linux/fs/direct-io.c

1425 lines
41 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* fs/direct-io.c
*
* Copyright (C) 2002, Linus Torvalds.
*
* O_DIRECT
*
* 04Jul2002 Andrew Morton
* Initial version
* 11Sep2002 janetinc@us.ibm.com
* added readv/writev support.
* 29Oct2002 Andrew Morton
* rewrote bio_add_page() support.
* 30Oct2002 pbadari@us.ibm.com
* added support for non-aligned IO.
* 06Nov2002 pbadari@us.ibm.com
* added asynchronous IO support.
* 21Jul2003 nathans@sgi.com
* added IO completion notifier.
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/bio.h>
#include <linux/wait.h>
#include <linux/err.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>
#include <linux/rwsem.h>
#include <linux/uio.h>
#include <linux/atomic.h>
#include <linux/prefetch.h>
/*
* How many user pages to map in one call to get_user_pages(). This determines
* the size of a structure in the slab cache
*/
#define DIO_PAGES 64
/*
* Flags for dio_complete()
*/
#define DIO_COMPLETE_ASYNC 0x01 /* This is async IO */
#define DIO_COMPLETE_INVALIDATE 0x02 /* Can invalidate pages */
/*
* This code generally works in units of "dio_blocks". A dio_block is
* somewhere between the hard sector size and the filesystem block size. it
* is determined on a per-invocation basis. When talking to the filesystem
* we need to convert dio_blocks to fs_blocks by scaling the dio_block quantity
* down by dio->blkfactor. Similarly, fs-blocksize quantities are converted
* to bio_block quantities by shifting left by blkfactor.
*
* If blkfactor is zero then the user's request was aligned to the filesystem's
* blocksize.
*/
/* dio_state only used in the submission path */
struct dio_submit {
struct bio *bio; /* bio under assembly */
unsigned blkbits; /* doesn't change */
unsigned blkfactor; /* When we're using an alignment which
is finer than the filesystem's soft
blocksize, this specifies how much
finer. blkfactor=2 means 1/4-block
alignment. Does not change */
unsigned start_zero_done; /* flag: sub-blocksize zeroing has
been performed at the start of a
write */
int pages_in_io; /* approximate total IO pages */
sector_t block_in_file; /* Current offset into the underlying
file in dio_block units. */
unsigned blocks_available; /* At block_in_file. changes */
int reap_counter; /* rate limit reaping */
sector_t final_block_in_request;/* doesn't change */
int boundary; /* prev block is at a boundary */
get_block_t *get_block; /* block mapping function */
dio_submit_t *submit_io; /* IO submition function */
loff_t logical_offset_in_bio; /* current first logical block in bio */
sector_t final_block_in_bio; /* current final block in bio + 1 */
sector_t next_block_for_io; /* next block to be put under IO,
in dio_blocks units */
/*
* Deferred addition of a page to the dio. These variables are
* private to dio_send_cur_page(), submit_page_section() and
* dio_bio_add_page().
*/
struct page *cur_page; /* The page */
unsigned cur_page_offset; /* Offset into it, in bytes */
unsigned cur_page_len; /* Nr of bytes at cur_page_offset */
sector_t cur_page_block; /* Where it starts */
loff_t cur_page_fs_offset; /* Offset in file */
struct iov_iter *iter;
/*
* Page queue. These variables belong to dio_refill_pages() and
* dio_get_page().
*/
unsigned head; /* next page to process */
unsigned tail; /* last valid page + 1 */
size_t from, to;
};
/* dio_state communicated between submission path and end_io */
struct dio {
int flags; /* doesn't change */
int op;
int op_flags;
blk_qc_t bio_cookie;
struct gendisk *bio_disk;
struct inode *inode;
loff_t i_size; /* i_size when submitted */
dio_iodone_t *end_io; /* IO completion function */
void *private; /* copy from map_bh.b_private */
/* BIO completion state */
spinlock_t bio_lock; /* protects BIO fields below */
int page_errors; /* errno from get_user_pages() */
int is_async; /* is IO async ? */
bool defer_completion; /* defer AIO completion to workqueue? */
bool should_dirty; /* if pages should be dirtied */
int io_error; /* IO error in completion path */
unsigned long refcount; /* direct_io_worker() and bios */
struct bio *bio_list; /* singly linked via bi_private */
struct task_struct *waiter; /* waiting task (NULL if none) */
/* AIO related stuff */
struct kiocb *iocb; /* kiocb */
ssize_t result; /* IO result */
dio: don't zero out the pages array inside struct dio Intel reported a performance regression caused by the following commit: commit 848c4dd5153c7a0de55470ce99a8e13a63b4703f Author: Zach Brown <zach.brown@oracle.com> Date: Mon Aug 20 17:12:01 2007 -0700 dio: zero struct dio with kzalloc instead of manually This patch uses kzalloc to zero all of struct dio rather than manually trying to track which fields we rely on being zero. It passed aio+dio stress testing and some bug regression testing on ext3. This patch was introduced by Linus in the conversation that lead up to Badari's minimal fix to manually zero .map_bh.b_state in commit: 6a648fa72161d1f6468dabd96c5d3c0db04f598a It makes the code a bit smaller. Maybe a couple fewer cachelines to load, if we're lucky: text data bss dec hex filename 3285925 568506 1304616 5159047 4eb887 vmlinux 3285797 568506 1304616 5158919 4eb807 vmlinux.patched I was unable to measure a stable difference in the number of cpu cycles spent in blockdev_direct_IO() when pushing aio+dio 256K reads at ~340MB/s. So the resulting intent of the patch isn't a performance gain but to avoid exposing ourselves to the risk of finding another field like .map_bh.b_state where we rely on zeroing but don't enforce it in the code. Zach surmised that zeroing out the page array was what caused most of the problem, and suggested the approach taken in the attached patch for resolving the issue. Intel re-tested with this patch and saw a 0.6% performance gain (the original regression was 0.5%). [akpm@linux-foundation.org: add comment] Signed-off-by: Jeff Moyer <jmoyer@redhat.com> Acked-by: Zach Brown <zach.brown@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:49 +00:00
/*
* pages[] (and any fields placed after it) are not zeroed out at
* allocation time. Don't add new fields after pages[] unless you
* wish that they not be zeroed.
*/
union {
struct page *pages[DIO_PAGES]; /* page buffer */
struct work_struct complete_work;/* deferred AIO completion */
};
} ____cacheline_aligned_in_smp;
static struct kmem_cache *dio_cache __read_mostly;
/*
* How many pages are in the queue?
*/
static inline unsigned dio_pages_present(struct dio_submit *sdio)
{
return sdio->tail - sdio->head;
}
/*
* Go grab and pin some userspace pages. Typically we'll get 64 at a time.
*/
static inline int dio_refill_pages(struct dio *dio, struct dio_submit *sdio)
{
ssize_t ret;
ret = iov_iter_get_pages(sdio->iter, dio->pages, LONG_MAX, DIO_PAGES,
&sdio->from);
if (ret < 0 && sdio->blocks_available && (dio->op == REQ_OP_WRITE)) {
remove ZERO_PAGE The commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note A last caveat: the ZERO_PAGE is now refcounted and managed with rmap (and thus mapcounted and count towards shared rss). These writes to the struct page could cause excessive cacheline bouncing on big systems. There are a number of ways this could be addressed if it is an issue. And indeed this cacheline bouncing has shown up on large SGI systems. There was a situation where an Altix system was essentially livelocked tearing down ZERO_PAGE pagetables when an HPC app aborted during startup. This situation can be avoided in userspace, but it does highlight the potential scalability problem with refcounting ZERO_PAGE, and corner cases where it can really hurt (we don't want the system to livelock!). There are several broad ways to fix this problem: 1. add back some special casing to avoid refcounting ZERO_PAGE 2. per-node or per-cpu ZERO_PAGES 3. remove the ZERO_PAGE completely I will argue for 3. The others should also fix the problem, but they result in more complex code than does 3, with little or no real benefit that I can see. Why? Inserting a ZERO_PAGE for anonymous read faults appears to be a false optimisation: if an application is performance critical, it would not be doing many read faults of new memory, or at least it could be expected to write to that memory soon afterwards. If cache or memory use is critical, it should not be working with a significant number of ZERO_PAGEs anyway (a more compact representation of zeroes should be used). As a sanity check -- mesuring on my desktop system, there are never many mappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not increase much without it. When running a make -j4 kernel compile on my dual core system, there are about 1,000 mappings to the ZERO_PAGE created per second, but about 1,000 ZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second is torn down without being COWed). So removing ZERO_PAGE will save 1,000 page faults per second when running kbuild, while keeping it only saves less than 1 page clearing operation per second. 1 page clear is cheaper than a thousand faults, presumably, so there isn't an obvious loss. Neither the logical argument nor these basic tests give a guarantee of no regressions. However, this is a reasonable opportunity to try to remove the ZERO_PAGE from the pagefault path. If it is found to cause regressions, we can reintroduce it and just avoid refcounting it. The /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don't see much use to them except on benchmarks. All other users of ZERO_PAGE are converted just to use ZERO_PAGE(0) for simplicity. We can look at replacing them all and maybe ripping out ZERO_PAGE completely when we are more satisfied with this solution. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus "snif" Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:24:40 +00:00
struct page *page = ZERO_PAGE(0);
/*
* A memory fault, but the filesystem has some outstanding
* mapped blocks. We need to use those blocks up to avoid
* leaking stale data in the file.
*/
if (dio->page_errors == 0)
dio->page_errors = ret;
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 12:29:47 +00:00
get_page(page);
2005-10-30 01:16:12 +00:00
dio->pages[0] = page;
sdio->head = 0;
sdio->tail = 1;
sdio->from = 0;
sdio->to = PAGE_SIZE;
return 0;
}
if (ret >= 0) {
iov_iter_advance(sdio->iter, ret);
ret += sdio->from;
sdio->head = 0;
sdio->tail = (ret + PAGE_SIZE - 1) / PAGE_SIZE;
sdio->to = ((ret - 1) & (PAGE_SIZE - 1)) + 1;
return 0;
}
return ret;
}
/*
* Get another userspace page. Returns an ERR_PTR on error. Pages are
* buffered inside the dio so that we can call get_user_pages() against a
* decent number of pages, less frequently. To provide nicer use of the
* L1 cache.
*/
static inline struct page *dio_get_page(struct dio *dio,
struct dio_submit *sdio)
{
if (dio_pages_present(sdio) == 0) {
int ret;
ret = dio_refill_pages(dio, sdio);
if (ret)
return ERR_PTR(ret);
BUG_ON(dio_pages_present(sdio) == 0);
}
return dio->pages[sdio->head];
}
/*
* Warn about a page cache invalidation failure during a direct io write.
