direct.c

linux 内核源代码

	if (put_dreq(dreq))
		nfs_direct_write_complete(dreq, inode);
}

static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
{
	struct nfs_write_data *data = calldata;
	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;

	/* Call the NFS version-specific code */
	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
		return;
	if (unlikely(task->tk_status < 0)) {
		dprintk("NFS: %5u commit failed with error %d.\n",
				task->tk_pid, task->tk_status);
		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
	} else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
		dprintk("NFS: %5u commit verify failed\n", task->tk_pid);
		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
	}

	dprintk("NFS: %5u commit returned %d\n", task->tk_pid, task->tk_status);
	nfs_direct_write_complete(dreq, data->inode);
}

static const struct rpc_call_ops nfs_commit_direct_ops = {
	.rpc_call_done = nfs_direct_commit_result,
	.rpc_release = nfs_commit_release,
};

static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
{
	struct nfs_write_data *data = dreq->commit_data;

	data->inode = dreq->inode;
	data->cred = dreq->ctx->cred;

	data->args.fh = NFS_FH(data->inode);
	data->args.offset = 0;
	data->args.count = 0;
	data->res.count = 0;
	data->res.fattr = &data->fattr;
	data->res.verf = &data->verf;

	rpc_init_task(&data->task, NFS_CLIENT(dreq->inode), RPC_TASK_ASYNC,
				&nfs_commit_direct_ops, data);
	NFS_PROTO(data->inode)->commit_setup(data, 0);

	data->task.tk_priority = RPC_PRIORITY_NORMAL;
	data->task.tk_cookie = (unsigned long)data->inode;
	/* Note: task.tk_ops->rpc_release will free dreq->commit_data */
	dreq->commit_data = NULL;

	dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);

	rpc_execute(&data->task);
}

static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
{
	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:
			if (dreq->commit_data != NULL)
				nfs_commit_free(dreq->commit_data);
			nfs_direct_free_writedata(dreq);
			nfs_zap_mapping(inode, inode->i_mapping);
			nfs_direct_complete(dreq);
	}
}

static void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
{
	dreq->commit_data = nfs_commit_alloc();
	if (dreq->commit_data != NULL)
		dreq->commit_data->req = (struct nfs_page *) dreq;
}
#else
static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
{
	dreq->commit_data = NULL;
}

static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
{
	nfs_direct_free_writedata(dreq);
	nfs_zap_mapping(inode, inode->i_mapping);
	nfs_direct_complete(dreq);
}
#endif

static void nfs_direct_write_result(struct rpc_task *task, void *calldata)
{
	struct nfs_write_data *data = calldata;
	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
	int status = task->tk_status;

	if (nfs_writeback_done(task, data) != 0)
		return;

	spin_lock(&dreq->lock);

	if (unlikely(status < 0)) {
		/* An error has occurred, so we should not commit */
		dreq->flags = 0;
		dreq->error = status;
	}
	if (unlikely(dreq->error != 0))
		goto out_unlock;

	dreq->count += data->res.count;

	if (data->res.verf->committed != NFS_FILE_SYNC) {
		switch (dreq->flags) {
			case 0:
				memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf));
				dreq->flags = NFS_ODIRECT_DO_COMMIT;
				break;
			case NFS_ODIRECT_DO_COMMIT:
				if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) {
					dprintk("NFS: %5u write verify failed\n", task->tk_pid);
					dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
				}
		}
	}
out_unlock:
	spin_unlock(&dreq->lock);
}

/*
 * NB: Return the value of the first error return code.  Subsequent
 *     errors after the first one are ignored.
 */
static void nfs_direct_write_release(void *calldata)
{
	struct nfs_write_data *data = calldata;
	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;

	if (put_dreq(dreq))
		nfs_direct_write_complete(dreq, data->inode);
}

static const struct rpc_call_ops nfs_write_direct_ops = {
	.rpc_call_done = nfs_direct_write_result,
	.rpc_release = nfs_direct_write_release,
};

/*
 * 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_segment(struct nfs_direct_req *dreq,
						 const struct iovec *iov,
						 loff_t pos, int sync)
{
	struct nfs_open_context *ctx = dreq->ctx;
	struct inode *inode = ctx->path.dentry->d_inode;
	unsigned long user_addr = (unsigned long)iov->iov_base;
	size_t count = iov->iov_len;
	size_t wsize = NFS_SERVER(inode)->wsize;
	unsigned int pgbase;
	int result;
	ssize_t started = 0;

	do {
		struct nfs_write_data *data;
		size_t bytes;

		pgbase = user_addr & ~PAGE_MASK;
		bytes = min(wsize,count);

		result = -ENOMEM;
		data = nfs_writedata_alloc(nfs_page_array_len(pgbase, bytes));
		if (unlikely(!data))
			break;

		down_read(&current->mm->mmap_sem);
		result = get_user_pages(current, current->mm, user_addr,
					data->npages, 0, 0, data->pagevec, NULL);
		up_read(&current->mm->mmap_sem);
		if (result < 0) {
			nfs_writedata_release(data);
			break;
		}
		if ((unsigned)result < data->npages) {
			bytes = result * PAGE_SIZE;
			if (bytes <= pgbase) {
				nfs_direct_release_pages(data->pagevec, result);
				nfs_writedata_release(data);
				break;
			}
			bytes -= pgbase;
			data->npages = result;
		}

		get_dreq(dreq);

		list_move_tail(&data->pages, &dreq->rewrite_list);

