2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* linux/fs/nfs/write.c
|
|
|
|
*
|
2006-12-14 03:23:48 +07:00
|
|
|
* Write file data over NFS.
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
|
|
|
* 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/writeback.h>
|
2007-03-17 04:38:26 +07:00
|
|
|
#include <linux/swap.h>
|
2009-08-10 19:54:13 +07:00
|
|
|
#include <linux/migrate.h>
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#include <linux/sunrpc/clnt.h>
|
|
|
|
#include <linux/nfs_fs.h>
|
|
|
|
#include <linux/nfs_mount.h>
|
|
|
|
#include <linux/nfs_page.h>
|
2006-10-20 13:28:16 +07:00
|
|
|
#include <linux/backing-dev.h>
|
2011-05-27 03:00:52 +07:00
|
|
|
#include <linux/export.h>
|
2006-10-20 13:28:16 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#include <asm/uaccess.h>
|
|
|
|
|
|
|
|
#include "delegation.h"
|
2006-12-05 12:35:38 +07:00
|
|
|
#include "internal.h"
|
2006-03-21 01:44:14 +07:00
|
|
|
#include "iostat.h"
|
2009-04-01 20:22:26 +07:00
|
|
|
#include "nfs4_fs.h"
|
2009-08-10 19:54:13 +07:00
|
|
|
#include "fscache.h"
|
2011-03-01 08:34:14 +07:00
|
|
|
#include "pnfs.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#define NFSDBG_FACILITY NFSDBG_PAGECACHE
|
|
|
|
|
|
|
|
#define MIN_POOL_WRITE (32)
|
|
|
|
#define MIN_POOL_COMMIT (4)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Local function declarations
|
|
|
|
*/
|
nfs: don't ignore return value from nfs_pageio_add_request
Ignoring the return value from nfs_pageio_add_request can cause deadlocks.
In read path:
call nfs_pageio_add_request from readpage_async_filler
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_pageio_add_request to return 0, *WITHOUT* adding the original
request.
BUT, since return code is ignored, readpage_async_filler assumes it has
been added, and does nothing further, leaving page locked.
do_generic_mapping_read will eventually call lock_page, resulting in deadlock
In write path:
page is marked dirty by generic_perform_write
nfs_writepages is called
call nfs_pageio_add_request from nfs_page_async_flush
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_page_async_flush to return 0, *WITHOUT* adding the original
request, yet marking the request as locked (PG_BUSY) and in writeback,
clearing dirty marks.
The next time a write is done to the page, deadlock will result as
nfs_write_end calls nfs_update_request
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-19 22:24:39 +07:00
|
|
|
static void nfs_redirty_request(struct nfs_page *req);
|
2012-04-21 01:47:47 +07:00
|
|
|
static const struct rpc_call_ops nfs_write_common_ops;
|
2006-03-21 01:44:27 +07:00
|
|
|
static const struct rpc_call_ops nfs_commit_ops;
|
2012-04-21 01:47:48 +07:00
|
|
|
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops;
|
2012-04-21 01:47:54 +07:00
|
|
|
static const struct nfs_commit_completion_ops nfs_commit_completion_ops;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-12-07 11:33:20 +07:00
|
|
|
static struct kmem_cache *nfs_wdata_cachep;
|
2006-03-21 01:44:37 +07:00
|
|
|
static mempool_t *nfs_wdata_mempool;
|
2012-04-21 01:47:39 +07:00
|
|
|
static struct kmem_cache *nfs_cdata_cachep;
|
2005-04-17 05:20:36 +07:00
|
|
|
static mempool_t *nfs_commit_mempool;
|
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
struct nfs_commit_data *nfs_commitdata_alloc(void)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-08-01 06:45:16 +07:00
|
|
|
struct nfs_commit_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOIO);
|
2005-12-01 06:09:02 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
if (p) {
|
|
|
|
memset(p, 0, sizeof(*p));
|
|
|
|
INIT_LIST_HEAD(&p->pages);
|
|
|
|
}
|
|
|
|
return p;
|
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_commitdata_alloc);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
void nfs_commit_free(struct nfs_commit_data *p)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
mempool_free(p, nfs_commit_mempool);
|
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_commit_free);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
struct nfs_write_header *nfs_writehdr_alloc(void)
|
2006-03-21 01:44:37 +07:00
|
|
|
{
|
2012-08-01 06:45:16 +07:00
|
|
|
struct nfs_write_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOIO);
|
2006-03-21 01:44:37 +07:00
|
|
|
|
|
|
|
if (p) {
|
2012-04-21 01:47:44 +07:00
|
|
|
struct nfs_pgio_header *hdr = &p->header;
|
|
|
|
|
2006-03-21 01:44:37 +07:00
|
|
|
memset(p, 0, sizeof(*p));
|
2012-04-21 01:47:44 +07:00
|
|
|
INIT_LIST_HEAD(&hdr->pages);
|
2012-04-21 01:47:47 +07:00
|
|
|
INIT_LIST_HEAD(&hdr->rpc_list);
|
|
|
|
spin_lock_init(&hdr->lock);
|
|
|
|
atomic_set(&hdr->refcnt, 0);
|
2012-06-06 05:32:03 +07:00
|
|
|
hdr->verf = &p->verf;
|
2006-03-21 01:44:37 +07:00
|
|
|
}
|
|
|
|
return p;
|
|
|
|
}
|
2012-07-31 03:05:25 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_writehdr_alloc);
|
2006-03-21 01:44:37 +07:00
|
|
|
|
2012-04-21 01:47:57 +07:00
|
|
|
static struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr,
|
|
|
|
unsigned int pagecount)
|
2012-04-21 01:47:47 +07:00
|
|
|
{
|
|
|
|
struct nfs_write_data *data, *prealloc;
|
|
|
|
|
|
|
|
prealloc = &container_of(hdr, struct nfs_write_header, header)->rpc_data;
|
|
|
|
if (prealloc->header == NULL)
|
|
|
|
data = prealloc;
|
|
|
|
else
|
|
|
|
data = kzalloc(sizeof(*data), GFP_KERNEL);
|
|
|
|
if (!data)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (nfs_pgarray_set(&data->pages, pagecount)) {
|
|
|
|
data->header = hdr;
|
|
|
|
atomic_inc(&hdr->refcnt);
|
|
|
|
} else {
|
|
|
|
if (data != prealloc)
|
|
|
|
kfree(data);
|
|
|
|
data = NULL;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
return data;
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:44 +07:00
|
|
|
void nfs_writehdr_free(struct nfs_pgio_header *hdr)
|
2006-03-21 01:44:37 +07:00
|
|
|
{
|
2012-04-21 01:47:44 +07:00
|
|
|
struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header);
|
|
|
|
mempool_free(whdr, nfs_wdata_mempool);
|
2006-03-21 01:44:37 +07:00
|
|
|
}
|
2012-07-31 03:05:25 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_writehdr_free);
|
2006-03-21 01:44:37 +07:00
|
|
|
|
2011-07-14 02:59:19 +07:00
|
|
|
void nfs_writedata_release(struct nfs_write_data *wdata)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:47 +07:00
|
|
|
struct nfs_pgio_header *hdr = wdata->header;
|
|
|
|
struct nfs_write_header *write_header = container_of(hdr, struct nfs_write_header, header);
|
|
|
|
|
2008-02-20 08:04:20 +07:00
|
|
|
put_nfs_open_context(wdata->args.context);
|
2012-04-21 01:47:45 +07:00
|
|
|
if (wdata->pages.pagevec != wdata->pages.page_array)
|
|
|
|
kfree(wdata->pages.pagevec);
|
2013-01-05 00:47:04 +07:00
|
|
|
if (wdata == &write_header->rpc_data) {
|
2012-04-21 01:47:47 +07:00
|
|
|
wdata->header = NULL;
|
2013-01-05 00:47:04 +07:00
|
|
|
wdata = NULL;
|
|
|
|
}
|
2012-04-21 01:47:47 +07:00
|
|
|
if (atomic_dec_and_test(&hdr->refcnt))
|
2012-04-21 01:47:48 +07:00
|
|
|
hdr->completion_ops->completion(hdr);
|
2013-01-05 00:47:04 +07:00
|
|
|
/* Note: we only free the rpc_task after callbacks are done.
|
|
|
|
* See the comment in rpc_free_task() for why
|
|
|
|
*/
|
|
|
|
kfree(wdata);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-07-31 03:05:25 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_writedata_release);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2007-07-26 01:09:54 +07:00
|
|
|
static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error)
|
|
|
|
{
|
|
|
|
ctx->error = error;
|
|
|
|
smp_wmb();
|
|
|
|
set_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
|
|
|
|
}
|
|
|
|
|
2012-08-01 06:45:10 +07:00
|
|
|
static struct nfs_page *
|
|
|
|
nfs_page_find_request_locked(struct nfs_inode *nfsi, struct page *page)
|
2006-12-05 12:35:35 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req = NULL;
|
|
|
|
|
2012-08-01 06:45:10 +07:00
|
|
|
if (PagePrivate(page))
|
2006-12-05 12:35:35 +07:00
|
|
|
req = (struct nfs_page *)page_private(page);
|
2012-08-01 06:45:10 +07:00
|
|
|
else if (unlikely(PageSwapCache(page))) {
|
|
|
|
struct nfs_page *freq, *t;
|
|
|
|
|
|
|
|
/* Linearly search the commit list for the correct req */
|
|
|
|
list_for_each_entry_safe(freq, t, &nfsi->commit_info.list, wb_list) {
|
|
|
|
if (freq->wb_page == page) {
|
|
|
|
req = freq;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2006-12-05 12:35:35 +07:00
|
|
|
}
|
2012-08-01 06:45:10 +07:00
|
|
|
|
|
|
|
if (req)
|
|
|
|
kref_get(&req->wb_kref);
|
|
|
|
|
2006-12-05 12:35:35 +07:00
|
|
|
return req;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct nfs_page *nfs_page_find_request(struct page *page)
|
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2006-12-05 12:35:35 +07:00
|
|
|
struct nfs_page *req = NULL;
|
|
|
|
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
2012-08-01 06:45:10 +07:00
|
|
|
req = nfs_page_find_request_locked(NFS_I(inode), page);
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2006-12-05 12:35:35 +07:00
|
|
|
return req;
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 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)
|
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2008-06-11 23:21:19 +07:00
|
|
|
loff_t end, i_size;
|
|
|
|
pgoff_t end_index;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-06-11 23:21:19 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
|
|
|
i_size = i_size_read(inode);
|
|
|
|
end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
|
2012-08-01 06:45:06 +07:00
|
|
|
if (i_size > 0 && page_file_index(page) < end_index)
|
2008-06-11 23:21:19 +07:00
|
|
|
goto out;
|
2012-08-01 06:45:06 +07:00
|
|
|
end = page_file_offset(page) + ((loff_t)offset+count);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (i_size >= end)
|
2008-06-11 23:21:19 +07:00
|
|
|
goto out;
|
2005-04-17 05:20:36 +07:00
|
|
|
i_size_write(inode, end);
|
2008-06-11 23:21:19 +07:00
|
|
|
nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
|
|
|
|
out:
|
|
|
|
spin_unlock(&inode->i_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2007-02-07 02:07:15 +07:00
|
|
|
/* A writeback failed: mark the page as bad, and invalidate the page cache */
|
|
|
|
static void nfs_set_pageerror(struct page *page)
|
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
nfs_zap_mapping(page_file_mapping(page)->host, page_file_mapping(page));
|
2007-02-07 02:07:15 +07:00
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 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)
|
|
|
|
{
|
|
|
|
if (PageUptodate(page))
|
|
|
|
return;
|
|
|
|
if (base != 0)
|
|
|
|
return;
|
2006-12-05 12:35:38 +07:00
|
|
|
if (count != nfs_page_length(page))
|
2005-04-17 05:20:36 +07:00
|
|
|
return;
|
2006-12-05 12:35:38 +07:00
|
|
|
SetPageUptodate(page);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int wb_priority(struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
if (wbc->for_reclaim)
|
2007-04-03 06:29:52 +07:00
|
|
|
return FLUSH_HIGHPRI | FLUSH_STABLE;
|
2009-12-03 19:54:25 +07:00
|
|
|
if (wbc->for_kupdate || wbc->for_background)
|
2011-03-22 04:02:00 +07:00
|
|
|
return FLUSH_LOWPRI | FLUSH_COND_STABLE;
|
|
|
|
return FLUSH_COND_STABLE;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2007-03-17 04:38:26 +07:00
|
|
|
/*
|
|
|
|
* NFS congestion control
|
|
|
|
*/
|
|
|
|
|
|
|
|
int nfs_congestion_kb;
|
|
|
|
|
|
|
|
#define NFS_CONGESTION_ON_THRESH (nfs_congestion_kb >> (PAGE_SHIFT-10))
|
|
|
|
#define NFS_CONGESTION_OFF_THRESH \
|
|
|
|
(NFS_CONGESTION_ON_THRESH - (NFS_CONGESTION_ON_THRESH >> 2))
|
|
|
|
|
2012-10-16 02:02:01 +07:00
|
|
|
static void nfs_set_page_writeback(struct page *page)
|
2007-03-17 04:38:26 +07:00
|
|
|
{
|
2012-10-16 02:02:01 +07:00
|
|
|
struct nfs_server *nfss = NFS_SERVER(page_file_mapping(page)->host);
|
2007-04-15 06:10:12 +07:00
|
|
|
int ret = test_set_page_writeback(page);
|
|
|
|
|
2012-10-16 02:02:01 +07:00
|
|
|
WARN_ON_ONCE(ret != 0);
|
2007-03-17 04:38:26 +07:00
|
|
|
|
2012-10-16 02:02:01 +07:00
|
|
|
if (atomic_long_inc_return(&nfss->writeback) >
|
|
|
|
NFS_CONGESTION_ON_THRESH) {
|
|
|
|
set_bdi_congested(&nfss->backing_dev_info,
|
|
|
|
BLK_RW_ASYNC);
|
2007-03-17 04:38:26 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nfs_end_page_writeback(struct page *page)
|
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2007-03-17 04:38:26 +07:00
|
|
|
struct nfs_server *nfss = NFS_SERVER(inode);
|
|
|
|
|
|
|
|
end_page_writeback(page);
|
2007-10-17 13:25:41 +07:00
|
|
|
if (atomic_long_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
|
2009-07-09 19:52:32 +07:00
|
|
|
clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
|
2007-03-17 04:38:26 +07:00
|
|
|
}
|
|
|
|
|
2010-07-31 02:31:57 +07:00
|
|
|
static struct nfs_page *nfs_find_and_lock_request(struct page *page, bool nonblock)
|
2006-12-05 12:35:41 +07:00
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2006-12-05 12:35:41 +07:00
|
|
|
struct nfs_page *req;
|
|
|
|
int ret;
|
|
|
|
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
2009-08-10 19:54:13 +07:00
|
|
|
for (;;) {
|
2012-08-01 06:45:10 +07:00
|
|
|
req = nfs_page_find_request_locked(NFS_I(inode), page);
|
2009-08-10 19:54:13 +07:00
|
|
|
if (req == NULL)
|
|
|
|
break;
|
2012-05-10 00:19:15 +07:00
|
|
|
if (nfs_lock_request(req))
|
2006-12-05 12:35:41 +07:00
|
|
|
break;
|
|
|
|
/* Note: If we hold the page lock, as is the case in nfs_writepage,
|
2012-05-10 00:19:15 +07:00
|
|
|
* then the call to nfs_lock_request() will always
|
2006-12-05 12:35:41 +07:00
|
|
|
* succeed provided that someone hasn't already marked the
|
|
|
|
* request as dirty (in which case we don't care).
