/*
* linux/fs/nfs/write.c
*
- * Writing file data over NFS.
- *
- * We do it like this: When a (user) process wishes to write data to an
- * NFS file, a write request is allocated that contains the RPC task data
- * plus some info on the page to be written, and added to the inode's
- * write chain. If the process writes past the end of the page, an async
- * RPC call to write the page is scheduled immediately; otherwise, the call
- * is delayed for a few seconds.
- *
- * Just like readahead, no async I/O is performed if wsize < PAGE_SIZE.
- *
- * Write requests are kept on the inode's writeback list. Each entry in
- * that list references the page (portion) to be written. When the
- * cache timeout has expired, the RPC task is woken up, and tries to
- * lock the page. As soon as it manages to do so, the request is moved
- * from the writeback list to the writelock list.
- *
- * Note: we must make sure never to confuse the inode passed in the
- * write_page request with the one in page->inode. As far as I understand
- * it, these are different when doing a swap-out.
- *
- * To understand everything that goes on here and in the NFS read code,
- * one should be aware that a page is locked in exactly one of the following
- * cases:
- *
- * - A write request is in progress.
- * - A user process is in generic_file_write/nfs_update_page
- * - A user process is in generic_file_read
- *
- * Also note that because of the way pages are invalidated in
- * nfs_revalidate_inode, the following assertions hold:
- *
- * - If a page is dirty, there will be no read requests (a page will
- * not be re-read unless invalidated by nfs_revalidate_inode).
- * - If the page is not uptodate, there will be no pending write
- * requests, and no process will be in nfs_update_page.
- *
- * FIXME: Interaction with the vmscan routines is not optimal yet.
- * Either vmscan must be made nfs-savvy, or we need a different page
- * reclaim concept that supports something like FS-independent
- * buffer_heads with a b_ops-> field.
+ * Write file data over NFS.
*
* Copyright (C) 1996, 1997, Olaf Kirch <okir@monad.swb.de>
*/
#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/writeback.h>
+#include <linux/swap.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
struct page *,
unsigned int, unsigned int);
static void nfs_mark_request_dirty(struct nfs_page *req);
-static int nfs_wait_on_write_congestion(struct address_space *, int);
-static int nfs_wait_on_requests(struct inode *, unsigned long, unsigned int);
static long nfs_flush_mapping(struct address_space *mapping, struct writeback_control *wbc, int how);
static const struct rpc_call_ops nfs_write_partial_ops;
static const struct rpc_call_ops nfs_write_full_ops;
static const struct rpc_call_ops nfs_commit_ops;
-static kmem_cache_t *nfs_wdata_cachep;
+static struct kmem_cache *nfs_wdata_cachep;
static mempool_t *nfs_wdata_mempool;
static mempool_t *nfs_commit_mempool;
-static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion);
-
struct nfs_write_data *nfs_commit_alloc(void)
{
- struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS);
+ struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOFS);
if (p) {
memset(p, 0, sizeof(*p));
struct nfs_write_data *nfs_writedata_alloc(size_t len)
{
unsigned int pagecount = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
- struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);
+ struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, GFP_NOFS);
if (p) {
memset(p, 0, sizeof(*p));
i_size_write(inode, end);
}
+/* A writeback failed: mark the page as bad, and invalidate the page cache */
+static void nfs_set_pageerror(struct page *page)
+{
+ SetPageError(page);
+ nfs_zap_mapping(page->mapping->host, page->mapping);
+}
+
/* We can set the PG_uptodate flag if we see that a write request
* covers the full page.