		data->req = (struct nfs_page *) dreq;
		data->inode = inode;
		data->cred = ctx->cred;
		data->args.fh = NFS_FH(inode);
		data->args.context = ctx;
		data->args.offset = pos;
		data->args.pgbase = pgbase;
		data->args.pages = data->pagevec;
		data->args.count = bytes;
		data->res.fattr = &data->fattr;
		data->res.count = bytes;
		data->res.verf = &data->verf;

		rpc_init_task(&data->task, NFS_CLIENT(inode), RPC_TASK_ASYNC,
				&nfs_write_direct_ops, data);
		NFS_PROTO(inode)->write_setup(data, sync);

		data->task.tk_priority = RPC_PRIORITY_NORMAL;
		data->task.tk_cookie = (unsigned long) inode;

		rpc_execute(&data->task);

		dprintk("NFS: %5u initiated direct write call "
			"(req %s/%Ld, %zu 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);

		started += bytes;
		user_addr += bytes;
		pos += bytes;

		/* FIXME: Remove this useless math from the final patch */
		pgbase += bytes;
		pgbase &= ~PAGE_MASK;
		BUG_ON(pgbase != (user_addr & ~PAGE_MASK));

		count -= bytes;
	} while (count != 0);

	if (started)
		return started;
	return result < 0 ? (ssize_t) result : -EFAULT;
}

static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
					       const struct iovec *iov,
					       unsigned long nr_segs,
					       loff_t pos, int sync)
{
	ssize_t result = 0;
	size_t requested_bytes = 0;
	unsigned long seg;

	get_dreq(dreq);

	for (seg = 0; seg < nr_segs; seg++) {
		const struct iovec *vec = &iov[seg];
		result = nfs_direct_write_schedule_segment(dreq, vec,
							   pos, sync);
		if (result < 0)
			break;
		requested_bytes += result;
		if ((size_t)result < vec->iov_len)
			break;
		pos += vec->iov_len;
	}

	if (put_dreq(dreq))
		nfs_direct_write_complete(dreq, dreq->inode);

	if (requested_bytes != 0)
		return 0;

	if (result < 0)
		return result;
	return -EIO;
}

static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
				unsigned long nr_segs, loff_t pos,
				size_t count)
{
	ssize_t result = 0;
	sigset_t oldset;
	struct inode *inode = iocb->ki_filp->f_mapping->host;
	struct rpc_clnt *clnt = NFS_CLIENT(inode);
	struct nfs_direct_req *dreq;
	size_t wsize = NFS_SERVER(inode)->wsize;
	int sync = 0;

	dreq = nfs_direct_req_alloc();
	if (!dreq)
		return -ENOMEM;
	nfs_alloc_commit_data(dreq);

	if (dreq->commit_data == NULL || count < wsize)
		sync = FLUSH_STABLE;

	dreq->inode = inode;
	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
	if (!is_sync_kiocb(iocb))
		dreq->iocb = iocb;

	rpc_clnt_sigmask(clnt, &oldset);
	result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync);
	if (!result)
		result = nfs_direct_wait(dreq);
	rpc_clnt_sigunmask(clnt, &oldset);
	nfs_direct_req_release(dreq);

	return result;
}

/**
 * nfs_file_direct_read - file direct read operation for NFS files
 * @iocb: target I/O control block
 * @iov: vector of user buffers into which to read data
 * @nr_segs: size of iov vector
 * @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.  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, const struct iovec *iov,
				unsigned long nr_segs, loff_t pos)
{
	ssize_t retval = -EINVAL;
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	size_t count;

	count = iov_length(iov, nr_segs);
	nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);

	dprintk("nfs: direct read(%s/%s, %zd@%Ld)\n",
		file->f_path.dentry->d_parent->d_name.name,
		file->f_path.dentry->d_name.name,
		count, (long long) pos);

	retval = 0;
	if (!count)
		goto out;

	retval = nfs_sync_mapping(mapping);
	if (retval)
		goto out;

	retval = nfs_direct_read(iocb, iov, nr_segs, pos);
	if (retval > 0)
		iocb->ki_pos = pos + retval;

out:
	return retval;
}

/**
 * nfs_file_direct_write - file direct write operation for NFS files
 * @iocb: target I/O control block
 * @iov: vector of user buffers from which to write data
 * @nr_segs: size of iov vector
 * @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
 * semaphore and updating the i_size.  The NFS server will set
 * the new i_size and this client must read the updated size
 * 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
 * readers of this file.
 *
 * 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 struct iovec *iov,
				unsigned long nr_segs, loff_t pos)
{
	ssize_t retval = -EINVAL;
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	size_t count;

	count = iov_length(iov, nr_segs);
	nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);

	dfprintk(VFS, "nfs: direct write(%s/%s, %zd@%Ld)\n",
		file->f_path.dentry->d_parent->d_name.name,
		file->f_path.dentry->d_name.name,
		count, (long long) pos);

	retval = generic_write_checks(file, &pos, &count, 0);
	if (retval)
		goto out;

	retval = -EINVAL;
	if ((ssize_t) count < 0)
		goto out;
	retval = 0;
	if (!count)
		goto out;

	retval = nfs_sync_mapping(mapping);
	if (retval)
		goto out;

	retval = nfs_direct_write(iocb, iov, nr_segs, pos, count);

	if (retval > 0)
		iocb->ki_pos = pos + retval;

out:
	return retval;
}

/**
 * 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|
							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)
{
	kmem_cache_destroy(nfs_direct_cachep);
}