|
|
|
|
*/
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2010-07-31 02:31:57 +07:00
|
|
|
if (!nonblock)
|
|
|
|
ret = nfs_wait_on_request(req);
|
|
|
|
else
|
|
|
|
ret = -EAGAIN;
|
2006-12-05 12:35:41 +07:00
|
|
|
nfs_release_request(req);
|
|
|
|
if (ret != 0)
|
2009-08-10 19:54:13 +07:00
|
|
|
return ERR_PTR(ret);
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
2006-12-05 12:35:41 +07:00
|
|
|
}
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2009-08-10 19:54:13 +07:00
|
|
|
return req;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find an associated nfs write request, and prepare to flush it out
|
|
|
|
* May return an error if the user signalled nfs_wait_on_request().
|
|
|
|
*/
|
|
|
|
static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
|
2010-07-31 02:31:57 +07:00
|
|
|
struct page *page, bool nonblock)
|
2009-08-10 19:54:13 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
int ret = 0;
|
|
|
|
|
2010-07-31 02:31:57 +07:00
|
|
|
req = nfs_find_and_lock_request(page, nonblock);
|
2009-08-10 19:54:13 +07:00
|
|
|
if (!req)
|
|
|
|
goto out;
|
|
|
|
ret = PTR_ERR(req);
|
|
|
|
if (IS_ERR(req))
|
|
|
|
goto out;
|
|
|
|
|
2012-10-16 02:02:01 +07:00
|
|
|
nfs_set_page_writeback(page);
|
|
|
|
WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
|
2009-08-10 19:54:13 +07:00
|
|
|
|
2012-10-16 02:02:01 +07:00
|
|
|
ret = 0;
|
nfs: don't ignore return value from nfs_pageio_add_request
Ignoring the return value from nfs_pageio_add_request can cause deadlocks.
In read path:
call nfs_pageio_add_request from readpage_async_filler
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_pageio_add_request to return 0, *WITHOUT* adding the original
request.
BUT, since return code is ignored, readpage_async_filler assumes it has
been added, and does nothing further, leaving page locked.
do_generic_mapping_read will eventually call lock_page, resulting in deadlock
In write path:
page is marked dirty by generic_perform_write
nfs_writepages is called
call nfs_pageio_add_request from nfs_page_async_flush
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_page_async_flush to return 0, *WITHOUT* adding the original
request, yet marking the request as locked (PG_BUSY) and in writeback,
clearing dirty marks.
The next time a write is done to the page, deadlock will result as
nfs_write_end calls nfs_update_request
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-19 22:24:39 +07:00
|
|
|
if (!nfs_pageio_add_request(pgio, req)) {
|
|
|
|
nfs_redirty_request(req);
|
2009-08-10 19:54:13 +07:00
|
|
|
ret = pgio->pg_error;
|
nfs: don't ignore return value from nfs_pageio_add_request
Ignoring the return value from nfs_pageio_add_request can cause deadlocks.
In read path:
call nfs_pageio_add_request from readpage_async_filler
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_pageio_add_request to return 0, *WITHOUT* adding the original
request.
BUT, since return code is ignored, readpage_async_filler assumes it has
been added, and does nothing further, leaving page locked.
do_generic_mapping_read will eventually call lock_page, resulting in deadlock
In write path:
page is marked dirty by generic_perform_write
nfs_writepages is called
call nfs_pageio_add_request from nfs_page_async_flush
assume at this point that there are requests already in desc, that
can't be merged with the current request.
so nfs_pageio_doio is fired up to clear out desc.
assume something goes wrong in setting up the io, so desc->pg_error is set.
This causes nfs_page_async_flush to return 0, *WITHOUT* adding the original
request, yet marking the request as locked (PG_BUSY) and in writeback,
clearing dirty marks.
The next time a write is done to the page, deadlock will result as
nfs_write_end calls nfs_update_request
Signed-off-by: Fred Isaman <iisaman@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-19 22:24:39 +07:00
|
|
|
}
|
2009-08-10 19:54:13 +07:00
|
|
|
out:
|
|
|
|
return ret;
|
2006-12-05 12:35:41 +07:00
|
|
|
}
|
|
|
|
|
2007-07-23 06:27:32 +07:00
|
|
|
static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, struct nfs_pageio_descriptor *pgio)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2010-07-31 02:31:57 +07:00
|
|
|
int ret;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-03-21 01:44:14 +07:00
|
|
|
nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
|
|
|
|
nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);
|
|
|
|
|
2012-08-01 06:45:06 +07:00
|
|
|
nfs_pageio_cond_complete(pgio, page_file_index(page));
|
2010-10-27 04:21:26 +07:00
|
|
|
ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
|
2010-07-31 02:31:57 +07:00
|
|
|
if (ret == -EAGAIN) {
|
|
|
|
redirty_page_for_writepage(wbc, page);
|
|
|
|
ret = 0;
|
|
|
|
}
|
|
|
|
return ret;
|
2007-07-23 06:27:32 +07:00
|
|
|
}
|
2007-05-20 21:18:27 +07:00
|
|
|
|
2007-07-23 06:27:32 +07:00
|
|
|
/*
|
|
|
|
* Write an mmapped page to the server.
|
|
|
|
*/
|
|
|
|
static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
struct nfs_pageio_descriptor pgio;
|
|
|
|
int err;
|
2006-12-05 12:35:38 +07:00
|
|
|
|
2012-08-01 06:45:06 +07:00
|
|
|
NFS_PROTO(page_file_mapping(page)->host)->write_pageio_init(&pgio,
|
2012-06-21 02:53:48 +07:00
|
|
|
page->mapping->host,
|
|
|
|
wb_priority(wbc),
|
|
|
|
&nfs_async_write_completion_ops);
|
2007-07-23 06:27:32 +07:00
|
|
|
err = nfs_do_writepage(page, wbc, &pgio);
|
|
|
|
nfs_pageio_complete(&pgio);
|
|
|
|
if (err < 0)
|
|
|
|
return err;
|
|
|
|
if (pgio.pg_error < 0)
|
|
|
|
return pgio.pg_error;
|
|
|
|
return 0;
|
2006-12-05 12:35:41 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int nfs_writepage(struct page *page, struct writeback_control *wbc)
|
|
|
|
{
|
2007-07-23 06:27:32 +07:00
|
|
|
int ret;
|
2006-12-05 12:35:41 +07:00
|
|
|
|
2007-07-23 06:27:32 +07:00
|
|
|
ret = nfs_writepage_locked(page, wbc);
|
2005-04-17 05:20:36 +07:00
|
|
|
unlock_page(page);
|
2007-07-23 06:27:32 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nfs_writepages_callback(struct page *page, struct writeback_control *wbc, void *data)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = nfs_do_writepage(page, wbc, data);
|
|
|
|
unlock_page(page);
|
|
|
|
return ret;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
struct inode *inode = mapping->host;
|
2009-03-12 01:10:30 +07:00
|
|
|
unsigned long *bitlock = &NFS_I(inode)->flags;
|
2007-04-03 06:29:52 +07:00
|
|
|
struct nfs_pageio_descriptor pgio;
|
2005-04-17 05:20:36 +07:00
|
|
|
int err;
|
|
|
|
|
2009-03-12 01:10:30 +07:00
|
|
|
/* Stop dirtying of new pages while we sync */
|
|
|
|
err = wait_on_bit_lock(bitlock, NFS_INO_FLUSHING,
|
|
|
|
nfs_wait_bit_killable, TASK_KILLABLE);
|
|
|
|
if (err)
|
|
|
|
goto out_err;
|
|
|
|
|
2006-03-21 01:44:14 +07:00
|
|
|
nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);
|
|
|
|
|
2012-06-21 02:53:48 +07:00
|
|
|
NFS_PROTO(inode)->write_pageio_init(&pgio, inode, wb_priority(wbc), &nfs_async_write_completion_ops);
|
2007-07-23 06:27:32 +07:00
|
|
|
err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio);
|
2007-04-03 06:29:52 +07:00
|
|
|
nfs_pageio_complete(&pgio);
|
2009-03-12 01:10:30 +07:00
|
|
|
|
|
|
|
clear_bit_unlock(NFS_INO_FLUSHING, bitlock);
|
|
|
|
smp_mb__after_clear_bit();
|
|
|
|
wake_up_bit(bitlock, NFS_INO_FLUSHING);
|
|
|
|
|
2007-07-23 06:27:32 +07:00
|
|
|
if (err < 0)
|
2009-03-12 01:10:30 +07:00
|
|
|
goto out_err;
|
|
|
|
err = pgio.pg_error;
|
|
|
|
if (err < 0)
|
|
|
|
goto out_err;
|
2007-04-03 06:29:52 +07:00
|
|
|
return 0;
|
2009-03-12 01:10:30 +07:00
|
|
|
out_err:
|
|
|
|
return err;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Insert a write request into an inode
|
|
|
|
*/
|
2012-03-09 05:29:35 +07:00
|
|
|
static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
struct nfs_inode *nfsi = NFS_I(inode);
|
2008-06-13 23:12:32 +07:00
|
|
|
|
|
|
|
/* Lock the request! */
|
2012-05-10 00:19:15 +07:00
|
|
|
nfs_lock_request(req);
|
2008-06-13 23:12:32 +07:00
|
|
|
|
|
|
|
spin_lock(&inode->i_lock);
|
2012-06-21 02:53:43 +07:00
|
|
|
if (!nfsi->npages && NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
|
2011-10-18 06:08:46 +07:00
|
|
|
inode->i_version++;
|
2012-08-01 06:45:10 +07:00
|
|
|
/*
|
|
|
|
* Swap-space should not get truncated. Hence no need to plug the race
|
|
|
|
* with invalidate/truncate.
|
|
|
|
*/
|
|
|
|
if (likely(!PageSwapCache(req->wb_page))) {
|
|
|
|
set_bit(PG_MAPPED, &req->wb_flags);
|
|
|
|
SetPagePrivate(req->wb_page);
|
|
|
|
set_page_private(req->wb_page, (unsigned long)req);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
nfsi->npages++;
|
2007-06-18 00:26:38 +07:00
|
|
|
kref_get(&req->wb_kref);
|
2008-06-13 23:12:32 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2007-03-17 04:38:26 +07:00
|
|
|
* Remove a write request from an inode
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
static void nfs_inode_remove_request(struct nfs_page *req)
|
|
|
|
{
|
2011-06-23 05:40:12 +07:00
|
|
|
struct inode *inode = req->wb_context->dentry->d_inode;
|
2005-04-17 05:20:36 +07:00
|
|
|
struct nfs_inode *nfsi = NFS_I(inode);
|
|
|
|
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
2012-08-01 06:45:10 +07:00
|
|
|
if (likely(!PageSwapCache(req->wb_page))) {
|
|
|
|
set_page_private(req->wb_page, 0);
|
|
|
|
ClearPagePrivate(req->wb_page);
|
|
|
|
clear_bit(PG_MAPPED, &req->wb_flags);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
nfsi->npages--;
|
2011-03-26 01:15:11 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
nfs_release_request(req);
|
|
|
|
}
|
|
|
|
|
2006-12-05 12:35:42 +07:00
|
|
|
static void
|
2008-03-19 22:24:38 +07:00
|
|
|
nfs_mark_request_dirty(struct nfs_page *req)
|
2006-12-05 12:35:42 +07:00
|
|
|
{
|
|
|
|
__set_page_dirty_nobuffers(req->wb_page);
|
|
|
|
}
|
|
|
|
|
2012-07-31 03:05:25 +07:00
|
|
|
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
|
2012-03-16 04:16:40 +07:00
|
|
|
/**
|
|
|
|
* nfs_request_add_commit_list - add request to a commit list
|
|
|
|
* @req: pointer to a struct nfs_page
|
2012-04-21 01:47:53 +07:00
|
|
|
* @dst: commit list head
|
|
|
|
* @cinfo: holds list lock and accounting info
|
2012-03-16 04:16:40 +07:00
|
|
|
*
|
2012-04-21 01:47:53 +07:00
|
|
|
* This sets the PG_CLEAN bit, updates the cinfo count of
|
2012-03-16 04:16:40 +07:00
|
|
|
* number of outstanding requests requiring a commit as well as
|
|
|
|
* the MM page stats.
