write.c

Linux Kernel 2.6.9 for OMAP1710

	while (!list_empty(&list)) {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del(&data->pages);
		nfs_writedata_free(data);
	}
	nfs_mark_request_dirty(req);
	nfs_unlock_request(req);
	return -ENOMEM;
}

/*
 * Create an RPC task for the given write request and kick it.
 * The page must have been locked by the caller.
 *
 * It may happen that the page we're passed is not marked dirty.
 * This is the case if nfs_updatepage detects a conflicting request
 * that has been written but not committed.
 */
static int nfs_flush_one(struct list_head *head, struct inode *inode, int how)
{
	struct nfs_page		*req;
	struct page		**pages;
	struct nfs_write_data	*data;
	unsigned int		count;

	if (NFS_SERVER(inode)->wsize < PAGE_CACHE_SIZE)
		return nfs_flush_multi(head, inode, how);

	data = nfs_writedata_alloc();
	if (!data)
		goto out_bad;

	pages = data->pagevec;
	count = 0;
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &data->pages);
		ClearPageError(req->wb_page);
		SetPageWriteback(req->wb_page);
		*pages++ = req->wb_page;
		count += req->wb_bytes;
	}
	req = nfs_list_entry(data->pages.next);

	data->complete = nfs_writeback_done_full;
	/* Set up the argument struct */
	nfs_write_rpcsetup(req, data, count, 0, how);

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		struct nfs_page *req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
		nfs_unlock_request(req);
	}
	return -ENOMEM;
}

int
nfs_flush_list(struct list_head *head, int wpages, int how)
{
	LIST_HEAD(one_request);
	struct nfs_page		*req;
	int			error = 0;
	unsigned int		pages = 0;

	while (!list_empty(head)) {
		pages += nfs_coalesce_requests(head, &one_request, wpages);
		req = nfs_list_entry(one_request.next);
		error = nfs_flush_one(&one_request, req->wb_context->dentry->d_inode, how);
		if (error < 0)
			break;
	}
	if (error >= 0)
		return pages;

	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
		nfs_unlock_request(req);
	}
	return error;
}

/*
 * Handle a write reply that flushed part of a page.
 */
static void nfs_writeback_done_partial(struct nfs_write_data *data, int status)
{
	struct nfs_page		*req = data->req;
	struct page		*page = req->wb_page;

	dprintk("NFS: write (%s/%Ld %d@%Ld)",
		req->wb_context->dentry->d_inode->i_sb->s_id,
		(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
		req->wb_bytes,
		(long long)req_offset(req));

	if (status < 0) {
		ClearPageUptodate(page);
		SetPageError(page);
		req->wb_context->error = status;
		dprintk(", error = %d\n", status);
	} else {
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->verf.committed < NFS_FILE_SYNC) {
			if (!NFS_NEED_COMMIT(req)) {
				nfs_defer_commit(req);
				memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
				dprintk(" defer commit\n");
			} else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) {
				nfs_defer_reschedule(req);
				dprintk(" server reboot detected\n");
			}
		} else
#endif
			dprintk(" OK\n");
	}

	if (atomic_dec_and_test(&req->wb_complete))
		nfs_writepage_release(req);
}

/*
 * Handle a write reply that flushes a whole page.
 *
 * FIXME: There is an inherent race with invalidate_inode_pages and
 *	  writebacks since the page->count is kept > 1 for as long
 *	  as the page has a write request pending.
 */
static void nfs_writeback_done_full(struct nfs_write_data *data, int status)
{
	struct nfs_page		*req;
	struct page		*page;

	/* Update attributes as result of writeback. */
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
		page = req->wb_page;

		dprintk("NFS: write (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));

		if (status < 0) {
			ClearPageUptodate(page);
			SetPageError(page);
			req->wb_context->error = status;
			end_page_writeback(page);
			nfs_inode_remove_request(req);
			dprintk(", error = %d\n", status);
			goto next;
		}
		end_page_writeback(page);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->args.stable != NFS_UNSTABLE || data->verf.committed == NFS_FILE_SYNC) {
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
		nfs_mark_request_commit(req);
		dprintk(" marked for commit\n");
#else
		nfs_inode_remove_request(req);
#endif
	next:
		nfs_unlock_request(req);
	}
}

