path: root/fs/nfs/write.c

/*
 * linux/fs/nfs/write.c
 *
 * Writing file data over NFS.
 *
 * We do it like this: When a (user) process wishes to write data to an
 * NFS file, a write request is allocated that contains the RPC task data
 * plus some info on the page to be written, and added to the inode's
 * write chain. If the process writes past the end of the page, an async
 * RPC call to write the page is scheduled immediately; otherwise, the call
 * is delayed for a few seconds.
 *
 * Just like readahead, no async I/O is performed if wsize < PAGE_SIZE.
 *
 * Write requests are kept on the inode's writeback list. Each entry in
 * that list references the page (portion) to be written. When the
 * cache timeout has expired, the RPC task is woken up, and tries to
 * lock the page. As soon as it manages to do so, the request is moved
 * from the writeback list to the writelock list.
 *
 * Note: we must make sure never to confuse the inode passed in the
 * write_page request with the one in page->inode. As far as I understand
 * it, these are different when doing a swap-out.
 *
 * To understand everything that goes on here and in the NFS read code,
 * one should be aware that a page is locked in exactly one of the following
 * cases:
 *
 *  -	A write request is in progress.
 *  -	A user process is in generic_file_write/nfs_update_page
 *  -	A user process is in generic_file_read
 *
 * Also note that because of the way pages are invalidated in
 * nfs_revalidate_inode, the following assertions hold:
 *
 *  -	If a page is dirty, there will be no read requests (a page will
 *	not be re-read unless invalidated by nfs_revalidate_inode).
 *  -	If the page is not uptodate, there will be no pending write
 *	requests, and no process will be in nfs_update_page.
 *
 * FIXME: Interaction with the vmscan routines is not optimal yet.
 * Either vmscan must be made nfs-savvy, or we need a different page
 * reclaim concept that supports something like FS-independent
 * buffer_heads with a b_ops-> field.
 *
 * Copyright (C) 1996, 1997, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/mpage.h>
#include <linux/writeback.h>

#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
#include <asm/uaccess.h>
#include <linux/smp_lock.h>

#include "delegation.h"
#include "iostat.h"

#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

#define MIN_POOL_WRITE		(32)
#define MIN_POOL_COMMIT		(4)

/*
 * Local function declarations
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context*,
					    struct inode *,
					    struct page *,
					    unsigned int, unsigned int);
static int nfs_wait_on_write_congestion(struct address_space *, int);
static int nfs_wait_on_requests(struct inode *, unsigned long, unsigned int);
static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
			   unsigned int npages, int how);
static const struct rpc_call_ops nfs_write_partial_ops;
static const struct rpc_call_ops nfs_write_full_ops;
static const struct rpc_call_ops nfs_commit_ops;

static kmem_cache_t *nfs_wdata_cachep;
static mempool_t *nfs_wdata_mempool;
static mempool_t *nfs_commit_mempool;

static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion);

struct nfs_write_data *nfs_commit_alloc(void)
{
	struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS);

	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

void nfs_commit_free(struct nfs_write_data *p)
{
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
	mempool_free(p, nfs_commit_mempool);
}

struct nfs_write_data *nfs_writedata_alloc(size_t len)
{
	unsigned int pagecount = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);

	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
		p->npages = pagecount;
		if (pagecount <= ARRAY_SIZE(p->page_array))
			p->pagevec = p->page_array;
		else {
			p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS);
			if (!p->pagevec) {
				mempool_free(p, nfs_wdata_mempool);
				p = NULL;
			}
		}
	}
	return p;
}

static void nfs_writedata_free(struct nfs_write_data *p)
{
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
	mempool_free(p, nfs_wdata_mempool);
}

void nfs_writedata_release(void *wdata)
{
	nfs_writedata_free(wdata);
}

/* Adjust the file length if we're writing beyond the end */
static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int count)
{
	struct inode *inode = page->mapping->host;
	loff_t end, i_size = i_size_read(inode);
	unsigned long end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;

	if (i_size > 0 && page->index < end_index)
		return;
	end = ((loff_t)page->index << PAGE_CACHE_SHIFT) + ((loff_t)offset+count);
	if (i_size >= end)
		return;
	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
	i_size_write(inode, end);
}

/* We can set the PG_uptodate flag if we see that a write request
 * covers the full page.
 */
static void nfs_mark_uptodate(struct page *page, unsigned int base, unsigned int count)
{
	loff_t end_offs;

	if (PageUptodate(page))
		return;
	if (base != 0)
		return;
	if (count == PAGE_CACHE_SIZE) {
		SetPageUptodate(page);
		return;
	}

	end_offs = i_size_read(page->mapping->host) - 1;
	if (end_offs < 0)
		return;
	/* Is this the last page? */
	if (page->index != (unsigned long)(end_offs >> PAGE_CACHE_SHIFT))
		return;
	/* This is the last page: set PG_uptodate if we cover the entire
	 * extent of the data, then zero the rest of the page.
	 */
	if (count == (unsigned int)(end_offs & (PAGE_CACHE_SIZE - 1)) + 1) {
		memclear_highpage_flush(page, count, PAGE_CACHE_SIZE - count);
		SetPageUptodate(page);
	}
}

/*
 * Write a page synchronously.
 * Offset is the data offset within the page.
 */
static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode,
		struct page *page, unsigned