linux_dsm_epyc7002/fs/nfs/write.c

2119 lines
55 KiB
C
Raw Normal View History

/*
* linux/fs/nfs/write.c
*
* Write file data over NFS.
*
* 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>
#include <linux/swap.h>
#include <linux/migrate.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
#include <linux/backing-dev.h>
#include <linux/export.h>
#include <linux/freezer.h>
#include <linux/wait.h>
#include <asm/uaccess.h>
#include "delegation.h"
#include "internal.h"
#include "iostat.h"
#include "nfs4_fs.h"
#include "fscache.h"
#include "pnfs.h"
#include "nfstrace.h"
#define NFSDBG_FACILITY NFSDBG_PAGECACHE
#define MIN_POOL_WRITE (32)
#define MIN_POOL_COMMIT (4)
/*
* Local function declarations
*/
2008-03-19 22:24:39 +07:00
static void nfs_redirty_request(struct nfs_page *req);
static const struct rpc_call_ops nfs_commit_ops;
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops;
static const struct nfs_commit_completion_ops nfs_commit_completion_ops;
static const struct nfs_rw_ops nfs_rw_write_ops;
static void nfs_clear_request_commit(struct nfs_page *req);
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
struct inode *inode);
static struct nfs_page *
nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
struct page *page);
static struct kmem_cache *nfs_wdata_cachep;
static mempool_t *nfs_wdata_mempool;
static struct kmem_cache *nfs_cdata_cachep;
static mempool_t *nfs_commit_mempool;
struct nfs_commit_data *nfs_commitdata_alloc(void)
{
struct nfs_commit_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOIO);
if (p) {
memset(p, 0, sizeof(*p));
INIT_LIST_HEAD(&p->pages);
}
return p;
}
EXPORT_SYMBOL_GPL(nfs_commitdata_alloc);
void nfs_commit_free(struct nfs_commit_data *p)
{
mempool_free(p, nfs_commit_mempool);
}
EXPORT_SYMBOL_GPL(nfs_commit_free);
static struct nfs_pgio_header *nfs_writehdr_alloc(void)
{
struct nfs_pgio_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOIO);
if (p)
memset(p, 0, sizeof(*p));
return p;
}
static void nfs_writehdr_free(struct nfs_pgio_header *hdr)
{
mempool_free(hdr, nfs_wdata_mempool);
}
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);
}
/*
* nfs_page_find_head_request_locked - find head request associated with @page
*
* must be called while holding the inode lock.
*
* returns matching head request with reference held, or NULL if not found.
*/
static struct nfs_page *
nfs_page_find_head_request_locked(struct nfs_inode *nfsi, struct page *page)
{
struct nfs_page *req = NULL;
if (PagePrivate(page))
req = (struct nfs_page *)page_private(page);
else if (unlikely(PageSwapCache(page)))
req = nfs_page_search_commits_for_head_request_locked(nfsi,
page);
if (req) {
WARN_ON_ONCE(req->wb_head != req);
kref_get(&req->wb_kref);
}
return req;
}
/*
* nfs_page_find_head_request - find head request associated with @page
*
* returns matching head request with reference held, or NULL if not found.
*/
static struct nfs_page *nfs_page_find_head_request(struct page *page)
{
struct inode *inode = page_file_mapping(page)->host;
struct nfs_page *req = NULL;
spin_lock(&inode->i_lock);
req = nfs_page_find_head_request_locked(NFS_I(inode), page);
spin_unlock(&inode->i_lock);
return req;
}
/* Adjust the file length if we're writing beyond the end */
static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int count)
{
struct inode *inode = page_file_mapping(page)->host;
loff_t end, i_size;
pgoff_t end_index;
spin_lock(&inode->i_lock);
i_size = i_size_read(inode);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
end_index = (i_size - 1) >> PAGE_SHIFT;
if (i_size > 0 && page_index(page) < end_index)
goto out;
end = page_file_offset(page) + ((loff_t)offset+count);
if (i_size >= end)
goto out;
i_size_write(inode, end);
nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
out:
spin_unlock(&inode->i_lock);
}
/* A writeback failed: mark the page as bad, and invalidate the page cache */
static void nfs_set_pageerror(struct page *page)
{
nfs_zap_mapping(page_file_mapping(page)->host, page_file_mapping(page));
}
/*
* nfs_page_group_search_locked
* @head - head request of page group
* @page_offset - offset into page
*
* Search page group with head @head to find a request that contains the
* page offset @page_offset.
*
* Returns a pointer to the first matching nfs request, or NULL if no
* match is found.
*
* Must be called with the page group lock held
*/
static struct nfs_page *
nfs_page_group_search_locked(struct nfs_page *head, unsigned int page_offset)
{
struct nfs_page *req;
WARN_ON_ONCE(head != head->wb_head);
WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &head->wb_head->wb_flags));
req = head;
do {
if (page_offset >= req->wb_pgbase &&
page_offset < (req->wb_pgbase + req->wb_bytes))
return req;
req = req->wb_this_page;
} while (req != head);
return NULL;
}
/*
* nfs_page_group_covers_page
* @head - head request of page group
*
* Return true if the page group with head @head covers the whole page,
* returns false otherwise
*/
static bool nfs_page_group_covers_page(struct nfs_page *req)
{
struct nfs_page *tmp;
unsigned int pos = 0;
unsigned int len = nfs_page_length(req->wb_page);
nfs_page_group_lock(req, false);
do {
tmp = nfs_page_group_search_locked(req->wb_head, pos);
if (tmp) {
/* no way this should happen */
WARN_ON_ONCE(tmp->wb_pgbase != pos);
pos += tmp->wb_bytes - (pos - tmp->wb_pgbase);
}
} while (tmp && pos < len);
nfs_page_group_unlock(req);
WARN_ON_ONCE(pos > len);
return pos == len;
}
/* We can set the PG_uptodate flag if we see that a write request
* covers the full page.
*/
static void nfs_mark_uptodate(struct nfs_page *req)
{
if (PageUptodate(req->wb_page))
return;
if (!nfs_page_group_covers_page(req))
return;
SetPageUptodate(req->wb_page);
}
static int wb_priority(struct writeback_control *wbc)
{
int ret = 0;
if (wbc->sync_mode == WB_SYNC_ALL)
ret = FLUSH_COND_STABLE;
return ret;
}
/*
* 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)
{
struct nfs_server *nfss = NFS_SERVER(page_file_mapping(page)->host);
int ret = test_set_page_writeback(page);
WARN_ON_ONCE(ret != 0);
if (atomic_long_inc_return(&nfss->writeback) >
NFS_CONGESTION_ON_THRESH) {
set_bdi_congested(&nfss->backing_dev_info,
BLK_RW_ASYNC);
}
}
static void nfs_end_page_writeback(struct nfs_page *req)
{
struct inode *inode = page_file_mapping(req->wb_page)->host;
struct nfs_server *nfss = NFS_SERVER(inode);
if (!nfs_page_group_sync_on_bit(req, PG_WB_END))
return;
end_page_writeback(req->wb_page);
if (atomic_long_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
}
/* nfs_page_group_clear_bits
* @req - an nfs request
* clears all page group related bits from @req
*/
static void
nfs_page_group_clear_bits(struct nfs_page *req)
{
clear_bit(PG_TEARDOWN, &req->wb_flags);
clear_bit(PG_UNLOCKPAGE, &req->wb_flags);
clear_bit(PG_UPTODATE, &req->wb_flags);
clear_bit(PG_WB_END, &req->wb_flags);
clear_bit(PG_REMOVE, &req->wb_flags);
}
/*
* nfs_unroll_locks_and_wait - unlock all newly locked reqs and wait on @req
*
* this is a helper function for nfs_lock_and_join_requests
*
* @inode - inode associated with request page group, must be holding inode lock
* @head - head request of page group, must be holding head lock
* @req - request that couldn't lock and needs to wait on the req bit lock
* @nonblock - if true, don't actually wait
*
* NOTE: this must be called holding page_group bit lock and inode spin lock
* and BOTH will be released before returning.
*
* returns 0 on success, < 0 on error.
*/
static int
nfs_unroll_locks_and_wait(struct inode *inode, struct nfs_page *head,
struct nfs_page *req, bool nonblock)
__releases(&inode->i_lock)
{
struct nfs_page *tmp;
int ret;
/* relinquish all the locks successfully grabbed this run */
for (tmp = head ; tmp != req; tmp = tmp->wb_this_page)
nfs_unlock_request(tmp);
WARN_ON_ONCE(test_bit(PG_TEARDOWN, &req->wb_flags));
/* grab a ref on the request that will be waited on */
kref_get(&req->wb_kref);
nfs_page_group_unlock(head);
spin_unlock(&inode->i_lock);
/* release ref from nfs_page_find_head_request_locked */
nfs_release_request(head);
if (!nonblock)
ret = nfs_wait_on_request(req);
else
ret = -EAGAIN;
nfs_release_request(req);
return ret;
}
/*
* nfs_destroy_unlinked_subrequests - destroy recently unlinked subrequests
*
* @destroy_list - request list (using wb_this_page) terminated by @old_head
* @old_head - the old head of the list
*
* All subrequests must be locked and removed from all lists, so at this point
* they are only "active" in this function, and possibly in nfs_wait_on_request
* with a reference held by some other context.
*/
static void
nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
struct nfs_page *old_head)
{
while (destroy_list) {
struct nfs_page *subreq = destroy_list;
destroy_list = (subreq->wb_this_page == old_head) ?
NULL : subreq->wb_this_page;
WARN_ON_ONCE(old_head != subreq->wb_head);
/* make sure old group is not used */
subreq->wb_head = subreq;
subreq->wb_this_page = subreq;
/* subreq is now totally disconnected from page group or any
* write / commit lists. last chance to wake any waiters */
nfs_unlock_request(subreq);
if (!test_bit(PG_TEARDOWN, &subreq->wb_flags)) {
/* release ref on old head request */
nfs_release_request(old_head);
nfs_page_group_clear_bits(subreq);
/* release the PG_INODE_REF reference */
if (test_and_clear_bit(PG_INODE_REF, &subreq->wb_flags))
nfs_release_request(subreq);
else
WARN_ON_ONCE(1);
} else {
WARN_ON_ONCE(test_bit(PG_CLEAN, &subreq->wb_flags));
/* zombie requests have already released the last
* reference and were waiting on the rest of the
* group to complete. Since it's no longer part of a
* group, simply free the request */
nfs_page_group_clear_bits(subreq);
nfs_free_request(subreq);
}
}
}
/*
* nfs_lock_and_join_requests - join all subreqs to the head req and return
* a locked reference, cancelling any pending
* operations for this page.