|
|
|
|
*
|
2012-04-21 01:47:53 +07:00
|
|
|
* The caller must _not_ hold the cinfo->lock, but must be
|
2012-03-16 04:16:40 +07:00
|
|
|
* holding the nfs_page lock.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2012-03-16 04:16:40 +07:00
|
|
|
void
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_request_add_commit_list(struct nfs_page *req, struct list_head *dst,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2008-06-14 00:25:22 +07:00
|
|
|
set_bit(PG_CLEAN, &(req)->wb_flags);
|
2012-04-21 01:47:53 +07:00
|
|
|
spin_lock(cinfo->lock);
|
|
|
|
nfs_list_add_request(req, dst);
|
|
|
|
cinfo->mds->ncommit++;
|
|
|
|
spin_unlock(cinfo->lock);
|
2012-04-21 01:47:56 +07:00
|
|
|
if (!cinfo->dreq) {
|
|
|
|
inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
|
2012-08-01 06:45:06 +07:00
|
|
|
inc_bdi_stat(page_file_mapping(req->wb_page)->backing_dev_info,
|
2012-04-21 01:47:56 +07:00
|
|
|
BDI_RECLAIMABLE);
|
|
|
|
__mark_inode_dirty(req->wb_context->dentry->d_inode,
|
|
|
|
I_DIRTY_DATASYNC);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-03-16 04:16:40 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_request_add_commit_list);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* nfs_request_remove_commit_list - Remove request from a commit list
|
|
|
|
* @req: pointer to a nfs_page
|
2012-04-21 01:47:53 +07:00
|
|
|
* @cinfo: holds list lock and accounting info
|
2012-03-16 04:16:40 +07:00
|
|
|
*
|
2012-04-21 01:47:53 +07:00
|
|
|
* This clears the PG_CLEAN bit, and updates the cinfo's count of
|
2012-03-16 04:16:40 +07:00
|
|
|
* number of outstanding requests requiring a commit
|
|
|
|
* It does not update the MM page stats.
|
|
|
|
*
|
2012-04-21 01:47:53 +07:00
|
|
|
* The caller _must_ hold the cinfo->lock and the nfs_page lock.
|
2012-03-16 04:16:40 +07:00
|
|
|
*/
|
|
|
|
void
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_request_remove_commit_list(struct nfs_page *req,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2012-03-16 04:16:40 +07:00
|
|
|
{
|
|
|
|
if (!test_and_clear_bit(PG_CLEAN, &(req)->wb_flags))
|
|
|
|
return;
|
|
|
|
nfs_list_remove_request(req);
|
2012-04-21 01:47:53 +07:00
|
|
|
cinfo->mds->ncommit--;
|
2012-03-16 04:16:40 +07:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);
|
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
cinfo->lock = &inode->i_lock;
|
|
|
|
cinfo->mds = &NFS_I(inode)->commit_info;
|
|
|
|
cinfo->ds = pnfs_get_ds_info(inode);
|
2012-04-21 01:47:55 +07:00
|
|
|
cinfo->dreq = NULL;
|
2012-04-21 01:47:54 +07:00
|
|
|
cinfo->completion_ops = &nfs_commit_completion_ops;
|
2012-04-21 01:47:53 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
void nfs_init_cinfo(struct nfs_commit_info *cinfo,
|
|
|
|
struct inode *inode,
|
|
|
|
struct nfs_direct_req *dreq)
|
|
|
|
{
|
2012-04-21 01:47:57 +07:00
|
|
|
if (dreq)
|
|
|
|
nfs_init_cinfo_from_dreq(cinfo, dreq);
|
|
|
|
else
|
|
|
|
nfs_init_cinfo_from_inode(cinfo, inode);
|
2012-04-21 01:47:53 +07:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(nfs_init_cinfo);
|
2012-03-16 04:16:40 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Add a request to the inode's commit list.
|
|
|
|
*/
|
2012-04-21 01:47:57 +07:00
|
|
|
void
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2012-03-16 04:16:40 +07:00
|
|
|
{
|
2012-04-21 01:47:53 +07:00
|
|
|
if (pnfs_mark_request_commit(req, lseg, cinfo))
|
2012-03-16 04:16:40 +07:00
|
|
|
return;
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_request_add_commit_list(req, &cinfo->mds->list, cinfo);
|
2012-03-16 04:16:40 +07:00
|
|
|
}
|
2007-04-21 03:12:34 +07:00
|
|
|
|
2012-03-09 05:29:35 +07:00
|
|
|
static void
|
|
|
|
nfs_clear_page_commit(struct page *page)
|
|
|
|
{
|
|
|
|
dec_zone_page_state(page, NR_UNSTABLE_NFS);
|
2012-08-01 06:45:06 +07:00
|
|
|
dec_bdi_stat(page_file_mapping(page)->backing_dev_info, BDI_RECLAIMABLE);
|
2012-03-09 05:29:35 +07:00
|
|
|
}
|
|
|
|
|
2012-03-16 04:16:40 +07:00
|
|
|
static void
|
2008-06-14 00:25:22 +07:00
|
|
|
nfs_clear_request_commit(struct nfs_page *req)
|
|
|
|
{
|
2012-03-16 04:16:40 +07:00
|
|
|
if (test_bit(PG_CLEAN, &req->wb_flags)) {
|
|
|
|
struct inode *inode = req->wb_context->dentry->d_inode;
|
2012-04-21 01:47:53 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2008-06-14 00:25:22 +07:00
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_init_cinfo_from_inode(&cinfo, inode);
|
|
|
|
if (!pnfs_clear_request_commit(req, &cinfo)) {
|
|
|
|
spin_lock(cinfo.lock);
|
|
|
|
nfs_request_remove_commit_list(req, &cinfo);
|
|
|
|
spin_unlock(cinfo.lock);
|
2012-03-16 04:16:40 +07:00
|
|
|
}
|
2012-03-09 05:29:35 +07:00
|
|
|
nfs_clear_page_commit(req->wb_page);
|
2008-06-14 00:25:22 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-04-21 03:12:34 +07:00
|
|
|
static inline
|
|
|
|
int nfs_write_need_commit(struct nfs_write_data *data)
|
|
|
|
{
|
2011-03-23 20:27:44 +07:00
|
|
|
if (data->verf.committed == NFS_DATA_SYNC)
|
2012-04-21 01:47:44 +07:00
|
|
|
return data->header->lseg == NULL;
|
|
|
|
return data->verf.committed != NFS_FILE_SYNC;
|
2007-04-21 03:12:34 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
#else
|
2012-05-01 01:30:22 +07:00
|
|
|
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
void nfs_init_cinfo(struct nfs_commit_info *cinfo,
|
|
|
|
struct inode *inode,
|
|
|
|
struct nfs_direct_req *dreq)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:57 +07:00
|
|
|
void
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2007-04-21 03:12:34 +07:00
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2012-03-16 04:16:40 +07:00
|
|
|
static void
|
2008-06-14 00:25:22 +07:00
|
|
|
nfs_clear_request_commit(struct nfs_page *req)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2007-04-21 03:12:34 +07:00
|
|
|
static inline
|
|
|
|
int nfs_write_need_commit(struct nfs_write_data *data)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
#endif
|
|
|
|
|
2012-04-21 01:47:48 +07:00
|
|
|
static void nfs_write_completion(struct nfs_pgio_header *hdr)
|
2007-04-21 03:12:34 +07:00
|
|
|
{
|
2012-04-21 01:47:53 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2012-04-21 01:47:47 +07:00
|
|
|
unsigned long bytes = 0;
|
|
|
|
|
|
|
|
if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
|
|
|
|
goto out;
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_init_cinfo_from_inode(&cinfo, hdr->inode);
|
2012-04-21 01:47:47 +07:00
|
|
|
while (!list_empty(&hdr->pages)) {
|
|
|
|
struct nfs_page *req = nfs_list_entry(hdr->pages.next);
|
|
|
|
|
|
|
|
bytes += req->wb_bytes;
|
|
|
|
nfs_list_remove_request(req);
|
|
|
|
if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) &&
|
|
|
|
(hdr->good_bytes < bytes)) {
|
2012-05-10 00:37:43 +07:00
|
|
|
nfs_set_pageerror(req->wb_page);
|
2012-04-21 01:47:47 +07:00
|
|
|
nfs_context_set_write_error(req->wb_context, hdr->error);
|
|
|
|
goto remove_req;
|
|
|
|
}
|
|
|
|
if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) {
|
|
|
|
nfs_mark_request_dirty(req);
|
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) {
|
2012-06-08 22:56:09 +07:00
|
|
|
memcpy(&req->wb_verf, &hdr->verf->verifier, sizeof(req->wb_verf));
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_mark_request_commit(req, hdr->lseg, &cinfo);
|
2012-04-21 01:47:47 +07:00
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
remove_req:
|
|
|
|
nfs_inode_remove_request(req);
|
|
|
|
next:
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_request(req);
|
2012-05-10 00:37:43 +07:00
|
|
|
nfs_end_page_writeback(req->wb_page);
|
2012-05-10 01:30:35 +07:00
|
|
|
nfs_release_request(req);
|
2012-04-21 01:47:47 +07:00
|
|
|
}
|
|
|
|
out:
|
|
|
|
hdr->release(hdr);
|
2007-04-21 03:12:34 +07:00
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-07-31 03:05:25 +07:00
|
|
|
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
|
2012-04-21 01:47:53 +07:00
|
|
|
static unsigned long
|
|
|
|
nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
|
2009-03-12 01:10:29 +07:00
|
|
|
{
|
2012-04-21 01:47:53 +07:00
|
|
|
return cinfo->mds->ncommit;
|
2012-03-09 05:29:35 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
/* cinfo->lock held by caller */
|
2012-04-21 01:47:57 +07:00
|
|
|
int
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
|
|
|
|
struct nfs_commit_info *cinfo, int max)
|
2012-03-09 05:29:35 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req, *tmp;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(req, tmp, src, wb_list) {
|
2012-03-16 04:16:40 +07:00
|
|
|
if (!nfs_lock_request(req))
|
|
|
|
continue;
|
2012-05-10 00:19:15 +07:00
|
|
|
kref_get(&req->wb_kref);
|
2012-04-21 01:47:53 +07:00
|
|
|
if (cond_resched_lock(cinfo->lock))
|
2012-03-17 22:59:30 +07:00
|
|
|
list_safe_reset_next(req, tmp, wb_list);
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_request_remove_commit_list(req, cinfo);
|
2012-03-16 04:16:40 +07:00
|
|
|
nfs_list_add_request(req, dst);
|
|
|
|
ret++;
|
2012-04-21 01:47:57 +07:00
|
|
|
if ((ret == max) && !cinfo->dreq)
|
2012-03-16 04:16:40 +07:00
|
|
|
break;
|
2012-03-09 05:29:35 +07:00
|
|
|
}
|
|
|
|
return ret;
|
2009-03-12 01:10:29 +07:00
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* nfs_scan_commit - Scan an inode for commit requests
|
|
|
|
* @inode: NFS inode to scan
|
2012-04-21 01:47:53 +07:00
|
|
|
* @dst: mds destination list
|
|
|
|
* @cinfo: mds and ds lists of reqs ready to commit
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
|
|
|
* Moves requests from the inode's 'commit' request list.
|
|
|
|
* The requests are *not* checked to ensure that they form a contiguous set.
|
|
|
|
*/
|
2012-04-21 01:47:57 +07:00
|
|
|
int
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_scan_commit(struct inode *inode, struct list_head *dst,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-03-09 05:29:35 +07:00
|
|
|
int ret = 0;
|
2009-03-12 01:10:29 +07:00
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
spin_lock(cinfo->lock);
|
|
|
|
if (cinfo->mds->ncommit > 0) {
|
2012-03-16 04:16:40 +07:00
|
|
|
const int max = INT_MAX;
|
2012-03-09 05:29:35 +07:00
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
ret = nfs_scan_commit_list(&cinfo->mds->list, dst,
|
|
|
|
cinfo, max);
|
|
|
|
ret += pnfs_scan_commit_lists(inode, cinfo, max - ret);
|
2012-03-09 05:29:35 +07:00
|
|
|
}
|
2012-04-21 01:47:53 +07:00
|
|
|
spin_unlock(cinfo->lock);
|
2010-02-20 07:53:39 +07:00
|
|
|
return ret;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-03-09 05:29:35 +07:00
|
|
|
|
2006-03-21 01:44:51 +07:00
|
|
|
#else
|
2012-04-21 01:47:53 +07:00
|
|
|
static unsigned long nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
|
2009-03-12 01:10:29 +07:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:57 +07:00
|
|
|
int nfs_scan_commit(struct inode *inode, struct list_head *dst,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2006-03-21 01:44:51 +07:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
2008-06-13 23:12:32 +07:00
|
|
|
* Search for an existing write request, and attempt to update
|
|
|
|
* it to reflect a new dirty region on a given page.
