xfs_buf.c

LINUX 2.6.17.4的源码

	xfs_buftarg_t		*target,
	xfs_off_t		ioff,
	size_t			isize,
	xfs_buf_flags_t		flags)
{
	xfs_buf_t		*bp;

	flags |= XBF_READ;

	bp = xfs_buf_get_flags(target, ioff, isize, flags);
	if (bp) {
		if (!XFS_BUF_ISDONE(bp)) {
			XB_TRACE(bp, "read", (unsigned long)flags);
			XFS_STATS_INC(xb_get_read);
			xfs_buf_iostart(bp, flags);
		} else if (flags & XBF_ASYNC) {
			XB_TRACE(bp, "read_async", (unsigned long)flags);
			/*
			 * Read ahead call which is already satisfied,
			 * drop the buffer
			 */
			goto no_buffer;
		} else {
			XB_TRACE(bp, "read_done", (unsigned long)flags);
			/* We do not want read in the flags */
			bp->b_flags &= ~XBF_READ;
		}
	}

	return bp;

 no_buffer:
	if (flags & (XBF_LOCK | XBF_TRYLOCK))
		xfs_buf_unlock(bp);
	xfs_buf_rele(bp);
	return NULL;
}

/*
 *	If we are not low on memory then do the readahead in a deadlock
 *	safe manner.
 */
void
xfs_buf_readahead(
	xfs_buftarg_t		*target,
	xfs_off_t		ioff,
	size_t			isize,
	xfs_buf_flags_t		flags)
{
	struct backing_dev_info *bdi;

	bdi = target->bt_mapping->backing_dev_info;
	if (bdi_read_congested(bdi))
		return;

	flags |= (XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD);
	xfs_buf_read_flags(target, ioff, isize, flags);
}

xfs_buf_t *
xfs_buf_get_empty(
	size_t			len,
	xfs_buftarg_t		*target)
{
	xfs_buf_t		*bp;

	bp = xfs_buf_allocate(0);
	if (bp)
		_xfs_buf_initialize(bp, target, 0, len, 0);
	return bp;
}

static inline struct page *
mem_to_page(
	void			*addr)
{
	if (((unsigned long)addr < VMALLOC_START) ||
	    ((unsigned long)addr >= VMALLOC_END)) {
		return virt_to_page(addr);
	} else {
		return vmalloc_to_page(addr);
	}
}

int
xfs_buf_associate_memory(
	xfs_buf_t		*bp,
	void			*mem,
	size_t			len)
{
	int			rval;
	int			i = 0;
	size_t			ptr;
	size_t			end, end_cur;
	off_t			offset;
	int			page_count;

	page_count = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
	offset = (off_t) mem - ((off_t)mem & PAGE_CACHE_MASK);
	if (offset && (len > PAGE_CACHE_SIZE))
		page_count++;

	/* Free any previous set of page pointers */
	if (bp->b_pages)
		_xfs_buf_free_pages(bp);

	bp->b_pages = NULL;
	bp->b_addr = mem;

	rval = _xfs_buf_get_pages(bp, page_count, 0);
	if (rval)
		return rval;

	bp->b_offset = offset;
	ptr = (size_t) mem & PAGE_CACHE_MASK;
	end = PAGE_CACHE_ALIGN((size_t) mem + len);
	end_cur = end;
	/* set up first page */
	bp->b_pages[0] = mem_to_page(mem);

	ptr += PAGE_CACHE_SIZE;
	bp->b_page_count = ++i;
	while (ptr < end) {
		bp->b_pages[i] = mem_to_page((void *)ptr);
		bp->b_page_count = ++i;
		ptr += PAGE_CACHE_SIZE;
	}
	bp->b_locked = 0;

	bp->b_count_desired = bp->b_buffer_length = len;
	bp->b_flags |= XBF_MAPPED;

	return 0;
}

xfs_buf_t *
xfs_buf_get_noaddr(
	size_t			len,
	xfs_buftarg_t		*target)
{
	size_t			malloc_len = len;
	xfs_buf_t		*bp;
	void			*data;
	int			error;

	bp = xfs_buf_allocate(0);
	if (unlikely(bp == NULL))
		goto fail;
	_xfs_buf_initialize(bp, target, 0, len, 0);

 try_again:
	data = kmem_alloc(malloc_len, KM_SLEEP | KM_MAYFAIL);
	if (unlikely(data == NULL))
		goto fail_free_buf;

	/* check whether alignment matches.. */
	if ((__psunsigned_t)data !=
	    ((__psunsigned_t)data & ~target->bt_smask)) {
		/* .. else double the size and try again */
		kmem_free(data, malloc_len);
		malloc_len <<= 1;
		goto try_again;
	}

	error = xfs_buf_associate_memory(bp, data, len);
	if (error)
		goto fail_free_mem;
	bp->b_flags |= _XBF_KMEM_ALLOC;

	xfs_buf_unlock(bp);