*
* @page - the page used to lookup the "page group" of nfs_page structures
* @nonblock - if true, don't block waiting for request locks
*
* This function joins all sub requests to the head request by first
* locking all requests in the group, cancelling any pending operations
* and finally updating the head request to cover the whole range covered by
* the (former) group. All subrequests are removed from any write or commit
* lists, unlinked from the group and destroyed.
*
* Returns a locked, referenced pointer to the head request - which after
* this call is guaranteed to be the only request associated with the page.
* Returns NULL if no requests are found for @page, or a ERR_PTR if an
* error was encountered.
*/
static struct nfs_page *
nfs_lock_and_join_requests(struct page *page, bool nonblock)
{
struct inode *inode = page_file_mapping(page)->host;
struct nfs_page *head, *subreq;
struct nfs_page *destroy_list = NULL;
unsigned int total_bytes;
int ret;
try_again:
total_bytes = 0;
WARN_ON_ONCE(destroy_list);
spin_lock(&inode->i_lock);
/*
* A reference is taken only on the head request which acts as a
* reference to the whole page group - the group will not be destroyed
* until the head reference is released.
*/
head = nfs_page_find_head_request_locked(NFS_I(inode), page);
if (!head) {
spin_unlock(&inode->i_lock);
return NULL;
}
/* holding inode lock, so always make a non-blocking call to try the
* page group lock */
ret = nfs_page_group_lock(head, true);
if (ret < 0) {
spin_unlock(&inode->i_lock);
if (!nonblock && ret == -EAGAIN) {
nfs_page_group_lock_wait(head);
nfs_release_request(head);
goto try_again;
}
nfs_release_request(head);
return ERR_PTR(ret);
}
/* lock each request in the page group */
subreq = head;
do {
/*
* Subrequests are always contiguous, non overlapping
* and in order - but may be repeated (mirrored writes).
*/
if (subreq->wb_offset == (head->wb_offset + total_bytes)) {
/* keep track of how many bytes this group covers */
total_bytes += subreq->wb_bytes;
} else if (WARN_ON_ONCE(subreq->wb_offset < head->wb_offset ||
((subreq->wb_offset + subreq->wb_bytes) >
(head->wb_offset + total_bytes)))) {
nfs_page_group_unlock(head);
spin_unlock(&inode->i_lock);
return ERR_PTR(-EIO);
}
if (!nfs_lock_request(subreq)) {
/* releases page group bit lock and
* inode spin lock and all references */
ret = nfs_unroll_locks_and_wait(inode, head,
subreq, nonblock);
if (ret == 0)
goto try_again;
return ERR_PTR(ret);
}
subreq = subreq->wb_this_page;
} while (subreq != head);
/* Now that all requests are locked, make sure they aren't on any list.
* Commit list removal accounting is done after locks are dropped */
subreq = head;
do {
nfs_clear_request_commit(subreq);
subreq = subreq->wb_this_page;
} while (subreq != head);
/* unlink subrequests from head, destroy them later */
if (head->wb_this_page != head) {
/* destroy list will be terminated by head */
destroy_list = head->wb_this_page;
head->wb_this_page = head;
/* change head request to cover whole range that
* the former page group covered */
head->wb_bytes = total_bytes;
}
/*
* prepare head request to be added to new pgio descriptor
*/
nfs_page_group_clear_bits(head);
/*
* some part of the group was still on the inode list - otherwise
* the group wouldn't be involved in async write.
* grab a reference for the head request, iff it needs one.
*/
if (!test_and_set_bit(PG_INODE_REF, &head->wb_flags))
kref_get(&head->wb_kref);
nfs_page_group_unlock(head);
/* drop lock to clean uprequests on destroy list */
spin_unlock(&inode->i_lock);
nfs_destroy_unlinked_subrequests(destroy_list, head);
/* still holds ref on head from nfs_page_find_head_request_locked
* and still has lock on head from lock loop */
return head;
}
static void nfs_write_error_remove_page(struct nfs_page *req)
{
nfs_unlock_request(req);
nfs_end_page_writeback(req);
nfs_release_request(req);
generic_error_remove_page(page_file_mapping(req->wb_page),
req->wb_page);
}
/*
* 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,
struct page *page, bool nonblock,
bool launder)
{
struct nfs_page *req;
int ret = 0;
req = nfs_lock_and_join_requests(page, nonblock);
if (!req)
goto out;
ret = PTR_ERR(req);
if (IS_ERR(req))
goto out;
nfs_set_page_writeback(page);
WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
ret = 0;
2008-03-19 22:24:39 +07:00
if (!nfs_pageio_add_request(pgio, req)) {
ret = pgio->pg_error;
/*
* Remove the problematic req upon fatal errors
* in launder case, while other dirty pages can
* still be around until they get flushed.
*/
if (nfs_error_is_fatal(ret)) {
nfs_context_set_write_error(req->wb_context, ret);
if (launder) {
nfs_write_error_remove_page(req);
goto out;
}
}
nfs_redirty_request(req);
ret = -EAGAIN;
} else
nfs_add_stats(page_file_mapping(page)->host,
NFSIOS_WRITEPAGES, 1);
out:
return ret;
}
static int nfs_do_writepage(struct page *page, struct writeback_control *wbc,
struct nfs_pageio_descriptor *pgio, bool launder)
{
int ret;
nfs_pageio_cond_complete(pgio, page_index(page));
ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE,
launder);
if (ret == -EAGAIN) {
redirty_page_for_writepage(wbc, page);
ret = 0;
}
return ret;
}
/*
* Write an mmapped page to the server.
*/
static int nfs_writepage_locked(struct page *page,
struct writeback_control *wbc,
bool launder)
{
struct nfs_pageio_descriptor pgio;
struct inode *inode = page_file_mapping(page)->host;
int err;
nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
nfs_pageio_init_write(&pgio, inode, 0,
false, &nfs_async_write_completion_ops);
err = nfs_do_writepage(page, wbc, &pgio, launder);
nfs_pageio_complete(&pgio);
if (err < 0)
return err;
if (pgio.pg_error < 0)
return pgio.pg_error;
return 0;
}
int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
int ret;
ret = nfs_writepage_locked(page, wbc, false);
unlock_page(page);
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, false);
unlock_page(page);
return ret;
}
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
struct inode *inode = mapping->host;
struct nfs_pageio_descriptor pgio;
int err;
nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);
nfs_pageio_init_write(&pgio, inode, wb_priority(wbc), false,
&nfs_async_write_completion_ops);
err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio);
nfs_pageio_complete(&pgio);
2009-03-12 01:10:30 +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;
return 0;
2009-03-12 01:10:30 +07:00
out_err:
return err;
}
/*
* Insert a write request into an inode
*/
static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
{
struct nfs_inode *nfsi = NFS_I(inode);
nfs: add support for multiple nfs reqs per page Add "page groups" - a circular list of nfs requests (struct nfs_page) that all reference the same page. This gives nfs read and write paths the ability to account for sub-page regions independently. This somewhat follows the design of struct buffer_head's sub-page accounting. Only "head" requests are ever added/removed from the inode list in the buffered write path. "head" and "sub" requests are treated the same through the read path and the rest of the write/commit path. Requests are given an extra reference across the life of the list. Page groups are never rejoined after being split. If the read/write request fails and the client falls back to another path (ie revert to MDS in PNFS case), the already split requests are pushed through the recoalescing code again, which may split them further and then coalesce them into properly sized requests on the wire. Fragmentation shouldn't be a problem with the current design, because we flush all requests in page group when a non-contiguous request is added, so the only time resplitting should occur is on a resend of a read or write. This patch lays the groundwork for sub-page splitting, but does not actually do any splitting. For now all page groups have one request as pg_test functions don't yet split pages. There are several related patches that are needed support multiple requests per page group. Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-05-15 22:56:45 +07:00
WARN_ON_ONCE(req->wb_this_page != req);
/* Lock the request! */
nfs_lock_request(req);
spin_lock(&inode->i_lock);
if (!nfsi->nrequests &&
NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
inode->i_version++;
/*
* 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);
}
nfsi->nrequests++;
/* this a head request for a page group - mark it as having an
* extra reference so sub groups can follow suit.
* This flag also informs pgio layer when to bump nrequests when
* adding subrequests. */
WARN_ON(test_and_set_bit(PG_INODE_REF, &req->wb_flags));
kref_get(&req->wb_kref);
spin_unlock(&inode->i_lock);
}
/*
* Remove a write request from an inode
*/
static void nfs_inode_remove_request(struct nfs_page *req)
{
struct inode *inode = d_inode(req->wb_context->dentry);
struct nfs_inode *nfsi = NFS_I(inode);
struct nfs_page *head;
if (nfs_page_group_sync_on_bit(req, PG_REMOVE)) {
head = req->wb_head;
spin_lock(&inode->i_lock);
if (likely(head->wb_page && !PageSwapCache(head->wb_page))) {
set_page_private(head->wb_page, 0);
ClearPagePrivate(head->wb_page);
NFS: avoid deadlocks with loop-back mounted NFS filesystems. Support for loop-back mounted NFS filesystems is useful when NFS is used to access shared storage in a high-availability cluster. If the node running the NFS server fails, some other node can mount the filesystem and start providing NFS service. If that node already had the filesystem NFS mounted, it will now have it loop-back mounted. nfsd can suffer a deadlock when allocating memory and entering direct reclaim. While direct reclaim does not write to the NFS filesystem it can send and wait for a COMMIT through nfs_release_page(). This patch modifies nfs_release_page() to wait a limited time for the commit to complete - one second. If the commit doesn't complete in this time, nfs_release_page() will fail. This means it might now fail in some cases where it wouldn't before. These cases are only when 'gfp' includes '__GFP_WAIT'. nfs_release_page() is only called by try_to_release_page(), and that can only be called on an NFS page with required 'gfp' flags from - page_cache_pipe_buf_steal() in splice.c - shrink_page_list() in vmscan.c - invalidate_inode_pages2_range() in truncate.c The first two handle failure quite safely. The last is only called after ->launder_page() has been called, and that will have waited for the commit to finish already. So aborting if the commit takes longer than 1 second is perfectly safe. Signed-off-by: NeilBrown <neilb@suse.de> Acked-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-09-24 08:28:32 +07:00
smp_mb__after_atomic();
wake_up_page(head->wb_page, PG_private);
clear_bit(PG_MAPPED, &head->wb_flags);
}
nfsi->nrequests--;
spin_unlock(&inode->i_lock);
} else {
spin_lock(&inode->i_lock);
nfsi->nrequests--;
spin_unlock(&inode->i_lock);
}
if (test_and_clear_bit(PG_INODE_REF, &req->wb_flags))
nfs_release_request(req);
}
static void
nfs_mark_request_dirty(struct nfs_page *req)
{
if (req->wb_page)
__set_page_dirty_nobuffers(req->wb_page);
}
/*
* nfs_page_search_commits_for_head_request_locked
*
* Search through commit lists on @inode for the head request for @page.