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
2008-06-13 23:12:32 +07:00
|
|
|
* If the attempt fails, then the existing request is flushed out
|
|
|
|
* to disk.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2008-06-13 23:12:32 +07:00
|
|
|
static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
|
|
|
|
struct page *page,
|
|
|
|
unsigned int offset,
|
|
|
|
unsigned int bytes)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2008-06-13 23:12:32 +07:00
|
|
|
struct nfs_page *req;
|
|
|
|
unsigned int rqend;
|
|
|
|
unsigned int end;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (!PagePrivate(page))
|
|
|
|
return NULL;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
end = offset + bytes;
|
2008-06-13 23:12:32 +07:00
|
|
|
spin_lock(&inode->i_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
for (;;) {
|
2012-08-01 06:45:10 +07:00
|
|
|
req = nfs_page_find_request_locked(NFS_I(inode), page);
|
2008-06-13 23:12:32 +07:00
|
|
|
if (req == NULL)
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
rqend = req->wb_offset + req->wb_bytes;
|
|
|
|
/*
|
|
|
|
* Tell the caller to flush out the request if
|
|
|
|
* the offsets are non-contiguous.
|
|
|
|
* Note: nfs_flush_incompatible() will already
|
|
|
|
* have flushed out requests having wrong owners.
|
|
|
|
*/
|
2008-06-14 00:25:22 +07:00
|
|
|
if (offset > rqend
|
2008-06-13 23:12:32 +07:00
|
|
|
|| end < req->wb_offset)
|
|
|
|
goto out_flushme;
|
|
|
|
|
2012-05-10 00:19:15 +07:00
|
|
|
if (nfs_lock_request(req))
|
2005-04-17 05:20:36 +07:00
|
|
|
break;
|
|
|
|
|
2008-06-13 23:12:32 +07:00
|
|
|
/* The request is locked, so wait and then retry */
|
2007-07-02 20:57:54 +07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2008-06-13 23:12:32 +07:00
|
|
|
error = nfs_wait_on_request(req);
|
|
|
|
nfs_release_request(req);
|
|
|
|
if (error != 0)
|
|
|
|
goto out_err;
|
|
|
|
spin_lock(&inode->i_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Okay, the request matches. Update the region */
|
|
|
|
if (offset < req->wb_offset) {
|
|
|
|
req->wb_offset = offset;
|
|
|
|
req->wb_pgbase = offset;
|
|
|
|
}
|
|
|
|
if (end > rqend)
|
|
|
|
req->wb_bytes = end - req->wb_offset;
|
2008-06-13 23:12:32 +07:00
|
|
|
else
|
|
|
|
req->wb_bytes = rqend - req->wb_offset;
|
|
|
|
out_unlock:
|
|
|
|
spin_unlock(&inode->i_lock);
|
2012-04-06 02:26:36 +07:00
|
|
|
if (req)
|
|
|
|
nfs_clear_request_commit(req);
|
2008-06-13 23:12:32 +07:00
|
|
|
return req;
|
|
|
|
out_flushme:
|
|
|
|
spin_unlock(&inode->i_lock);
|
|
|
|
nfs_release_request(req);
|
|
|
|
error = nfs_wb_page(inode, page);
|
|
|
|
out_err:
|
|
|
|
return ERR_PTR(error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to update an existing write request, or create one if there is none.
|
|
|
|
*
|
|
|
|
* Note: Should always be called with the Page Lock held to prevent races
|
|
|
|
* if we have to add a new request. Also assumes that the caller has
|
|
|
|
* already called nfs_flush_incompatible() if necessary.
|
|
|
|
*/
|
|
|
|
static struct nfs_page * nfs_setup_write_request(struct nfs_open_context* ctx,
|
|
|
|
struct page *page, unsigned int offset, unsigned int bytes)
|
|
|
|
{
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2008-06-13 23:12:32 +07:00
|
|
|
struct nfs_page *req;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-06-13 23:12:32 +07:00
|
|
|
req = nfs_try_to_update_request(inode, page, offset, bytes);
|
|
|
|
if (req != NULL)
|
|
|
|
goto out;
|
|
|
|
req = nfs_create_request(ctx, inode, page, offset, bytes);
|
|
|
|
if (IS_ERR(req))
|
|
|
|
goto out;
|
2012-03-09 05:29:35 +07:00
|
|
|
nfs_inode_add_request(inode, req);
|
2008-06-11 05:31:00 +07:00
|
|
|
out:
|
2007-10-19 04:08:05 +07:00
|
|
|
return req;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2008-06-13 23:12:32 +07:00
|
|
|
static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page,
|
|
|
|
unsigned int offset, unsigned int count)
|
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
|
|
|
|
req = nfs_setup_write_request(ctx, page, offset, count);
|
|
|
|
if (IS_ERR(req))
|
|
|
|
return PTR_ERR(req);
|
|
|
|
/* Update file length */
|
|
|
|
nfs_grow_file(page, offset, count);
|
|
|
|
nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
|
2010-04-10 06:07:08 +07:00
|
|
|
nfs_mark_request_dirty(req);
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_and_release_request(req);
|
2008-06-13 23:12:32 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
int nfs_flush_incompatible(struct file *file, struct page *page)
|
|
|
|
{
|
2007-08-11 04:44:32 +07:00
|
|
|
struct nfs_open_context *ctx = nfs_file_open_context(file);
|
2012-08-14 05:54:45 +07:00
|
|
|
struct nfs_lock_context *l_ctx;
|
2005-04-17 05:20:36 +07:00
|
|
|
struct nfs_page *req;
|
2006-12-05 12:35:40 +07:00
|
|
|
int do_flush, status;
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Look for a request corresponding to this page. If there
|
|
|
|
* is one, and it belongs to another file, we flush it out
|
|
|
|
* before we try to copy anything into the page. Do this
|
|
|
|
* due to the lack of an ACCESS-type call in NFSv2.
|
|
|
|
* Also do the same if we find a request from an existing
|
|
|
|
* dropped page.
|
|
|
|
*/
|
2006-12-05 12:35:40 +07:00
|
|
|
do {
|
|
|
|
req = nfs_page_find_request(page);
|
|
|
|
if (req == NULL)
|
|
|
|
return 0;
|
2012-08-14 05:54:45 +07:00
|
|
|
l_ctx = req->wb_lock_context;
|
|
|
|
do_flush = req->wb_page != page || req->wb_context != ctx;
|
|
|
|
if (l_ctx) {
|
|
|
|
do_flush |= l_ctx->lockowner.l_owner != current->files
|
|
|
|
|| l_ctx->lockowner.l_pid != current->tgid;
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
nfs_release_request(req);
|
2006-12-05 12:35:40 +07:00
|
|
|
if (!do_flush)
|
|
|
|
return 0;
|
2012-08-01 06:45:06 +07:00
|
|
|
status = nfs_wb_page(page_file_mapping(page)->host, page);
|
2006-12-05 12:35:40 +07:00
|
|
|
} while (status == 0);
|
|
|
|
return status;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
NFS: Fix a potential file corruption issue when writing
If the inode is flagged as having an invalid mapping, then we can't rely on
the PageUptodate() flag. Ensure that we don't use the "anti-fragmentation"
write optimisation in nfs_updatepage(), since that will cause NFS to write
out areas of the page that are no longer guaranteed to be up to date.
A potential corruption could occur in the following scenario:
client 1 client 2
=============== ===============
fd=open("f",O_CREAT|O_WRONLY,0644);
write(fd,"fubar\n",6); // cache last page
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"foo\n",4);
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"bar\n",4);
close(fd);
-----
The bug may lead to the file "f" reading 'fubar\n\0\0\0\nbar\n' because
client 2 does not update the cached page after re-opening the file for
write. Instead it keeps it marked as PageUptodate() until someone calls
invaldate_inode_pages2() (typically by calling read()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-02-08 05:24:07 +07:00
|
|
|
/*
|
|
|
|
* If the page cache is marked as unsafe or invalid, then we can't rely on
|
|
|
|
* the PageUptodate() flag. In this case, we will need to turn off
|
|
|
|
* write optimisations that depend on the page contents being correct.
|
|
|
|
*/
|
2012-04-29 23:50:01 +07:00
|
|
|
static bool nfs_write_pageuptodate(struct page *page, struct inode *inode)
|
NFS: Fix a potential file corruption issue when writing
If the inode is flagged as having an invalid mapping, then we can't rely on
the PageUptodate() flag. Ensure that we don't use the "anti-fragmentation"
write optimisation in nfs_updatepage(), since that will cause NFS to write
out areas of the page that are no longer guaranteed to be up to date.
A potential corruption could occur in the following scenario:
client 1 client 2
=============== ===============
fd=open("f",O_CREAT|O_WRONLY,0644);
write(fd,"fubar\n",6); // cache last page
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"foo\n",4);
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"bar\n",4);
close(fd);
-----
The bug may lead to the file "f" reading 'fubar\n\0\0\0\nbar\n' because
client 2 does not update the cached page after re-opening the file for
write. Instead it keeps it marked as PageUptodate() until someone calls
invaldate_inode_pages2() (typically by calling read()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-02-08 05:24:07 +07:00
|
|
|
{
|
2012-04-29 23:50:01 +07:00
|
|
|
if (nfs_have_delegated_attributes(inode))
|
|
|
|
goto out;
|
2012-12-10 21:25:48 +07:00
|
|
|
if (NFS_I(inode)->cache_validity & (NFS_INO_INVALID_DATA|NFS_INO_REVAL_PAGECACHE))
|
2012-04-29 23:50:01 +07:00
|
|
|
return false;
|
|
|
|
out:
|
|
|
|
return PageUptodate(page) != 0;
|
NFS: Fix a potential file corruption issue when writing
If the inode is flagged as having an invalid mapping, then we can't rely on
the PageUptodate() flag. Ensure that we don't use the "anti-fragmentation"
write optimisation in nfs_updatepage(), since that will cause NFS to write
out areas of the page that are no longer guaranteed to be up to date.
A potential corruption could occur in the following scenario:
client 1 client 2
=============== ===============
fd=open("f",O_CREAT|O_WRONLY,0644);
write(fd,"fubar\n",6); // cache last page
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"foo\n",4);
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"bar\n",4);
close(fd);
-----
The bug may lead to the file "f" reading 'fubar\n\0\0\0\nbar\n' because
client 2 does not update the cached page after re-opening the file for
write. Instead it keeps it marked as PageUptodate() until someone calls
invaldate_inode_pages2() (typically by calling read()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-02-08 05:24:07 +07:00
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Update and possibly write a cached page of an NFS file.
|
|
|
|
*
|
|
|
|
* XXX: Keep an eye on generic_file_read to make sure it doesn't do bad
|
|
|
|
* things with a page scheduled for an RPC call (e.g. invalidate it).
|
|
|
|
*/
|
|
|
|
int nfs_updatepage(struct file *file, struct page *page,
|
|
|
|
unsigned int offset, unsigned int count)
|
|
|
|
{
|
2007-08-11 04:44:32 +07:00
|
|
|
struct nfs_open_context *ctx = nfs_file_open_context(file);
|
2012-08-01 06:45:06 +07:00
|
|
|
struct inode *inode = page_file_mapping(page)->host;
|
2005-04-17 05:20:36 +07:00
|
|
|
int status = 0;
|
|
|
|
|
2006-03-21 01:44:14 +07:00
|
|
|
nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);
|
|
|
|
|
2008-06-12 04:56:05 +07:00
|
|
|
dprintk("NFS: nfs_updatepage(%s/%s %d@%lld)\n",
|
2006-12-08 17:36:40 +07:00
|
|
|
file->f_path.dentry->d_parent->d_name.name,
|
|
|
|
file->f_path.dentry->d_name.name, count,
|
2012-08-01 06:45:06 +07:00
|
|
|
(long long)(page_file_offset(page) + offset));
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* If we're not using byte range locks, and we know the page
|
NFS: Fix a potential file corruption issue when writing
If the inode is flagged as having an invalid mapping, then we can't rely on
the PageUptodate() flag. Ensure that we don't use the "anti-fragmentation"
write optimisation in nfs_updatepage(), since that will cause NFS to write
out areas of the page that are no longer guaranteed to be up to date.
A potential corruption could occur in the following scenario:
client 1 client 2
=============== ===============
fd=open("f",O_CREAT|O_WRONLY,0644);
write(fd,"fubar\n",6); // cache last page
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"foo\n",4);
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"bar\n",4);
close(fd);
-----
The bug may lead to the file "f" reading 'fubar\n\0\0\0\nbar\n' because
client 2 does not update the cached page after re-opening the file for
write. Instead it keeps it marked as PageUptodate() until someone calls
invaldate_inode_pages2() (typically by calling read()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-02-08 05:24:07 +07:00
|
|
|
* is up to date, it may be more efficient to extend the write
|
|
|
|
* to cover the entire page in order to avoid fragmentation
|
|
|
|
* inefficiencies.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
NFS: Fix a potential file corruption issue when writing
If the inode is flagged as having an invalid mapping, then we can't rely on
the PageUptodate() flag. Ensure that we don't use the "anti-fragmentation"
write optimisation in nfs_updatepage(), since that will cause NFS to write
out areas of the page that are no longer guaranteed to be up to date.