*/
return 0;
}
+/*
+ * 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))
+
+static void nfs_set_page_writeback(struct page *page)
+{
+ if (!test_set_page_writeback(page)) {
+ struct inode *inode = page->mapping->host;
+ struct nfs_server *nfss = NFS_SERVER(inode);
+
+ if (atomic_inc_return(&nfss->writeback) >
+ NFS_CONGESTION_ON_THRESH)
+ set_bdi_congested(&nfss->backing_dev_info, WRITE);
+ }
+}
+
+static void nfs_end_page_writeback(struct page *page)
+{
+ struct inode *inode = page->mapping->host;
+ struct nfs_server *nfss = NFS_SERVER(inode);
+
+ end_page_writeback(page);
+ if (atomic_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH) {
+ clear_bdi_congested(&nfss->backing_dev_info, WRITE);
+ congestion_end(WRITE);
+ }
+}
+
/*
* Find an associated nfs write request, and prepare to flush it out
* Returns 1 if there was no write request, or if the request was
spin_unlock(req_lock);
if (test_and_set_bit(PG_FLUSHING, &req->wb_flags) == 0) {
nfs_mark_request_dirty(req);
- set_page_writeback(page);
+ nfs_set_page_writeback(page);
}
ret = test_bit(PG_NEED_FLUSH, &req->wb_flags);
nfs_unlock_request(req);
err = 0;
out:
if (!wbc->for_writepages)
- nfs_flush_mapping(page->mapping, wbc, wb_priority(wbc));
+ nfs_flush_mapping(page->mapping, wbc, FLUSH_STABLE|wb_priority(wbc));
return err;
}
return err;
}
-/*
- * Note: causes nfs_update_request() to block on the assumption
- * that the writeback is generated due to memory pressure.
- */
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
- struct backing_dev_info *bdi = mapping->backing_dev_info;
struct inode *inode = mapping->host;
int err;
err = generic_writepages(mapping, wbc);
if (err)
return err;
- while (test_and_set_bit(BDI_write_congested, &bdi->state) != 0) {
- if (wbc->nonblocking)
- return 0;
- nfs_wait_on_write_congestion(mapping, 0);
- }
err = nfs_flush_mapping(mapping, wbc, wb_priority(wbc));
if (err < 0)
goto out;
nfs_add_stats(inode, NFSIOS_WRITEPAGES, err);
- if (!wbc->nonblocking && wbc->sync_mode == WB_SYNC_ALL) {
- err = nfs_wait_on_requests(inode, 0, 0);
- if (err < 0)
- goto out;
- }
- err = nfs_commit_inode(inode, wb_priority(wbc));
- if (err > 0)
- err = 0;
+ err = 0;
out:
- clear_bit(BDI_write_congested, &bdi->state);
- wake_up_all(&nfs_write_congestion);
- congestion_end(WRITE);
return err;
}
}
/*
- * Insert a write request into an inode
+ * Remove a write request from an inode
*/
static void nfs_inode_remove_request(struct nfs_page *req)
{
nfs_list_add_request(req, &nfsi->dirty);
nfsi->ndirty++;
spin_unlock(&nfsi->req_lock);
- mark_inode_dirty(inode);
+ __mark_inode_dirty(inode, I_DIRTY_PAGES);
}
static void
nfsi->ncommit++;
spin_unlock(&nfsi->req_lock);
inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
- mark_inode_dirty(inode);
+ __mark_inode_dirty(inode, I_DIRTY_DATASYNC);
}
#endif
return res;
}
-static int nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int npages)
-{
- struct nfs_inode *nfsi = NFS_I(inode);
- int ret;
-
- spin_lock(&nfsi->req_lock);
- ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
- spin_unlock(&nfsi->req_lock);
- return ret;
-}
-
static void nfs_cancel_dirty_list(struct list_head *head)
{
struct nfs_page *req;
}
#endif
-static int nfs_wait_on_write_congestion(struct address_space *mapping, int intr)
+static int nfs_wait_on_write_congestion(struct address_space *mapping)
{
+ struct inode *inode = mapping->host;
struct backing_dev_info *bdi = mapping->backing_dev_info;
- DEFINE_WAIT(wait);
int ret = 0;
might_sleep();
if (!bdi_write_congested(bdi))
return 0;
- nfs_inc_stats(mapping->host, NFSIOS_CONGESTIONWAIT);
+ nfs_inc_stats(inode, NFSIOS_CONGESTIONWAIT);
- if (intr) {
- struct rpc_clnt *clnt = NFS_CLIENT(mapping->host);
+ do {
+ struct rpc_clnt *clnt = NFS_CLIENT(inode);
sigset_t oldset;
rpc_clnt_sigmask(clnt, &oldset);
- prepare_to_wait(&nfs_write_congestion, &wait, TASK_INTERRUPTIBLE);
- if (bdi_write_congested(bdi)) {
- if (signalled())
- ret = -ERESTARTSYS;
- else
- schedule();
- }
+ ret = congestion_wait_interruptible(WRITE, HZ/10);
rpc_clnt_sigunmask(clnt, &oldset);
- } else {
- prepare_to_wait(&nfs_write_congestion, &wait, TASK_UNINTERRUPTIBLE);
- if (bdi_write_congested(bdi))
- schedule();
- }
- finish_wait(&nfs_write_congestion, &wait);
+ if (ret == -ERESTARTSYS)
+ break;
+ ret = 0;
+ } while (bdi_write_congested(bdi));
+
return ret;
}
-
/*
* Try to update any existing write request, or create one if there is none.