* Must be called while holding the inode (which is cinfo) lock.
*
* Returns the head request if found, or NULL if not found.
*/
static struct nfs_page *
nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
struct page *page)
{
struct nfs_page *freq, *t;
struct nfs_commit_info cinfo;
struct inode *inode = &nfsi->vfs_inode;
nfs_init_cinfo_from_inode(&cinfo, inode);
/* search through pnfs commit lists */
freq = pnfs_search_commit_reqs(inode, &cinfo, page);
if (freq)
return freq->wb_head;
/* Linearly search the commit list for the correct request */
list_for_each_entry_safe(freq, t, &cinfo.mds->list, wb_list) {
if (freq->wb_page == page)
return freq->wb_head;
}
return NULL;
}
/**
* nfs_request_add_commit_list_locked - add request to a commit list
* @req: pointer to a struct nfs_page
* @dst: commit list head
* @cinfo: holds list lock and accounting info
*
* This sets the PG_CLEAN bit, updates the cinfo count of
* number of outstanding requests requiring a commit as well as
* the MM page stats.
*
* The caller must hold cinfo->inode->i_lock, and the nfs_page lock.
*/
void
nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst,
struct nfs_commit_info *cinfo)
{
set_bit(PG_CLEAN, &req->wb_flags);
nfs_list_add_request(req, dst);
cinfo->mds->ncommit++;
}
EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked);
/**
* nfs_request_add_commit_list - add request to a commit list
* @req: pointer to a struct nfs_page
* @dst: commit list head
* @cinfo: holds list lock and accounting info
*
* This sets the PG_CLEAN bit, updates the cinfo count of
* number of outstanding requests requiring a commit as well as
* the MM page stats.
*
* The caller must _not_ hold the cinfo->lock, but must be
* holding the nfs_page lock.
*/
void
nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo)
{
spin_lock(&cinfo->inode->i_lock);
nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo);
spin_unlock(&cinfo->inode->i_lock);
if (req->wb_page)
nfs_mark_page_unstable(req->wb_page, cinfo);
}
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
* @cinfo: holds list lock and accounting info
*
* This clears the PG_CLEAN bit, and updates the cinfo's count of
* number of outstanding requests requiring a commit
* It does not update the MM page stats.
*
* The caller _must_ hold the cinfo->lock and the nfs_page lock.
*/
void
nfs_request_remove_commit_list(struct nfs_page *req,
struct nfs_commit_info *cinfo)
{
if (!test_and_clear_bit(PG_CLEAN, &(req)->wb_flags))
return;
nfs_list_remove_request(req);
cinfo->mds->ncommit--;
}
EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
struct inode *inode)
{
cinfo->inode = inode;
cinfo->mds = &NFS_I(inode)->commit_info;
cinfo->ds = pnfs_get_ds_info(inode);
cinfo->dreq = NULL;
cinfo->completion_ops = &nfs_commit_completion_ops;
}
void nfs_init_cinfo(struct nfs_commit_info *cinfo,
struct inode *inode,
struct nfs_direct_req *dreq)
{
if (dreq)
nfs_init_cinfo_from_dreq(cinfo, dreq);
else
nfs_init_cinfo_from_inode(cinfo, inode);
}
EXPORT_SYMBOL_GPL(nfs_init_cinfo);
/*
* Add a request to the inode's commit list.
*/
void
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
struct nfs_commit_info *cinfo, u32 ds_commit_idx)
{
if (pnfs_mark_request_commit(req, lseg, cinfo, ds_commit_idx))
return;
nfs_request_add_commit_list(req, cinfo);
}
static void
nfs_clear_page_commit(struct page *page)
{
dec_node_page_state(page, NR_UNSTABLE_NFS);
dec_wb_stat(&inode_to_bdi(page_file_mapping(page)->host)->wb,
WB_RECLAIMABLE);
}
/* Called holding inode (/cinfo) lock */
static void
nfs_clear_request_commit(struct nfs_page *req)
{
if (test_bit(PG_CLEAN, &req->wb_flags)) {
struct inode *inode = d_inode(req->wb_context->dentry);
struct nfs_commit_info cinfo;
nfs_init_cinfo_from_inode(&cinfo, inode);
if (!pnfs_clear_request_commit(req, &cinfo)) {
nfs_request_remove_commit_list(req, &cinfo);
}
nfs_clear_page_commit(req->wb_page);
}
}
int nfs_write_need_commit(struct nfs_pgio_header *hdr)
{
if (hdr->verf.committed == NFS_DATA_SYNC)
return hdr->lseg == NULL;
return hdr->verf.committed != NFS_FILE_SYNC;
}
static void nfs_write_completion(struct nfs_pgio_header *hdr)
{
struct nfs_commit_info cinfo;
unsigned long bytes = 0;
if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
goto out;
nfs_init_cinfo_from_inode(&cinfo, hdr->inode);
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)) {
nfs_set_pageerror(req->wb_page);
nfs_context_set_write_error(req->wb_context, hdr->error);
goto remove_req;
}
if (nfs_write_need_commit(hdr)) {
memcpy(&req->wb_verf, &hdr->verf.verifier, sizeof(req->wb_verf));
nfs_mark_request_commit(req, hdr->lseg, &cinfo,
hdr->pgio_mirror_idx);
goto next;
}
remove_req:
nfs_inode_remove_request(req);
next:
nfs_unlock_request(req);
nfs_end_page_writeback(req);
nfs_release_request(req);
}
out:
hdr->release(hdr);
}
unsigned long
nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
{
return cinfo->mds->ncommit;
}
/* cinfo->inode->i_lock held by caller */
int
nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
struct nfs_commit_info *cinfo, int max)
{
struct nfs_page *req, *tmp;
int ret = 0;
list_for_each_entry_safe(req, tmp, src, wb_list) {
if (!nfs_lock_request(req))
continue;
kref_get(&req->wb_kref);
if (cond_resched_lock(&cinfo->inode->i_lock))
list_safe_reset_next(req, tmp, wb_list);
nfs_request_remove_commit_list(req, cinfo);
nfs_list_add_request(req, dst);
ret++;
if ((ret == max) && !cinfo->dreq)
break;
}
return ret;
}
/*
* nfs_scan_commit - Scan an inode for commit requests
* @inode: NFS inode to scan
* @dst: mds destination list
* @cinfo: mds and ds lists of reqs ready to commit
*
* Moves requests from the inode's 'commit' request list.
* The requests are *not* checked to ensure that they form a contiguous set.
*/
int
nfs_scan_commit(struct inode *inode, struct list_head *dst,
struct nfs_commit_info *cinfo)
{
int ret = 0;
spin_lock(&cinfo->inode->i_lock);
if (cinfo->mds->ncommit > 0) {
const int max = INT_MAX;
ret = nfs_scan_commit_list(&cinfo->mds->list, dst,
cinfo, max);
ret += pnfs_scan_commit_lists(inode, cinfo, max - ret);
}
spin_unlock(&cinfo->inode->i_lock);
return ret;
}
/*
* Search for an existing write request, and attempt to update
* it to reflect a new dirty region on a given page.
*
* If the attempt fails, then the existing request is flushed out
* to disk.
*/
static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
struct page *page,
unsigned int offset,
unsigned int bytes)
{
struct nfs_page *req;
unsigned int rqend;
unsigned int end;
int error;
if (!PagePrivate(page))
return NULL;
end = offset + bytes;
spin_lock(&inode->i_lock);
for (;;) {
req = nfs_page_find_head_request_locked(NFS_I(inode), page);
if (req == NULL)
goto out_unlock;
nfs: add support for multiple nfs reqs per page Add "page groups" - a circular list of nfs requests (struct nfs_page) that all reference the same page. This gives nfs read and write paths the ability to account for sub-page regions independently. This somewhat follows the design of struct buffer_head's sub-page accounting. Only "head" requests are ever added/removed from the inode list in the buffered write path. "head" and "sub" requests are treated the same through the read path and the rest of the write/commit path. Requests are given an extra reference across the life of the list. Page groups are never rejoined after being split. If the read/write request fails and the client falls back to another path (ie revert to MDS in PNFS case), the already split requests are pushed through the recoalescing code again, which may split them further and then coalesce them into properly sized requests on the wire. Fragmentation shouldn't be a problem with the current design, because we flush all requests in page group when a non-contiguous request is added, so the only time resplitting should occur is on a resend of a read or write. This patch lays the groundwork for sub-page splitting, but does not actually do any splitting. For now all page groups have one request as pg_test functions don't yet split pages. There are several related patches that are needed support multiple requests per page group. Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-05-15 22:56:45 +07:00
/* should be handled by nfs_flush_incompatible */
WARN_ON_ONCE(req->wb_head != req);
WARN_ON_ONCE(req->wb_this_page != req);
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.
