec array - * - */ -ssize_t -nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, - loff_t file_offset, unsigned long nr_segs) -{ - ssize_t result = -EINVAL; - struct file *file = iocb->ki_filp; - struct nfs_open_context *ctx; - struct dentry *dentry = file->f_dentry; - struct inode *inode = dentry->d_inode; + nfs_pageio_complete(&desc); /* - * No support for async yet + * If no bytes were started, return the error, and let the + * generic layer handle the completion. */ - if (!is_sync_kiocb(iocb)) - return result; - - ctx = (struct nfs_open_context *)file->private_data; - switch (rw) { - case READ: - dprintk("NFS: direct_IO(read) (%s) off/no(%Lu/%lu)\n", - dentry->d_name.name, file_offset, nr_segs); - - result = nfs_direct_read(inode, ctx, iov, - file_offset, nr_segs); - break; - case WRITE: - dprintk("NFS: direct_IO(write) (%s) off/no(%Lu/%lu)\n", - dentry->d_name.name, file_offset, nr_segs); - - result = nfs_direct_write(inode, ctx, iov, - file_offset, nr_segs); - break; - default: - break; + if (requested_bytes == 0) { + inode_dio_done(inode); + nfs_direct_req_release(dreq); + return result < 0 ? result : -EIO; } - return result; + + if (put_dreq(dreq)) + nfs_direct_complete(dreq, false); + return 0; } /** * nfs_file_direct_read - file direct read operation for NFS files * @iocb: target I/O control block - * @buf: user's buffer into which to read data - * count: number of bytes to read - * pos: byte offset in file where reading starts + * @iter: vector of user buffers into which to read data + * @pos: byte offset in file where reading starts * * We use this function for direct reads instead of calling * generic_file_aio_read() in order to avoid gfar's check to see if * the request starts before the end of the file. For that check * to work, we must generate a GETATTR before each direct read, and * even then there is a window between the GETATTR and the subsequent - * READ where the file size could change. So our preference is simply + * READ where the file size could change. Our preference is simply * to do all reads the application wants, and the server will take * care of managing the end of file boundary. - * + * * This function also eliminates unnecessarily updating the file's * atime locally, as the NFS server sets the file's atime, and this * client must read the updated atime from the server back into its * cache. */ -ssize_t -nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t pos) +ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter, + loff_t pos, bool uio) { - ssize_t retval = -EINVAL; - loff_t *ppos = &iocb->ki_pos; struct file *file = iocb->ki_filp; - struct nfs_open_context *ctx = - (struct nfs_open_context *) file->private_data; struct address_space *mapping = file->f_mapping; struct inode *inode = mapping->host; - struct iovec iov = { - .iov_base = buf, - .iov_len = count, - }; + struct nfs_direct_req *dreq; + struct nfs_lock_context *l_ctx; + ssize_t result = -EINVAL; + size_t count = iov_iter_count(iter); + nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count); - dprintk("nfs: direct read(%s/%s, %lu@%lu)\n", - file->f_dentry->d_parent->d_name.name, - file->f_dentry->d_name.name, - (unsigned long) count, (unsigned long) pos); + dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n", + file, count, (long long) pos); - if (!is_sync_kiocb(iocb)) - goto out; - if (count < 0) - goto out; - retval = -EFAULT; - if (!access_ok(VERIFY_WRITE, iov.iov_base, iov.iov_len)) - goto out; - retval = 0; + result = 0; if (!count) goto out; - if (mapping->nrpages) { - retval = filemap_fdatawrite(mapping); - if (retval == 0) - retval = nfs_wb_all(inode); - if (retval == 0) - retval = filemap_fdatawait(mapping); - if (retval) - goto out; + mutex_lock(&inode->i_mutex); + result = nfs_sync_mapping(mapping); + if (result) + goto out_unlock; + + task_io_account_read(count); + + result = -ENOMEM; + dreq = nfs_direct_req_alloc(); + if (dreq == NULL) + goto out_unlock; + + dreq->inode = inode; + dreq->bytes_left = count; + dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); + l_ctx = nfs_get_lock_context(dreq->ctx); + if (IS_ERR(l_ctx)) { + result = PTR_ERR(l_ctx); + goto out_release; } + dreq->l_ctx = l_ctx; + if (!is_sync_kiocb(iocb)) + dreq->iocb = iocb; + + NFS_I(inode)->read_io += count; + result = nfs_direct_read_schedule_iovec(dreq, iter, pos); - retval = nfs_direct_read(inode, ctx, &iov, pos, 1); - if (retval > 0) - *ppos = pos + retval; + mutex_unlock(&inode->i_mutex); + + if (!result) { + result = nfs_direct_wait(dreq); + if (result > 0) + iocb->ki_pos = pos + result; + } + + nfs_direct_req_release(dreq); + return result; +out_release: + nfs_direct_req_release(dreq); +out_unlock: + mutex_unlock(&inode->i_mutex); out: - return retval; + return result; +} + +#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4) +static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) +{ + struct nfs_pageio_descriptor desc; + struct nfs_page *req, *tmp; + LIST_HEAD(reqs); + struct nfs_commit_info cinfo; + LIST_HEAD(failed); + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo); + spin_lock(cinfo.lock); + nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0); + spin_unlock(cinfo.lock); + + dreq->count = 0; + get_dreq(dreq); + + nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false, + &nfs_direct_write_completion_ops); + desc.pg_dreq = dreq; + + list_for_each_entry_safe(req, tmp, &reqs, wb_list) { + if (!nfs_pageio_add_request(&desc, req)) { + nfs_list_remove_request(req); + nfs_list_add_request(req, &failed); + spin_lock(cinfo.lock); + dreq->flags = 0; + dreq->error = -EIO; + spin_unlock(cinfo.lock); + } + nfs_release_request(req); + } + nfs_pageio_complete(&desc); + + while (!list_empty(&failed)) { + req = nfs_list_entry(failed.next); + nfs_list_remove_request(req); + nfs_unlock_and_release_request(req); + } + + if (put_dreq(dreq)) + nfs_direct_write_complete(dreq, dreq->inode); +} + +static void nfs_direct_commit_complete(struct nfs_commit_data *data) +{ + struct nfs_direct_req *dreq = data->dreq; + struct nfs_commit_info cinfo; + struct nfs_page *req; + int status = data->task.tk_status; + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + if (status < 0) { + dprintk("NFS: %5u commit failed with error %d.\n", + data->task.tk_pid, status); + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + } else if (nfs_direct_cmp_commit_data_verf(dreq, data)) { + dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid); + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + } + + dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status); + while (!list_empty(&data->pages)) { + req = nfs_list_entry(data->pages.next); + nfs_list_remove_request(req); + if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) { + /* Note the rewrite will go through mds */ + nfs_mark_request_commit(req, NULL, &cinfo); + } else + nfs_release_request(req); + nfs_unlock_and_release_request(req); + } + + if (atomic_dec_and_test(&cinfo.mds->rpcs_out)) + nfs_direct_write_complete(dreq, data->inode); +} + +static void nfs_direct_error_cleanup(struct nfs_inode *nfsi) +{ + /* There is no lock to clear */ +} + +static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = { + .completion = nfs_direct_commit_complete, + .error_cleanup = nfs_direct_error_cleanup, +}; + +static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) +{ + int res; + struct nfs_commit_info cinfo; + LIST_HEAD(mds_list); + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + nfs_scan_commit(dreq->inode, &mds_list, &cinfo); + res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo); + if (res < 0) /* res == -ENOMEM */ + nfs_direct_write_reschedule(dreq); +} + +static void nfs_direct_write_schedule_work(struct work_struct *work) +{ + struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work); + int flags = dreq->flags; + + dreq->flags = 0; + switch (flags) { + case NFS_ODIRECT_DO_COMMIT: + nfs_direct_commit_schedule(dreq); + break; + case NFS_ODIRECT_RESCHED_WRITES: + nfs_direct_write_reschedule(dreq); + break; + default: + nfs_direct_complete(dreq, true); + } +} + +static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) +{ + schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */ +} + +#else +static void nfs_direct_write_schedule_work(struct work_struct *work) +{ +} + +static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) +{ + nfs_direct_complete(dreq, true); +} +#endif + +static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) +{ + struct nfs_direct_req *dreq = hdr->dreq; + struct nfs_commit_info cinfo; + int bit = -1; + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out_put; + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + + spin_lock(&dreq->lock); + + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + dreq->flags = 0; + dreq->error = hdr->error; + } + if (dreq->error != 0) + bit = NFS_IOHDR_ERROR; + else { + dreq->count += hdr->good_bytes; + if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) { + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + bit = NFS_IOHDR_NEED_RESCHED; + } else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) { + if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) + bit = NFS_IOHDR_NEED_RESCHED; + else if (dreq->flags == 0) { + nfs_direct_set_hdr_verf(dreq, hdr); + bit = NFS_IOHDR_NEED_COMMIT; + dreq->flags = NFS_ODIRECT_DO_COMMIT; + } else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) { + if (nfs_direct_set_or_cmp_hdr_verf(dreq, hdr)) { + dreq->flags = + NFS_ODIRECT_RESCHED_WRITES; + bit = NFS_IOHDR_NEED_RESCHED; + } else + bit = NFS_IOHDR_NEED_COMMIT; + } + } + } + spin_unlock(&dreq->lock); + + while (!list_empty(&hdr->pages)) { + + req = nfs_list_entry(hdr->pages.next); + nfs_list_remove_request(req); + switch (bit) { + case NFS_IOHDR_NEED_RESCHED: + case NFS_IOHDR_NEED_COMMIT: + kref_get(&req->wb_kref); + nfs_mark_request_commit(req, hdr->lseg, &cinfo); + } + nfs_unlock_and_release_request(req); + } + +out_put: + if (put_dreq(dreq)) + nfs_direct_write_complete(dreq, hdr->inode); + hdr->release(hdr); +} + +static void nfs_write_sync_pgio_error(struct list_head *head) +{ + struct nfs_page *req; + + while (!list_empty(head)) { + req = nfs_list_entry(head->next); + nfs_list_remove_request(req); + nfs_unlock_and_release_request(req); + } +} + +static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { + .error_cleanup = nfs_write_sync_pgio_error, + .init_hdr = nfs_direct_pgio_init, + .completion = nfs_direct_write_completion, +}; + + +/* + * NB: Return the value of the first error return code. Subsequent + * errors after the first one are ignored. + */ +/* + * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE + * operation. If nfs_writedata_alloc() or get_user_pages() fails, + * bail and stop sending more writes. Write length accounting is + * handled automatically by nfs_direct_write_result(). Otherwise, if + * no requests have been sent, just return an error. + */ +static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, + struct iov_iter *iter, + loff_t pos) +{ + struct nfs_pageio_descriptor desc; + struct inode *inode = dreq->inode; + ssize_t result = 0; + size_t requested_bytes = 0; + size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE); + + nfs_pageio_init_write(&desc, inode, FLUSH_COND_STABLE, false, + &nfs_direct_write_completion_ops); + desc.pg_dreq = dreq; + get_dreq(dreq); + atomic_inc(&inode->i_dio_count); + + NFS_I(inode)->write_io += iov_iter_count(iter); + while (iov_iter_count(iter)) { + struct page **pagevec; + size_t bytes; + size_t pgbase; + unsigned npages, i; + + result = iov_iter_get_pages_alloc(iter, &pagevec, + wsize, &pgbase); + if (result < 0) + break; + + bytes = result; + iov_iter_advance(iter, bytes); + npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; + for (i = 0; i < npages; i++) { + struct nfs_page *req; + unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); + + req = nfs_create_request(dreq->ctx, pagevec[i], NULL, + pgbase, req_len); + if (IS_ERR(req)) { + result = PTR_ERR(req); + break; + } + nfs_lock_request(req); + req->wb_index = pos >> PAGE_SHIFT; + req->wb_offset = pos & ~PAGE_MASK; + if (!nfs_pageio_add_request(&desc, req)) { + result = desc.pg_error; + nfs_unlock_and_release_request(req); + break; + } + pgbase = 0; + bytes -= req_len; + requested_bytes += req_len; + pos += req_len; + dreq->bytes_left -= req_len; + } + nfs_direct_release_pages(pagevec, npages); + kvfree(pagevec); + if (result < 0) + break; + } + nfs_pageio_complete(&desc); + + /* + * If no bytes were started, return the error, and let the + * generic layer handle the completion. + */ + if (requested_bytes == 0) { + inode_dio_done(inode); + nfs_direct_req_release(dreq); + return result < 0 ? result : -EIO; + } + + if (put_dreq(dreq)) + nfs_direct_write_complete(dreq, dreq->inode); + return 0; } /** * nfs_file_direct_write - file direct write operation for NFS files * @iocb: target I/O control block - * @buf: user's buffer from which to write data - * count: number of bytes to write - * pos: byte offset in file where writing starts + * @iter: vector of user buffers from which to write data + * @pos: byte offset in file where writing starts * * We use this function for direct writes instead of calling * generic_file_aio_write() in order to avoid taking the inode @@ -710,9 +893,6 @@ out: * back into its cache. We let the server do generic write * parameter checking and report problems. * - * We