* In order to match, the request's credentials must match those of
static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
struct page *page, unsigned int offset, unsigned int bytes)
{
- struct inode *inode = page->mapping->host;
+ struct address_space *mapping = page->mapping;
+ struct inode *inode = mapping->host;
struct nfs_inode *nfsi = NFS_I(inode);
struct nfs_page *req, *new = NULL;
unsigned long rqend, end;
end = offset + bytes;
- if (nfs_wait_on_write_congestion(page->mapping, NFS_SERVER(inode)->flags & NFS_MOUNT_INTR))
+ if (nfs_wait_on_write_congestion(mapping))
return ERR_PTR(-ERESTARTSYS);
for (;;) {
/* Loop over all inode entries and see if we find
nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);
dprintk("NFS: nfs_updatepage(%s/%s %d@%Ld)\n",
- file->f_dentry->d_parent->d_name.name,
- file->f_dentry->d_name.name, count,
+ file->f_path.dentry->d_parent->d_name.name,
+ file->f_path.dentry->d_name.name, count,
(long long)(page_offset(page) +offset));
/* If we're not using byte range locks, and we know the page
dprintk("NFS: nfs_updatepage returns %d (isize %Ld)\n",
status, (long long)i_size_read(inode));
if (status < 0)
- ClearPageUptodate(page);
+ nfs_set_pageerror(page);
return status;
}
static void nfs_writepage_release(struct nfs_page *req)
{
- end_page_writeback(req->wb_page);
+ nfs_end_page_writeback(req->wb_page);
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
if (!PageError(req->wb_page)) {
data->task.tk_priority = flush_task_priority(how);
data->task.tk_cookie = (unsigned long)inode;
- dprintk("NFS: %4d initiated write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
+ dprintk("NFS: %5u initiated write call "
+ "(req %s/%Ld, %u bytes @ offset %Lu)\n",
data->task.tk_pid,
inode->i_sb->s_id,
(long long)NFS_FILEID(inode),
return;
if (task->tk_status < 0) {
- ClearPageUptodate(page);
- SetPageError(page);
+ nfs_set_pageerror(page);
req->wb_context->error = task->tk_status;
dprintk(", error = %d\n", task->tk_status);
} else {
(long long)req_offset(req));
if (task->tk_status < 0) {
- ClearPageUptodate(page);
- SetPageError(page);
+ nfs_set_pageerror(page);
req->wb_context->error = task->tk_status;
- end_page_writeback(page);
+ nfs_end_page_writeback(page);
nfs_inode_remove_request(req);
dprintk(", error = %d\n", task->tk_status);
goto next;
}
- end_page_writeback(page);
+ nfs_end_page_writeback(page);
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
if (data->args.stable != NFS_UNSTABLE || data->verf.committed == NFS_FILE_SYNC) {
struct nfs_writeres *resp = &data->res;
int status;
- dprintk("NFS: %4d nfs_writeback_done (status %d)\n",
+ dprintk("NFS: %5u nfs_writeback_done (status %d)\n",
task->tk_pid, task->tk_status);
/*
data->task.tk_priority = flush_task_priority(how);
data->task.tk_cookie = (unsigned long)inode;
- dprintk("NFS: %4d initiated commit call\n", data->task.tk_pid);
+ dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
}
/*
struct nfs_write_data *data = calldata;
struct nfs_page *req;
- dprintk("NFS: %4d nfs_commit_done (status %d)\n",
+ dprintk("NFS: %5u nfs_commit_done (status %d)\n",
task->tk_pid, task->tk_status);
/* Call the NFS version-specific code */
if (ret < 0)
goto out;
}
+ if (!PagePrivate(page))
+ return 0;
ret = nfs_sync_mapping_wait(page->mapping, &wbc, how);
if (ret >= 0)
return 0;
if (nfs_commit_mempool == NULL)
return -ENOMEM;
+ /*
+ * 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;
+
return 0;
}
projects / powerpc.git / blobdiff