*/
if (offset > rqend
|| end < req->wb_offset)
goto out_flushme;
if (nfs_lock_request(req))
break;
/* The request is locked, so wait and then retry */
spin_unlock(&inode->i_lock);
error = nfs_wait_on_request(req);
nfs_release_request(req);
if (error != 0)
goto out_err;
spin_lock(&inode->i_lock);
}
/* 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;
else
req->wb_bytes = rqend - req->wb_offset;
out_unlock:
if (req)
nfs_clear_request_commit(req);
spin_unlock(&inode->i_lock);
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)
{
struct inode *inode = page_file_mapping(page)->host;
struct nfs_page *req;
req = nfs_try_to_update_request(inode, page, offset, bytes);
if (req != NULL)
goto out;
nfs: add support for multiple nfs reqs per page Add "page groups" - a circular list of nfs requests (struct nfs_page) that all reference the same page. This gives nfs read and write paths the ability to account for sub-page regions independently. This somewhat follows the design of struct buffer_head's sub-page accounting. Only "head" requests are ever added/removed from the inode list in the buffered write path. "head" and "sub" requests are treated the same through the read path and the rest of the write/commit path. Requests are given an extra reference across the life of the list. Page groups are never rejoined after being split. If the read/write request fails and the client falls back to another path (ie revert to MDS in PNFS case), the already split requests are pushed through the recoalescing code again, which may split them further and then coalesce them into properly sized requests on the wire. Fragmentation shouldn't be a problem with the current design, because we flush all requests in page group when a non-contiguous request is added, so the only time resplitting should occur is on a resend of a read or write. This patch lays the groundwork for sub-page splitting, but does not actually do any splitting. For now all page groups have one request as pg_test functions don't yet split pages. There are several related patches that are needed support multiple requests per page group. Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-05-15 22:56:45 +07:00
req = nfs_create_request(ctx, page, NULL, offset, bytes);
if (IS_ERR(req))
goto out;
nfs_inode_add_request(inode, req);
out:
return req;
}
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(req);
nfs_mark_request_dirty(req);
nfs_unlock_and_release_request(req);
return 0;
}
int nfs_flush_incompatible(struct file *file, struct page *page)
{
struct nfs_open_context *ctx = nfs_file_open_context(file);
struct nfs_lock_context *l_ctx;
struct file_lock_context *flctx = file_inode(file)->i_flctx;
struct nfs_page *req;
int do_flush, status;
/*
* 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.
*/
do {
req = nfs_page_find_head_request(page);
if (req == NULL)
return 0;
l_ctx = req->wb_lock_context;
do_flush = req->wb_page != page ||
!nfs_match_open_context(req->wb_context, ctx);
nfs: add support for multiple nfs reqs per page Add "page groups" - a circular list of nfs requests (struct nfs_page) that all reference the same page. This gives nfs read and write paths the ability to account for sub-page regions independently. This somewhat follows the design of struct buffer_head's sub-page accounting. Only "head" requests are ever added/removed from the inode list in the buffered write path. "head" and "sub" requests are treated the same through the read path and the rest of the write/commit path. Requests are given an extra reference across the life of the list. Page groups are never rejoined after being split. If the read/write request fails and the client falls back to another path (ie revert to MDS in PNFS case), the already split requests are pushed through the recoalescing code again, which may split them further and then coalesce them into properly sized requests on the wire. Fragmentation shouldn't be a problem with the current design, because we flush all requests in page group when a non-contiguous request is added, so the only time resplitting should occur is on a resend of a read or write. This patch lays the groundwork for sub-page splitting, but does not actually do any splitting. For now all page groups have one request as pg_test functions don't yet split pages. There are several related patches that are needed support multiple requests per page group. Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-05-15 22:56:45 +07:00
/* for now, flush if more than 1 request in page_group */
do_flush |= req->wb_this_page != req;
if (l_ctx && flctx &&
!(list_empty_careful(&flctx->flc_posix) &&
list_empty_careful(&flctx->flc_flock))) {
do_flush |= l_ctx->lockowner.l_owner != current->files
|| l_ctx->lockowner.l_pid != current->tgid;
}
nfs_release_request(req);
if (!do_flush)
return 0;
status = nfs_wb_page(page_file_mapping(page)->host, page);
} while (status == 0);
return status;
}
/*
* Avoid buffered writes when a open context credential's key would
* expire soon.
*
* Returns -EACCES if the key will expire within RPC_KEY_EXPIRE_FAIL.
*
* Return 0 and set a credential flag which triggers the inode to flush
* and performs NFS_FILE_SYNC writes if the key will expired within
* RPC_KEY_EXPIRE_TIMEO.
*/
int
nfs_key_timeout_notify(struct file *filp, struct inode *inode)
{
struct nfs_open_context *ctx = nfs_file_open_context(filp);
struct rpc_auth *auth = NFS_SERVER(inode)->client->cl_auth;
return rpcauth_key_timeout_notify(auth, ctx->cred);
}
/*
* Test if the open context credential key is marked to expire soon.
*/
sunrpc: move NO_CRKEY_TIMEOUT to the auth->au_flags A generic_cred can be used to look up a unx_cred or a gss_cred, so it's not really safe to use the the generic_cred->acred->ac_flags to store the NO_CRKEY_TIMEOUT flag. A lookup for a unx_cred triggered while the KEY_EXPIRE_SOON flag is already set will cause both NO_CRKEY_TIMEOUT and KEY_EXPIRE_SOON to be set in the ac_flags, leaving the user associated with the auth_cred to be in a state where they're perpetually doing 4K NFS_FILE_SYNC writes. This can be reproduced as follows: 1. Mount two NFS filesystems, one with sec=krb5 and one with sec=sys. They do not need to be the same export, nor do they even need to be from the same NFS server. Also, v3 is fine. $ sudo mount -o v3,sec=krb5 server1:/export /mnt/krb5 $ sudo mount -o v3,sec=sys server2:/export /mnt/sys 2. As the normal user, before accessing the kerberized mount, kinit with a short lifetime (but not so short that renewing the ticket would leave you within the 4-minute window again by the time the original ticket expires), e.g. $ kinit -l 10m -r 60m 3. Do some I/O to the kerberized mount and verify that the writes are wsize, UNSTABLE: $ dd if=/dev/zero of=/mnt/krb5/file bs=1M count=1 4. Wait until you're within 4 minutes of key expiry, then do some more I/O to the kerberized mount to ensure that RPC_CRED_KEY_EXPIRE_SOON gets set. Verify that the writes are 4K, FILE_SYNC: $ dd if=/dev/zero of=/mnt/krb5/file bs=1M count=1 5. Now do some I/O to the sec=sys mount. This will cause RPC_CRED_NO_CRKEY_TIMEOUT to be set: $ dd if=/dev/zero of=/mnt/sys/file bs=1M count=1 6. Writes for that user will now be permanently 4K, FILE_SYNC for that user, regardless of which mount is being written to, until you reboot the client. Renewing the kerberos ticket (assuming it hasn't already expired) will have no effect. Grabbing a new kerberos ticket at this point will have no effect either. Move the flag to the auth->au_flags field (which is currently unused) and rename it slightly to reflect that it's no longer associated with the auth_cred->ac_flags. Add the rpc_auth to the arg list of rpcauth_cred_key_to_expire and check the au_flags there too. Finally, add the inode to the arg list of nfs_ctx_key_to_expire so we can determine the rpc_auth to pass to rpcauth_cred_key_to_expire. Signed-off-by: Scott Mayhew <smayhew@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2016-06-08 02:14:48 +07:00
bool nfs_ctx_key_to_expire(struct nfs_open_context *ctx, struct inode *inode)
{
sunrpc: move NO_CRKEY_TIMEOUT to the auth->au_flags A generic_cred can be used to look up a unx_cred or a gss_cred, so it's not really safe to use the the generic_cred->acred->ac_flags to store the NO_CRKEY_TIMEOUT flag. A lookup for a unx_cred triggered while the KEY_EXPIRE_SOON flag is already set will cause both NO_CRKEY_TIMEOUT and KEY_EXPIRE_SOON to be set in the ac_flags, leaving the user associated with the auth_cred to be in a state where they're perpetually doing 4K NFS_FILE_SYNC writes. This can be reproduced as follows: 1. Mount two NFS filesystems, one with sec=krb5 and one with sec=sys. They do not need to be the same export, nor do they even need to be from the same NFS server. Also, v3 is fine. $ sudo mount -o v3,sec=krb5 server1:/export /mnt/krb5 $ sudo mount -o v3,sec=sys server2:/export /mnt/sys 2. As the normal user, before accessing the kerberized mount, kinit with a short lifetime (but not so short that renewing the ticket would leave you within the 4-minute window again by the time the original ticket expires), e.g. $ kinit -l 10m -r 60m 3. Do some I/O to the kerberized mount and verify that the writes are wsize, UNSTABLE: $ dd if=/dev/zero of=/mnt/krb5/file bs=1M count=1 4. Wait until you're within 4 minutes of key expiry, then do some more I/O to the kerberized mount to ensure that RPC_CRED_KEY_EXPIRE_SOON gets set. Verify that the writes are 4K, FILE_SYNC: $ dd if=/dev/zero of=/mnt/krb5/file bs=1M count=1 5. Now do some I/O to the sec=sys mount. This will cause RPC_CRED_NO_CRKEY_TIMEOUT to be set: $ dd if=/dev/zero of=/mnt/sys/file bs=1M count=1 6. Writes for that user will now be permanently 4K, FILE_SYNC for that user, regardless of which mount is being written to, until you reboot the client. Renewing the kerberos ticket (assuming it hasn't already expired) will have no effect. Grabbing a new kerberos ticket at this point will have no effect either. Move the flag to the auth->au_flags field (which is currently unused) and rename it slightly to reflect that it's no longer associated with the auth_cred->ac_flags. Add the rpc_auth to the arg list of rpcauth_cred_key_to_expire and check the au_flags there too. Finally, add the inode to the arg list of nfs_ctx_key_to_expire so we can determine the rpc_auth to pass to rpcauth_cred_key_to_expire. Signed-off-by: Scott Mayhew <smayhew@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2016-06-08 02:14:48 +07:00
struct rpc_auth *auth = NFS_SERVER(inode)->client->cl_auth;
return rpcauth_cred_key_to_expire(auth, ctx->cred);
}
/*
* 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.