*/
void dio_warn_stale_pagecache(struct file *filp)
{
static DEFINE_RATELIMIT_STATE(_rs, 86400 * HZ, DEFAULT_RATELIMIT_BURST);
char pathname[128];
struct inode *inode = file_inode(filp);
char *path;
errseq_set(&inode->i_mapping->wb_err, -EIO);
if (__ratelimit(&_rs)) {
path = file_path(filp, pathname, sizeof(pathname));
if (IS_ERR(path))
path = "(unknown)";
pr_crit("Page cache invalidation failure on direct I/O. Possible data corruption due to collision with buffered I/O!\n");
pr_crit("File: %s PID: %d Comm: %.20s\n", path, current->pid,
current->comm);
}
}
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
/**
* dio_complete() - called when all DIO BIO I/O has been completed
* @offset: the byte offset in the file of the completed operation
*
* This drops i_dio_count, lets interested parties know that a DIO operation
* has completed, and calculates the resulting return code for the operation.
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
*
* It lets the filesystem know if it registered an interest earlier via
* get_block. Pass the private field of the map buffer_head so that
* filesystems can use it to hold additional state between get_block calls and
* dio_complete.
*/
static ssize_t dio_complete(struct dio *dio, ssize_t ret, unsigned int flags)
{
loff_t offset = dio->iocb->ki_pos;
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
ssize_t transferred = 0;
int err;
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
[PATCH] dio: only call aio_complete() after returning -EIOCBQUEUED The only time it is safe to call aio_complete() is when the ->ki_retry function returns -EIOCBQUEUED to the AIO core. direct_io_worker() has historically done this by relying on its caller to translate positive return codes into -EIOCBQUEUED for the aio case. It did this by trying to keep conditionals in sync. direct_io_worker() knew when finished_one_bio() was going to call aio_complete(). It would reverse the test and wait and free the dio in the cases it thought that finished_one_bio() wasn't going to. Not surprisingly, it ended up getting it wrong. 'ret' could be a negative errno from the submission path but it failed to communicate this to finished_one_bio(). direct_io_worker() would return < 0, it's callers wouldn't raise -EIOCBQUEUED, and aio_complete() would be called. In the future finished_one_bio()'s tests wouldn't reflect this and aio_complete() would be called for a second time which can manifest as an oops. The previous cleanups have whittled the sync and async completion paths down to the point where we can collapse them and clearly reassert the invariant that we must only call aio_complete() after returning -EIOCBQUEUED. direct_io_worker() will only return -EIOCBQUEUED when it is not the last to drop the dio refcount and the aio bio completion path will only call aio_complete() when it is the last to drop the dio refcount. direct_io_worker() can ensure that it is the last to drop the reference count by waiting for bios to drain. It does this for sync ops, of course, and for partial dio writes that must fall back to buffered and for aio ops that saw errors during submission. This means that operations that end up waiting, even if they were issued as aio ops, will not call aio_complete() from dio. Instead we return the return code of the operation and let the aio core call aio_complete(). This is purposely done to fix a bug where AIO DIO file extensions would call aio_complete() before their callers have a chance to update i_size. Now that direct_io_worker() is explicitly returning -EIOCBQUEUED its callers no longer have to translate for it. XFS needs to be careful not to free resources that will be used during AIO completion if -EIOCBQUEUED is returned. We maintain the previous behaviour of trying to write fs metadata for O_SYNC aio+dio writes. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:05 +00:00
/*
* AIO submission can race with bio completion to get here while
* expecting to have the last io completed by bio completion.
* In that case -EIOCBQUEUED is in fact not an error we want
* to preserve through this call.
*/
if (ret == -EIOCBQUEUED)
ret = 0;
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
if (dio->result) {
transferred = dio->result;
/* Check for short read case */
if ((dio->op == REQ_OP_READ) &&
((offset + transferred) > dio->i_size))
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
transferred = dio->i_size - offset;
/* ignore EFAULT if some IO has been done */
if (unlikely(ret == -EFAULT) && transferred)
ret = 0;
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
}
if (ret == 0)
ret = dio->page_errors;
if (ret == 0)
ret = dio->io_error;
if (ret == 0)
ret = transferred;
fs: invalidate page cache after end_io() in dio completion Commit 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO") moved page cache invalidation from iomap_dio_rw() to iomap_dio_complete() for iomap based direct write path, but before the dio->end_io() call, and it re-introdued the bug fixed by commit c771c14baa33 ("iomap: invalidate page caches should be after iomap_dio_complete() in direct write"). I found this because fstests generic/418 started failing on XFS with v4.14-rc3 kernel, which is the regression test for this specific bug. So similarly, fix it by moving dio->end_io() (which does the unwritten extent conversion) before page cache invalidation, to make sure next buffer read reads the final real allocations not unwritten extents. I also add some comments about why should end_io() go first in case we get it wrong again in the future. Note that, there's no such problem in the non-iomap based direct write path, because we didn't remove the page cache invalidation after the ->direct_IO() in generic_file_direct_write() call, but I decided to fix dio_complete() too so we don't leave a landmine there, also be consistent with iomap_dio_complete(). Fixes: 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO") Signed-off-by: Eryu Guan <eguan@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Jan Kara <jack@suse.cz> Reviewed-by: Lukas Czerner <lczerner@redhat.com>
2017-10-13 16:47:46 +00:00
if (dio->end_io) {
// XXX: ki_pos??
err = dio->end_io(dio->iocb, offset, ret, dio->private);
if (err)
ret = err;
}
/*
* Try again to invalidate clean pages which might have been cached by
* non-direct readahead, or faulted in by get_user_pages() if the source
* of the write was an mmap'ed region of the file we're writing. Either
* one is a pretty crazy thing to do, so we don't support it 100%. If
* this invalidation fails, tough, the write still worked...
fs: invalidate page cache after end_io() in dio completion Commit 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO") moved page cache invalidation from iomap_dio_rw() to iomap_dio_complete() for iomap based direct write path, but before the dio->end_io() call, and it re-introdued the bug fixed by commit c771c14baa33 ("iomap: invalidate page caches should be after iomap_dio_complete() in direct write"). I found this because fstests generic/418 started failing on XFS with v4.14-rc3 kernel, which is the regression test for this specific bug. So similarly, fix it by moving dio->end_io() (which does the unwritten extent conversion) before page cache invalidation, to make sure next buffer read reads the final real allocations not unwritten extents. I also add some comments about why should end_io() go first in case we get it wrong again in the future. Note that, there's no such problem in the non-iomap based direct write path, because we didn't remove the page cache invalidation after the ->direct_IO() in generic_file_direct_write() call, but I decided to fix dio_complete() too so we don't leave a landmine there, also be consistent with iomap_dio_complete(). Fixes: 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO") Signed-off-by: Eryu Guan <eguan@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Jan Kara <jack@suse.cz> Reviewed-by: Lukas Czerner <lczerner@redhat.com>
2017-10-13 16:47:46 +00:00
*
* And this page cache invalidation has to be after dio->end_io(), as
* some filesystems convert unwritten extents to real allocations in
* end_io() when necessary, otherwise a racing buffer read would cache
* zeros from unwritten extents.
*/
if (flags & DIO_COMPLETE_INVALIDATE &&
ret > 0 && dio->op == REQ_OP_WRITE &&
dio->inode->i_mapping->nrpages) {
err = invalidate_inode_pages2_range(dio->inode->i_mapping,
offset >> PAGE_SHIFT,
(offset + ret - 1) >> PAGE_SHIFT);
if (err)
dio_warn_stale_pagecache(dio->iocb->ki_filp);
}
inode_dio_end(dio->inode);
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-15 23:05:48 +00:00
if (flags & DIO_COMPLETE_ASYNC) {
/*
* generic_write_sync expects ki_pos to have been updated
* already, but the submission path only does this for
* synchronous I/O.
*/
dio->iocb->ki_pos += transferred;
2018-11-30 15:35:14 +00:00
if (ret > 0 && dio->op == REQ_OP_WRITE)
ret = generic_write_sync(dio->iocb, ret);
dio->iocb->ki_complete(dio->iocb, ret, 0);
}
kmem_cache_free(dio_cache, dio);
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
return ret;
}
static void dio_aio_complete_work(struct work_struct *work)
{
struct dio *dio = container_of(work, struct dio, complete_work);
dio_complete(dio, 0, DIO_COMPLETE_ASYNC | DIO_COMPLETE_INVALIDATE);
}
static blk_status_t dio_bio_complete(struct dio *dio, struct bio *bio);
/*
* Asynchronous IO callback.
*/
static void dio_bio_end_aio(struct bio *bio)
{
struct dio *dio = bio->bi_private;
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
unsigned long remaining;
unsigned long flags;
bool defer_completion = false;
/* cleanup the bio */
dio_bio_complete(dio, bio);
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
spin_lock_irqsave(&dio->bio_lock, flags);
remaining = --dio->refcount;
if (remaining == 1 && dio->waiter)
wake_up_process(dio->waiter);
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
spin_unlock_irqrestore(&dio->bio_lock, flags);
[PATCH] dio: only call aio_complete() after returning -EIOCBQUEUED The only time it is safe to call aio_complete() is when the ->ki_retry function returns -EIOCBQUEUED to the AIO core. direct_io_worker() has historically done this by relying on its caller to translate positive return codes into -EIOCBQUEUED for the aio case. It did this by trying to keep conditionals in sync. direct_io_worker() knew when finished_one_bio() was going to call aio_complete(). It would reverse the test and wait and free the dio in the cases it thought that finished_one_bio() wasn't going to. Not surprisingly, it ended up getting it wrong. 'ret' could be a negative errno from the submission path but it failed to communicate this to finished_one_bio(). direct_io_worker() would return < 0, it's callers wouldn't raise -EIOCBQUEUED, and aio_complete() would be called. In the future finished_one_bio()'s tests wouldn't reflect this and aio_complete() would be called for a second time which can manifest as an oops. The previous cleanups have whittled the sync and async completion paths down to the point where we can collapse them and clearly reassert the invariant that we must only call aio_complete() after returning -EIOCBQUEUED. direct_io_worker() will only return -EIOCBQUEUED when it is not the last to drop the dio refcount and the aio bio completion path will only call aio_complete() when it is the last to drop the dio refcount. direct_io_worker() can ensure that it is the last to drop the reference count by waiting for bios to drain. It does this for sync ops, of course, and for partial dio writes that must fall back to buffered and for aio ops that saw errors during submission. This means that operations that end up waiting, even if they were issued as aio ops, will not call aio_complete() from dio. Instead we return the return code of the operation and let the aio core call aio_complete(). This is purposely done to fix a bug where AIO DIO file extensions would call aio_complete() before their callers have a chance to update i_size. Now that direct_io_worker() is explicitly returning -EIOCBQUEUED its callers no longer have to translate for it. XFS needs to be careful not to free resources that will be used during AIO completion if -EIOCBQUEUED is returned. We maintain the previous behaviour of trying to write fs metadata for O_SYNC aio+dio writes. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:05 +00:00
if (remaining == 0) {
/*
* Defer completion when defer_completion is set or
* when the inode has pages mapped and this is AIO write.
* We need to invalidate those pages because there is a
* chance they contain stale data in the case buffered IO
* went in between AIO submission and completion into the
* same region.
