diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 345aa5c0f382..0c542ec92d5b 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -122,19 +122,25 @@ ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_ return -EINVAL; } -static void nfs_direct_dirty_pages(struct page **pages, int npages) +static void nfs_direct_dirty_pages(struct page **pages, unsigned int pgbase, size_t count) { - int i; + unsigned int npages; + unsigned int i; + + if (count == 0) + return; + pages += (pgbase >> PAGE_SHIFT); + npages = (count + (pgbase & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT; for (i = 0; i < npages; i++) { struct page *page = pages[i]; if (!PageCompound(page)) - set_page_dirty_lock(page); + set_page_dirty(page); } } -static void nfs_direct_release_pages(struct page **pages, int npages) +static void nfs_direct_release_pages(struct page **pages, unsigned int npages) { - int i; + unsigned int i; for (i = 0; i < npages; i++) page_cache_release(pages[i]); } @@ -224,17 +230,18 @@ static void nfs_direct_read_result(struct rpc_task *task, void *calldata) if (nfs_readpage_result(task, data) != 0) return; - nfs_direct_dirty_pages(data->pagevec, data->npages); - nfs_direct_release_pages(data->pagevec, data->npages); - spin_lock(&dreq->lock); - - if (likely(task->tk_status >= 0)) - dreq->count += data->res.count; - else + if (unlikely(task->tk_status < 0)) { dreq->error = task->tk_status; - - spin_unlock(&dreq->lock); + spin_unlock(&dreq->lock); + } else { + dreq->count += data->res.count; + spin_unlock(&dreq->lock); + nfs_direct_dirty_pages(data->pagevec, + data->args.pgbase, + data->res.count); + } + nfs_direct_release_pages(data->pagevec, data->npages); if (put_dreq(dreq)) nfs_direct_complete(dreq); @@ -279,9 +286,12 @@ static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned lo result = get_user_pages(current, current->mm, user_addr, data->npages, 1, 0, data->pagevec, NULL); up_read(¤t->mm->mmap_sem); - if (unlikely(result < data->npages)) { - if (result > 0) - nfs_direct_release_pages(data->pagevec, result); + if (result < 0) { + nfs_readdata_release(data); + break; + } + if ((unsigned)result < data->npages) { + nfs_direct_release_pages(data->pagevec, result); nfs_readdata_release(data); break; } @@ -610,9 +620,12 @@ static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned l result = get_user_pages(current, current->mm, user_addr, data->npages, 0, 0, data->pagevec, NULL); up_read(¤t->mm->mmap_sem); - if (unlikely(result < data->npages)) { - if (result > 0) - nfs_direct_release_pages(data->pagevec, result); + if (result < 0) { + nfs_writedata_release(data); + break; + } + if ((unsigned)result < data->npages) { + nfs_direct_release_pages(data->pagevec, result); nfs_writedata_release(data); break; } diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index cbdd1c6aaa94..c5bb51a29e80 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -355,6 +355,26 @@ void nfs_pageio_complete(struct nfs_pageio_descriptor *desc) nfs_pageio_doio(desc); } +/** + * nfs_pageio_cond_complete - Conditional I/O completion + * @desc: pointer to io descriptor + * @index: page index + * + * It is important to ensure that processes don't try to take locks + * on non-contiguous ranges of pages as that might deadlock. This + * function should be called before attempting to wait on a locked + * nfs_page. It will complete the I/O if the page index 'index' + * is not contiguous with the existing list of pages in 'desc'. + */ +void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index) +{ + if (!list_empty(&desc->pg_list)) { + struct nfs_page *prev = nfs_list_entry(desc->pg_list.prev); + if (index != prev->wb_index + 1) + nfs_pageio_doio(desc); + } +} + #define NFS_SCAN_MAXENTRIES 16 /** * nfs_scan_list - Scan a list for matching requests diff --git a/fs/nfs/write.c b/fs/nfs/write.c index b084c03ce493..af344a158e01 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -273,8 +273,6 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, * request as dirty (in which case we don't care). */ spin_unlock(req_lock); - /* Prevent deadlock! */ - nfs_pageio_complete(pgio); ret = nfs_wait_on_request(req); nfs_release_request(req); if (ret != 0) @@ -321,6 +319,8 @@ static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc pgio = &mypgio; } + nfs_pageio_cond_complete(pgio, page->index); + err = nfs_page_async_flush(pgio, page); if (err <= 0) goto out; @@ -329,6 +329,8 @@ static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc if (!offset) goto out; + nfs_pageio_cond_complete(pgio, page->index); + ctx = nfs_find_open_context(inode, NULL, FMODE_WRITE); if (ctx == NULL) { err = -EBADF; diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 41afab6b5f09..bd193af80162 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -81,6 +81,7 @@ extern void nfs_pageio_init(struct nfs_pageio_descriptor *desc, extern int nfs_pageio_add_request(struct nfs_pageio_descriptor *, struct nfs_page *); 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);