*/
static bool nfs_write_pageuptodate(struct page *page, struct inode *inode)
{
NFS: fix the handling of NFS_INO_INVALID_DATA flag in nfs_revalidate_mapping There is a possible race in how the nfs_invalidate_mapping function is handled. Currently, we go and invalidate the pages in the file and then clear NFS_INO_INVALID_DATA. The problem is that it's possible for a stale page to creep into the mapping after the page was invalidated (i.e., via readahead). If another writer comes along and sets the flag after that happens but before invalidate_inode_pages2 returns then we could clear the flag without the cache having been properly invalidated. So, we must clear the flag first and then invalidate the pages. Doing this however, opens another race: It's possible to have two concurrent read() calls that end up in nfs_revalidate_mapping at the same time. The first one clears the NFS_INO_INVALID_DATA flag and then goes to call nfs_invalidate_mapping. Just before calling that though, the other task races in, checks the flag and finds it cleared. At that point, it trusts that the mapping is good and gets the lock on the page, allowing the read() to be satisfied from the cache even though the data is no longer valid. These effects are easily manifested by running diotest3 from the LTP test suite on NFS. That program does a series of DIO writes and buffered reads. The operations are serialized and page-aligned but the existing code fails the test since it occasionally allows a read to come out of the cache incorrectly. While mixing direct and buffered I/O isn't recommended, I believe it's possible to hit this in other ways that just use buffered I/O, though that situation is much harder to reproduce. The problem is that the checking/clearing of that flag and the invalidation of the mapping really need to be atomic. Fix this by serializing concurrent invalidations with a bitlock. At the same time, we also need to allow other places that check NFS_INO_INVALID_DATA to check whether we might be in the middle of invalidating the file, so fix up a couple of places that do that to look for the new NFS_INO_INVALIDATING flag. Doing this requires us to be careful not to set the bitlock unnecessarily, so this code only does that if it believes it will be doing an invalidation. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-01-28 01:46:15 +07:00
struct nfs_inode *nfsi = NFS_I(inode);
if (nfs_have_delegated_attributes(inode))
goto out;
if (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
NFS: fix the handling of NFS_INO_INVALID_DATA flag in nfs_revalidate_mapping There is a possible race in how the nfs_invalidate_mapping function is handled. Currently, we go and invalidate the pages in the file and then clear NFS_INO_INVALID_DATA. The problem is that it's possible for a stale page to creep into the mapping after the page was invalidated (i.e., via readahead). If another writer comes along and sets the flag after that happens but before invalidate_inode_pages2 returns then we could clear the flag without the cache having been properly invalidated. So, we must clear the flag first and then invalidate the pages. Doing this however, opens another race: It's possible to have two concurrent read() calls that end up in nfs_revalidate_mapping at the same time. The first one clears the NFS_INO_INVALID_DATA flag and then goes to call nfs_invalidate_mapping. Just before calling that though, the other task races in, checks the flag and finds it cleared. At that point, it trusts that the mapping is good and gets the lock on the page, allowing the read() to be satisfied from the cache even though the data is no longer valid. These effects are easily manifested by running diotest3 from the LTP test suite on NFS. That program does a series of DIO writes and buffered reads. The operations are serialized and page-aligned but the existing code fails the test since it occasionally allows a read to come out of the cache incorrectly. While mixing direct and buffered I/O isn't recommended, I believe it's possible to hit this in other ways that just use buffered I/O, though that situation is much harder to reproduce. The problem is that the checking/clearing of that flag and the invalidation of the mapping really need to be atomic. Fix this by serializing concurrent invalidations with a bitlock. At the same time, we also need to allow other places that check NFS_INO_INVALID_DATA to check whether we might be in the middle of invalidating the file, so fix up a couple of places that do that to look for the new NFS_INO_INVALIDATING flag. Doing this requires us to be careful not to set the bitlock unnecessarily, so this code only does that if it believes it will be doing an invalidation. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-01-28 01:46:15 +07:00
return false;
smp_rmb();
NFS: fix the handling of NFS_INO_INVALID_DATA flag in nfs_revalidate_mapping There is a possible race in how the nfs_invalidate_mapping function is handled. Currently, we go and invalidate the pages in the file and then clear NFS_INO_INVALID_DATA. The problem is that it's possible for a stale page to creep into the mapping after the page was invalidated (i.e., via readahead). If another writer comes along and sets the flag after that happens but before invalidate_inode_pages2 returns then we could clear the flag without the cache having been properly invalidated. So, we must clear the flag first and then invalidate the pages. Doing this however, opens another race: It's possible to have two concurrent read() calls that end up in nfs_revalidate_mapping at the same time. The first one clears the NFS_INO_INVALID_DATA flag and then goes to call nfs_invalidate_mapping. Just before calling that though, the other task races in, checks the flag and finds it cleared. At that point, it trusts that the mapping is good and gets the lock on the page, allowing the read() to be satisfied from the cache even though the data is no longer valid. These effects are easily manifested by running diotest3 from the LTP test suite on NFS. That program does a series of DIO writes and buffered reads. The operations are serialized and page-aligned but the existing code fails the test since it occasionally allows a read to come out of the cache incorrectly. While mixing direct and buffered I/O isn't recommended, I believe it's possible to hit this in other ways that just use buffered I/O, though that situation is much harder to reproduce. The problem is that the checking/clearing of that flag and the invalidation of the mapping really need to be atomic. Fix this by serializing concurrent invalidations with a bitlock. At the same time, we also need to allow other places that check NFS_INO_INVALID_DATA to check whether we might be in the middle of invalidating the file, so fix up a couple of places that do that to look for the new NFS_INO_INVALIDATING flag. Doing this requires us to be careful not to set the bitlock unnecessarily, so this code only does that if it believes it will be doing an invalidation. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-01-28 01:46:15 +07:00
if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags))
return false;
out:
if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
return false;
return PageUptodate(page) != 0;
}
static bool
is_whole_file_wrlock(struct file_lock *fl)
{
return fl->fl_start == 0 && fl->fl_end == OFFSET_MAX &&
fl->fl_type == F_WRLCK;
}
/* If we know the page is up to date, and we're not using byte range locks (or
* if we have the whole file locked for writing), it may be more efficient to
* extend the write to cover the entire page in order to avoid fragmentation
* inefficiencies.
*
* If the file is opened for synchronous writes then we can just skip the rest
* of the checks.
*/
static int nfs_can_extend_write(struct file *file, struct page *page, struct inode *inode)
{
int ret;
struct file_lock_context *flctx = inode->i_flctx;
struct file_lock *fl;
if (file->f_flags & O_DSYNC)
return 0;
if (!nfs_write_pageuptodate(page, inode))
return 0;
if (NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
return 1;
if (!flctx || (list_empty_careful(&flctx->flc_flock) &&
list_empty_careful(&flctx->flc_posix)))
return 1;
/* Check to see if there are whole file write locks */
ret = 0;
spin_lock(&flctx->flc_lock);
if (!list_empty(&flctx->flc_posix)) {
fl = list_first_entry(&flctx->flc_posix, struct file_lock,
fl_list);
if (is_whole_file_wrlock(fl))
ret = 1;
} else if (!list_empty(&flctx->flc_flock)) {
fl = list_first_entry(&flctx->flc_flock, struct file_lock,
fl_list);
if (fl->fl_type == F_WRLCK)
ret = 1;
}
spin_unlock(&flctx->flc_lock);
return ret;
}
/*
* 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)
{
struct nfs_open_context *ctx = nfs_file_open_context(file);
struct inode *inode = page_file_mapping(page)->host;
int status = 0;
nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);
dprintk("NFS: nfs_updatepage(%pD2 %d@%lld)\n",
file, count, (long long)(page_file_offset(page) + offset));
if (!count)
goto out;
if (nfs_can_extend_write(file, page, inode)) {
count = max(count + offset, nfs_page_length(page));
offset = 0;
}
status = nfs_writepage_setup(ctx, page, offset, count);
if (status < 0)
nfs_set_pageerror(page);
else
__set_page_dirty_nobuffers(page);
out:
dprintk("NFS: nfs_updatepage returns %d (isize %lld)\n",
status, (long long)i_size_read(inode));
return status;
}
static int flush_task_priority(int how)
{
switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
case FLUSH_HIGHPRI:
return RPC_PRIORITY_HIGH;
case FLUSH_LOWPRI:
return RPC_PRIORITY_LOW;
}
return RPC_PRIORITY_NORMAL;
}
static void nfs_initiate_write(struct nfs_pgio_header *hdr,
struct rpc_message *msg,
const struct nfs_rpc_ops *rpc_ops,
struct rpc_task_setup *task_setup_data, int how)
{
int priority = flush_task_priority(how);
task_setup_data->priority = priority;
rpc_ops->write_setup(hdr, msg);
nfs4_state_protect_write(NFS_SERVER(hdr->inode)->nfs_client,
&task_setup_data->rpc_client, msg, hdr);
}
/* 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);
set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
nfs_unlock_request(req);
nfs_end_page_writeback(req);
nfs_release_request(req);
}
static void nfs_async_write_error(struct list_head *head)
{
struct nfs_page *req;
while (!list_empty(head)) {
req = nfs_list_entry(head->next);
nfs_list_remove_request(req);
nfs_redirty_request(req);
}
}
static void nfs_async_write_reschedule_io(struct nfs_pgio_header *hdr)
{
nfs_async_write_error(&hdr->pages);
}
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops = {
.error_cleanup = nfs_async_write_error,
.completion = nfs_write_completion,
.reschedule_io = nfs_async_write_reschedule_io,
};
void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
struct inode *inode, int ioflags, bool force_mds,
const struct nfs_pgio_completion_ops *compl_ops)
{
struct nfs_server *server = NFS_SERVER(inode);
const struct nfs_pageio_ops *pg_ops = &nfs_pgio_rw_ops;
#ifdef CONFIG_NFS_V4_1
if (server->pnfs_curr_ld && !force_mds)
pg_ops = server->pnfs_curr_ld->pg_write_ops;
#endif
nfs_pageio_init(pgio, inode, pg_ops, compl_ops, &nfs_rw_write_ops,
server->wsize, ioflags);
}
EXPORT_SYMBOL_GPL(nfs_pageio_init_write);
void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio)
{
struct nfs_pgio_mirror *mirror;
if (pgio->pg_ops && pgio->pg_ops->pg_cleanup)
pgio->pg_ops->pg_cleanup(pgio);
pgio->pg_ops = &nfs_pgio_rw_ops;
nfs_pageio_stop_mirroring(pgio);
mirror = &pgio->pg_mirrors[0];
mirror->pg_bsize = NFS_SERVER(pgio->pg_inode)->wsize;
}
EXPORT_SYMBOL_GPL(nfs_pageio_reset_write_mds);
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);
}
/*
* Special version of should_remove_suid() that ignores capabilities.
*/
static int nfs_should_remove_suid(const struct inode *inode)
{
umode_t mode = inode->i_mode;
int kill = 0;
/* suid always must be killed */
if (unlikely(mode & S_ISUID))
kill = ATTR_KILL_SUID;
/*
* sgid without any exec bits is just a mandatory locking mark; leave
* it alone. If some exec bits are set, it's a real sgid; kill it.