*/
if (dio->result)
defer_completion = dio->defer_completion ||
(dio->op == REQ_OP_WRITE &&
dio->inode->i_mapping->nrpages);
if (defer_completion) {
INIT_WORK(&dio->complete_work, dio_aio_complete_work);
queue_work(dio->inode->i_sb->s_dio_done_wq,
&dio->complete_work);
} else {
dio_complete(dio, 0, DIO_COMPLETE_ASYNC);
}
[PATCH] dio: only call aio_complete() after returning -EIOCBQUEUED The only time it is safe to call aio_complete() is when the ->ki_retry function returns -EIOCBQUEUED to the AIO core. direct_io_worker() has historically done this by relying on its caller to translate positive return codes into -EIOCBQUEUED for the aio case. It did this by trying to keep conditionals in sync. direct_io_worker() knew when finished_one_bio() was going to call aio_complete(). It would reverse the test and wait and free the dio in the cases it thought that finished_one_bio() wasn't going to. Not surprisingly, it ended up getting it wrong. 'ret' could be a negative errno from the submission path but it failed to communicate this to finished_one_bio(). direct_io_worker() would return < 0, it's callers wouldn't raise -EIOCBQUEUED, and aio_complete() would be called. In the future finished_one_bio()'s tests wouldn't reflect this and aio_complete() would be called for a second time which can manifest as an oops. The previous cleanups have whittled the sync and async completion paths down to the point where we can collapse them and clearly reassert the invariant that we must only call aio_complete() after returning -EIOCBQUEUED. direct_io_worker() will only return -EIOCBQUEUED when it is not the last to drop the dio refcount and the aio bio completion path will only call aio_complete() when it is the last to drop the dio refcount. direct_io_worker() can ensure that it is the last to drop the reference count by waiting for bios to drain. It does this for sync ops, of course, and for partial dio writes that must fall back to buffered and for aio ops that saw errors during submission. This means that operations that end up waiting, even if they were issued as aio ops, will not call aio_complete() from dio. Instead we return the return code of the operation and let the aio core call aio_complete(). This is purposely done to fix a bug where AIO DIO file extensions would call aio_complete() before their callers have a chance to update i_size. Now that direct_io_worker() is explicitly returning -EIOCBQUEUED its callers no longer have to translate for it. XFS needs to be careful not to free resources that will be used during AIO completion if -EIOCBQUEUED is returned. We maintain the previous behaviour of trying to write fs metadata for O_SYNC aio+dio writes. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:05 +00:00
}
}
/*
* The BIO completion handler simply queues the BIO up for the process-context
* handler.
*
* During I/O bi_private points at the dio. After I/O, bi_private is used to
* implement a singly-linked list of completed BIOs, at dio->bio_list.
*/
static void dio_bio_end_io(struct bio *bio)
{
struct dio *dio = bio->bi_private;
unsigned long flags;
spin_lock_irqsave(&dio->bio_lock, flags);
bio->bi_private = dio->bio_list;
dio->bio_list = bio;
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
if (--dio->refcount == 1 && dio->waiter)
wake_up_process(dio->waiter);
spin_unlock_irqrestore(&dio->bio_lock, flags);
}
/**
* dio_end_io - handle the end io action for the given bio
* @bio: The direct io bio thats being completed
*
* This is meant to be called by any filesystem that uses their own dio_submit_t
* so that the DIO specific endio actions are dealt with after the filesystem
* has done it's completion work.
*/
void dio_end_io(struct bio *bio)
{
struct dio *dio = bio->bi_private;
if (dio->is_async)
dio_bio_end_aio(bio);
else
dio_bio_end_io(bio);
}
EXPORT_SYMBOL_GPL(dio_end_io);
static inline void
dio_bio_alloc(struct dio *dio, struct dio_submit *sdio,
struct block_device *bdev,
sector_t first_sector, int nr_vecs)
{
struct bio *bio;
/*
* bio_alloc() is guaranteed to return a bio when allowed to sleep and
* we request a valid number of vectors.
*/
bio = bio_alloc(GFP_KERNEL, nr_vecs);
bio_set_dev(bio, bdev);
block: Abstract out bvec iterator Immutable biovecs are going to require an explicit iterator. To implement immutable bvecs, a later patch is going to add a bi_bvec_done member to this struct; for now, this patch effectively just renames things. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Alasdair Kergon <agk@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Boaz Harrosh <bharrosh@panasas.com> Cc: Benny Halevy <bhalevy@tonian.com> Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Chris Mason <chris.mason@fusionio.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jaegeuk Kim <jaegeuk.kim@samsung.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: Joern Engel <joern@logfs.org> Cc: Prasad Joshi <prasadjoshi.linux@gmail.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: KONISHI Ryusuke <konishi.ryusuke@lab.ntt.co.jp> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Ben Myers <bpm@sgi.com> Cc: xfs@oss.sgi.com Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Len Brown <len.brown@intel.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Cc: Ian Campbell <Ian.Campbell@citrix.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchand@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Peng Tao <tao.peng@emc.com> Cc: Andy Adamson <andros@netapp.com> Cc: fanchaoting <fanchaoting@cn.fujitsu.com> Cc: Jie Liu <jeff.liu@oracle.com> Cc: Sunil Mushran <sunil.mushran@gmail.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Namjae Jeon <namjae.jeon@samsung.com> Cc: Pankaj Kumar <pankaj.km@samsung.com> Cc: Dan Magenheimer <dan.magenheimer@oracle.com> Cc: Mel Gorman <mgorman@suse.de>6
2013-10-11 22:44:27 +00:00
bio->bi_iter.bi_sector = first_sector;
bio_set_op_attrs(bio, dio->op, dio->op_flags);
if (dio->is_async)
bio->bi_end_io = dio_bio_end_aio;
else
bio->bi_end_io = dio_bio_end_io;
bio->bi_write_hint = dio->iocb->ki_hint;
sdio->bio = bio;
sdio->logical_offset_in_bio = sdio->cur_page_fs_offset;
}
/*
* In the AIO read case we speculatively dirty the pages before starting IO.
* During IO completion, any of these pages which happen to have been written
* back will be redirtied by bio_check_pages_dirty().
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
*
* bios hold a dio reference between submit_bio and ->end_io.
*/
static inline void dio_bio_submit(struct dio *dio, struct dio_submit *sdio)
{
struct bio *bio = sdio->bio;
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
unsigned long flags;
bio->bi_private = dio;
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
spin_lock_irqsave(&dio->bio_lock, flags);
dio->refcount++;
spin_unlock_irqrestore(&dio->bio_lock, flags);
if (dio->is_async && dio->op == REQ_OP_READ && dio->should_dirty)
bio_set_pages_dirty(bio);
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
dio->bio_disk = bio->bi_disk;
if (sdio->submit_io) {
sdio->submit_io(bio, dio->inode, sdio->logical_offset_in_bio);
dio->bio_cookie = BLK_QC_T_NONE;
} else
dio->bio_cookie = submit_bio(bio);
sdio->bio = NULL;
sdio->boundary = 0;
sdio->logical_offset_in_bio = 0;
}
/*
* Release any resources in case of a failure
*/
static inline void dio_cleanup(struct dio *dio, struct dio_submit *sdio)
{
while (sdio->head < sdio->tail)
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 12:29:47 +00:00
put_page(dio->pages[sdio->head++]);
}
/*
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
* Wait for the next BIO to complete. Remove it and return it. NULL is
* returned once all BIOs have been completed. This must only be called once
* all bios have been issued so that dio->refcount can only decrease. This
* requires that that the caller hold a reference on the dio.
*/
static struct bio *dio_await_one(struct dio *dio)
{
unsigned long flags;
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
struct bio *bio = NULL;
spin_lock_irqsave(&dio->bio_lock, flags);
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
/*
* Wait as long as the list is empty and there are bios in flight. bio
* completion drops the count, maybe adds to the list, and wakes while
* holding the bio_lock so we don't need set_current_state()'s barrier
* and can call it after testing our condition.
*/
while (dio->refcount > 1 && dio->bio_list == NULL) {
__set_current_state(TASK_UNINTERRUPTIBLE);
dio->waiter = current;
spin_unlock_irqrestore(&dio->bio_lock, flags);
if (!(dio->iocb->ki_flags & IOCB_HIPRI) ||
!blk_poll(dio->bio_disk->queue, dio->bio_cookie, true))
io_schedule();
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
/* wake up sets us TASK_RUNNING */
spin_lock_irqsave(&dio->bio_lock, flags);
dio->waiter = NULL;
}
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
if (dio->bio_list) {
bio = dio->bio_list;
dio->bio_list = bio->bi_private;
}
spin_unlock_irqrestore(&dio->bio_lock, flags);
return bio;
}
/*
* Process one completed BIO. No locks are held.
