write.c

Linux Kernel 2.6.9 for OMAP1710

/*
 * 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 <[e-mail]>
 */

#include <linux/config.h>
#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 <linux/mempool.h>

#include "delegation.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 void nfs_writeback_done_partial(struct nfs_write_data *, int);
static void nfs_writeback_done_full(struct nfs_write_data *, 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 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);

static __inline__ struct nfs_write_data *nfs_writedata_alloc(void)
{
	struct nfs_write_data	*p;
	p = (struct nfs_write_data *)mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

static __inline__ void nfs_writedata_free(struct nfs_write_data *p)
{
	mempool_free(p, nfs_wdata_mempool);
}

static void nfs_writedata_release(struct rpc_task *task)
{
	struct nfs_write_data	*wdata = (struct nfs_write_data *)task->tk_calldata;
	nfs_writedata_free(wdata);
}

static __inline__ struct nfs_write_data *nfs_commit_alloc(void)
{
	struct nfs_write_data	*p;
	p = (struct nfs_write_data *)mempool_alloc(nfs_commit_mempool, SLAB_NOFS);
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

static __inline__ void nfs_commit_free(struct nfs_write_data *p)
{
	mempool_free(p, nfs_commit_mempool);
}

/* 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;
	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 int offset, unsigned int count,
		int how)
{
	unsigned int	wsize = NFS_SERVER(inode)->wsize;
	int		result, written = 0;
	struct nfs_write_data *wdata;

	wdata = kmalloc(sizeof(*wdata), GFP_NOFS);
	if (!wdata)
		return -ENOMEM;

	memset(wdata, 0, sizeof(*wdata));
	wdata->flags = how;
	wdata->cred = ctx->cred;
	wdata->inode = inode;
	wdata->args.fh = NFS_FH(inode);
	wdata->args.context = ctx;
	wdata->args.pages = &page;
	wdata->args.stable = NFS_FILE_SYNC;
	wdata->args.pgbase = offset;
	wdata->args.count = wsize;
	wdata->res.fattr = &wdata->fattr;
	wdata->res.verf = &wdata->verf;

	dprintk("NFS:      nfs_writepage_sync(%s/%Ld %d@%Ld)\n",
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count, (long long)(page_offset(page) + offset));

	nfs_begin_data_update(inode);
	do {
		if (count < wsize)
			wdata->args.count = count;
		wdata->args.offset = page_offset(page) + wdata->args.pgbase;

		result = NFS_PROTO(inode)->write(wdata);

		if (result < 0) {
			/* Must mark the page invalid after I/O error */
			ClearPageUptodate(page);
			goto io_error;
		}
		if (result < wdata->args.count)
			printk(KERN_WARNING "NFS: short write, count=%u, result=%d\n",
					wdata->args.count, result);

		wdata->args.offset += result;
	        wdata->args.pgbase += result;
		written += result;
		count -= result;
	} while (count);
	/* Update file length */
	nfs_grow_file(page, offset, written);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, written);

	if (PageError(page))
		ClearPageError(page);

io_error:
	nfs_end_data_update_defer(inode);

	kfree(wdata);
	return written ? written : result;
}

static int nfs_writepage_async(struct nfs_open_context *ctx,
		struct inode *inode, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_page	*req;
	int		status;

	req = nfs_update_request(ctx, inode, page, offset, count);
	status = (IS_ERR(req)) ? PTR_ERR(req) : 0;
	if (status < 0)
		goto out;
	/* Update file length */
	nfs_grow_file(page, offset, count);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, count);
	nfs_unlock_request(req);
 out:
	return status;
}

static int wb_priority(struct writeback_control *wbc)
{
	if (wbc->for_reclaim)
		return FLUSH_HIGHPRI;
	if (wbc->for_kupdate)
		return FLUSH_LOWPRI;
	return 0;
}

/*
 * Write an mmapped page to the server.
 */
int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
	struct nfs_open_context *ctx;
	struct inode *inode = page->mapping->host;
	unsigned long end_index;
	unsigned offset = PAGE_CACHE_SIZE;
	loff_t i_size = i_size_read(inode);
	int inode_referenced = 0;
	int priority = wb_priority(wbc);
	int err;

	/*
	 * Note: We need to ensure that we have a reference to the inode
	 *       if we are to do asynchronous writes. If not, waiting
	 *       in nfs_wait_on_request() may deadlock with clear_inode().
	 *
	 *       If igrab() fails here, then it is in any case safe to
	 *       call nfs_wb_page(), since there will be no pending writes.
	 */
	if (igrab(inode) != 0)
		inode_referenced = 1;
	end_index = i_size >> PAGE_CACHE_SHIFT;