/*
 * This function is called when the WRITE call is complete.
 */
void nfs_writeback_done(struct rpc_task *task)
{
	struct nfs_write_data	*data = (struct nfs_write_data *) task->tk_calldata;
	struct nfs_writeargs	*argp = &data->args;
	struct nfs_writeres	*resp = &data->res;

	dprintk("NFS: %4d nfs_writeback_done (status %d)\n",
		task->tk_pid, task->tk_status);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (resp->verf->committed < argp->stable && task->tk_status >= 0) {
		/* We tried a write call, but the server did not
		 * commit data to stable storage even though we
		 * requested it.
		 * Note: There is a known bug in Tru64 < 5.0 in which
		 *	 the server reports NFS_DATA_SYNC, but performs
		 *	 NFS_FILE_SYNC. We therefore implement this checking
		 *	 as a dprintk() in order to avoid filling syslog.
		 */
		static unsigned long    complain;

		if (time_before(complain, jiffies)) {
			dprintk("NFS: faulty NFS server %s:"
				" (committed = %d) != (stable = %d)\n",
				NFS_SERVER(data->inode)->hostname,
				resp->verf->committed, argp->stable);
			complain = jiffies + 300 * HZ;
		}
	}
#endif
	/* Is this a short write? */
	if (task->tk_status >= 0 && resp->count < argp->count) {
		static unsigned long    complain;

		/* Has the server at least made some progress? */
		if (resp->count != 0) {
			/* Was this an NFSv2 write or an NFSv3 stable write? */
			if (resp->verf->committed != NFS_UNSTABLE) {
				/* Resend from where the server left off */
				argp->offset += resp->count;
				argp->pgbase += resp->count;
				argp->count -= resp->count;
			} else {
				/* Resend as a stable write in order to avoid
				 * headaches in the case of a server crash.
				 */
				argp->stable = NFS_FILE_SYNC;
			}
			rpc_restart_call(task);
			return;
		}
		if (time_before(complain, jiffies)) {
			printk(KERN_WARNING
			       "NFS: Server wrote less than requested.\n");
			complain = jiffies + 300 * HZ;
		}
		/* Can't do anything about it except throw an error. */
		task->tk_status = -EIO;
	}

	/*
	 * Process the nfs_page list
	 */
	data->complete(data, task->tk_status);
}


#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
static void nfs_commit_release(struct rpc_task *task)
{
	struct nfs_write_data	*wdata = (struct nfs_write_data *)task->tk_calldata;
	nfs_commit_free(wdata);
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_commit_rpcsetup(struct list_head *head,
		struct nfs_write_data *data, int how)
{
	struct rpc_task		*task = &data->task;
	struct nfs_page		*first, *last;
	struct inode		*inode;
	loff_t			start, end, len;

	/* Set up the RPC argument and reply structs
	 * NB: take care not to mess about with data->commit et al. */

	list_splice_init(head, &data->pages);
	first = nfs_list_entry(data->pages.next);
	last = nfs_list_entry(data->pages.prev);
	inode = first->wb_context->dentry->d_inode;

	/*
	 * Determine the offset range of requests in the COMMIT call.
	 * We rely on the fact that data->pages is an ordered list...
	 */
	start = req_offset(first);
	end = req_offset(last) + last->wb_bytes;
	len = end - start;
	/* If 'len' is not a 32-bit quantity, pass '0' in the COMMIT call */
	if (end >= i_size_read(inode) || len < 0 || len > (~((u32)0) >> 1))
		len = 0;

	data->inode	  = inode;
	data->cred	  = first->wb_context->cred;

	data->args.fh     = NFS_FH(data->inode);
	data->args.offset = start;
	data->args.count  = len;
	data->res.count   = len;
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
	