*/
static blk_status_t dio_bio_complete(struct dio *dio, struct bio *bio)
{
blk_status_t err = bio->bi_status;
bool should_dirty = dio->op == REQ_OP_READ && dio->should_dirty;
if (err) {
if (err == BLK_STS_AGAIN && (bio->bi_opf & REQ_NOWAIT))
dio->io_error = -EAGAIN;
else
dio->io_error = -EIO;
}
if (dio->is_async && should_dirty) {
block: fix use-after-free in dio_bio_complete kasan reported the following error when i ran xfstest: [ 701.826854] ================================================================== [ 701.826864] BUG: KASAN: use-after-free in dio_bio_complete+0x41a/0x600 at addr ffff880080b95f94 [ 701.826870] Read of size 4 by task loop2/3874 [ 701.826879] page:ffffea000202e540 count:0 mapcount:0 mapping: (null) index:0x0 [ 701.826890] flags: 0x100000000000000() [ 701.826895] page dumped because: kasan: bad access detected [ 701.826904] CPU: 3 PID: 3874 Comm: loop2 Tainted: G B W L 4.5.0-rc1-next-20160129 #83 [ 701.826910] Hardware name: LENOVO 23205NG/23205NG, BIOS G2ET95WW (2.55 ) 07/09/2013 [ 701.826917] ffff88008fadf800 ffff88008fadf758 ffffffff81ca67bb 0000000041b58ab3 [ 701.826941] ffffffff830d1e74 ffffffff81ca6724 ffff88008fadf748 ffffffff8161c05c [ 701.826963] 0000000000000282 ffff88008fadf800 ffffed0010172bf2 ffffea000202e540 [ 701.826987] Call Trace: [ 701.826997] [<ffffffff81ca67bb>] dump_stack+0x97/0xdc [ 701.827005] [<ffffffff81ca6724>] ? _atomic_dec_and_lock+0xc4/0xc4 [ 701.827014] [<ffffffff8161c05c>] ? __dump_page+0x32c/0x490 [ 701.827023] [<ffffffff816b0d03>] kasan_report_error+0x5f3/0x8b0 [ 701.827033] [<ffffffff817c302a>] ? dio_bio_complete+0x41a/0x600 [ 701.827040] [<ffffffff816b1119>] __asan_report_load4_noabort+0x59/0x80 [ 701.827048] [<ffffffff817c302a>] ? dio_bio_complete+0x41a/0x600 [ 701.827053] [<ffffffff817c302a>] dio_bio_complete+0x41a/0x600 [ 701.827057] [<ffffffff81bd19c8>] ? blk_queue_exit+0x108/0x270 [ 701.827060] [<ffffffff817c32b0>] dio_bio_end_aio+0xa0/0x4d0 [ 701.827063] [<ffffffff817c3210>] ? dio_bio_complete+0x600/0x600 [ 701.827067] [<ffffffff81bd2806>] ? blk_account_io_completion+0x316/0x5d0 [ 701.827070] [<ffffffff81bafe89>] bio_endio+0x79/0x200 [ 701.827074] [<ffffffff81bd2c9f>] blk_update_request+0x1df/0xc50 [ 701.827078] [<ffffffff81c02c27>] blk_mq_end_request+0x57/0x120 [ 701.827081] [<ffffffff81c03670>] __blk_mq_complete_request+0x310/0x590 [ 701.827084] [<ffffffff812348d8>] ? set_next_entity+0x2f8/0x2ed0 [ 701.827088] [<ffffffff8124b34d>] ? put_prev_entity+0x22d/0x2a70 [ 701.827091] [<ffffffff81c0394b>] blk_mq_complete_request+0x5b/0x80 [ 701.827094] [<ffffffff821e2a33>] loop_queue_work+0x273/0x19d0 [ 701.827098] [<ffffffff811f6578>] ? finish_task_switch+0x1c8/0x8e0 [ 701.827101] [<ffffffff8129d058>] ? trace_hardirqs_on_caller+0x18/0x6c0 [ 701.827104] [<ffffffff821e27c0>] ? lo_read_simple+0x890/0x890 [ 701.827108] [<ffffffff8129dd60>] ? debug_check_no_locks_freed+0x350/0x350 [ 701.827111] [<ffffffff811f63b0>] ? __hrtick_start+0x130/0x130 [ 701.827115] [<ffffffff82a0c8f6>] ? __schedule+0x936/0x20b0 [ 701.827118] [<ffffffff811dd6bd>] ? kthread_worker_fn+0x3ed/0x8d0 [ 701.827121] [<ffffffff811dd4ed>] ? kthread_worker_fn+0x21d/0x8d0 [ 701.827125] [<ffffffff8129d058>] ? trace_hardirqs_on_caller+0x18/0x6c0 [ 701.827128] [<ffffffff811dd57f>] kthread_worker_fn+0x2af/0x8d0 [ 701.827132] [<ffffffff811dd2d0>] ? __init_kthread_worker+0x170/0x170 [ 701.827135] [<ffffffff82a1ea46>] ? _raw_spin_unlock_irqrestore+0x36/0x60 [ 701.827138] [<ffffffff811dd2d0>] ? __init_kthread_worker+0x170/0x170 [ 701.827141] [<ffffffff811dd2d0>] ? __init_kthread_worker+0x170/0x170 [ 701.827144] [<ffffffff811dd00b>] kthread+0x24b/0x3a0 [ 701.827148] [<ffffffff811dcdc0>] ? kthread_create_on_node+0x4c0/0x4c0 [ 701.827151] [<ffffffff8129d70d>] ? trace_hardirqs_on+0xd/0x10 [ 701.827155] [<ffffffff8116d41d>] ? do_group_exit+0xdd/0x350 [ 701.827158] [<ffffffff811dcdc0>] ? kthread_create_on_node+0x4c0/0x4c0 [ 701.827161] [<ffffffff82a1f52f>] ret_from_fork+0x3f/0x70 [ 701.827165] [<ffffffff811dcdc0>] ? kthread_create_on_node+0x4c0/0x4c0 [ 701.827167] Memory state around the buggy address: [ 701.827170] ffff880080b95e80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff [ 701.827172] ffff880080b95f00: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff [ 701.827175] >ffff880080b95f80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff [ 701.827177] ^ [ 701.827179] ffff880080b96000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff [ 701.827182] ffff880080b96080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff [ 701.827183] ================================================================== The problem is that bio_check_pages_dirty calls bio_put, so we must not access bio fields after bio_check_pages_dirty. Fixes: 9b81c842355ac96097ba ("block: don't access bio->bi_error after bio_put()"). Signed-off-by: Mike Krinkin <krinkin.m.u@gmail.com> Cc: stable@vger.kernel.org Signed-off-by: Jens Axboe <axboe@fb.com>
2016-01-30 16:09:59 +00:00
bio_check_pages_dirty(bio); /* transfers ownership */
} else {
bio_release_pages(bio, should_dirty);
bio_put(bio);
}
block: don't access bio->bi_error after bio_put() Commit 4246a0b6 ("block: add a bi_error field to struct bio") has added a few dereferences of 'bio' after a call to bio_put(). This causes use-after-frees such as: [521120.719695] BUG: KASan: use after free in dio_bio_complete+0x2b3/0x320 at addr ffff880f36b38714 [521120.720638] Read of size 4 by task mount.ocfs2/9644 [521120.721212] ============================================================================= [521120.722056] BUG kmalloc-256 (Not tainted): kasan: bad access detected [521120.722968] ----------------------------------------------------------------------------- [521120.722968] [521120.723915] Disabling lock debugging due to kernel taint [521120.724539] INFO: Slab 0xffffea003cdace00 objects=32 used=25 fp=0xffff880f36b38600 flags=0x46fffff80004080 [521120.726037] INFO: Object 0xffff880f36b38700 @offset=1792 fp=0xffff880f36b38800 [521120.726037] [521120.726974] Bytes b4 ffff880f36b386f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.727898] Object ffff880f36b38700: 00 88 b3 36 0f 88 ff ff 00 00 d8 de 0b 88 ff ff ...6............ [521120.728822] Object ffff880f36b38710: 02 00 00 f0 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.729705] Object ffff880f36b38720: 01 00 00 00 00 00 00 00 00 00 00 00 01 00 00 00 ................ [521120.730623] Object ffff880f36b38730: 00 00 00 00 00 00 00 00 01 00 00 00 00 02 00 00 ................ [521120.731621] Object ffff880f36b38740: 00 02 00 00 01 00 00 00 d0 f7 87 ad ff ff ff ff ................ [521120.732776] Object ffff880f36b38750: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.733640] Object ffff880f36b38760: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.734508] Object ffff880f36b38770: 01 00 03 00 01 00 00 00 88 87 b3 36 0f 88 ff ff ...........6.... [521120.735385] Object ffff880f36b38780: 00 73 22 ad 02 88 ff ff 40 13 e0 3c 00 ea ff ff .s".....@..<.... [521120.736667] Object ffff880f36b38790: 00 02 00 00 00 04 00 00 00 00 00 00 00 00 00 00 ................ [521120.737596] Object ffff880f36b387a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.738524] Object ffff880f36b387b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.739388] Object ffff880f36b387c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.740277] Object ffff880f36b387d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.741187] Object ffff880f36b387e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.742233] Object ffff880f36b387f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [521120.743229] CPU: 41 PID: 9644 Comm: mount.ocfs2 Tainted: G B 4.2.0-rc6-next-20150810-sasha-00039-gf909086 #2420 [521120.744274] ffff880f36b38000 ffff880d89c8f638 ffffffffb6e9ba8a ffff880101c0e5c0 [521120.745025] ffff880d89c8f668 ffffffffad76a313 ffff880101c0e5c0 ffffea003cdace00 [521120.745908] ffff880f36b38700 ffff880f36b38798 ffff880d89c8f690 ffffffffad772854 [521120.747063] Call Trace: [521120.747520] dump_stack (lib/dump_stack.c:52) [521120.748053] print_trailer (mm/slub.c:653) [521120.748582] object_err (mm/slub.c:660) [521120.749079] kasan_report_error (include/linux/kasan.h:20 mm/kasan/report.c:152 mm/kasan/report.c:194) [521120.750834] __asan_report_load4_noabort (mm/kasan/report.c:250) [521120.753580] dio_bio_complete (fs/direct-io.c:478) [521120.755752] do_blockdev_direct_IO (fs/direct-io.c:494 fs/direct-io.c:1291) [521120.759765] __blockdev_direct_IO (fs/direct-io.c:1322) [521120.761658] blkdev_direct_IO (fs/block_dev.c:162) [521120.762993] generic_file_read_iter (mm/filemap.c:1738) [521120.767405] blkdev_read_iter (fs/block_dev.c:1649) [521120.768556] __vfs_read (fs/read_write.c:423 fs/read_write.c:434) [521120.772126] vfs_read (fs/read_write.c:454) [521120.773118] SyS_pread64 (fs/read_write.c:607 fs/read_write.c:594) [521120.776062] entry_SYSCALL_64_fastpath (arch/x86/entry/entry_64.S:186) [521120.777375] Memory state around the buggy address: [521120.778118] ffff880f36b38600: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [521120.779211] ffff880f36b38680: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [521120.780315] >ffff880f36b38700: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [521120.781465] ^ [521120.782083] ffff880f36b38780: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [521120.783717] ffff880f36b38800: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [521120.784818] ================================================================== This patch fixes a few of those places that I caught while auditing the patch, but the original patch should be audited further for more occurences of this issue since I'm not too familiar with the code. Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2015-08-10 23:05:18 +00:00
return err;
}
/*
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
* Wait on and process all in-flight BIOs. This must only be called once
* all bios have been issued so that the refcount can only decrease.
* This just waits for all bios to make it through dio_bio_complete. IO
* errors are propagated through dio->io_error and should be propagated via
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
* dio_complete().
*/
[PATCH] dio: centralize completion in dio_complete() There have been a lot of bugs recently due to the way direct_io_worker() tries to decide how to finish direct IO operations. In the worst examples it has failed to call aio_complete() at all (hang) or called it too many times (oops). This set of patches cleans up the completion phase with the goal of removing the complexity that lead to these bugs. We end up with one path that calculates the result of the operation after all off the bios have completed. We decide when to generate a result of the operation using that path based on the final release of a refcount on the dio structure. I tried to progress towards the final state in steps that were relatively easy to understand. Each step should compile but I only tested the final result of having all the patches applied. I've tested these on low end PC drives with aio-stress, the direct IO tests I could manage to get running in LTP, orasim, and some home-brew functional tests. In http://lkml.org/lkml/2006/9/21/103 IBM reports success with ext2 and ext3 running DIO LTP tests. They found that XFS bug which has since been addressed in the patch series. This patch: The mechanics which decide the result of a direct IO operation were duplicated in the sync and async paths. The async path didn't check page_errors which can manifest as silently returning success when the final pointer in an operation faults and its matching file region is filled with zeros. The sync path and async path differed in whether they passed errors to the caller's dio->end_io operation. The async path was passing errors to it which trips an assertion in XFS, though it is apparently harmless. This centralizes the completion phase of dio ops in one place. AIO will now return EFAULT consistently and all paths fall back to the previously sync behaviour of passing the number of bytes 'transferred' to the dio->end_io callback, regardless of errors. dio_await_completion() doesn't have to propogate EIO from non-uptodate bios now that it's being propogated through dio_complete() via dio->io_error. This lets it return void which simplifies its sole caller. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:54 +00:00
static void dio_await_completion(struct dio *dio)
{
[PATCH] dio: formalize bio counters as a dio reference count Previously we had two confusing counts of bio progress. 'bio_count' was decremented as bios were processed and freed by the dio core. It was used to indicate final completion of the dio operation. 'bios_in_flight' reflected how many bios were between submit_bio() and bio->end_io. It was used by the sync path to decide when to wake up and finish completing bios and was ignored by the async path. This patch collapses the two notions into one notion of a dio reference count. bios hold a dio reference when they're between submit_bio and bio->end_io. Since bios_in_flight was only used in the sync path it is now equivalent to dio->refcount - 1 which accounts for direct_io_worker() holding a reference for the duration of the operation. dio_bio_complete() -> finished_one_bio() was called from the sync path after finding bios on the list that the bio->end_io function had deposited. finished_one_bio() can not drop the dio reference on behalf of these bios now because bio->end_io already has. The is_async test in finished_one_bio() meant that it never actually did anything other than drop the bio_count for sync callers. So we remove its refcount decrement, don't call it from dio_bio_complete(), and hoist its call up into the async dio_bio_complete() caller after an explicit refcount decrement. It is renamed dio_complete_aio() to reflect the remaining work it actually does. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:20:59 +00:00
struct bio *bio;
do {
bio = dio_await_one(dio);
if (bio)
dio_bio_complete(dio, bio);
} while (bio);
}
/*
* A really large O_DIRECT read or write can generate a lot of BIOs. So
* to keep the memory consumption sane we periodically reap any completed BIOs
* during the BIO generation phase.