A potential corruption could occur in the following scenario:
client 1 client 2
=============== ===============
fd=open("f",O_CREAT|O_WRONLY,0644);
write(fd,"fubar\n",6); // cache last page
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"foo\n",4);
close(fd);
fd=open("f",O_WRONLY|O_APPEND);
write(fd,"bar\n",4);
close(fd);
-----
The bug may lead to the file "f" reading 'fubar\n\0\0\0\nbar\n' because
client 2 does not update the cached page after re-opening the file for
write. Instead it keeps it marked as PageUptodate() until someone calls
invaldate_inode_pages2() (typically by calling read()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-02-08 05:24:07 +07:00
|
|
|
if (nfs_write_pageuptodate(page, inode) &&
|
|
|
|
inode->i_flock == NULL &&
|
2009-10-27 17:05:28 +07:00
|
|
|
!(file->f_flags & O_DSYNC)) {
|
2006-12-05 12:35:38 +07:00
|
|
|
count = max(count + offset, nfs_page_length(page));
|
2005-04-17 05:20:36 +07:00
|
|
|
offset = 0;
|
|
|
|
}
|
|
|
|
|
2006-12-05 12:35:39 +07:00
|
|
|
status = nfs_writepage_setup(ctx, page, offset, count);
|
2008-06-06 03:02:35 +07:00
|
|
|
if (status < 0)
|
|
|
|
nfs_set_pageerror(page);
|
2011-10-18 08:22:55 +07:00
|
|
|
else
|
|
|
|
__set_page_dirty_nobuffers(page);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-06-12 04:56:05 +07:00
|
|
|
dprintk("NFS: nfs_updatepage returns %d (isize %lld)\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
status, (long long)i_size_read(inode));
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2007-07-15 02:40:00 +07:00
|
|
|
static int flush_task_priority(int how)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
|
|
|
|
case FLUSH_HIGHPRI:
|
|
|
|
return RPC_PRIORITY_HIGH;
|
|
|
|
case FLUSH_LOWPRI:
|
|
|
|
return RPC_PRIORITY_LOW;
|
|
|
|
}
|
|
|
|
return RPC_PRIORITY_NORMAL;
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:41 +07:00
|
|
|
int nfs_initiate_write(struct rpc_clnt *clnt,
|
|
|
|
struct nfs_write_data *data,
|
2011-03-03 22:13:41 +07:00
|
|
|
const struct rpc_call_ops *call_ops,
|
2012-04-28 04:53:44 +07:00
|
|
|
int how, int flags)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:44 +07:00
|
|
|
struct inode *inode = data->header->inode;
|
2007-07-15 02:40:00 +07:00
|
|
|
int priority = flush_task_priority(how);
|
2007-10-26 05:42:54 +07:00
|
|
|
struct rpc_task *task;
|
2007-07-15 02:40:00 +07:00
|
|
|
struct rpc_message msg = {
|
|
|
|
.rpc_argp = &data->args,
|
|
|
|
.rpc_resp = &data->res,
|
2012-04-21 01:47:44 +07:00
|
|
|
.rpc_cred = data->header->cred,
|
2007-07-15 02:40:00 +07:00
|
|
|
};
|
2007-07-15 02:39:59 +07:00
|
|
|
struct rpc_task_setup task_setup_data = {
|
2011-03-03 22:13:41 +07:00
|
|
|
.rpc_client = clnt,
|
2007-10-26 05:42:54 +07:00
|
|
|
.task = &data->task,
|
2007-07-15 02:40:00 +07:00
|
|
|
.rpc_message = &msg,
|
2007-07-15 02:39:59 +07:00
|
|
|
.callback_ops = call_ops,
|
|
|
|
.callback_data = data,
|
2008-02-20 08:04:23 +07:00
|
|
|
.workqueue = nfsiod_workqueue,
|
2012-04-28 04:53:44 +07:00
|
|
|
.flags = RPC_TASK_ASYNC | flags,
|
2007-07-15 02:40:00 +07:00
|
|
|
.priority = priority,
|
2007-07-15 02:39:59 +07:00
|
|
|
};
|
2010-04-10 06:54:50 +07:00
|
|
|
int ret = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-03-03 22:13:41 +07:00
|
|
|
/* Set up the initial task struct. */
|
|
|
|
NFS_PROTO(inode)->write_setup(data, &msg);
|
|
|
|
|
|
|
|
dprintk("NFS: %5u initiated write call "
|
|
|
|
"(req %s/%lld, %u bytes @ offset %llu)\n",
|
|
|
|
data->task.tk_pid,
|
|
|
|
inode->i_sb->s_id,
|
|
|
|
(long long)NFS_FILEID(inode),
|
|
|
|
data->args.count,
|
|
|
|
(unsigned long long)data->args.offset);
|
|
|
|
|
|
|
|
task = rpc_run_task(&task_setup_data);
|
|
|
|
if (IS_ERR(task)) {
|
|
|
|
ret = PTR_ERR(task);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (how & FLUSH_SYNC) {
|
|
|
|
ret = rpc_wait_for_completion_task(task);
|
|
|
|
if (ret == 0)
|
|
|
|
ret = task->tk_status;
|
|
|
|
}
|
|
|
|
rpc_put_task(task);
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
2011-03-03 22:13:47 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_initiate_write);
|
2011-03-03 22:13:41 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up the argument/result storage required for the RPC call.
|
|
|
|
*/
|
2012-04-21 01:47:47 +07:00
|
|
|
static void nfs_write_rpcsetup(struct nfs_write_data *data,
|
2011-03-03 22:13:41 +07:00
|
|
|
unsigned int count, unsigned int offset,
|
2012-04-21 01:47:53 +07:00
|
|
|
int how, struct nfs_commit_info *cinfo)
|
2011-03-03 22:13:41 +07:00
|
|
|
{
|
2012-04-21 01:47:47 +07:00
|
|
|
struct nfs_page *req = data->header->req;
|
2011-03-03 22:13:41 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* Set up the RPC argument and reply structs
|
|
|
|
* NB: take care not to mess about with data->commit et al. */
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
data->args.fh = NFS_FH(data->header->inode);
|
2005-04-17 05:20:36 +07:00
|
|
|
data->args.offset = req_offset(req) + offset;
|
2011-06-16 22:35:46 +07:00
|
|
|
/* pnfs_set_layoutcommit needs this */
|
|
|
|
data->mds_offset = data->args.offset;
|
2005-04-17 05:20:36 +07:00
|
|
|
data->args.pgbase = req->wb_pgbase + offset;
|
2012-04-21 01:47:45 +07:00
|
|
|
data->args.pages = data->pages.pagevec;
|
2005-04-17 05:20:36 +07:00
|
|
|
data->args.count = count;
|
2008-02-20 08:04:20 +07:00
|
|
|
data->args.context = get_nfs_open_context(req->wb_context);
|
2010-06-26 03:35:53 +07:00
|
|
|
data->args.lock_context = req->wb_lock_context;
|
2007-07-15 02:40:00 +07:00
|
|
|
data->args.stable = NFS_UNSTABLE;
|
2011-07-13 00:42:01 +07:00
|
|
|
switch (how & (FLUSH_STABLE | FLUSH_COND_STABLE)) {
|
|
|
|
case 0:
|
|
|
|
break;
|
|
|
|
case FLUSH_COND_STABLE:
|
2012-04-21 01:47:53 +07:00
|
|
|
if (nfs_reqs_to_commit(cinfo))
|
2011-07-13 00:42:01 +07:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
data->args.stable = NFS_FILE_SYNC;
|
2007-07-15 02:40:00 +07:00
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
data->res.fattr = &data->fattr;
|
|
|
|
data->res.count = count;
|
|
|
|
data->res.verf = &data->verf;
|
2005-10-28 09:12:38 +07:00
|
|
|
nfs_fattr_init(&data->fattr);
|
2011-07-13 00:42:02 +07:00
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-07-13 00:42:02 +07:00
|
|
|
static int nfs_do_write(struct nfs_write_data *data,
|
|
|
|
const struct rpc_call_ops *call_ops,
|
|
|
|
int how)
|
|
|
|
{
|
2012-04-21 01:47:44 +07:00
|
|
|
struct inode *inode = data->header->inode;
|
2011-03-03 22:13:45 +07:00
|
|
|
|
2012-04-28 04:53:44 +07:00
|
|
|
return nfs_initiate_write(NFS_CLIENT(inode), data, call_ops, how, 0);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2011-07-13 00:42:02 +07:00
|
|
|
static int nfs_do_multiple_writes(struct list_head *head,
|
|
|
|
const struct rpc_call_ops *call_ops,
|
|
|
|
int how)
|
|
|
|
{
|
|
|
|
struct nfs_write_data *data;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
while (!list_empty(head)) {
|
|
|
|
int ret2;
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
data = list_first_entry(head, struct nfs_write_data, list);
|
2011-07-13 00:42:02 +07:00
|
|
|
list_del_init(&data->list);
|
|
|
|
|
2011-07-14 02:59:19 +07:00
|
|
|
ret2 = nfs_do_write(data, call_ops, how);
|
2011-07-13 00:42:02 +07:00
|
|
|
if (ret == 0)
|
|
|
|
ret = ret2;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-03-19 22:24:38 +07:00
|
|
|
/* If a nfs_flush_* function fails, it should remove reqs from @head and
|
|
|
|
* call this on each, which will prepare them to be retried on next
|
|
|
|
* writeback using standard nfs.
|
|
|
|
*/
|
|
|
|
static void nfs_redirty_request(struct nfs_page *req)
|
|
|
|
{
|
|
|
|
nfs_mark_request_dirty(req);
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_request(req);
|
2012-05-10 00:37:43 +07:00
|
|
|
nfs_end_page_writeback(req->wb_page);
|
2012-05-10 01:30:35 +07:00
|
|
|
nfs_release_request(req);
|
2008-03-19 22:24:38 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:48 +07:00
|
|
|
static void nfs_async_write_error(struct list_head *head)
|
2012-04-21 01:47:47 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
|
|
|
|
while (!list_empty(head)) {
|
|
|
|
req = nfs_list_entry(head->next);
|
|
|
|
nfs_list_remove_request(req);
|
|
|
|
nfs_redirty_request(req);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:48 +07:00
|
|
|
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops = {
|
|
|
|
.error_cleanup = nfs_async_write_error,
|
|
|
|
.completion = nfs_write_completion,
|
|
|
|
};
|
|
|
|
|
2012-05-01 23:07:22 +07:00
|
|
|
static void nfs_flush_error(struct nfs_pageio_descriptor *desc,
|
|
|
|
struct nfs_pgio_header *hdr)
|
|
|
|
{
|
|
|
|
set_bit(NFS_IOHDR_REDO, &hdr->flags);
|
|
|
|
while (!list_empty(&hdr->rpc_list)) {
|
|
|
|
struct nfs_write_data *data = list_first_entry(&hdr->rpc_list,
|
|
|
|
struct nfs_write_data, list);
|
|
|
|
list_del(&data->list);
|
|
|
|
nfs_writedata_release(data);
|
|
|
|
}
|
|
|
|
desc->pg_completion_ops->error_cleanup(&desc->pg_list);
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Generate multiple small requests to write out a single
|
|
|
|
* contiguous dirty area on one page.