	XB_TRACE(bp, "no_daddr", data);
	return bp;
 fail_free_mem:
	kmem_free(data, malloc_len);
 fail_free_buf:
	xfs_buf_free(bp);
 fail:
	return NULL;
}

/*
 *	Increment reference count on buffer, to hold the buffer concurrently
 *	with another thread which may release (free) the buffer asynchronously.
 *	Must hold the buffer already to call this function.
 */
void
xfs_buf_hold(
	xfs_buf_t		*bp)
{
	atomic_inc(&bp->b_hold);
	XB_TRACE(bp, "hold", 0);
}

/*
 *	Releases a hold on the specified buffer.  If the
 *	the hold count is 1, calls xfs_buf_free.
 */
void
xfs_buf_rele(
	xfs_buf_t		*bp)
{
	xfs_bufhash_t		*hash = bp->b_hash;

	XB_TRACE(bp, "rele", bp->b_relse);

	if (unlikely(!hash)) {
		ASSERT(!bp->b_relse);
		if (atomic_dec_and_test(&bp->b_hold))
			xfs_buf_free(bp);
		return;
	}

	if (atomic_dec_and_lock(&bp->b_hold, &hash->bh_lock)) {
		if (bp->b_relse) {
			atomic_inc(&bp->b_hold);
			spin_unlock(&hash->bh_lock);
			(*(bp->b_relse)) (bp);
		} else if (bp->b_flags & XBF_FS_MANAGED) {
			spin_unlock(&hash->bh_lock);
		} else {
			ASSERT(!(bp->b_flags & (XBF_DELWRI|_XBF_DELWRI_Q)));
			list_del_init(&bp->b_hash_list);
			spin_unlock(&hash->bh_lock);
			xfs_buf_free(bp);
		}
	} else {
		/*
		 * Catch reference count leaks
		 */
		ASSERT(atomic_read(&bp->b_hold) >= 0);
	}
}


/*
 *	Mutual exclusion on buffers.  Locking model:
 *
 *	Buffers associated with inodes for which buffer locking
 *	is not enabled are not protected by semaphores, and are
 *	assumed to be exclusively owned by the caller.  There is a
 *	spinlock in the buffer, used by the caller when concurrent
 *	access is possible.
 */

/*
 *	Locks a buffer object, if it is not already locked.
 *	Note that this in no way locks the underlying pages, so it is only
 *	useful for synchronizing concurrent use of buffer objects, not for
 *	synchronizing independent access to the underlying pages.
 */
int
xfs_buf_cond_lock(
	xfs_buf_t		*bp)
{
	int			locked;

	locked = down_trylock(&bp->b_sema) == 0;
	if (locked) {
		XB_SET_OWNER(bp);
	}
	XB_TRACE(bp, "cond_lock", (long)locked);
	return locked ? 0 : -EBUSY;
}

#if defined(DEBUG) || defined(XFS_BLI_TRACE)
int
xfs_buf_lock_value(
	xfs_buf_t		*bp)
{
	return atomic_read(&bp->b_sema.count);
}
#endif

/*
 *	Locks a buffer object.
 *	Note that this in no way locks the underlying pages, so it is only
 *	useful for synchronizing concurrent use of buffer objects, not for
 *	synchronizing independent access to the underlying pages.
 */
void
xfs_buf_lock(
	xfs_buf_t		*bp)
{
	XB_TRACE(bp, "lock", 0);
	if (atomic_read(&bp->b_io_remaining))
		blk_run_address_space(bp->b_target->bt_mapping);
	down(&bp->b_sema);
	XB_SET_OWNER(bp);
	XB_TRACE(bp, "locked", 0);
}

/*
 *	Releases the lock on the buffer object.
 *	If the buffer is marked delwri but is not queued, do so before we
 *	unlock the buffer as we need to set flags correctly.  We also need to
 *	take a reference for the delwri queue because the unlocker is going to
 *	drop their's and they don't know we just queued it.
 */
void
xfs_buf_unlock(
	xfs_buf_t		*bp)
{
	if ((bp->b_flags & (XBF_DELWRI|_XBF_DELWRI_Q)) == XBF_DELWRI) {
		atomic_inc(&bp->b_hold);
		bp->b_flags |= XBF_ASYNC;
		xfs_buf_delwri_queue(bp, 0);
	}