also avoid an unnecessary invocation of generic_osync_inode(), - * as it is fairly meaningless to sync the metadata of an NFS file. - * * We eliminate local atime updates, see direct read above. * * We avoid unnecessary page cache invalidations for normal cached @@ -721,83 +901,121 @@ out: * Note that O_APPEND is not supported for NFS direct writes, as there * is no atomic O_APPEND write facility in the NFS protocol. */ -ssize_t -nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t pos) +ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter, + loff_t pos, bool uio) { - ssize_t retval = -EINVAL; - loff_t *ppos = &iocb->ki_pos; - unsigned long limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur; + ssize_t result = -EINVAL; struct file *file = iocb->ki_filp; - struct nfs_open_context *ctx = - (struct nfs_open_context *) file->private_data; struct address_space *mapping = file->f_mapping; struct inode *inode = mapping->host; - struct iovec iov = { - .iov_base = (char __user *)buf, - .iov_len = count, - }; + struct nfs_direct_req *dreq; + struct nfs_lock_context *l_ctx; + loff_t end; + size_t count = iov_iter_count(iter); + end = (pos + count - 1) >> PAGE_CACHE_SHIFT; - dfprintk(VFS, "nfs: direct write(%s/%s(%ld), %lu@%lu)\n", - file->f_dentry->d_parent->d_name.name, - file->f_dentry->d_name.name, inode->i_ino, - (unsigned long) count, (unsigned long) pos); + nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count); - if (!is_sync_kiocb(iocb)) - goto out; - if (count < 0) - goto out; - if (pos < 0) + dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n", + file, count, (long long) pos); + + result = generic_write_checks(file, &pos, &count, 0); + if (result) goto out; - retval = -EFAULT; - if (!access_ok(VERIFY_READ, iov.iov_base, iov.iov_len)) + + result = -EINVAL; + if ((ssize_t) count < 0) goto out; - retval = -EFBIG; - if (limit != RLIM_INFINITY) { - if (pos >= limit) { - send_sig(SIGXFSZ, current, 0); - goto out; - } - if (count > limit - (unsigned long) pos) - count = limit - (unsigned long) pos; - } - retval = 0; + result = 0; if (!count) goto out; + mutex_lock(&inode->i_mutex); + + result = nfs_sync_mapping(mapping); + if (result) + goto out_unlock; + + if (mapping->nrpages) { + result = invalidate_inode_pages2_range(mapping, + pos >> PAGE_CACHE_SHIFT, end); + if (result) + goto out_unlock; + } + + task_io_account_write(count); + + result = -ENOMEM; + dreq = nfs_direct_req_alloc(); + if (!dreq) + goto out_unlock; + + dreq->inode = inode; + dreq->bytes_left = count; + dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); + l_ctx = nfs_get_lock_context(dreq->ctx); + if (IS_ERR(l_ctx)) { + result = PTR_ERR(l_ctx); + goto out_release; + } + dreq->l_ctx = l_ctx; + if (!is_sync_kiocb(iocb)) + dreq->iocb = iocb; + + result = nfs_direct_write_schedule_iovec(dreq, iter, pos); + if (mapping->nrpages) { - retval = filemap_fdatawrite(mapping); - if (retval == 0) - retval = nfs_wb_all(inode); - if (retval == 0) - retval = filemap_fdatawait(mapping); - if (retval) - goto out; + invalidate_inode_pages2_range(mapping, + pos >> PAGE_CACHE_SHIFT, end); } - retval = nfs_direct_write(inode, ctx, &iov, pos, 1); - if (mapping->nrpages) - invalidate_inode_pages2(mapping); - if (retval > 0) - *ppos = pos + retval; + mutex_unlock(&inode->i_mutex); + if (!result) { + result = nfs_direct_wait(dreq); + if (result > 0) { + struct inode *inode = mapping->host; + + iocb->ki_pos = pos + result; + spin_lock(&inode->i_lock); + if (i_size_read(inode) < iocb->ki_pos) + i_size_write(inode, iocb->ki_pos); + spin_unlock(&inode->i_lock); + } + } + nfs_direct_req_release(dreq); + return result; + +out_release: + nfs_direct_req_release(dreq); +out_unlock: + mutex_unlock(&inode->i_mutex); out: - return retval; + return result; } -int nfs_init_directcache(void) +/** + * nfs_init_directcache - create a slab cache for nfs_direct_req structures + * + */ +int __init nfs_init_directcache(void) { nfs_direct_cachep = kmem_cache_create("nfs_direct_cache", sizeof(struct nfs_direct_req), - 0, SLAB_RECLAIM_ACCOUNT, - NULL, NULL); + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD), + NULL); if (nfs_direct_cachep == NULL) return -ENOMEM; return 0; } +/** + * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures + * + */ void nfs_destroy_directcache(void) { - if (kmem_cache_destroy(nfs_direct_cachep)) - printk(KERN_INFO "nfs_direct_cache: not all structures were freed\n"); + kmem_cache_destroy(nfs_direct_cachep); } |