*
* This also helps to limit the peak amount of pinned userspace memory.
*/
static inline int dio_bio_reap(struct dio *dio, struct dio_submit *sdio)
{
int ret = 0;
if (sdio->reap_counter++ >= 64) {
while (dio->bio_list) {
unsigned long flags;
struct bio *bio;
int ret2;
spin_lock_irqsave(&dio->bio_lock, flags);
bio = dio->bio_list;
dio->bio_list = bio->bi_private;
spin_unlock_irqrestore(&dio->bio_lock, flags);
ret2 = blk_status_to_errno(dio_bio_complete(dio, bio));
if (ret == 0)
ret = ret2;
}
sdio->reap_counter = 0;
}
return ret;
}
/*
* Create workqueue for deferred direct IO completions. We allocate the
* workqueue when it's first needed. This avoids creating workqueue for
* filesystems that don't need it and also allows us to create the workqueue
* late enough so the we can include s_id in the name of the workqueue.
*/
int sb_init_dio_done_wq(struct super_block *sb)
{
struct workqueue_struct *old;
struct workqueue_struct *wq = alloc_workqueue("dio/%s",
WQ_MEM_RECLAIM, 0,
sb->s_id);
if (!wq)
return -ENOMEM;
/*
* This has to be atomic as more DIOs can race to create the workqueue
*/
old = cmpxchg(&sb->s_dio_done_wq, NULL, wq);
/* Someone created workqueue before us? Free ours... */
if (old)
destroy_workqueue(wq);
return 0;
}
static int dio_set_defer_completion(struct dio *dio)
{
struct super_block *sb = dio->inode->i_sb;
if (dio->defer_completion)
return 0;
dio->defer_completion = true;
if (!sb->s_dio_done_wq)
return sb_init_dio_done_wq(sb);
return 0;
}
/*
* Call into the fs to map some more disk blocks. We record the current number
* of available blocks at sdio->blocks_available. These are in units of the
* fs blocksize, i_blocksize(inode).
*
* The fs is allowed to map lots of blocks at once. If it wants to do that,
* it uses the passed inode-relative block number as the file offset, as usual.
*
* get_block() is passed the number of i_blkbits-sized blocks which direct_io
* has remaining to do. The fs should not map more than this number of blocks.
*
* If the fs has mapped a lot of blocks, it should populate bh->b_size to
* indicate how much contiguous disk space has been made available at
* bh->b_blocknr.
*
* If *any* of the mapped blocks are new, then the fs must set buffer_new().
* This isn't very efficient...
*
* In the case of filesystem holes: the fs may return an arbitrarily-large
* hole by returning an appropriate value in b_size and by clearing
* buffer_mapped(). However the direct-io code will only process holes one
* block at a time - it will repeatedly call get_block() as it walks the hole.
*/
static int get_more_blocks(struct dio *dio, struct dio_submit *sdio,
struct buffer_head *map_bh)
{
int ret;
sector_t fs_startblk; /* Into file, in filesystem-sized blocks */
sector_t fs_endblk; /* Into file, in filesystem-sized blocks */
unsigned long fs_count; /* Number of filesystem-sized blocks */
int create;
unsigned int i_blkbits = sdio->blkbits + sdio->blkfactor;
loff_t i_size;
/*
* If there was a memory error and we've overwritten all the
* mapped blocks then we can now return that memory error
*/
ret = dio->page_errors;
if (ret == 0) {
BUG_ON(sdio->block_in_file >= sdio->final_block_in_request);
fs_startblk = sdio->block_in_file >> sdio->blkfactor;
fs_endblk = (sdio->final_block_in_request - 1) >>
sdio->blkfactor;
fs_count = fs_endblk - fs_startblk + 1;
map_bh->b_state = 0;
map_bh->b_size = fs_count << i_blkbits;
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
/*
direct-io: fix direct write stale data exposure from concurrent buffered read Currently direct writes inside i_size on a DIO_SKIP_HOLES filesystem are not allowed to allocate blocks(get_more_blocks() sets 'create' to 0 before calling get_block() callback), if it's a sparse file, direct writes fall back to buffered writes to avoid stale data exposure from concurrent buffered read. But there're two cases that can result in stale data exposure are not correctly detected. 1. The detection for "writing inside i_size" is not sufficient, writes can be treated as "extending writes" wrongly. For example, direct write 1FSB (file system block) to a 1FSB sparse file on ext2/3/4, starting from offset 0, in this case it's writing inside i_size, but 'create' is non-zero, because 'block_in_file' and '(i_size_read(inode) >> blkbits' are both zero. 2. Direct writes starting from or beyong i_size (not inside i_size) also could trigger block allocation and expose stale data. For example, consider a sparse file with i_size of 2k, and a write to offset 2k or 3k into the file, with a filesystem block size of 4k. (Thanks to Jeff Moyer for pointing this case out in his review.) The first problem can be demostrated by running ltp-aiodio test ADSP045 many times. When testing on extN filesystems, I see test failures occasionally, buffered read could read non-zero (stale) data. ADSP045: dio_sparse -a 4k -w 4k -s 2k -n 1 dio_sparse 0 TINFO : Dirtying free blocks dio_sparse 0 TINFO : Starting I/O tests non zero buffer at buf[0] => 0xffffffaa,ffffffaa,ffffffaa,ffffffaa non-zero read at offset 0 dio_sparse 0 TINFO : Killing childrens(s) dio_sparse 1 TFAIL : dio_sparse.c:191: 1 children(s) exited abnormally The second problem can also be reproduced easily by a hacked dio_sparse program, which accepts an option to specify the write offset. What we should really do is to disable block allocation for writes that could result in filling holes inside i_size. Link: http://lkml.kernel.org/r/1463156728-13357-1-git-send-email-guaneryu@gmail.com Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Eryu Guan <guaneryu@gmail.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-27 21:27:18 +00:00
* For writes that could fill holes inside i_size on a
* DIO_SKIP_HOLES filesystem we forbid block creations: only
* overwrites are permitted. We will return early to the caller
* once we see an unmapped buffer head returned, and the caller
* will fall back to buffered I/O.
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
*
* Otherwise the decision is left to the get_blocks method,
* which may decide to handle it or also return an unmapped
* buffer head.
*/
create = dio->op == REQ_OP_WRITE;
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
if (dio->flags & DIO_SKIP_HOLES) {
i_size = i_size_read(dio->inode);
if (i_size && fs_startblk <= (i_size - 1) >> i_blkbits)
create = 0;
}
ret = (*sdio->get_block)(dio->inode, fs_startblk,
map_bh, create);
/* Store for completion */
dio->private = map_bh->b_private;
if (ret == 0 && buffer_defer_completion(map_bh))
ret = dio_set_defer_completion(dio);
}
return ret;
}
/*
* There is no bio. Make one now.
*/
static inline int dio_new_bio(struct dio *dio, struct dio_submit *sdio,
sector_t start_sector, struct buffer_head *map_bh)
{
sector_t sector;
int ret, nr_pages;
ret = dio_bio_reap(dio, sdio);
if (ret)
goto out;
sector = start_sector << (sdio->blkbits - 9);
nr_pages = min(sdio->pages_in_io, BIO_MAX_PAGES);
BUG_ON(nr_pages <= 0);
dio_bio_alloc(dio, sdio, map_bh->b_bdev, sector, nr_pages);
sdio->boundary = 0;
out:
return ret;
}
/*
* Attempt to put the current chunk of 'cur_page' into the current BIO. If
* that was successful then update final_block_in_bio and take a ref against
* the just-added page.
*
* Return zero on success. Non-zero means the caller needs to start a new BIO.
*/
static inline int dio_bio_add_page(struct dio_submit *sdio)
{
int ret;
ret = bio_add_page(sdio->bio, sdio->cur_page,
sdio->cur_page_len, sdio->cur_page_offset);
if (ret == sdio->cur_page_len) {
/*
* Decrement count only, if we are done with this page
*/
if ((sdio->cur_page_len + sdio->cur_page_offset) == PAGE_SIZE)
sdio->pages_in_io--;
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 12:29:47 +00:00
get_page(sdio->cur_page);
sdio->final_block_in_bio = sdio->cur_page_block +
(sdio->cur_page_len >> sdio->blkbits);
ret = 0;
} else {
ret = 1;
}
return ret;
}
/*
* Put cur_page under IO. The section of cur_page which is described by
* cur_page_offset,cur_page_len is put into a BIO. The section of cur_page
* starts on-disk at cur_page_block.
*
* We take a ref against the page here (on behalf of its presence in the bio).
*
* The caller of this function is responsible for removing cur_page from the
* dio, and for dropping the refcount which came from that presence.
*/
static inline int dio_send_cur_page(struct dio *dio, struct dio_submit *sdio,
struct buffer_head *map_bh)
{
int ret = 0;
if (sdio->bio) {
loff_t cur_offset = sdio->cur_page_fs_offset;
loff_t bio_next_offset = sdio->logical_offset_in_bio +
block: Abstract out bvec iterator Immutable biovecs are going to require an explicit iterator. To implement immutable bvecs, a later patch is going to add a bi_bvec_done member to this struct; for now, this patch effectively just renames things. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Alasdair Kergon <agk@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Boaz Harrosh <bharrosh@panasas.com> Cc: Benny Halevy <bhalevy@tonian.com> Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Chris Mason <chris.mason@fusionio.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jaegeuk Kim <jaegeuk.kim@samsung.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: Joern Engel <joern@logfs.org> Cc: Prasad Joshi <prasadjoshi.linux@gmail.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: KONISHI Ryusuke <konishi.ryusuke@lab.ntt.co.jp> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Ben Myers <bpm@sgi.com> Cc: xfs@oss.sgi.com Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Len Brown <len.brown@intel.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Cc: Ian Campbell <Ian.Campbell@citrix.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchand@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Peng Tao <tao.peng@emc.com> Cc: Andy Adamson <andros@netapp.com> Cc: fanchaoting <fanchaoting@cn.fujitsu.com> Cc: Jie Liu <jeff.liu@oracle.com> Cc: Sunil Mushran <sunil.mushran@gmail.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Namjae Jeon <namjae.jeon@samsung.com> Cc: Pankaj Kumar <pankaj.km@samsung.com> Cc: Dan Magenheimer <dan.magenheimer@oracle.com> Cc: Mel Gorman <mgorman@suse.de>6
2013-10-11 22:44:27 +00:00
sdio->bio->bi_iter.bi_size;
/*
* See whether this new request is contiguous with the old.