|
|
|
|
*/
|
2012-04-21 01:47:47 +07:00
|
|
|
static int nfs_flush_multi(struct nfs_pageio_descriptor *desc,
|
|
|
|
struct nfs_pgio_header *hdr)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:47 +07:00
|
|
|
struct nfs_page *req = hdr->req;
|
2005-04-17 05:20:36 +07:00
|
|
|
struct page *page = req->wb_page;
|
|
|
|
struct nfs_write_data *data;
|
2011-07-13 00:42:02 +07:00
|
|
|
size_t wsize = desc->pg_bsize, nbytes;
|
2006-09-08 23:48:54 +07:00
|
|
|
unsigned int offset;
|
2005-04-17 05:20:36 +07:00
|
|
|
int requests = 0;
|
2012-04-21 01:47:53 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-03-22 04:02:00 +07:00
|
|
|
if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
|
2012-04-21 01:47:53 +07:00
|
|
|
(desc->pg_moreio || nfs_reqs_to_commit(&cinfo) ||
|
2011-03-22 04:02:00 +07:00
|
|
|
desc->pg_count > wsize))
|
|
|
|
desc->pg_ioflags &= ~FLUSH_COND_STABLE;
|
|
|
|
|
|
|
|
|
2011-07-13 00:42:02 +07:00
|
|
|
offset = 0;
|
2011-03-03 22:13:48 +07:00
|
|
|
nbytes = desc->pg_count;
|
2006-09-08 23:48:54 +07:00
|
|
|
do {
|
|
|
|
size_t len = min(nbytes, wsize);
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
data = nfs_writedata_alloc(hdr, 1);
|
2012-05-01 23:07:22 +07:00
|
|
|
if (!data) {
|
|
|
|
nfs_flush_error(desc, hdr);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2012-04-21 01:47:45 +07:00
|
|
|
data->pages.pagevec[0] = page;
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_write_rpcsetup(data, len, offset, desc->pg_ioflags, &cinfo);
|
2012-04-21 01:47:47 +07:00
|
|
|
list_add(&data->list, &hdr->rpc_list);
|
2005-04-17 05:20:36 +07:00
|
|
|
requests++;
|
2006-09-08 23:48:54 +07:00
|
|
|
nbytes -= len;
|
2011-07-13 00:42:02 +07:00
|
|
|
offset += len;
|
2006-09-08 23:48:54 +07:00
|
|
|
} while (nbytes != 0);
|
2012-05-01 23:07:22 +07:00
|
|
|
nfs_list_remove_request(req);
|
|
|
|
nfs_list_add_request(req, &hdr->pages);
|
2012-04-21 01:47:47 +07:00
|
|
|
desc->pg_rpc_callops = &nfs_write_common_ops;
|
2012-05-01 23:07:22 +07:00
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
2012-04-21 01:47:47 +07:00
|
|
|
static int nfs_flush_one(struct nfs_pageio_descriptor *desc,
|
|
|
|
struct nfs_pgio_header *hdr)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
struct page **pages;
|
|
|
|
struct nfs_write_data *data;
|
2011-03-03 22:13:48 +07:00
|
|
|
struct list_head *head = &desc->pg_list;
|
2012-04-21 01:47:53 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
data = nfs_writedata_alloc(hdr, nfs_page_array_len(desc->pg_base,
|
|
|
|
desc->pg_count));
|
|
|
|
if (!data) {
|
2012-05-01 23:07:22 +07:00
|
|
|
nfs_flush_error(desc, hdr);
|
|
|
|
return -ENOMEM;
|
2011-03-03 22:13:44 +07:00
|
|
|
}
|
2012-04-21 01:47:47 +07:00
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
|
2012-04-21 01:47:45 +07:00
|
|
|
pages = data->pages.pagevec;
|
2005-04-17 05:20:36 +07:00
|
|
|
while (!list_empty(head)) {
|
|
|
|
req = nfs_list_entry(head->next);
|
|
|
|
nfs_list_remove_request(req);
|
2012-04-21 01:47:47 +07:00
|
|
|
nfs_list_add_request(req, &hdr->pages);
|
2005-04-17 05:20:36 +07:00
|
|
|
*pages++ = req->wb_page;
|
|
|
|
}
|
|
|
|
|
2011-03-22 04:02:00 +07:00
|
|
|
if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
|
2012-04-21 01:47:53 +07:00
|
|
|
(desc->pg_moreio || nfs_reqs_to_commit(&cinfo)))
|
2011-03-22 04:02:00 +07:00
|
|
|
desc->pg_ioflags &= ~FLUSH_COND_STABLE;
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* Set up the argument struct */
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_write_rpcsetup(data, desc->pg_count, 0, desc->pg_ioflags, &cinfo);
|
2012-04-21 01:47:47 +07:00
|
|
|
list_add(&data->list, &hdr->rpc_list);
|
|
|
|
desc->pg_rpc_callops = &nfs_write_common_ops;
|
2012-05-01 23:07:22 +07:00
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
int nfs_generic_flush(struct nfs_pageio_descriptor *desc,
|
|
|
|
struct nfs_pgio_header *hdr)
|
2011-07-14 02:59:19 +07:00
|
|
|
{
|
|
|
|
if (desc->pg_bsize < PAGE_CACHE_SIZE)
|
2012-04-21 01:47:47 +07:00
|
|
|
return nfs_flush_multi(desc, hdr);
|
|
|
|
return nfs_flush_one(desc, hdr);
|
2011-07-14 02:59:19 +07:00
|
|
|
}
|
2012-07-31 03:05:25 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_generic_flush);
|
2011-07-14 02:59:19 +07:00
|
|
|
|
|
|
|
static int nfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc)
|
2011-06-11 00:30:23 +07:00
|
|
|
{
|
2012-04-21 01:47:47 +07:00
|
|
|
struct nfs_write_header *whdr;
|
|
|
|
struct nfs_pgio_header *hdr;
|
2011-07-13 00:42:02 +07:00
|
|
|
int ret;
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
whdr = nfs_writehdr_alloc();
|
|
|
|
if (!whdr) {
|
2012-04-28 01:31:47 +07:00
|
|
|
desc->pg_completion_ops->error_cleanup(&desc->pg_list);
|
2012-04-21 01:47:47 +07:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
hdr = &whdr->header;
|
|
|
|
nfs_pgheader_init(desc, hdr, nfs_writehdr_free);
|
|
|
|
atomic_inc(&hdr->refcnt);
|
|
|
|
ret = nfs_generic_flush(desc, hdr);
|
2011-07-13 00:42:02 +07:00
|
|
|
if (ret == 0)
|
2012-04-21 01:47:47 +07:00
|
|
|
ret = nfs_do_multiple_writes(&hdr->rpc_list,
|
|
|
|
desc->pg_rpc_callops,
|
|
|
|
desc->pg_ioflags);
|
|
|
|
if (atomic_dec_and_test(&hdr->refcnt))
|
2012-04-21 01:47:48 +07:00
|
|
|
hdr->completion_ops->completion(hdr);
|
2011-07-13 00:42:02 +07:00
|
|
|
return ret;
|
2011-06-11 00:30:23 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static const struct nfs_pageio_ops nfs_pageio_write_ops = {
|
|
|
|
.pg_test = nfs_generic_pg_test,
|
|
|
|
.pg_doio = nfs_generic_pg_writepages,
|
|
|
|
};
|
|
|
|
|
2012-06-21 02:53:48 +07:00
|
|
|
void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
|
2012-04-21 01:47:48 +07:00
|
|
|
struct inode *inode, int ioflags,
|
|
|
|
const struct nfs_pgio_completion_ops *compl_ops)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:48 +07:00
|
|
|
nfs_pageio_init(pgio, inode, &nfs_pageio_write_ops, compl_ops,
|
2011-06-11 00:30:23 +07:00
|
|
|
NFS_SERVER(inode)->wsize, ioflags);
|
|
|
|
}
|
2012-07-31 03:05:23 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_pageio_init_write);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-07-14 02:59:19 +07:00
|
|
|
void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio)
|
|
|
|
{
|
|
|
|
pgio->pg_ops = &nfs_pageio_write_ops;
|
|
|
|
pgio->pg_bsize = NFS_SERVER(pgio->pg_inode)->wsize;
|
|
|
|
}
|
2011-07-14 02:59:57 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_pageio_reset_write_mds);
|
2011-07-14 02:59:19 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2009-04-01 20:22:26 +07:00
|
|
|
void nfs_write_prepare(struct rpc_task *task, void *calldata)
|
|
|
|
{
|
|
|
|
struct nfs_write_data *data = calldata;
|
2012-04-21 01:47:44 +07:00
|
|
|
NFS_PROTO(data->header->inode)->write_rpc_prepare(task, data);
|
2013-03-19 06:45:14 +07:00
|
|
|
if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
|
|
|
|
rpc_exit(task, -EIO);
|
2009-04-01 20:22:26 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
void nfs_commit_prepare(struct rpc_task *task, void *calldata)
|
|
|
|
{
|
|
|
|
struct nfs_commit_data *data = calldata;
|
|
|
|
|
|
|
|
NFS_PROTO(data->inode)->commit_rpc_prepare(task, data);
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
2012-04-21 01:47:47 +07:00
|
|
|
static void nfs_writeback_done_common(struct rpc_task *task, void *calldata)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2006-03-21 01:44:27 +07:00
|
|
|
struct nfs_write_data *data = calldata;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-04-16 03:56:39 +07:00
|
|
|
nfs_writeback_done(task, data);
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
static void nfs_writeback_release_common(void *calldata)
|
2008-04-16 03:56:39 +07:00
|
|
|
{
|
|
|
|
struct nfs_write_data *data = calldata;
|
2012-04-21 01:47:44 +07:00
|
|
|
struct nfs_pgio_header *hdr = data->header;
|
2012-01-06 20:57:46 +07:00
|
|
|
int status = data->task.tk_status;
|
2006-03-21 01:44:27 +07:00
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
if ((status >= 0) && nfs_write_need_commit(data)) {
|
|
|
|
spin_lock(&hdr->lock);
|
|
|
|
if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags))
|
|
|
|
; /* Do nothing */
|
|
|
|
else if (!test_and_set_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags))
|
2012-06-06 05:32:03 +07:00
|
|
|
memcpy(hdr->verf, &data->verf, sizeof(*hdr->verf));
|
|
|
|
else if (memcmp(hdr->verf, &data->verf, sizeof(*hdr->verf)))
|
2012-04-21 01:47:47 +07:00
|
|
|
set_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags);
|
|
|
|
spin_unlock(&hdr->lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-04-21 01:47:44 +07:00
|
|
|
nfs_writedata_release(data);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
static const struct rpc_call_ops nfs_write_common_ops = {
|
2009-04-01 20:22:26 +07:00
|
|
|
.rpc_call_prepare = nfs_write_prepare,
|
2012-04-21 01:47:47 +07:00
|
|
|
.rpc_call_done = nfs_writeback_done_common,
|
|
|
|
.rpc_release = nfs_writeback_release_common,
|
2006-03-21 01:44:27 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* This function is called when the WRITE call is complete.
|
|
|
|
*/
|
2011-02-11 22:42:38 +07:00
|
|
|
void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
struct nfs_writeargs *argp = &data->args;
|
|
|
|
struct nfs_writeres *resp = &data->res;
|
2012-04-21 01:47:44 +07:00
|
|
|
struct inode *inode = data->header->inode;
|
2006-03-21 01:44:27 +07:00
|
|
|
int status;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2007-02-01 00:14:01 +07:00
|
|
|
dprintk("NFS: %5u nfs_writeback_done (status %d)\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
task->tk_pid, task->tk_status);
|
|
|
|
|
2006-09-21 01:33:04 +07:00
|
|
|
/*
|
|
|
|
* ->write_done will attempt to use post-op attributes to detect
|
|
|
|
* conflicting writes by other clients. A strict interpretation
|
|
|
|
* of close-to-open would allow us to continue caching even if
|
|
|
|
* another writer had changed the file, but some applications
|
|
|
|
* depend on tighter cache coherency when writing.
|
|
|
|
*/
|
2012-04-21 01:47:44 +07:00
|
|
|
status = NFS_PROTO(inode)->write_done(task, data);
|
2006-03-21 01:44:27 +07:00
|
|
|
if (status != 0)
|
2011-02-11 22:42:38 +07:00
|
|
|
return;
|
2012-04-21 01:47:44 +07:00
|
|
|
nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, resp->count);
|
2006-03-21 01:44:14 +07:00
|
|
|
|
2012-07-31 03:05:25 +07:00
|
|
|
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
|
2005-04-17 05:20:36 +07:00
|
|
|
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;
|
|
|
|
|
2011-03-03 22:13:47 +07:00
|
|
|
/* Note this will print the MDS for a DS write */
|
2005-04-17 05:20:36 +07:00
|
|
|
if (time_before(complain, jiffies)) {
|
2008-06-12 04:56:05 +07:00
|
|
|
dprintk("NFS: faulty NFS server %s:"
|
2005-04-17 05:20:36 +07:00
|
|
|
" (committed = %d) != (stable = %d)\n",
|
2012-04-21 01:47:44 +07:00
|
|
|
NFS_SERVER(inode)->nfs_client->cl_hostname,
|
2005-04-17 05:20:36 +07:00
|
|
|
resp->verf->committed, argp->stable);
|
|
|
|
complain = jiffies + 300 * HZ;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
2012-04-21 01:47:47 +07:00
|
|
|
if (task->tk_status < 0)
|
|
|
|
nfs_set_pgio_error(data->header, task->tk_status, argp->offset);
|
|
|
|
else if (resp->count < argp->count) {
|
2005-04-17 05:20:36 +07:00
|
|
|
static unsigned long complain;
|
|
|
|
|
2012-04-21 01:47:47 +07:00
|
|
|
/* This a short write! */
|
2012-04-21 01:47:44 +07:00
|
|
|
nfs_inc_stats(inode, NFSIOS_SHORTWRITE);
|
2006-03-21 01:44:14 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* Has the server at least made some progress? */
|
2012-04-21 01:47:47 +07:00
|
|
|
if (resp->count == 0) {
|
|
|
|
if (time_before(complain, jiffies)) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"NFS: Server wrote zero bytes, expected %u.\n",
|
|
|
|
argp->count);
|
|
|
|
complain = jiffies + 300 * HZ;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-04-21 01:47:47 +07:00
|
|
|
nfs_set_pgio_error(data->header, -EIO, argp->offset);
|
|
|
|
task->tk_status = -EIO;
|
2011-02-11 22:42:38 +07:00
|
|
|
return;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-04-21 01:47:47 +07:00
|
|
|
/* Was this an NFSv2 write or an NFSv3 stable write? */
|
|
|
|
if (resp->verf->committed != NFS_UNSTABLE) {
|
|
|
|
/* Resend from where the server left off */
|
|
|
|
data->mds_offset += resp->count;
|
|
|
|
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;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-04-21 01:47:47 +07:00
|
|
|
rpc_restart_call_prepare(task);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2012-07-31 03:05:25 +07:00
|
|
|
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
|
2010-04-23 02:35:57 +07:00
|
|
|
static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
|
|
|
|
{
|
2011-03-22 02:37:01 +07:00
|
|
|
int ret;
|
|
|
|
|
2010-04-23 02:35:57 +07:00
|
|
|
if (!test_and_set_bit(NFS_INO_COMMIT, &nfsi->flags))
|
|
|
|
return 1;
|
2011-03-22 02:37:01 +07:00
|
|
|
if (!may_wait)
|
|
|
|
return 0;
|
|
|
|
ret = out_of_line_wait_on_bit_lock(&nfsi->flags,
|
|
|
|
NFS_INO_COMMIT,
|
|
|
|
nfs_wait_bit_killable,
|
|
|
|
TASK_KILLABLE);
|
|
|
|
return (ret < 0) ? ret : 1;
|
2010-04-23 02:35:57 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:54 +07:00
|
|
|
static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
|
2010-04-23 02:35:57 +07:00
|
|
|
{
|
|
|
|
clear_bit(NFS_INO_COMMIT, &nfsi->flags);
|
|
|
|
smp_mb__after_clear_bit();
|
|
|
|
wake_up_bit(&nfsi->flags, NFS_INO_COMMIT);
|
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
void nfs_commitdata_release(struct nfs_commit_data *data)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:39 +07:00
|
|
|
put_nfs_open_context(data->context);
|
|
|
|
nfs_commit_free(data);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_commitdata_release);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
|
2011-03-23 20:27:45 +07:00
|
|
|
const struct rpc_call_ops *call_ops,
|
2012-04-28 04:53:44 +07:00
|
|
|
int how, int flags)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2007-10-26 05:42:54 +07:00
|
|
|
struct rpc_task *task;
|
2011-03-23 20:27:45 +07:00
|
|
|
int priority = flush_task_priority(how);
|
2007-07-15 02:40:00 +07:00
|
|
|
struct rpc_message msg = {
|
|
|
|
.rpc_argp = &data->args,
|
|
|
|
.rpc_resp = &data->res,
|
2011-03-23 20:27:45 +07:00
|
|
|
.rpc_cred = data->cred,
|
2007-07-15 02:40:00 +07:00
|
|
|
};
|
2007-07-15 02:39:59 +07:00
|
|
|
struct rpc_task_setup task_setup_data = {
|
2007-10-26 05:42:54 +07:00
|
|
|
.task = &data->task,
|
2011-03-23 20:27:45 +07:00
|
|
|
.rpc_client = clnt,
|
2007-07-15 02:40:00 +07:00
|
|
|
.rpc_message = &msg,
|
2011-03-23 20:27:45 +07:00
|
|
|
.callback_ops = call_ops,
|
2007-07-15 02:39:59 +07:00
|
|
|
.callback_data = data,
|
2008-02-20 08:04:23 +07:00
|
|
|
.workqueue = nfsiod_workqueue,
|
2012-04-28 04:53:44 +07:00
|
|
|
.flags = RPC_TASK_ASYNC | flags,
|
2007-07-15 02:40:00 +07:00
|
|
|
.priority = priority,
|
2007-07-15 02:39:59 +07:00
|
|
|
};
|
2011-03-23 20:27:45 +07:00
|
|
|
/* Set up the initial task struct. */
|
|
|
|
NFS_PROTO(data->inode)->commit_setup(data, &msg);
|
|
|
|
|
|
|
|
dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
|
|
|
|
|
|
|
|
task = rpc_run_task(&task_setup_data);
|
|
|
|
if (IS_ERR(task))
|
|
|
|
return PTR_ERR(task);
|
|
|
|
if (how & FLUSH_SYNC)
|
|
|
|
rpc_wait_for_completion_task(task);
|
|
|
|
rpc_put_task(task);
|
|
|
|
return 0;
|
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_initiate_commit);
|
2011-03-23 20:27:45 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up the argument/result storage required for the RPC call.