	NFS_PROTO(inode)->commit_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;
	data->task.tk_calldata = data;
	/* Release requests */
	data->task.tk_release = nfs_commit_release;
	
	dprintk("NFS: %4d initiated commit call\n", task->tk_pid);
}

/*
 * Commit dirty pages
 */
int
nfs_commit_list(struct list_head *head, int how)
{
	struct nfs_write_data	*data;
	struct nfs_page         *req;

	data = nfs_commit_alloc();

	if (!data)
		goto out_bad;

	/* Set up the argument struct */
	nfs_commit_rpcsetup(head, data, how);

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_commit(req);
		nfs_unlock_request(req);
	}
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
void
nfs_commit_done(struct rpc_task *task)
{
	struct nfs_write_data	*data = (struct nfs_write_data *)task->tk_calldata;
	struct nfs_page		*req;
	int res = 0;

        dprintk("NFS: %4d nfs_commit_done (status %d)\n",
                                task->tk_pid, task->tk_status);

	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);

		dprintk("NFS: commit (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));
		if (task->tk_status < 0) {
			req->wb_context->error = task->tk_status;
			nfs_inode_remove_request(req);
			dprintk(", error = %d\n", task->tk_status);
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
		if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) {
			/* We have a match */
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		/* We have a mismatch. Write the page again */
		dprintk(" mismatch\n");
		nfs_mark_request_dirty(req);
	next:
		nfs_unlock_request(req);
		res++;
	}
	sub_page_state(nr_unstable,res);
}
#endif

int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
		   unsigned int npages, int how)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
	int			res,
				error = 0;

	spin_lock(&nfsi->req_lock);
	res = nfs_scan_dirty(inode, &head, idx_start, npages);
	spin_unlock(&nfsi->req_lock);
	if (res)
		error = nfs_flush_list(&head, NFS_SERVER(inode)->wpages, how);
	if (error < 0)
		return error;
	return res;
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
int nfs_commit_inode(struct inode *inode, unsigned long idx_start,
		    unsigned int npages, int how)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
	int			res,
				error = 0;

	spin_lock(&nfsi->req_lock);
	res = nfs_scan_commit(inode, &head, idx_start, npages);
	if (res) {
		res += nfs_scan_commit(inode, &head, 0, 0);
		spin_unlock(&nfsi->req_lock);
		error = nfs_commit_list(&head, how);
	} else
		spin_unlock(&nfsi->req_lock);
	if (error < 0)
		return error;
	return res;
}
#endif

int nfs_sync_inode(struct inode *inode, unsigned long idx_start,
		  unsigned int npages, int how)
{
	int	error,
		wait;

	wait = how & FLUSH_WAIT;
	how &= ~FLUSH_WAIT;

	do {
		error = 0;
		if (wait)
			error = nfs_wait_on_requests(inode, idx_start, npages);
		if (error == 0)
			error = nfs_flush_inode(inode, idx_start, npages, how);
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (error == 0)
			error = nfs_commit_inode(inode, idx_start, npages, how);
#endif
	} while (error > 0);
	return error;
}

int nfs_init_writepagecache(void)
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
					     sizeof(struct nfs_write_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL, NULL);
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

	nfs_wdata_mempool = mempool_create(MIN_POOL_WRITE,
					   mempool_alloc_slab,
					   mempool_free_slab,
					   nfs_wdata_cachep);
	if (nfs_wdata_mempool == NULL)
		return -ENOMEM;

	nfs_commit_mempool = mempool_create(MIN_POOL_COMMIT,
					   mempool_alloc_slab,
					   mempool_free_slab,
					   nfs_wdata_cachep);
	if (nfs_commit_mempool == NULL)
		return -ENOMEM;

	return 0;
}

void nfs_destroy_writepagecache(void)
{
	mempool_destroy(nfs_commit_mempool);
	mempool_destroy(nfs_wdata_mempool);
	if (kmem_cache_destroy(nfs_wdata_cachep))
		printk(KERN_INFO "nfs_write_data: not all structures were freed\n");
}