*/
if (unlikely((mode & S_ISGID) && (mode & S_IXGRP)))
kill |= ATTR_KILL_SGID;
if (unlikely(kill && S_ISREG(mode)))
return kill;
return 0;
}
static void nfs_writeback_check_extend(struct nfs_pgio_header *hdr,
struct nfs_fattr *fattr)
{
struct nfs_pgio_args *argp = &hdr->args;
struct nfs_pgio_res *resp = &hdr->res;
u64 size = argp->offset + resp->count;
if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
fattr->size = size;
if (nfs_size_to_loff_t(fattr->size) < i_size_read(hdr->inode)) {
fattr->valid &= ~NFS_ATTR_FATTR_SIZE;
return;
}
if (size != fattr->size)
return;
/* Set attribute barrier */
nfs_fattr_set_barrier(fattr);
/* ...and update size */
fattr->valid |= NFS_ATTR_FATTR_SIZE;
}
void nfs_writeback_update_inode(struct nfs_pgio_header *hdr)
{
struct nfs_fattr *fattr = &hdr->fattr;
struct inode *inode = hdr->inode;
spin_lock(&inode->i_lock);
nfs_writeback_check_extend(hdr, fattr);
nfs_post_op_update_inode_force_wcc_locked(inode, fattr);
spin_unlock(&inode->i_lock);
}
EXPORT_SYMBOL_GPL(nfs_writeback_update_inode);
/*
* This function is called when the WRITE call is complete.
*/
static int nfs_writeback_done(struct rpc_task *task,
struct nfs_pgio_header *hdr,
struct inode *inode)
{
int status;
/*
* ->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.
*/
status = NFS_PROTO(inode)->write_done(task, hdr);
if (status != 0)
return status;
nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, hdr->res.count);
if (hdr->res.verf->committed < hdr->args.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;
/* Note this will print the MDS for a DS write */
if (time_before(complain, jiffies)) {
dprintk("NFS: faulty NFS server %s:"
" (committed = %d) != (stable = %d)\n",
NFS_SERVER(inode)->nfs_client->cl_hostname,
hdr->res.verf->committed, hdr->args.stable);
complain = jiffies + 300 * HZ;
}
}
/* Deal with the suid/sgid bit corner case */
if (nfs_should_remove_suid(inode))
nfs_mark_for_revalidate(inode);
return 0;
}
/*
* This function is called when the WRITE call is complete.
*/
static void nfs_writeback_result(struct rpc_task *task,
struct nfs_pgio_header *hdr)
{
struct nfs_pgio_args *argp = &hdr->args;
struct nfs_pgio_res *resp = &hdr->res;
if (resp->count < argp->count) {
static unsigned long complain;
/* This a short write! */
nfs_inc_stats(hdr->inode, NFSIOS_SHORTWRITE);
/* Has the server at least made some progress? */
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;
}
nfs_set_pgio_error(hdr, -EIO, argp->offset);
task->tk_status = -EIO;
return;
}
NFSv4.1/pnfs: Retry through MDS when getting bad length of data If non rpc-based layout driver return bad length of data, nfs retries by calling rpc_restart_call_prepare() that cause an NULL reference panic. This patch lets nfs retry through MDS for non rpc-based layout driver return bad length of data. [13034.883329] BUG: unable to handle kernel NULL pointer dereference at (null) [13034.884902] IP: [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc] [13034.886558] PGD 0 [13034.888126] Oops: 0000 [#1] KASAN [13034.889710] Modules linked in: blocklayoutdriver(OE) nfsv4(OE) nfs(OE) fscache(E) nfsd(OE) xfs libcrc32c coretemp btrfs crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel ppdev vmw_balloon auth_rpcgss shpchp nfs_acl lockd vmw_vmci parport_pc xor raid6_pq grace parport sunrpc i2c_piix4 vmwgfx drm_kms_helper ttm drm mptspi e1000 serio_raw scsi_transport_spi mptscsih mptbase ata_generic pata_acpi [last unloaded: fscache] [13034.898260] CPU: 0 PID: 10112 Comm: kworker/0:1 Tainted: G OE 4.3.0-rc5+ #279 [13034.899932] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/02/2015 [13034.903342] Workqueue: events bl_read_cleanup [blocklayoutdriver] [13034.905059] task: ffff88006a9148c0 ti: ffff880035e90000 task.ti: ffff880035e90000 [13034.906827] RIP: 0010:[<ffffffffa00db372>] [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc] [13034.910522] RSP: 0018:ffff880035e97b58 EFLAGS: 00010282 [13034.912378] RAX: fffffbfff04a5a94 RBX: ffff880068fe4858 RCX: 0000000000000003 [13034.914339] RDX: dffffc0000000000 RSI: 0000000000000003 RDI: 0000000000000282 [13034.916236] RBP: ffff880035e97b68 R08: 0000000000000001 R09: 0000000000000001 [13034.918229] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000 [13034.920007] R13: ffff880068fe4858 R14: ffff880068fe4a60 R15: 0000000000001000 [13034.921845] FS: 0000000000000000(0000) GS:ffffffff82247000(0000) knlGS:0000000000000000 [13034.923645] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [13034.925525] CR2: 0000000000000000 CR3: 00000000063dd000 CR4: 00000000001406f0 [13034.932808] Stack: [13034.934813] ffff880068fe4780 0000000000001000 ffff880035e97ba8 ffffffffa08800d2 [13034.936675] ffffffffa088029d ffff880068fe4780 ffff880068fe4858 ffffffffa089c0a0 [13034.938593] ffff880068fe47e0 ffff88005d59faf0 ffff880035e97be0 ffffffffa087e08f [13034.940454] Call Trace: [13034.942388] [<ffffffffa08800d2>] nfs_readpage_result+0x112/0x200 [nfs] [13034.944317] [<ffffffffa088029d>] ? nfs_readpage_done+0xdd/0x160 [nfs] [13034.946267] [<ffffffffa087e08f>] nfs_pgio_result+0x9f/0x120 [nfs] [13034.948166] [<ffffffffa09266cc>] pnfs_ld_read_done+0x7c/0x1e0 [nfsv4] [13034.950247] [<ffffffffa03b07ee>] bl_read_cleanup+0x2e/0x60 [blocklayoutdriver] [13034.952156] [<ffffffff810ebf62>] process_one_work+0x412/0x870 [13034.954102] [<ffffffff810ebe84>] ? process_one_work+0x334/0x870 [13034.955949] [<ffffffff810ebb50>] ? queue_delayed_work_on+0x40/0x40 [13034.957985] [<ffffffff810ec441>] worker_thread+0x81/0x6a0 [13034.959817] [<ffffffff810ec3c0>] ? process_one_work+0x870/0x870 [13034.961785] [<ffffffff810f43bd>] kthread+0x17d/0x1a0 [13034.963544] [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330 [13034.965479] [<ffffffff81100428>] ? finish_task_switch+0x88/0x220 [13034.967223] [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330 [13034.968929] [<ffffffff81b6ae5f>] ret_from_fork+0x3f/0x70 [13034.970534] [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330 [13034.972176] Code: c7 43 50 40 84 0d a0 e8 3d fe 1c e1 48 8d 7b 58 c7 83 e4 00 00 00 00 00 00 00 e8 ca fe 1c e1 4c 8b 63 58 4c 89 e7 e8 be fe 1c e1 <49> 83 3c 24 00 74 12 48 c7 43 50 f0 a2 0e a0 b8 01 00 00 00 5b [13034.977148] RIP [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc] [13034.978780] RSP <ffff880035e97b58> [13034.980399] CR2: 0000000000000000 Signed-off-by: Kinglong Mee <kinglongmee@gmail.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2015-10-16 16:23:29 +07:00
/* For non rpc-based layout drivers, retry-through-MDS */
if (!task->tk_ops) {
hdr->pnfs_error = -EAGAIN;
return;
}
/* Was this an NFSv2 write or an NFSv3 stable write? */
if (resp->verf->committed != NFS_UNSTABLE) {
/* Resend from where the server left off */
hdr->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;
}
rpc_restart_call_prepare(task);
}
}
static int wait_on_commit(struct nfs_mds_commit_info *cinfo)
{
return wait_on_atomic_t(&cinfo->rpcs_out,
nfs_wait_atomic_killable, TASK_KILLABLE);
}
static void nfs_commit_begin(struct nfs_mds_commit_info *cinfo)
{
atomic_inc(&cinfo->rpcs_out);
}
static void nfs_commit_end(struct nfs_mds_commit_info *cinfo)
{
if (atomic_dec_and_test(&cinfo->rpcs_out))
wake_up_atomic_t(&cinfo->rpcs_out);
}
void nfs_commitdata_release(struct nfs_commit_data *data)
{
put_nfs_open_context(data->context);
nfs_commit_free(data);
}
EXPORT_SYMBOL_GPL(nfs_commitdata_release);
int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
const struct nfs_rpc_ops *nfs_ops,
const struct rpc_call_ops *call_ops,
int how, int flags)
{
struct rpc_task *task;
int priority = flush_task_priority(how);
struct rpc_message msg = {
.rpc_argp = &data->args,
.rpc_resp = &data->res,
.rpc_cred = data->cred,
};
struct rpc_task_setup task_setup_data = {
.task = &data->task,
.rpc_client = clnt,
.rpc_message = &msg,
.callback_ops = call_ops,
.callback_data = data,
.workqueue = nfsiod_workqueue,
.flags = RPC_TASK_ASYNC | flags,
.priority = priority,
};
/* Set up the initial task struct. */
nfs_ops->commit_setup(data, &msg);
dprintk("NFS: initiated commit call\n");
nfs4_state_protect(NFS_SERVER(data->inode)->nfs_client,
NFS_SP4_MACH_CRED_COMMIT, &task_setup_data.rpc_client, &msg);
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;
}
EXPORT_SYMBOL_GPL(nfs_initiate_commit);
static loff_t nfs_get_lwb(struct list_head *head)
{
loff_t lwb = 0;
struct nfs_page *req;
list_for_each_entry(req, head, wb_list)
if (lwb < (req_offset(req) + req->wb_bytes))
lwb = req_offset(req) + req->wb_bytes;
return lwb;
}
/*
* Set up the argument/result storage required for the RPC call.