|
|
|
|
*/
|
2012-04-21 01:47:39 +07:00
|
|
|
void nfs_init_commit(struct nfs_commit_data *data,
|
2012-04-21 01:47:54 +07:00
|
|
|
struct list_head *head,
|
|
|
|
struct pnfs_layout_segment *lseg,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2011-03-23 20:27:45 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *first = nfs_list_entry(head->next);
|
2011-06-23 05:40:12 +07:00
|
|
|
struct inode *inode = first->wb_context->dentry->d_inode;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* 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);
|
|
|
|
|
|
|
|
data->inode = inode;
|
2011-03-23 20:27:45 +07:00
|
|
|
data->cred = first->wb_context->cred;
|
2011-03-23 20:27:52 +07:00
|
|
|
data->lseg = lseg; /* reference transferred */
|
2011-03-23 20:27:45 +07:00
|
|
|
data->mds_ops = &nfs_commit_ops;
|
2012-04-21 01:47:54 +07:00
|
|
|
data->completion_ops = cinfo->completion_ops;
|
2012-04-21 01:47:55 +07:00
|
|
|
data->dreq = cinfo->dreq;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
data->args.fh = NFS_FH(data->inode);
|
2005-06-23 00:16:31 +07:00
|
|
|
/* Note: we always request a commit of the entire inode */
|
|
|
|
data->args.offset = 0;
|
|
|
|
data->args.count = 0;
|
2012-04-21 01:47:39 +07:00
|
|
|
data->context = get_nfs_open_context(first->wb_context);
|
2005-04-17 05:20:36 +07:00
|
|
|
data->res.fattr = &data->fattr;
|
|
|
|
data->res.verf = &data->verf;
|
2005-10-28 09:12:38 +07:00
|
|
|
nfs_fattr_init(&data->fattr);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_init_commit);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-03-23 20:27:53 +07:00
|
|
|
void nfs_retry_commit(struct list_head *page_list,
|
2012-04-21 01:47:53 +07:00
|
|
|
struct pnfs_layout_segment *lseg,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2011-03-23 20:27:47 +07:00
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
|
|
|
|
while (!list_empty(page_list)) {
|
|
|
|
req = nfs_list_entry(page_list->next);
|
|
|
|
nfs_list_remove_request(req);
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_mark_request_commit(req, lseg, cinfo);
|
2012-04-21 01:47:56 +07:00
|
|
|
if (!cinfo->dreq) {
|
|
|
|
dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
|
2012-08-01 06:45:06 +07:00
|
|
|
dec_bdi_stat(page_file_mapping(req->wb_page)->backing_dev_info,
|
2012-04-21 01:47:56 +07:00
|
|
|
BDI_RECLAIMABLE);
|
|
|
|
}
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_and_release_request(req);
|
2011-03-23 20:27:47 +07:00
|
|
|
}
|
|
|
|
}
|
2011-03-23 20:27:53 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_retry_commit);
|
2011-03-23 20:27:47 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Commit dirty pages
|
|
|
|
*/
|
|
|
|
static int
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_commit_list(struct inode *inode, struct list_head *head, int how,
|
|
|
|
struct nfs_commit_info *cinfo)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:39 +07:00
|
|
|
struct nfs_commit_data *data;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-04-16 03:56:39 +07:00
|
|
|
data = nfs_commitdata_alloc();
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (!data)
|
|
|
|
goto out_bad;
|
|
|
|
|
|
|
|
/* Set up the argument struct */
|
2012-04-21 01:47:54 +07:00
|
|
|
nfs_init_commit(data, head, NULL, cinfo);
|
|
|
|
atomic_inc(&cinfo->mds->rpcs_out);
|
2012-04-28 04:53:44 +07:00
|
|
|
return nfs_initiate_commit(NFS_CLIENT(inode), data, data->mds_ops,
|
|
|
|
how, 0);
|
2005-04-17 05:20:36 +07:00
|
|
|
out_bad:
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_retry_commit(head, NULL, cinfo);
|
2012-04-21 01:47:54 +07:00
|
|
|
cinfo->completion_ops->error_cleanup(NFS_I(inode));
|
2005-04-17 05:20:36 +07:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* COMMIT call returned
|
|
|
|
*/
|
2006-03-21 01:44:27 +07:00
|
|
|
static void nfs_commit_done(struct rpc_task *task, void *calldata)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-04-21 01:47:39 +07:00
|
|
|
struct nfs_commit_data *data = calldata;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2007-02-01 00:14:01 +07:00
|
|
|
dprintk("NFS: %5u nfs_commit_done (status %d)\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
task->tk_pid, task->tk_status);
|
|
|
|
|
2006-03-21 01:44:27 +07:00
|
|
|
/* Call the NFS version-specific code */
|
2011-04-12 23:29:15 +07:00
|
|
|
NFS_PROTO(data->inode)->commit_done(task, data);
|
2008-04-16 03:56:39 +07:00
|
|
|
}
|
|
|
|
|
2012-04-21 01:47:54 +07:00
|
|
|
static void nfs_commit_release_pages(struct nfs_commit_data *data)
|
2008-04-16 03:56:39 +07:00
|
|
|
{
|
2011-03-23 20:27:48 +07:00
|
|
|
struct nfs_page *req;
|
2008-04-16 03:56:39 +07:00
|
|
|
int status = data->task.tk_status;
|
2012-04-21 01:47:54 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2006-03-21 01:44:27 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
while (!list_empty(&data->pages)) {
|
|
|
|
req = nfs_list_entry(data->pages.next);
|
|
|
|
nfs_list_remove_request(req);
|
2012-03-09 05:29:35 +07:00
|
|
|
nfs_clear_page_commit(req->wb_page);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-06-12 04:56:05 +07:00
|
|
|
dprintk("NFS: commit (%s/%lld %d@%lld)",
|
2011-06-23 05:40:12 +07:00
|
|
|
req->wb_context->dentry->d_sb->s_id,
|
|
|
|
(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
|
2005-04-17 05:20:36 +07:00
|
|
|
req->wb_bytes,
|
|
|
|
(long long)req_offset(req));
|
2008-04-16 03:56:39 +07:00
|
|
|
if (status < 0) {
|
|
|
|
nfs_context_set_write_error(req->wb_context, status);
|
2005-04-17 05:20:36 +07:00
|
|
|
nfs_inode_remove_request(req);
|
2008-04-16 03:56:39 +07:00
|
|
|
dprintk(", error = %d\n", status);
|
2005-04-17 05:20:36 +07:00
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Okay, COMMIT succeeded, apparently. Check the verifier
|
|
|
|
* returned by the server against all stored verfs. */
|
2012-06-08 22:56:09 +07:00
|
|
|
if (!memcmp(&req->wb_verf, &data->verf.verifier, sizeof(req->wb_verf))) {
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 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");
|
2008-03-19 22:24:38 +07:00
|
|
|
nfs_mark_request_dirty(req);
|
2012-09-12 03:01:22 +07:00
|
|
|
set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
|
2005-04-17 05:20:36 +07:00
|
|
|
next:
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_and_release_request(req);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2012-04-21 01:47:54 +07:00
|
|
|
nfs_init_cinfo(&cinfo, data->inode, data->dreq);
|
|
|
|
if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
|
|
|
|
nfs_commit_clear_lock(NFS_I(data->inode));
|
2011-03-23 20:27:48 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static void nfs_commit_release(void *calldata)
|
|
|
|
{
|
2012-04-21 01:47:39 +07:00
|
|
|
struct nfs_commit_data *data = calldata;
|
2011-03-23 20:27:48 +07:00
|
|
|
|
2012-04-21 01:47:54 +07:00
|
|
|
data->completion_ops->completion(data);
|
2008-04-16 03:56:39 +07:00
|
|
|
nfs_commitdata_release(calldata);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2006-03-21 01:44:27 +07:00
|
|
|
|
|
|
|
static const struct rpc_call_ops nfs_commit_ops = {
|
2012-04-21 01:47:39 +07:00
|
|
|
.rpc_call_prepare = nfs_commit_prepare,
|
2006-03-21 01:44:27 +07:00
|
|
|
.rpc_call_done = nfs_commit_done,
|
|
|
|
.rpc_release = nfs_commit_release,
|
|
|
|
};
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:54 +07:00
|
|
|
static const struct nfs_commit_completion_ops nfs_commit_completion_ops = {
|
|
|
|
.completion = nfs_commit_release_pages,
|
|
|
|
.error_cleanup = nfs_commit_clear_lock,
|
|
|
|
};
|
|
|
|
|
2012-04-21 01:47:57 +07:00
|
|
|
int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
|
|
|
|
int how, struct nfs_commit_info *cinfo)
|
2012-04-21 01:47:52 +07:00
|
|
|
{
|
|
|
|
int status;
|
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
status = pnfs_commit_list(inode, head, how, cinfo);
|
2012-04-21 01:47:52 +07:00
|
|
|
if (status == PNFS_NOT_ATTEMPTED)
|
2012-04-21 01:47:53 +07:00
|
|
|
status = nfs_commit_list(inode, head, how, cinfo);
|
2012-04-21 01:47:52 +07:00
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2010-07-31 02:31:54 +07:00
|
|
|
int nfs_commit_inode(struct inode *inode, int how)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
LIST_HEAD(head);
|
2012-04-21 01:47:53 +07:00
|
|
|
struct nfs_commit_info cinfo;
|
2010-04-23 02:35:57 +07:00
|
|
|
int may_wait = how & FLUSH_SYNC;
|
2011-03-22 02:37:01 +07:00
|
|
|
int res;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-03-22 02:37:01 +07:00
|
|
|
res = nfs_commit_set_lock(NFS_I(inode), may_wait);
|
|
|
|
if (res <= 0)
|
2010-05-26 19:42:11 +07:00
|
|
|
goto out_mark_dirty;
|
2012-04-21 01:47:53 +07:00
|
|
|
nfs_init_cinfo_from_inode(&cinfo, inode);
|
|
|
|
res = nfs_scan_commit(inode, &head, &cinfo);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (res) {
|
2011-03-23 20:27:51 +07:00
|
|
|
int error;
|
|
|
|
|
2012-04-21 01:47:53 +07:00
|
|
|
error = nfs_generic_commit_list(inode, &head, how, &cinfo);
|
2005-06-23 00:16:31 +07:00
|
|
|
if (error < 0)
|
|
|
|
return error;
|
2011-03-22 02:37:01 +07:00
|
|
|
if (!may_wait)
|
2010-05-26 19:42:11 +07:00
|
|
|
goto out_mark_dirty;
|
2011-03-22 02:37:01 +07:00
|
|
|
error = wait_on_bit(&NFS_I(inode)->flags,
|
|
|
|
NFS_INO_COMMIT,
|
|
|
|
nfs_wait_bit_killable,
|
|
|
|
TASK_KILLABLE);
|
|
|
|
if (error < 0)
|
|
|
|
return error;
|
2010-04-23 02:35:57 +07:00
|
|
|
} else
|
|
|
|
nfs_commit_clear_lock(NFS_I(inode));
|
2010-05-26 19:42:11 +07:00
|
|
|
return res;
|
|
|
|
/* Note: If we exit without ensuring that the commit is complete,
|
|
|
|
* we must mark the inode as dirty. Otherwise, future calls to
|
|
|
|
* sync_inode() with the WB_SYNC_ALL flag set will fail to ensure
|
|
|
|
* that the data is on the disk.