	XB_CLEAR_OWNER(bp);
	up(&bp->b_sema);
	XB_TRACE(bp, "unlock", 0);
}


/*
 *	Pinning Buffer Storage in Memory
 *	Ensure that no attempt to force a buffer to disk will succeed.
 */
void
xfs_buf_pin(
	xfs_buf_t		*bp)
{
	atomic_inc(&bp->b_pin_count);
	XB_TRACE(bp, "pin", (long)bp->b_pin_count.counter);
}

void
xfs_buf_unpin(
	xfs_buf_t		*bp)
{
	if (atomic_dec_and_test(&bp->b_pin_count))
		wake_up_all(&bp->b_waiters);
	XB_TRACE(bp, "unpin", (long)bp->b_pin_count.counter);
}

int
xfs_buf_ispin(
	xfs_buf_t		*bp)
{
	return atomic_read(&bp->b_pin_count);
}

STATIC void
xfs_buf_wait_unpin(
	xfs_buf_t		*bp)
{
	DECLARE_WAITQUEUE	(wait, current);

	if (atomic_read(&bp->b_pin_count) == 0)
		return;

	add_wait_queue(&bp->b_waiters, &wait);
	for (;;) {
		set_current_state(TASK_UNINTERRUPTIBLE);
		if (atomic_read(&bp->b_pin_count) == 0)
			break;
		if (atomic_read(&bp->b_io_remaining))
			blk_run_address_space(bp->b_target->bt_mapping);
		schedule();
	}
	remove_wait_queue(&bp->b_waiters, &wait);
	set_current_state(TASK_RUNNING);
}

/*
 *	Buffer Utility Routines
 */

STATIC void
xfs_buf_iodone_work(
	void			*v)
{
	xfs_buf_t		*bp = (xfs_buf_t *)v;

	if (bp->b_iodone)
		(*(bp->b_iodone))(bp);
	else if (bp->b_flags & XBF_ASYNC)
		xfs_buf_relse(bp);
}

void
xfs_buf_ioend(
	xfs_buf_t		*bp,
	int			schedule)
{
	bp->b_flags &= ~(XBF_READ | XBF_WRITE);
	if (bp->b_error == 0)
		bp->b_flags |= XBF_DONE;

	XB_TRACE(bp, "iodone", bp->b_iodone);

	if ((bp->b_iodone) || (bp->b_flags & XBF_ASYNC)) {
		if (schedule) {
			INIT_WORK(&bp->b_iodone_work, xfs_buf_iodone_work, bp);
			queue_work(xfslogd_workqueue, &bp->b_iodone_work);
		} else {
			xfs_buf_iodone_work(bp);
		}
	} else {
		up(&bp->b_iodonesema);
	}
}

void
xfs_buf_ioerror(
	xfs_buf_t		*bp,
	int			error)
{
	ASSERT(error >= 0 && error <= 0xffff);
	bp->b_error = (unsigned short)error;
	XB_TRACE(bp, "ioerror", (unsigned long)error);
}

/*
 *	Initiate I/O on a buffer, based on the flags supplied.
 *	The b_iodone routine in the buffer supplied will only be called
 *	when all of the subsidiary I/O requests, if any, have been completed.
 */
int
xfs_buf_iostart(
	xfs_buf_t		*bp,
	xfs_buf_flags_t		flags)
{
	int			status = 0;

	XB_TRACE(bp, "iostart", (unsigned long)flags);

	if (flags & XBF_DELWRI) {
		bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_ASYNC);
		bp->b_flags |= flags & (XBF_DELWRI | XBF_ASYNC);
		xfs_buf_delwri_queue(bp, 1);
		return status;
	}

	bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_ASYNC | XBF_DELWRI | \
			XBF_READ_AHEAD | _XBF_RUN_QUEUES);
	bp->b_flags |= flags & (XBF_READ | XBF_WRITE | XBF_ASYNC | \
			XBF_READ_AHEAD | _XBF_RUN_QUEUES);

	BUG_ON(bp->b_bn == XFS_BUF_DADDR_NULL);

	/* For writes allow an alternate strategy routine to precede
	 * the actual I/O request (which may not be issued at all in
	 * a shutdown situation, for example).
	 */
	status = (flags & XBF_WRITE) ?
		xfs_buf_iostrategy(bp) : xfs_buf_iorequest(bp);

	/* Wait for I/O if we are not an async request.
	 * Note: async I/O request completion will release the buffer,
	 * and that can already be done by this point.  So using the
	 * buffer pointer from here on, after async I/O, is invalid.
	 */
	if (!status && !(flags & XBF_ASYNC))
		status = xfs_buf_iowait(bp);

	return status;
}

STATIC __inline__ int
_xfs_buf_iolocked(
	xfs_buf_t		*bp)
{
	ASSERT(bp->b_flags & (XBF_READ | XBF_WRITE));
	if (bp->b_flags & XBF_READ)
		return bp->b_locked;
	return 0;
}

STATIC __inline__ void
_xfs_buf_ioend(
	xfs_buf_t		*bp,
	int			schedule)
{
	if (atomic_dec_and_test(&bp->b_io_remaining) == 1) {
		bp->b_locked = 0;
		xfs_buf_ioend(bp, schedule);
	}
}