*/
void nfs_init_commit(struct nfs_commit_data *data,
struct list_head *head,
struct pnfs_layout_segment *lseg,
struct nfs_commit_info *cinfo)
{
struct nfs_page *first = nfs_list_entry(head->next);
struct inode *inode = d_inode(first->wb_context->dentry);
/* 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;
data->cred = first->wb_context->cred;
data->lseg = lseg; /* reference transferred */
/* only set lwb for pnfs commit */
if (lseg)
data->lwb = nfs_get_lwb(&data->pages);
data->mds_ops = &nfs_commit_ops;
data->completion_ops = cinfo->completion_ops;
data->dreq = cinfo->dreq;
data->args.fh = NFS_FH(data->inode);
/* Note: we always request a commit of the entire inode */
data->args.offset = 0;
data->args.count = 0;
data->context = get_nfs_open_context(first->wb_context);
data->res.fattr = &data->fattr;
data->res.verf = &data->verf;
nfs_fattr_init(&data->fattr);
}
EXPORT_SYMBOL_GPL(nfs_init_commit);
void nfs_retry_commit(struct list_head *page_list,
struct pnfs_layout_segment *lseg,
struct nfs_commit_info *cinfo,
u32 ds_commit_idx)
{
struct nfs_page *req;
while (!list_empty(page_list)) {
req = nfs_list_entry(page_list->next);
nfs_list_remove_request(req);
nfs_mark_request_commit(req, lseg, cinfo, ds_commit_idx);
if (!cinfo->dreq)
nfs_clear_page_commit(req->wb_page);
nfs_unlock_and_release_request(req);
}
}
EXPORT_SYMBOL_GPL(nfs_retry_commit);
static void
nfs_commit_resched_write(struct nfs_commit_info *cinfo,
struct nfs_page *req)
{
__set_page_dirty_nobuffers(req->wb_page);
}
/*
* Commit dirty pages
*/
static int
nfs_commit_list(struct inode *inode, struct list_head *head, int how,
struct nfs_commit_info *cinfo)
{
struct nfs_commit_data *data;
nfs: avoid race that crashes nfs_init_commit Since the patch "NFS: Allow multiple commit requests in flight per file" we can run multiple simultaneous commits on the same inode. This introduced a race over collecting pages to commit that made it possible to call nfs_init_commit() with an empty list - which causes crashes like the one below. The fix is to catch this race and avoid calling nfs_init_commit and initiate_commit when there is no work to do. Here is the crash: [600522.076832] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 [600522.078475] IP: [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs] [600522.078745] PGD 4272b1067 PUD 4272cb067 PMD 0 [600522.078972] Oops: 0000 [#1] SMP [600522.079204] Modules linked in: nfsv3 nfs_layout_flexfiles rpcsec_gss_krb5 nfsv4 dns_resolver nfs fscache dcdbas ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 xt_conntrack ebtable_nat ebtable_broute bridge stp llc ebtable_filter ebtables ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw ip6table_filter ip6_tables iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw vmw_vsock_vmci_transport vsock bonding ipmi_devintf ipmi_msghandler coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel ppdev vmw_balloon parport_pc parport acpi_cpufreq vmw_vmci i2c_piix4 shpchp nfsd auth_rpcgss nfs_acl lockd grace sunrpc xfs libcrc32c vmwgfx drm_kms_helper ttm drm crc32c_intel serio_raw vmxnet3 [600522.081380] vmw_pvscsi ata_generic pata_acpi [600522.081809] CPU: 3 PID: 15667 Comm: /usr/bin/python Not tainted 4.1.9-100.pd.88.el7.x86_64 #1 [600522.082281] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2014 [600522.082814] task: ffff8800bbbfa780 ti: ffff88042ae84000 task.ti: ffff88042ae84000 [600522.083378] RIP: 0010:[<ffffffffa0479e72>] [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs] [600522.083973] RSP: 0018:ffff88042ae87438 EFLAGS: 00010246 [600522.084571] RAX: 0000000000000000 RBX: ffff880003485e40 RCX: ffff88042ae87588 [600522.085188] RDX: 0000000000000000 RSI: ffff88042ae874b0 RDI: ffff880003485e40 [600522.085756] RBP: ffff88042ae87448 R08: ffff880003486010 R09: ffff88042ae874b0 [600522.086332] R10: 0000000000000000 R11: 0000000000000005 R12: ffff88042ae872d0 [600522.086905] R13: ffff88042ae874b0 R14: ffff880003485e40 R15: ffff88042704c840 [600522.087484] FS: 00007f4728ff2740(0000) GS:ffff88043fd80000(0000) knlGS:0000000000000000 [600522.088070] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [600522.088663] CR2: 0000000000000040 CR3: 000000042b6aa000 CR4: 00000000001406e0 [600522.089327] Stack: [600522.089926] 0000000000000001 ffff88042ae87588 ffff88042ae874f8 ffffffffa04f09fa [600522.090549] 0000000000017840 0000000000017840 ffff88042ae87588 ffff8803258d9930 [600522.091169] ffff88042ae87578 ffffffffa0563d80 0000000000000000 ffff88042704c840 [600522.091789] Call Trace: [600522.092420] [<ffffffffa04f09fa>] pnfs_generic_commit_pagelist+0x1da/0x320 [nfsv4] [600522.093052] [<ffffffffa0563d80>] ? ff_layout_commit_prepare_v3+0x30/0x30 [nfs_layout_flexfiles] [600522.093696] [<ffffffffa0562645>] ff_layout_commit_pagelist+0x15/0x20 [nfs_layout_flexfiles] [600522.094359] [<ffffffffa047bc78>] nfs_generic_commit_list+0xe8/0x120 [nfs] [600522.095032] [<ffffffffa047bd6a>] nfs_commit_inode+0xba/0x110 [nfs] [600522.095719] [<ffffffffa046ac54>] nfs_release_page+0x44/0xd0 [nfs] [600522.096410] [<ffffffff811a8122>] try_to_release_page+0x32/0x50 [600522.097109] [<ffffffff811bd4f1>] shrink_page_list+0x961/0xb30 [600522.097812] [<ffffffff811bdced>] shrink_inactive_list+0x1cd/0x550 [600522.098530] [<ffffffff811bea65>] shrink_lruvec+0x635/0x840 [600522.099250] [<ffffffff811bed60>] shrink_zone+0xf0/0x2f0 [600522.099974] [<ffffffff811bf312>] do_try_to_free_pages+0x192/0x470 [600522.100709] [<ffffffff811bf6ca>] try_to_free_pages+0xda/0x170 [600522.101464] [<ffffffff811b2198>] __alloc_pages_nodemask+0x588/0x970 [600522.102235] [<ffffffff811fbbd5>] alloc_pages_vma+0xb5/0x230 [600522.103000] [<ffffffff813a1589>] ? cpumask_any_but+0x39/0x50 [600522.103774] [<ffffffff811d6115>] wp_page_copy.isra.55+0x95/0x490 [600522.104558] [<ffffffff810e3438>] ? __wake_up+0x48/0x60 [600522.105357] [<ffffffff811d7d3b>] do_wp_page+0xab/0x4f0 [600522.106137] [<ffffffff810a1bbb>] ? release_task+0x36b/0x470 [600522.106902] [<ffffffff8126dbd7>] ? eventfd_ctx_read+0x67/0x1c0 [600522.107659] [<ffffffff811da2a8>] handle_mm_fault+0xc78/0x1900 [600522.108431] [<ffffffff81067ef1>] __do_page_fault+0x181/0x420 [600522.109173] [<ffffffff811446a6>] ? __audit_syscall_exit+0x1e6/0x280 [600522.109893] [<ffffffff810681c0>] do_page_fault+0x30/0x80 [600522.110594] [<ffffffff81024f36>] ? syscall_trace_leave+0xc6/0x120 [600522.111288] [<ffffffff81790a58>] page_fault+0x28/0x30 [600522.111947] Code: 5d c3 0f 1f 80 00 00 00 00 0f 1f 44 00 00 55 4c 8d 87 d0 01 00 00 48 89 e5 53 48 89 fb 48 83 ec 08 4c 8b 0e 49 8b 41 18 4c 39 ce <48> 8b 40 40 4c 8b 50 30 74 24 48 8b 87 d0 01 00 00 48 8b 7e 08 [600522.113343] RIP [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs] [600522.114003] RSP <ffff88042ae87438> [600522.114636] CR2: 0000000000000040 Fixes: af7cf057 (NFS: Allow multiple commit requests in flight per file) CC: stable@vger.kernel.org Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
2016-05-25 21:07:23 +07:00
/* another commit raced with us */
if (list_empty(head))
return 0;
data = nfs_commitdata_alloc();
if (!data)
goto out_bad;
/* Set up the argument struct */
nfs_init_commit(data, head, NULL, cinfo);
atomic_inc(&cinfo->mds->rpcs_out);
return nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(inode),
data->mds_ops, how, 0);
out_bad:
nfs_retry_commit(head, NULL, cinfo, 0);
return -ENOMEM;
}
int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf)
{
struct inode *inode = file_inode(file);
struct nfs_open_context *open;
struct nfs_commit_info cinfo;
struct nfs_page *req;
int ret;
open = get_nfs_open_context(nfs_file_open_context(file));
req = nfs_create_request(open, NULL, NULL, 0, i_size_read(inode));
if (IS_ERR(req)) {
ret = PTR_ERR(req);
goto out_put;
}
nfs_init_cinfo_from_inode(&cinfo, inode);
memcpy(&req->wb_verf, verf, sizeof(struct nfs_write_verifier));
nfs_request_add_commit_list(req, &cinfo);
ret = nfs_commit_inode(inode, FLUSH_SYNC);
if (ret > 0)
ret = 0;
nfs_free_request(req);
out_put:
put_nfs_open_context(open);
return ret;
}
EXPORT_SYMBOL_GPL(nfs_commit_file);
/*
* COMMIT call returned
*/
static void nfs_commit_done(struct rpc_task *task, void *calldata)
{
struct nfs_commit_data *data = calldata;
dprintk("NFS: %5u nfs_commit_done (status %d)\n",
task->tk_pid, task->tk_status);
/* Call the NFS version-specific code */
NFS_PROTO(data->inode)->commit_done(task, data);
}
static void nfs_commit_release_pages(struct nfs_commit_data *data)
{
struct nfs_page *req;
int status = data->task.tk_status;
struct nfs_commit_info cinfo;
struct nfs_server *nfss;
while (!list_empty(&data->pages)) {
req = nfs_list_entry(data->pages.next);
nfs_list_remove_request(req);
if (req->wb_page)
nfs_clear_page_commit(req->wb_page);
dprintk("NFS: commit (%s/%llu %d@%lld)",
req->wb_context->dentry->d_sb->s_id,
(unsigned long long)NFS_FILEID(d_inode(req->wb_context->dentry)),
req->wb_bytes,
(long long)req_offset(req));
if (status < 0) {
nfs_context_set_write_error(req->wb_context, status);
nfs_inode_remove_request(req);
dprintk(", error = %d\n", status);
goto next;
}
/* Okay, COMMIT succeeded, apparently. Check the verifier
* returned by the server against all stored verfs. */
if (!nfs_write_verifier_cmp(&req->wb_verf, &data->verf.verifier)) {
/* 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");
nfs_mark_request_dirty(req);
set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
next:
nfs_unlock_and_release_request(req);
}
nfss = NFS_SERVER(data->inode);
if (atomic_long_read(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
nfs_init_cinfo(&cinfo, data->inode, data->dreq);
nfs_commit_end(cinfo.mds);
}
static void nfs_commit_release(void *calldata)
{
struct nfs_commit_data *data = calldata;
data->completion_ops->completion(data);
nfs_commitdata_release(calldata);
}
static const struct rpc_call_ops nfs_commit_ops = {
.rpc_call_prepare = nfs_commit_prepare,
.rpc_call_done = nfs_commit_done,
.rpc_release = nfs_commit_release,
};
static const struct nfs_commit_completion_ops nfs_commit_completion_ops = {
.completion = nfs_commit_release_pages,
.resched_write = nfs_commit_resched_write,
};
int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
int how, struct nfs_commit_info *cinfo)
{
int status;
status = pnfs_commit_list(inode, head, how, cinfo);
if (status == PNFS_NOT_ATTEMPTED)
status = nfs_commit_list(inode, head, how, cinfo);
return status;
}
int nfs_commit_inode(struct inode *inode, int how)
{
LIST_HEAD(head);
struct nfs_commit_info cinfo;
int may_wait = how & FLUSH_SYNC;
int error = 0;
int res;
nfs_init_cinfo_from_inode(&cinfo, inode);
nfs_commit_begin(cinfo.mds);
res = nfs_scan_commit(inode, &head, &cinfo);
if (res)
error = nfs_generic_commit_list(inode, &head, how, &cinfo);
nfs_commit_end(cinfo.mds);
if (error < 0)
goto out_error;
if (!may_wait)
goto out_mark_dirty;
error = wait_on_commit(cinfo.mds);
if (error < 0)
return error;
return res;
out_error:
res = error;
/* 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);
return res;
}
EXPORT_SYMBOL_GPL(nfs_commit_inode);
int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
struct nfs_inode *nfsi = NFS_I(inode);
int flags = FLUSH_SYNC;
int ret = 0;
/* no commits means nothing needs to be done */
if (!nfsi->commit_info.ncommit)
return ret;
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.