*
* Btrfs cannot handle having logically non-contiguous requests
* submitted. For example if you have
*
* Logical: [0-4095][HOLE][8192-12287]
* Physical: [0-4095] [4096-8191]
*
* We cannot submit those pages together as one BIO. So if our
* current logical offset in the file does not equal what would
* be the next logical offset in the bio, submit the bio we
* have.
*/
if (sdio->final_block_in_bio != sdio->cur_page_block ||
cur_offset != bio_next_offset)
dio_bio_submit(dio, sdio);
}
if (sdio->bio == NULL) {
ret = dio_new_bio(dio, sdio, sdio->cur_page_block, map_bh);
if (ret)
goto out;
}
if (dio_bio_add_page(sdio) != 0) {
dio_bio_submit(dio, sdio);
ret = dio_new_bio(dio, sdio, sdio->cur_page_block, map_bh);
if (ret == 0) {
ret = dio_bio_add_page(sdio);
BUG_ON(ret != 0);
}
}
out:
return ret;
}
/*
* An autonomous function to put a chunk of a page under deferred IO.
*
* The caller doesn't actually know (or care) whether this piece of page is in
* a BIO, or is under IO or whatever. We just take care of all possible
* situations here. The separation between the logic of do_direct_IO() and
* that of submit_page_section() is important for clarity. Please don't break.
*
* The chunk of page starts on-disk at blocknr.
*
* We perform deferred IO, by recording the last-submitted page inside our
* private part of the dio structure. If possible, we just expand the IO
* across that page here.
*
* If that doesn't work out then we put the old page into the bio and add this
* page to the dio instead.
*/
static inline int
submit_page_section(struct dio *dio, struct dio_submit *sdio, struct page *page,
unsigned offset, unsigned len, sector_t blocknr,
struct buffer_head *map_bh)
{
int ret = 0;
if (dio->op == REQ_OP_WRITE) {
/*
* Read accounting is performed in submit_bio()
*/
task_io_account_write(len);
}
/*
* Can we just grow the current page's presence in the dio?
*/
if (sdio->cur_page == page &&
sdio->cur_page_offset + sdio->cur_page_len == offset &&
sdio->cur_page_block +
(sdio->cur_page_len >> sdio->blkbits) == blocknr) {
sdio->cur_page_len += len;
goto out;
}
/*
* If there's a deferred page already there then send it.
*/
if (sdio->cur_page) {
ret = dio_send_cur_page(dio, sdio, map_bh);
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 12:29:47 +00:00
put_page(sdio->cur_page);
sdio->cur_page = NULL;
if (ret)
return ret;
}
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 12:29:47 +00:00
get_page(page); /* It is in dio */
sdio->cur_page = page;
sdio->cur_page_offset = offset;
sdio->cur_page_len = len;
sdio->cur_page_block = blocknr;
sdio->cur_page_fs_offset = sdio->block_in_file << sdio->blkbits;
out:
/*
* If sdio->boundary then we want to schedule the IO now to
* avoid metadata seeks.
*/
if (sdio->boundary) {
ret = dio_send_cur_page(dio, sdio, map_bh);
if (sdio->bio)
dio_bio_submit(dio, sdio);
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 12:29:47 +00:00
put_page(sdio->cur_page);
sdio->cur_page = NULL;
}
return ret;
}
/*
* If we are not writing the entire block and get_block() allocated
* the block for us, we need to fill-in the unused portion of the
* block with zeros. This happens only if user-buffer, fileoffset or
* io length is not filesystem block-size multiple.
*
* `end' is zero if we're doing the start of the IO, 1 at the end of the
* IO.
*/
static inline void dio_zero_block(struct dio *dio, struct dio_submit *sdio,
int end, struct buffer_head *map_bh)
{
unsigned dio_blocks_per_fs_block;
unsigned this_chunk_blocks; /* In dio_blocks */
unsigned this_chunk_bytes;
struct page *page;
sdio->start_zero_done = 1;
if (!sdio->blkfactor || !buffer_new(map_bh))
return;
dio_blocks_per_fs_block = 1 << sdio->blkfactor;
this_chunk_blocks = sdio->block_in_file & (dio_blocks_per_fs_block - 1);
if (!this_chunk_blocks)
return;
/*
* We need to zero out part of an fs block. It is either at the
* beginning or the end of the fs block.
*/
if (end)
this_chunk_blocks = dio_blocks_per_fs_block - this_chunk_blocks;
this_chunk_bytes = this_chunk_blocks << sdio->blkbits;
remove ZERO_PAGE The commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note A last caveat: the ZERO_PAGE is now refcounted and managed with rmap (and thus mapcounted and count towards shared rss). These writes to the struct page could cause excessive cacheline bouncing on big systems. There are a number of ways this could be addressed if it is an issue. And indeed this cacheline bouncing has shown up on large SGI systems. There was a situation where an Altix system was essentially livelocked tearing down ZERO_PAGE pagetables when an HPC app aborted during startup. This situation can be avoided in userspace, but it does highlight the potential scalability problem with refcounting ZERO_PAGE, and corner cases where it can really hurt (we don't want the system to livelock!). There are several broad ways to fix this problem: 1. add back some special casing to avoid refcounting ZERO_PAGE 2. per-node or per-cpu ZERO_PAGES 3. remove the ZERO_PAGE completely I will argue for 3. The others should also fix the problem, but they result in more complex code than does 3, with little or no real benefit that I can see. Why? Inserting a ZERO_PAGE for anonymous read faults appears to be a false optimisation: if an application is performance critical, it would not be doing many read faults of new memory, or at least it could be expected to write to that memory soon afterwards. If cache or memory use is critical, it should not be working with a significant number of ZERO_PAGEs anyway (a more compact representation of zeroes should be used). As a sanity check -- mesuring on my desktop system, there are never many mappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not increase much without it. When running a make -j4 kernel compile on my dual core system, there are about 1,000 mappings to the ZERO_PAGE created per second, but about 1,000 ZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second is torn down without being COWed). So removing ZERO_PAGE will save 1,000 page faults per second when running kbuild, while keeping it only saves less than 1 page clearing operation per second. 1 page clear is cheaper than a thousand faults, presumably, so there isn't an obvious loss. Neither the logical argument nor these basic tests give a guarantee of no regressions. However, this is a reasonable opportunity to try to remove the ZERO_PAGE from the pagefault path. If it is found to cause regressions, we can reintroduce it and just avoid refcounting it. The /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don't see much use to them except on benchmarks. All other users of ZERO_PAGE are converted just to use ZERO_PAGE(0) for simplicity. We can look at replacing them all and maybe ripping out ZERO_PAGE completely when we are more satisfied with this solution. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus "snif" Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:24:40 +00:00
page = ZERO_PAGE(0);
if (submit_page_section(dio, sdio, page, 0, this_chunk_bytes,
sdio->next_block_for_io, map_bh))
return;
sdio->next_block_for_io += this_chunk_blocks;
}
/*
* Walk the user pages, and the file, mapping blocks to disk and generating
* a sequence of (page,offset,len,block) mappings. These mappings are injected
* into submit_page_section(), which takes care of the next stage of submission
*
* Direct IO against a blockdev is different from a file. Because we can
* happily perform page-sized but 512-byte aligned IOs. It is important that
* blockdev IO be able to have fine alignment and large sizes.
*
* So what we do is to permit the ->get_block function to populate bh.b_size
* with the size of IO which is permitted at this offset and this i_blkbits.
*
* For best results, the blockdev should be set up with 512-byte i_blkbits and
* it should set b_size to PAGE_SIZE or more inside get_block(). This gives
* fine alignment but still allows this function to work in PAGE_SIZE units.
*/
static int do_direct_IO(struct dio *dio, struct dio_submit *sdio,
struct buffer_head *map_bh)
{
const unsigned blkbits = sdio->blkbits;
const unsigned i_blkbits = blkbits + sdio->blkfactor;
int ret = 0;
while (sdio->block_in_file < sdio->final_block_in_request) {
struct page *page;
size_t from, to;
page = dio_get_page(dio, sdio);
if (IS_ERR(page)) {
ret = PTR_ERR(page);
goto out;
}
from = sdio->head ? 0 : sdio->from;
to = (sdio->head == sdio->tail - 1) ? sdio->to : PAGE_SIZE;
sdio->head++;
while (from < to) {
unsigned this_chunk_bytes; /* # of bytes mapped */
unsigned this_chunk_blocks; /* # of blocks */
unsigned u;
if (sdio->blocks_available == 0) {
/*
* Need to go and map some more disk
*/
unsigned long blkmask;
unsigned long dio_remainder;
ret = get_more_blocks(dio, sdio, map_bh);
if (ret) {
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 12:29:47 +00:00
put_page(page);
goto out;
}
if (!buffer_mapped(map_bh))
goto do_holes;
sdio->blocks_available =
map_bh->b_size >> blkbits;
sdio->next_block_for_io =
map_bh->b_blocknr << sdio->blkfactor;
if (buffer_new(map_bh)) {
clean_bdev_aliases(
map_bh->b_bdev,
map_bh->b_blocknr,
map_bh->b_size >> i_blkbits);
}
if (!sdio->blkfactor)
goto do_holes;
blkmask = (1 << sdio->blkfactor) - 1;
dio_remainder = (sdio->block_in_file & blkmask);
/*
* If we are at the start of IO and that IO
* starts partway into a fs-block,
* dio_remainder will be non-zero. If the IO
* is a read then we can simply advance the IO
* cursor to the first block which is to be
* read. But if the IO is a write and the
* block was newly allocated we cannot do that;
* the start of the fs block must be zeroed out
* on-disk
*/
if (!buffer_new(map_bh))
sdio->next_block_for_io += dio_remainder;
sdio->blocks_available -= dio_remainder;
}
do_holes:
/* Handle holes */
if (!buffer_mapped(map_bh)) {
loff_t i_size_aligned;
/* AKPM: eargh, -ENOTBLK is a hack */
if (dio->op == REQ_OP_WRITE) {
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 12:29:47 +00:00
put_page(page);
return -ENOTBLK;
}
/*
* Be sure to account for a partial block as the
* last block in the file
*/
i_size_aligned = ALIGN(i_size_read(dio->inode),
1 << blkbits);
if (sdio->block_in_file >=
i_size_aligned >> blkbits) {
/* We hit eof */
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 12:29:47 +00:00
put_page(page);
goto out;
}
zero_user(page, from, 1 << blkbits);
sdio->block_in_file++;
from += 1 << blkbits;
dio->result += 1 << blkbits;
goto next_block;
}
/*
* If we're performing IO which has an alignment which
* is finer than the underlying fs, go check to see if
* we must zero out the start of this block.