|
|
|
|
*/
|
|
|
|
out_mark_dirty:
|
|
|
|
__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
|
2005-04-17 05:20:36 +07:00
|
|
|
return res;
|
|
|
|
}
|
2010-02-20 07:46:56 +07:00
|
|
|
|
|
|
|
static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc)
|
|
|
|
{
|
2010-02-20 08:00:02 +07:00
|
|
|
struct nfs_inode *nfsi = NFS_I(inode);
|
|
|
|
int flags = FLUSH_SYNC;
|
|
|
|
int ret = 0;
|
2010-02-20 07:46:56 +07:00
|
|
|
|
2011-10-11 20:49:21 +07:00
|
|
|
/* no commits means nothing needs to be done */
|
2012-04-21 01:47:53 +07:00
|
|
|
if (!nfsi->commit_info.ncommit)
|
2011-10-11 20:49:21 +07:00
|
|
|
return ret;
|
|
|
|
|
2010-09-28 20:14:01 +07:00
|
|
|
if (wbc->sync_mode == WB_SYNC_NONE) {
|
|
|
|
/* Don't commit yet if this is a non-blocking flush and there
|
|
|
|
* are a lot of outstanding writes for this mapping.
|
|
|
|
*/
|
2012-04-21 01:47:53 +07:00
|
|
|
if (nfsi->commit_info.ncommit <= (nfsi->npages >> 1))
|
2010-09-28 20:14:01 +07:00
|
|
|
goto out_mark_dirty;
|
2010-02-20 08:00:02 +07:00
|
|
|
|
2010-09-28 20:14:01 +07:00
|
|
|
/* don't wait for the COMMIT response */
|
2010-02-20 08:00:02 +07:00
|
|
|
flags = 0;
|
2010-09-28 20:14:01 +07:00
|
|
|
}
|
|
|
|
|
2010-02-20 08:00:02 +07:00
|
|
|
ret = nfs_commit_inode(inode, flags);
|
|
|
|
if (ret >= 0) {
|
|
|
|
if (wbc->sync_mode == WB_SYNC_NONE) {
|
|
|
|
if (ret < wbc->nr_to_write)
|
|
|
|
wbc->nr_to_write -= ret;
|
|
|
|
else
|
|
|
|
wbc->nr_to_write = 0;
|
|
|
|
}
|
2010-02-20 07:46:56 +07:00
|
|
|
return 0;
|
2010-02-20 08:00:02 +07:00
|
|
|
}
|
|
|
|
out_mark_dirty:
|
2010-02-20 07:46:56 +07:00
|
|
|
__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
|
|
|
|
return ret;
|
|
|
|
}
|
2007-04-03 06:29:52 +07:00
|
|
|
#else
|
2010-02-20 07:46:56 +07:00
|
|
|
static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif
|
|
|
|
|
2010-02-20 07:46:56 +07:00
|
|
|
int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
|
|
|
|
{
|
2012-06-21 02:53:49 +07:00
|
|
|
return nfs_commit_unstable_pages(inode, wbc);
|
|
|
|
}
|
2012-07-31 03:05:25 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_write_inode);
|
2012-06-21 02:53:49 +07:00
|
|
|
|
2010-02-20 08:03:26 +07:00
|
|
|
/*
|
|
|
|
* flush the inode to disk.
|
|
|
|
*/
|
|
|
|
int nfs_wb_all(struct inode *inode)
|
2007-07-26 01:09:54 +07:00
|
|
|
{
|
|
|
|
struct writeback_control wbc = {
|
2009-03-12 01:10:30 +07:00
|
|
|
.sync_mode = WB_SYNC_ALL,
|
2007-07-26 01:09:54 +07:00
|
|
|
.nr_to_write = LONG_MAX,
|
2008-10-07 07:08:56 +07:00
|
|
|
.range_start = 0,
|
|
|
|
.range_end = LLONG_MAX,
|
2007-07-26 01:09:54 +07:00
|
|
|
};
|
|
|
|
|
2010-02-20 08:03:26 +07:00
|
|
|
return sync_inode(inode, &wbc);
|
2006-10-10 03:18:38 +07:00
|
|
|
}
|
2012-07-31 03:05:23 +07:00
|
|
|
EXPORT_SYMBOL_GPL(nfs_wb_all);
|
2006-10-10 03:18:38 +07:00
|
|
|
|
2007-08-28 21:29:36 +07:00
|
|
|
int nfs_wb_page_cancel(struct inode *inode, struct page *page)
|
|
|
|
{
|
|
|
|
struct nfs_page *req;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
for (;;) {
|
2010-04-28 05:33:54 +07:00
|
|
|
wait_on_page_writeback(page);
|
2007-08-28 21:29:36 +07:00
|
|
|
req = nfs_page_find_request(page);
|
|
|
|
if (req == NULL)
|
|
|
|
break;
|
2012-05-10 00:19:15 +07:00
|
|
|
if (nfs_lock_request(req)) {
|
2012-03-16 04:16:40 +07:00
|
|
|
nfs_clear_request_commit(req);
|
2007-08-28 21:29:36 +07:00
|
|
|
nfs_inode_remove_request(req);
|
|
|
|
/*
|
|
|
|
* In case nfs_inode_remove_request has marked the
|
|
|
|
* page as being dirty
|
|
|
|
*/
|
|
|
|
cancel_dirty_page(page, PAGE_CACHE_SIZE);
|
2012-05-10 01:04:55 +07:00
|
|
|
nfs_unlock_and_release_request(req);
|
2007-08-28 21:29:36 +07:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
ret = nfs_wait_on_request(req);
|
2010-01-27 03:41:34 +07:00
|
|
|
nfs_release_request(req);
|
2007-08-28 21:29:36 +07:00
|
|
|
if (ret < 0)
|
2010-02-20 08:03:21 +07:00
|
|
|
break;
|
2007-08-28 21:29:36 +07:00
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-02-20 08:03:28 +07:00
|
|
|
/*
|
|
|
|
* Write back all requests on one page - we do this before reading it.
|
|
|
|
*/
|
|
|
|
int nfs_wb_page(struct inode *inode, struct page *page)
|
2006-10-10 03:18:38 +07:00
|
|
|
{
|
2012-08-01 06:45:10 +07:00
|
|
|
loff_t range_start = page_file_offset(page);
|
2006-10-10 03:18:38 +07:00
|
|
|
loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
|
2006-12-05 12:35:41 +07:00
|
|
|
struct writeback_control wbc = {
|
|
|
|
.sync_mode = WB_SYNC_ALL,
|
2010-02-20 08:03:28 +07:00
|
|
|
.nr_to_write = 0,
|
2006-12-05 12:35:41 +07:00
|
|
|
.range_start = range_start,
|
|
|
|
.range_end = range_end,
|
|
|
|
};
|
|
|
|
int ret;
|
2006-10-10 03:18:38 +07:00
|
|
|
|
2010-05-26 19:42:24 +07:00
|
|
|
for (;;) {
|
2010-04-28 05:33:54 +07:00
|
|
|
wait_on_page_writeback(page);
|
2008-04-12 03:03:54 +07:00
|
|
|
if (clear_page_dirty_for_io(page)) {
|
|
|
|
ret = nfs_writepage_locked(page, &wbc);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out_error;
|
2010-05-26 19:42:24 +07:00
|
|
|
continue;
|
2010-02-20 08:03:28 +07:00
|
|
|
}
|
2010-05-26 19:42:24 +07:00
|
|
|
if (!PagePrivate(page))
|
|
|
|
break;
|
|
|
|
ret = nfs_commit_inode(inode, FLUSH_SYNC);
|
2010-04-28 05:33:54 +07:00
|
|
|
if (ret < 0)
|
2008-04-12 03:03:54 +07:00
|
|
|
goto out_error;
|
2010-02-20 08:03:28 +07:00
|
|
|
}
|
2008-04-12 03:03:54 +07:00
|
|
|
return 0;
|
|
|
|
out_error:
|
2006-12-05 12:35:41 +07:00
|
|
|
return ret;
|
2006-10-10 03:18:38 +07:00
|
|
|
}
|
|
|
|
|
2009-08-10 19:54:13 +07:00
|
|
|
#ifdef CONFIG_MIGRATION
|
|
|
|
int nfs_migrate_page(struct address_space *mapping, struct page *newpage,
|
2012-01-13 08:19:43 +07:00
|
|
|
struct page *page, enum migrate_mode mode)
|
2009-08-10 19:54:13 +07:00
|
|
|
{
|
2011-10-12 21:57:42 +07:00
|
|
|
/*
|
|
|
|
* If PagePrivate is set, then the page is currently associated with
|
|
|
|
* an in-progress read or write request. Don't try to migrate it.
|
|
|
|
*
|
|
|
|
* FIXME: we could do this in principle, but we'll need a way to ensure
|
|
|
|
* that we can safely release the inode reference while holding
|
|
|
|
* the page lock.
|
|
|
|
*/
|
|
|
|
if (PagePrivate(page))
|
|
|
|
return -EBUSY;
|
2009-08-10 19:54:13 +07:00
|
|
|
|
2012-12-05 20:34:49 +07:00
|
|
|
if (!nfs_fscache_release_page(page, GFP_KERNEL))
|
|
|
|
return -EBUSY;
|
2009-08-10 19:54:13 +07:00
|
|
|
|
2012-01-13 08:19:43 +07:00
|
|
|
return migrate_page(mapping, newpage, page, mode);
|
2009-08-10 19:54:13 +07:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
NFS: Split fs/nfs/inode.c
As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached
patch splits it up into a number of files:
(*) fs/nfs/inode.c
Strictly inode specific functions.
(*) fs/nfs/super.c
Superblock management functions for NFS and NFS4, normal access, clones
and referrals. The NFS4 superblock functions _could_ move out into a
separate conditionally compiled file, but it's probably not worth it as
there're so many common bits.
(*) fs/nfs/namespace.c
Some namespace-specific functions have been moved here.
(*) fs/nfs/nfs4namespace.c
NFS4-specific namespace functions (this could be merged into the previous
file). This file is conditionally compiled.
(*) fs/nfs/internal.h
Inter-file declarations, plus a few simple utility functions moved from
fs/nfs/inode.c.
Additionally, all the in-.c-file externs have been moved here, and those
files they were moved from now includes this file.
For the most part, the functions have not been changed, only some multiplexor
functions have changed significantly.
I've also:
(*) Added some extra banner comments above some functions.
(*) Rearranged the function order within the files to be more logical and
better grouped (IMO), though someone may prefer a different order.
(*) Reduced the number of #ifdefs in .c files.
(*) Added missing __init and __exit directives.
Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 20:34:33 +07:00
|
|
|
int __init nfs_init_writepagecache(void)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
|
2012-04-21 01:47:44 +07:00
|
|
|
sizeof(struct nfs_write_header),
|
2005-04-17 05:20:36 +07:00
|
|
|
0, SLAB_HWCACHE_ALIGN,
|
2007-07-20 08:11:58 +07:00
|
|
|
NULL);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (nfs_wdata_cachep == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2006-03-26 16:37:50 +07:00
|
|
|
nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
|
|
|
|
nfs_wdata_cachep);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (nfs_wdata_mempool == NULL)
|
2012-08-03 01:30:56 +07:00
|
|
|
goto out_destroy_write_cache;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-21 01:47:39 +07:00
|
|
|
nfs_cdata_cachep = kmem_cache_create("nfs_commit_data",
|
|
|
|
sizeof(struct nfs_commit_data),
|
|
|
|
0, SLAB_HWCACHE_ALIGN,
|
|
|
|
NULL);
|
|
|
|
if (nfs_cdata_cachep == NULL)
|
2012-08-03 01:30:56 +07:00
|
|
|
goto out_destroy_write_mempool;
|
2012-04-21 01:47:39 +07:00
|
|
|
|
2006-03-26 16:37:50 +07:00
|
|
|
nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
|
2012-11-12 08:27:37 +07:00
|
|
|
nfs_cdata_cachep);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (nfs_commit_mempool == NULL)
|
2012-08-03 01:30:56 +07:00
|
|
|
goto out_destroy_commit_cache;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2007-03-17 04:38:26 +07:00
|
|
|
/*
|
|
|
|
* NFS congestion size, scale with available memory.
|
|
|
|
*
|
|
|
|
* 64MB: 8192k
|
|
|
|
* 128MB: 11585k
|
|
|
|
* 256MB: 16384k
|
|
|
|
* 512MB: 23170k
|
|
|
|
* 1GB: 32768k
|
|
|
|
* 2GB: 46340k
|
|
|
|
* 4GB: 65536k
|
|
|
|
* 8GB: 92681k
|
|
|
|
* 16GB: 131072k
|
|
|
|
*
|
|
|
|
* This allows larger machines to have larger/more transfers.
|
|
|
|
* Limit the default to 256M
|
|
|
|
*/
|
|
|
|
nfs_congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
|
|
|
|
if (nfs_congestion_kb > 256*1024)
|
|
|
|
nfs_congestion_kb = 256*1024;
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
return 0;
|
2012-08-03 01:30:56 +07:00
|
|
|
|
|
|
|
out_destroy_commit_cache:
|
|
|
|
kmem_cache_destroy(nfs_cdata_cachep);
|
|
|
|
out_destroy_write_mempool:
|
|
|
|
mempool_destroy(nfs_wdata_mempool);
|
|
|
|
out_destroy_write_cache:
|
|
|
|
kmem_cache_destroy(nfs_wdata_cachep);
|
|
|
|
return -ENOMEM;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-06-28 02:59:15 +07:00
|
|
|
void nfs_destroy_writepagecache(void)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
mempool_destroy(nfs_commit_mempool);
|
2012-08-03 01:30:56 +07:00
|
|
|
kmem_cache_destroy(nfs_cdata_cachep);
|
2005-04-17 05:20:36 +07:00
|
|
|
mempool_destroy(nfs_wdata_mempool);
|
2006-09-27 15:49:40 +07:00
|
|
|
kmem_cache_destroy(nfs_wdata_cachep);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|