*/
if (nfsi->commit_info.ncommit <= (nfsi->nrequests >> 1))
goto out_mark_dirty;
/* don't wait for the COMMIT response */
flags = 0;
}
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;
}
return 0;
}
out_mark_dirty:
__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
return ret;
}
EXPORT_SYMBOL_GPL(nfs_write_inode);
/*
* Wrapper for filemap_write_and_wait_range()
*
* Needed for pNFS in order to ensure data becomes visible to the
* client.
*/
int nfs_filemap_write_and_wait_range(struct address_space *mapping,
loff_t lstart, loff_t lend)
{
int ret;
ret = filemap_write_and_wait_range(mapping, lstart, lend);
if (ret == 0)
ret = pnfs_sync_inode(mapping->host, true);
return ret;
}
EXPORT_SYMBOL_GPL(nfs_filemap_write_and_wait_range);
/*
* flush the inode to disk.
*/
int nfs_wb_all(struct inode *inode)
{
int ret;
trace_nfs_writeback_inode_enter(inode);
ret = filemap_write_and_wait(inode->i_mapping);
if (ret)
goto out;
ret = nfs_commit_inode(inode, FLUSH_SYNC);
if (ret < 0)
goto out;
pnfs_sync_inode(inode, true);
ret = 0;
out:
trace_nfs_writeback_inode_exit(inode, ret);
return ret;
}
EXPORT_SYMBOL_GPL(nfs_wb_all);
int nfs_wb_page_cancel(struct inode *inode, struct page *page)
{
struct nfs_page *req;
int ret = 0;
wait_on_page_writeback(page);
/* blocking call to cancel all requests and join to a single (head)
* request */
req = nfs_lock_and_join_requests(page, false);
if (IS_ERR(req)) {
ret = PTR_ERR(req);
} else if (req) {
/* all requests from this page have been cancelled by
* nfs_lock_and_join_requests, so just remove the head
* request from the inode / page_private pointer and
* release it */
nfs_inode_remove_request(req);
nfs_unlock_and_release_request(req);
}
return ret;
}
/*
* Write back all requests on one page - we do this before reading it.
*/
int nfs_wb_single_page(struct inode *inode, struct page *page, bool launder)
{
loff_t range_start = page_file_offset(page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
loff_t range_end = range_start + (loff_t)(PAGE_SIZE - 1);
struct writeback_control wbc = {
.sync_mode = WB_SYNC_ALL,
.nr_to_write = 0,
.range_start = range_start,
.range_end = range_end,
};
int ret;
trace_nfs_writeback_page_enter(inode);
for (;;) {
wait_on_page_writeback(page);
if (clear_page_dirty_for_io(page)) {
ret = nfs_writepage_locked(page, &wbc, launder);
if (ret < 0)
goto out_error;
continue;
}
ret = 0;
if (!PagePrivate(page))
break;
ret = nfs_commit_inode(inode, FLUSH_SYNC);
if (ret < 0)
goto out_error;
}
out_error:
trace_nfs_writeback_page_exit(inode, ret);
return ret;
}
#ifdef CONFIG_MIGRATION
int nfs_migrate_page(struct address_space *mapping, struct page *newpage,
struct page *page, enum migrate_mode mode)
{
/*
* 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;
NFS: nfs_migrate_page() does not wait for FS-Cache to finish with a page nfs_migrate_page() does not wait for FS-Cache to finish with a page, probably leading to the following bad-page-state: BUG: Bad page state in process python-bin pfn:17d39b page:ffffea00053649e8 flags:004000000000100c count:0 mapcount:0 mapping:(null) index:38686 (Tainted: G B ---------------- ) Pid: 31053, comm: python-bin Tainted: G B ---------------- 2.6.32-71.24.1.el6.x86_64 #1 Call Trace: [<ffffffff8111bfe7>] bad_page+0x107/0x160 [<ffffffff8111ee69>] free_hot_cold_page+0x1c9/0x220 [<ffffffff8111ef19>] __pagevec_free+0x59/0xb0 [<ffffffff8104b988>] ? flush_tlb_others_ipi+0x128/0x130 [<ffffffff8112230c>] release_pages+0x21c/0x250 [<ffffffff8115b92a>] ? remove_migration_pte+0x28a/0x2b0 [<ffffffff8115f3f8>] ? mem_cgroup_get_reclaim_stat_from_page+0x18/0x70 [<ffffffff81122687>] ____pagevec_lru_add+0x167/0x180 [<ffffffff811226f8>] __lru_cache_add+0x58/0x70 [<ffffffff81122731>] lru_cache_add_lru+0x21/0x40 [<ffffffff81123f49>] putback_lru_page+0x69/0x100 [<ffffffff8115c0bd>] migrate_pages+0x13d/0x5d0 [<ffffffff81122687>] ? ____pagevec_lru_add+0x167/0x180 [<ffffffff81152ab0>] ? compaction_alloc+0x0/0x370 [<ffffffff8115255c>] compact_zone+0x4cc/0x600 [<ffffffff8111cfac>] ? get_page_from_freelist+0x15c/0x820 [<ffffffff810672f4>] ? check_preempt_wakeup+0x1c4/0x3c0 [<ffffffff8115290e>] compact_zone_order+0x7e/0xb0 [<ffffffff81152a49>] try_to_compact_pages+0x109/0x170 [<ffffffff8111e94d>] __alloc_pages_nodemask+0x5ed/0x850 [<ffffffff814c9136>] ? thread_return+0x4e/0x778 [<ffffffff81150d43>] alloc_pages_vma+0x93/0x150 [<ffffffff81167ea5>] do_huge_pmd_anonymous_page+0x135/0x340 [<ffffffff814cb6f6>] ? rwsem_down_read_failed+0x26/0x30 [<ffffffff81136755>] handle_mm_fault+0x245/0x2b0 [<ffffffff814ce383>] do_page_fault+0x123/0x3a0 [<ffffffff814cbdf5>] page_fault+0x25/0x30 nfs_migrate_page() calls nfs_fscache_release_page() which doesn't actually wait - even if __GFP_WAIT is set. The reason that doesn't wait is that fscache_maybe_release_page() might deadlock the allocator as the work threads writing to the cache may all end up sleeping on memory allocation. However, I wonder if that is actually a problem. There are a number of things I can do to deal with this: (1) Make nfs_migrate_page() wait. (2) Make fscache_maybe_release_page() honour the __GFP_WAIT flag. (3) Set a timeout around the wait. (4) Make nfs_migrate_page() return an error if the page is still busy. For the moment, I'll select (2) and (4). Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Jeff Layton <jlayton@redhat.com>
2012-12-05 20:34:49 +07:00
if (!nfs_fscache_release_page(page, GFP_KERNEL))
return -EBUSY;
return migrate_page(mapping, newpage, page, mode);
}
#endif
int __init nfs_init_writepagecache(void)
{
nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
sizeof(struct nfs_pgio_header),
0, SLAB_HWCACHE_ALIGN,
NULL);
if (nfs_wdata_cachep == NULL)
return -ENOMEM;
nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
nfs_wdata_cachep);
if (nfs_wdata_mempool == NULL)
goto out_destroy_write_cache;
nfs_cdata_cachep = kmem_cache_create("nfs_commit_data",
sizeof(struct nfs_commit_data),
0, SLAB_HWCACHE_ALIGN,
NULL);
if (nfs_cdata_cachep == NULL)
goto out_destroy_write_mempool;
nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
nfs_cdata_cachep);
if (nfs_commit_mempool == NULL)
goto out_destroy_commit_cache;
/*
* 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;
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;
}
void nfs_destroy_writepagecache(void)
{
mempool_destroy(nfs_commit_mempool);
kmem_cache_destroy(nfs_cdata_cachep);
mempool_destroy(nfs_wdata_mempool);
kmem_cache_destroy(nfs_wdata_cachep);
}
static const struct nfs_rw_ops nfs_rw_write_ops = {
.rw_mode = FMODE_WRITE,
.rw_alloc_header = nfs_writehdr_alloc,
.rw_free_header = nfs_writehdr_free,
.rw_done = nfs_writeback_done,
.rw_result = nfs_writeback_result,
.rw_initiate = nfs_initiate_write,
};