*/
if (unlikely(sdio->blkfactor && !sdio->start_zero_done))
dio_zero_block(dio, sdio, 0, map_bh);
/*
* Work out, in this_chunk_blocks, how much disk we
* can add to this page
*/
this_chunk_blocks = sdio->blocks_available;
u = (to - from) >> blkbits;
if (this_chunk_blocks > u)
this_chunk_blocks = u;
u = sdio->final_block_in_request - sdio->block_in_file;
if (this_chunk_blocks > u)
this_chunk_blocks = u;
this_chunk_bytes = this_chunk_blocks << blkbits;
BUG_ON(this_chunk_bytes == 0);
if (this_chunk_blocks == sdio->blocks_available)
sdio->boundary = buffer_boundary(map_bh);
ret = submit_page_section(dio, sdio, page,
from,
this_chunk_bytes,
sdio->next_block_for_io,
map_bh);
if (ret) {
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 12:29:47 +00:00
put_page(page);
goto out;
}
sdio->next_block_for_io += this_chunk_blocks;
sdio->block_in_file += this_chunk_blocks;
from += this_chunk_bytes;
dio->result += this_chunk_bytes;
sdio->blocks_available -= this_chunk_blocks;
next_block:
BUG_ON(sdio->block_in_file > sdio->final_block_in_request);
if (sdio->block_in_file == sdio->final_block_in_request)
break;
}
/* Drop the ref which was taken in get_user_pages() */
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 12:29:47 +00:00
put_page(page);
}
out:
return ret;
}
static inline int drop_refcount(struct dio *dio)
{
int ret2;
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
unsigned long flags;
[PATCH] dio: only call aio_complete() after returning -EIOCBQUEUED The only time it is safe to call aio_complete() is when the ->ki_retry function returns -EIOCBQUEUED to the AIO core. direct_io_worker() has historically done this by relying on its caller to translate positive return codes into -EIOCBQUEUED for the aio case. It did this by trying to keep conditionals in sync. direct_io_worker() knew when finished_one_bio() was going to call aio_complete(). It would reverse the test and wait and free the dio in the cases it thought that finished_one_bio() wasn't going to. Not surprisingly, it ended up getting it wrong. 'ret' could be a negative errno from the submission path but it failed to communicate this to finished_one_bio(). direct_io_worker() would return < 0, it's callers wouldn't raise -EIOCBQUEUED, and aio_complete() would be called. In the future finished_one_bio()'s tests wouldn't reflect this and aio_complete() would be called for a second time which can manifest as an oops. The previous cleanups have whittled the sync and async completion paths down to the point where we can collapse them and clearly reassert the invariant that we must only call aio_complete() after returning -EIOCBQUEUED. direct_io_worker() will only return -EIOCBQUEUED when it is not the last to drop the dio refcount and the aio bio completion path will only call aio_complete() when it is the last to drop the dio refcount. direct_io_worker() can ensure that it is the last to drop the reference count by waiting for bios to drain. It does this for sync ops, of course, and for partial dio writes that must fall back to buffered and for aio ops that saw errors during submission. This means that operations that end up waiting, even if they were issued as aio ops, will not call aio_complete() from dio. Instead we return the return code of the operation and let the aio core call aio_complete(). This is purposely done to fix a bug where AIO DIO file extensions would call aio_complete() before their callers have a chance to update i_size. Now that direct_io_worker() is explicitly returning -EIOCBQUEUED its callers no longer have to translate for it. XFS needs to be careful not to free resources that will be used during AIO completion if -EIOCBQUEUED is returned. We maintain the previous behaviour of trying to write fs metadata for O_SYNC aio+dio writes. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:05 +00:00
/*
* Sync will always be dropping the final ref and completing the
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
* operation. AIO can if it was a broken operation described above or
* in fact if all the bios race to complete before we get here. In
* that case dio_complete() translates the EIOCBQUEUED into the proper
* return code that the caller will hand to ->complete().
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
*
* This is managed by the bio_lock instead of being an atomic_t so that
* completion paths can drop their ref and use the remaining count to
* decide to wake the submission path atomically.
[PATCH] dio: only call aio_complete() after returning -EIOCBQUEUED The only time it is safe to call aio_complete() is when the ->ki_retry function returns -EIOCBQUEUED to the AIO core. direct_io_worker() has historically done this by relying on its caller to translate positive return codes into -EIOCBQUEUED for the aio case. It did this by trying to keep conditionals in sync. direct_io_worker() knew when finished_one_bio() was going to call aio_complete(). It would reverse the test and wait and free the dio in the cases it thought that finished_one_bio() wasn't going to. Not surprisingly, it ended up getting it wrong. 'ret' could be a negative errno from the submission path but it failed to communicate this to finished_one_bio(). direct_io_worker() would return < 0, it's callers wouldn't raise -EIOCBQUEUED, and aio_complete() would be called. In the future finished_one_bio()'s tests wouldn't reflect this and aio_complete() would be called for a second time which can manifest as an oops. The previous cleanups have whittled the sync and async completion paths down to the point where we can collapse them and clearly reassert the invariant that we must only call aio_complete() after returning -EIOCBQUEUED. direct_io_worker() will only return -EIOCBQUEUED when it is not the last to drop the dio refcount and the aio bio completion path will only call aio_complete() when it is the last to drop the dio refcount. direct_io_worker() can ensure that it is the last to drop the reference count by waiting for bios to drain. It does this for sync ops, of course, and for partial dio writes that must fall back to buffered and for aio ops that saw errors during submission. This means that operations that end up waiting, even if they were issued as aio ops, will not call aio_complete() from dio. Instead we return the return code of the operation and let the aio core call aio_complete(). This is purposely done to fix a bug where AIO DIO file extensions would call aio_complete() before their callers have a chance to update i_size. Now that direct_io_worker() is explicitly returning -EIOCBQUEUED its callers no longer have to translate for it. XFS needs to be careful not to free resources that will be used during AIO completion if -EIOCBQUEUED is returned. We maintain the previous behaviour of trying to write fs metadata for O_SYNC aio+dio writes. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:05 +00:00
*/
[PATCH] dio: lock refcount operations The wait_for_more_bios() function name was poorly chosen. While looking to clean it up it I noticed that the dio struct refcounting between the bio completion and dio submission paths was racey. The bio submission path was simply freeing the dio struct if atomic_dec_and_test() indicated that it dropped the final reference. The aio bio completion path was dereferencing its dio struct pointer *after dropping its reference* based on the remaining number of references. These two paths could race and result in the aio bio completion path dereferencing a freed dio, though this was not observed in the wild. This moves the refcount under the bio lock so that bio completion can drop its reference and decide to wake all in one atomic step. Once testing and waking is locked dio_await_one() can test its sleeping condition and mark itself uninterruptible under the lock. It gets simpler and wait_for_more_bios() disappears. The addition of the interrupt masking spin lock acquiry in dio_bio_submit() looks alarming. This lock acquiry existed in that path before the recent dio completion patch set. We shouldn't expect significant performance regression from returning to the behaviour that existed before the completion clean up work. This passed 4k block ext3 O_DIRECT fsx and aio-stress on an SMP machine. Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Suparna Bhattacharya <suparna@in.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: <xfs-masters@oss.sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-10 10:21:07 +00:00
spin_lock_irqsave(&dio->bio_lock, flags);
ret2 = --dio->refcount;
spin_unlock_irqrestore(&dio->bio_lock, flags);
return ret2;
}
/*
* This is a library function for use by filesystem drivers.
*
* The locking rules are governed by the flags parameter:
* - if the flags value contains DIO_LOCKING we use a fancy locking
* scheme for dumb filesystems.
* For writes this function is called under i_mutex and returns with
* i_mutex held, for reads, i_mutex is not held on entry, but it is
* taken and dropped again before returning.
* - if the flags value does NOT contain DIO_LOCKING we don't use any
* internal locking but rather rely on the filesystem to synchronize
* direct I/O reads/writes versus each other and truncate.
*
* To help with locking against truncate we incremented the i_dio_count
* counter before starting direct I/O, and decrement it once we are done.
* Truncate can wait for it to reach zero to provide exclusion. It is
* expected that filesystem provide exclusion between new direct I/O
* and truncates. For DIO_LOCKING filesystems this is done by i_mutex,
* but other filesystems need to take care of this on their own.
*
* NOTE: if you pass "sdio" to anything by pointer make sure that function
* is always inlined. Otherwise gcc is unable to split the structure into
* individual fields and will generate much worse code. This is important
* for the whole file.
*/
static inline ssize_t
do_blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
struct block_device *bdev, struct iov_iter *iter,
get_block_t get_block, dio_iodone_t end_io,
dio_submit_t submit_io, int flags)
{
locking/atomics: COCCINELLE/treewide: Convert trivial ACCESS_ONCE() patterns to READ_ONCE()/WRITE_ONCE() Please do not apply this to mainline directly, instead please re-run the coccinelle script shown below and apply its output. For several reasons, it is desirable to use {READ,WRITE}_ONCE() in preference to ACCESS_ONCE(), and new code is expected to use one of the former. So far, there's been no reason to change most existing uses of ACCESS_ONCE(), as these aren't harmful, and changing them results in churn. However, for some features, the read/write distinction is critical to correct operation. To distinguish these cases, separate read/write accessors must be used. This patch migrates (most) remaining ACCESS_ONCE() instances to {READ,WRITE}_ONCE(), using the following coccinelle script: ---- // Convert trivial ACCESS_ONCE() uses to equivalent READ_ONCE() and // WRITE_ONCE() // $ make coccicheck COCCI=/home/mark/once.cocci SPFLAGS="--include-headers" MODE=patch virtual patch @ depends on patch @ expression E1, E2; @@ - ACCESS_ONCE(E1) = E2 + WRITE_ONCE(E1, E2) @ depends on patch @ expression E; @@ - ACCESS_ONCE(E) + READ_ONCE(E) ---- Signed-off-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: davem@davemloft.net Cc: linux-arch@vger.kernel.org Cc: mpe@ellerman.id.au Cc: shuah@kernel.org Cc: snitzer@redhat.com Cc: thor.thayer@linux.intel.com Cc: tj@kernel.org Cc: viro@zeniv.linux.org.uk Cc: will.deacon@arm.com Link: http://lkml.kernel.org/r/1508792849-3115-19-git-send-email-paulmck@linux.vnet.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-10-23 21:07:29 +00:00
unsigned i_blkbits = READ_ONCE(inode->i_blkbits);
unsigned blkbits = i_blkbits;
unsigned blocksize_mask = (1 << blkbits) - 1;
ssize_t retval = -EINVAL;
const size_t count = iov_iter_count(iter);
loff_t offset = iocb->ki_pos;
const loff_t end = offset + count;
struct dio *dio;
struct dio_submit sdio = { 0, };
struct buffer_head map_bh = { 0, };
struct blk_plug plug;
unsigned long align = offset | iov_iter_alignment(iter);
/*
* Avoid references to bdev if not absolutely needed to give
* the early prefetch in the caller enough time.