	/* Ensure we've flushed out any previous writes */
	nfs_wb_page_priority(inode, page, priority);

	/* easy case */
	if (page->index < end_index)
		goto do_it;
	/* things got complicated... */
	offset = i_size & (PAGE_CACHE_SIZE-1);

	/* OK, are we completely out? */
	err = 0; /* potential race with truncate - ignore */
	if (page->index >= end_index+1 || !offset)
		goto out;
do_it:
	ctx = nfs_find_open_context(inode, FMODE_WRITE);
	if (ctx == NULL) {
		err = -EBADF;
		goto out;
	}
	lock_kernel();
	if (!IS_SYNC(inode) && inode_referenced) {
		err = nfs_writepage_async(ctx, inode, page, 0, offset);
		if (err >= 0) {
			err = 0;
			if (wbc->for_reclaim)
				nfs_flush_inode(inode, 0, 0, FLUSH_STABLE);
		}
	} else {
		err = nfs_writepage_sync(ctx, inode, page, 0,
						offset, priority);
		if (err >= 0) {
			if (err != offset)
				redirty_page_for_writepage(wbc, page);
			err = 0;
		}
	}
	unlock_kernel();
	put_nfs_open_context(ctx);
out:
	unlock_page(page);
	if (inode_referenced)
		iput(inode);
	return err; 
}

/*
 * Note: causes nfs_update_request() to block on the assumption
 * 	 that the writeback is generated due to memory pressure.
 */
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	struct inode *inode = mapping->host;
	int err;

	err = generic_writepages(mapping, wbc);
	if (err)
		return err;
	while (test_and_set_bit(BDI_write_congested, &bdi->state) != 0) {
		if (wbc->nonblocking)
			return 0;
		nfs_wait_on_write_congestion(mapping, 0);
	}
	err = nfs_flush_inode(inode, 0, 0, wb_priority(wbc));
	if (err < 0)
		goto out;
	wbc->nr_to_write -= err;
	if (!wbc->nonblocking && wbc->sync_mode == WB_SYNC_ALL) {
		err = nfs_wait_on_requests(inode, 0, 0);
		if (err < 0)
			goto out;
	}
	err = nfs_commit_inode(inode, 0, 0, wb_priority(wbc));
	if (err > 0) {
		wbc->nr_to_write -= err;
		err = 0;
	}
out:
	clear_bit(BDI_write_congested, &bdi->state);
	wake_up_all(&nfs_write_congestion);
	return err;
}

/*
 * Insert a write request into an inode
 */
static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	int error;

	error = radix_tree_insert(&nfsi->nfs_page_tree, req->wb_index, req);
	BUG_ON(error == -EEXIST);
	if (error)
		return error;
	if (!nfsi->npages) {
		igrab(inode);
		nfs_begin_data_update(inode);
		if (nfs_have_delegation(inode, FMODE_WRITE))
			nfsi->change_attr++;
	}
	nfsi->npages++;
	atomic_inc(&req->wb_count);
	return 0;
}

/*
 * Insert a write request into an inode
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	BUG_ON (!NFS_WBACK_BUSY(req));

	spin_lock(&nfsi->req_lock);
	radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
	nfsi->npages--;
	if (!nfsi->npages) {
		spin_unlock(&nfsi->req_lock);
		nfs_end_data_update_defer(inode);
		iput(inode);
	} else
		spin_unlock(&nfsi->req_lock);
	nfs_clear_request(req);
	nfs_release_request(req);
}

/*
 * Find a request
 */
static inline struct nfs_page *
_nfs_find_request(struct inode *inode, unsigned long index)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page *req;

	req = (struct nfs_page*)radix_tree_lookup(&nfsi->nfs_page_tree, index);
	if (req)
		atomic_inc(&req->wb_count);
	return req;
}

static struct nfs_page *
nfs_find_request(struct inode *inode, unsigned long index)
{
	struct nfs_page		*req;
	struct nfs_inode	*nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
	req = _nfs_find_request(inode, index);
	spin_unlock(&nfsi->req_lock);
	return req;
}

/*
 * Add a request to the inode's dirty list.
 */
static void
nfs_mark_request_dirty(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
	nfs_list_add_request(req, &nfsi->dirty);
	nfsi->ndirty++;
	spin_unlock(&nfsi->req_lock);
	inc_page_state(nr_dirty);
	mark_inode_dirty(inode);
}

/*
 * Check if a request is dirty
 */
static inline int
nfs_dirty_request(struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
	return !list_empty(&req->wb_list) && req->wb_list_head == &nfsi->dirty;
}