STATIC int
xfs_buf_bio_end_io(
	struct bio		*bio,
	unsigned int		bytes_done,
	int			error)
{
	xfs_buf_t		*bp = (xfs_buf_t *)bio->bi_private;
	unsigned int		blocksize = bp->b_target->bt_bsize;
	struct bio_vec		*bvec = bio->bi_io_vec + bio->bi_vcnt - 1;

	if (bio->bi_size)
		return 1;

	if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
		bp->b_error = EIO;

	do {
		struct page	*page = bvec->bv_page;

		if (unlikely(bp->b_error)) {
			if (bp->b_flags & XBF_READ)
				ClearPageUptodate(page);
			SetPageError(page);
		} else if (blocksize >= PAGE_CACHE_SIZE) {
			SetPageUptodate(page);
		} else if (!PagePrivate(page) &&
				(bp->b_flags & _XBF_PAGE_CACHE)) {
			set_page_region(page, bvec->bv_offset, bvec->bv_len);
		}

		if (--bvec >= bio->bi_io_vec)
			prefetchw(&bvec->bv_page->flags);

		if (_xfs_buf_iolocked(bp)) {
			unlock_page(page);
		}
	} while (bvec >= bio->bi_io_vec);

	_xfs_buf_ioend(bp, 1);
	bio_put(bio);
	return 0;
}

STATIC void
_xfs_buf_ioapply(
	xfs_buf_t		*bp)
{
	int			i, rw, map_i, total_nr_pages, nr_pages;
	struct bio		*bio;
	int			offset = bp->b_offset;
	int			size = bp->b_count_desired;
	sector_t		sector = bp->b_bn;
	unsigned int		blocksize = bp->b_target->bt_bsize;
	int			locking = _xfs_buf_iolocked(bp);

	total_nr_pages = bp->b_page_count;
	map_i = 0;

	if (bp->b_flags & _XBF_RUN_QUEUES) {
		bp->b_flags &= ~_XBF_RUN_QUEUES;
		rw = (bp->b_flags & XBF_READ) ? READ_SYNC : WRITE_SYNC;
	} else {
		rw = (bp->b_flags & XBF_READ) ? READ : WRITE;
	}

	if (bp->b_flags & XBF_ORDERED) {
		ASSERT(!(bp->b_flags & XBF_READ));
		rw = WRITE_BARRIER;
	}

	/* Special code path for reading a sub page size buffer in --
	 * we populate up the whole page, and hence the other metadata
	 * in the same page.  This optimization is only valid when the
	 * filesystem block size is not smaller than the page size.
	 */
	if ((bp->b_buffer_length < PAGE_CACHE_SIZE) &&
	    (bp->b_flags & XBF_READ) && locking &&
	    (blocksize >= PAGE_CACHE_SIZE)) {
		bio = bio_alloc(GFP_NOIO, 1);

		bio->bi_bdev = bp->b_target->bt_bdev;
		bio->bi_sector = sector - (offset >> BBSHIFT);
		bio->bi_end_io = xfs_buf_bio_end_io;
		bio->bi_private = bp;

		bio_add_page(bio, bp->b_pages[0], PAGE_CACHE_SIZE, 0);
		size = 0;

		atomic_inc(&bp->b_io_remaining);

		goto submit_io;
	}

	/* Lock down the pages which we need to for the request */
	if (locking && (bp->b_flags & XBF_WRITE) && (bp->b_locked == 0)) {
		for (i = 0; size; i++) {
			int		nbytes = PAGE_CACHE_SIZE - offset;
			struct page	*page = bp->b_pages[i];

			if (nbytes > size)
				nbytes = size;

			lock_page(page);

			size -= nbytes;
			offset = 0;
		}
		offset = bp->b_offset;
		size = bp->b_count_desired;
	}

next_chunk:
	atomic_inc(&bp->b_io_remaining);
	nr_pages = BIO_MAX_SECTORS >> (PAGE_SHIFT - BBSHIFT);
	if (nr_pages > total_nr_pages)
		nr_pages = total_nr_pages;

	bio = bio_alloc(GFP_NOIO, nr_pages);
	bio->bi_bdev = bp->b_target->bt_bdev;
	bio->bi_sector = sector;
	bio->bi_end_io = xfs_buf_bio_end_io;
	bio->bi_private = bp;

	for (; size && nr_pages; nr_pages--, map_i++) {
		int	rbytes, nbytes = PAGE_CACHE_SIZE - offset;

		if (nbytes > size)
			nbytes = size;

		rbytes = bio_add_page(bio, bp->b_pages[map_i], nbytes, offset);
		if (rbytes < nbytes)
			break;

		offset = 0;
		sector += nbytes >> BBSHIFT;
		size -= nbytes;
		total_nr_pages--;
	}