*/
if (align & blocksize_mask) {
if (bdev)
blkbits = blksize_bits(bdev_logical_block_size(bdev));
blocksize_mask = (1 << blkbits) - 1;
if (align & blocksize_mask)
goto out;
}
/* watch out for a 0 len io from a tricksy fs */
if (iov_iter_rw(iter) == READ && !count)
return 0;
dio = kmem_cache_alloc(dio_cache, GFP_KERNEL);
retval = -ENOMEM;
if (!dio)
goto out;
dio: don't zero out the pages array inside struct dio Intel reported a performance regression caused by the following commit: commit 848c4dd5153c7a0de55470ce99a8e13a63b4703f Author: Zach Brown <zach.brown@oracle.com> Date: Mon Aug 20 17:12:01 2007 -0700 dio: zero struct dio with kzalloc instead of manually This patch uses kzalloc to zero all of struct dio rather than manually trying to track which fields we rely on being zero. It passed aio+dio stress testing and some bug regression testing on ext3. This patch was introduced by Linus in the conversation that lead up to Badari's minimal fix to manually zero .map_bh.b_state in commit: 6a648fa72161d1f6468dabd96c5d3c0db04f598a It makes the code a bit smaller. Maybe a couple fewer cachelines to load, if we're lucky: text data bss dec hex filename 3285925 568506 1304616 5159047 4eb887 vmlinux 3285797 568506 1304616 5158919 4eb807 vmlinux.patched I was unable to measure a stable difference in the number of cpu cycles spent in blockdev_direct_IO() when pushing aio+dio 256K reads at ~340MB/s. So the resulting intent of the patch isn't a performance gain but to avoid exposing ourselves to the risk of finding another field like .map_bh.b_state where we rely on zeroing but don't enforce it in the code. Zach surmised that zeroing out the page array was what caused most of the problem, and suggested the approach taken in the attached patch for resolving the issue. Intel re-tested with this patch and saw a 0.6% performance gain (the original regression was 0.5%). [akpm@linux-foundation.org: add comment] Signed-off-by: Jeff Moyer <jmoyer@redhat.com> Acked-by: Zach Brown <zach.brown@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:49 +00:00
/*
* Believe it or not, zeroing out the page array caused a .5%
* performance regression in a database benchmark. So, we take
* care to only zero out what's needed.
*/
memset(dio, 0, offsetof(struct dio, pages));
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
dio->flags = flags;
if (dio->flags & DIO_LOCKING) {
if (iov_iter_rw(iter) == READ) {
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
struct address_space *mapping =
iocb->ki_filp->f_mapping;
direct-io: cleanup blockdev_direct_IO locking Currently the locking in blockdev_direct_IO is a mess, we have three different locking types and very confusing checks for some of them. The most complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be used. This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case is unused anyway, and the write side is almost identical to DIO_NO_LOCKING. The difference is that DIO_NO_LOCKING always sets the create argument for the get_blocks callback to zero, but we can easily move that to the actual get_blocks callbacks. There are four users of the DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is fine with the new version, ocfs2 only errors out if create were ever set, and we can remove this dead code now, the block device code only ever uses create for an error message if we are fully beyond the device which can never happen, and last but not least XFS will need the new behavour for writes. Now we can replace the lock_type variable with a flags one, where no flag means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first flag. Separate out the check for not allowing to fill holes into a separate flag, although for now both flags always get set at the same time. Also revamp the documentation of the locking scheme to actually make sense. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alex Elder <aelder@sgi.com> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-16 00:47:50 +00:00
/* will be released by direct_io_worker */
inode_lock(inode);
retval = filemap_write_and_wait_range(mapping, offset,
end - 1);
if (retval) {
inode_unlock(inode);
kmem_cache_free(dio_cache, dio);
goto out;
}
}
}
/* Once we sampled i_size check for reads beyond EOF */
dio->i_size = i_size_read(inode);
if (iov_iter_rw(iter) == READ && offset >= dio->i_size) {
if (dio->flags & DIO_LOCKING)
inode_unlock(inode);
kmem_cache_free(dio_cache, dio);
retval = 0;
goto out;
}
/*
* For file extending writes updating i_size before data writeouts
* complete can expose uninitialized blocks in dumb filesystems.
* In that case we need to wait for I/O completion even if asked
* for an asynchronous write.
*/
if (is_sync_kiocb(iocb))
dio->is_async = false;
else if (iov_iter_rw(iter) == WRITE && end > i_size_read(inode))
dio->is_async = false;
else
dio->is_async = true;
dio->inode = inode;
if (iov_iter_rw(iter) == WRITE) {
dio->op = REQ_OP_WRITE;
dio->op_flags = REQ_SYNC | REQ_IDLE;
if (iocb->ki_flags & IOCB_NOWAIT)
dio->op_flags |= REQ_NOWAIT;
} else {
dio->op = REQ_OP_READ;
}
if (iocb->ki_flags & IOCB_HIPRI)
dio->op_flags |= REQ_HIPRI;
/*
* For AIO O_(D)SYNC writes we need to defer completions to a workqueue
* so that we can call ->fsync.
*/
if (dio->is_async && iov_iter_rw(iter) == WRITE) {
retval = 0;
if (iocb->ki_flags & IOCB_DSYNC)
retval = dio_set_defer_completion(dio);
else if (!dio->inode->i_sb->s_dio_done_wq) {
/*
* In case of AIO write racing with buffered read we
* need to defer completion. We can't decide this now,
* however the workqueue needs to be initialized here.
*/
retval = sb_init_dio_done_wq(dio->inode->i_sb);
}
if (retval) {
/*
* We grab i_mutex only for reads so we don't have
* to release it here
*/
kmem_cache_free(dio_cache, dio);
goto out;
}
}
/*
* Will be decremented at I/O completion time.
*/
inode_dio_begin(inode);
retval = 0;
sdio.blkbits = blkbits;
sdio.blkfactor = i_blkbits - blkbits;
sdio.block_in_file = offset >> blkbits;
sdio.get_block = get_block;
dio->end_io = end_io;
sdio.submit_io = submit_io;
sdio.final_block_in_bio = -1;
sdio.next_block_for_io = -1;
dio->iocb = iocb;
spin_lock_init(&dio->bio_lock);
dio->refcount = 1;
dio->should_dirty = iter_is_iovec(iter) && iov_iter_rw(iter) == READ;
sdio.iter = iter;
sdio.final_block_in_request = end >> blkbits;
/*
* In case of non-aligned buffers, we may need 2 more
* pages since we need to zero out first and last block.
*/
if (unlikely(sdio.blkfactor))
sdio.pages_in_io = 2;
sdio.pages_in_io += iov_iter_npages(iter, INT_MAX);
blk_start_plug(&plug);
retval = do_direct_IO(dio, &sdio, &map_bh);
if (retval)
dio_cleanup(dio, &sdio);
if (retval == -ENOTBLK) {
/*
* The remaining part of the request will be
* be handled by buffered I/O when we return
*/
retval = 0;
}
/*
* There may be some unwritten disk at the end of a part-written
* fs-block-sized block. Go zero that now.
*/
dio_zero_block(dio, &sdio, 1, &map_bh);
if (sdio.cur_page) {
ssize_t ret2;
ret2 = dio_send_cur_page(dio, &sdio, &map_bh);
if (retval == 0)
retval = ret2;
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 12:29:47 +00:00
put_page(sdio.cur_page);
sdio.cur_page = NULL;
}
if (sdio.bio)
dio_bio_submit(dio, &sdio);
blk_finish_plug(&plug);
/*
* It is possible that, we return short IO due to end of file.
* In that case, we need to release all the pages we got hold on.
*/
dio_cleanup(dio, &sdio);
/*
* All block lookups have been performed. For READ requests
* we can let i_mutex go now that its achieved its purpose
* of protecting us from looking up uninitialized blocks.
*/
if (iov_iter_rw(iter) == READ && (dio->flags & DIO_LOCKING))
inode_unlock(dio->inode);
/*
* The only time we want to leave bios in flight is when a successful
* partial aio read or full aio write have been setup. In that case
* bio completion will call aio_complete. The only time it's safe to
* call aio_complete is when we return -EIOCBQUEUED, so we key on that.
* This had *better* be the only place that raises -EIOCBQUEUED.
*/
BUG_ON(retval == -EIOCBQUEUED);
if (dio->is_async && retval == 0 && dio->result &&
(iov_iter_rw(iter) == READ || dio->result == count))
retval = -EIOCBQUEUED;
else
dio_await_completion(dio);
if (drop_refcount(dio) == 0) {
retval = dio_complete(dio, retval, DIO_COMPLETE_INVALIDATE);
} else
BUG_ON(retval != -EIOCBQUEUED);
fs: introduce new truncate sequence Introduce a new truncate calling sequence into fs/mm subsystems. Rather than setattr > vmtruncate > truncate, have filesystems call their truncate sequence from ->setattr if filesystem specific operations are required. vmtruncate is deprecated, and truncate_pagecache and inode_newsize_ok helpers introduced previously should be used. simple_setattr is introduced for simple in-ram filesystems to implement the new truncate sequence. Eventually all filesystems should be converted to implement a setattr, and the default code in notify_change should go away. simple_setsize is also introduced to perform just the ATTR_SIZE portion of simple_setattr (ie. changing i_size and trimming pagecache). To implement the new truncate sequence: - filesystem specific manipulations (eg freeing blocks) must be done in the setattr method rather than ->truncate. - vmtruncate can not be used by core code to trim blocks past i_size in the event of write failure after allocation, so this must be performed in the fs code. - convert usage of helpers block_write_begin, nobh_write_begin, cont_write_begin, and *blockdev_direct_IO* to use _newtrunc postfixed variants. These avoid calling vmtruncate to trim blocks (see previous). - inode_setattr should not be used. generic_setattr is a new function to be used to copy simple attributes into the generic inode. - make use of the better opportunity to handle errors with the new sequence. Big problem with the previous calling sequence: the filesystem is not called until i_size has already changed. This means it is not allowed to fail the call, and also it does not know what the previous i_size was. Also, generic code calling vmtruncate to truncate allocated blocks in case of error had no good way to return a meaningful error (or, for example, atomically handle block deallocation). Cc: Christoph Hellwig <hch@lst.de> Acked-by: Jan Kara <jack@suse.cz> Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2010-05-26 15:05:33 +00:00
out:
return retval;
}
ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
struct block_device *bdev, struct iov_iter *iter,
get_block_t get_block,
dio_iodone_t end_io, dio_submit_t submit_io,
int flags)
{
/*
* The block device state is needed in the end to finally
* submit everything. Since it's likely to be cache cold
* prefetch it here as first thing to hide some of the
* latency.
*
* Attempt to prefetch the pieces we likely need later.
*/
prefetch(&bdev->bd_disk->part_tbl);
prefetch(bdev->bd_queue);
prefetch((char *)bdev->bd_queue + SMP_CACHE_BYTES);
return do_blockdev_direct_IO(iocb, inode, bdev, iter, get_block,
end_io, submit_io, flags);
}
EXPORT_SYMBOL(__blockdev_direct_IO);
static __init int dio_init(void)
{
dio_cache = KMEM_CACHE(dio, SLAB_PANIC);
return 0;
}
module_init(dio_init)