Commit a1dd52af authored by Nitin Gupta's avatar Nitin Gupta Committed by Greg Kroah-Hartman
Browse files

Staging: ramzswap: Support generic I/O requests 

Currently, ramzwap devices (/dev/ramzswapX) can only
be used as swap disks since it was hard-coded to consider
only the first request in bio vector.

Now, we iterate over all the segments in an incoming
bio which allows us to handle all kinds of I/O requests.

ramzswap devices can still handle PAGE_SIZE aligned and
multiple of PAGE_SIZE sized I/O requests only. To ensure
that we get always get such requests only, we set following
request_queue attributes to PAGE_SIZE:
 - physical_block_size
 - logical_block_size
 - io_min
 - io_opt

Note: physical and logical block sizes were already set
equal to PAGE_SIZE and that seems to be sufficient to get
PAGE_SIZE aligned I/O.

Since we are no longer limited to handling swap requests
only, the next few patches rename ramzswap to zram. So,
the devices will then be called /dev/zram{0, 1, 2, ...}

Usage/Examples:
 1) Use as /tmp storage
 - mkfs.ext4 /dev/zram0
 - mount /dev/zram0 /tmp

 2) Use as swap:
 - mkswap /dev/zram0
 - swapon /dev/zram0 -p 10 # give highest priority to zram0

Performance:

 - I/O benchamark done with 'dd' command. Details can be
found here:
http://code.google.com/p/compcache/wiki/zramperf


Summary:
 - Maximum read speed (approx):
   - ram disk: 1200 MB/sec
   - zram disk: 600 MB/sec
 - Maximum write speed (approx):
   - ram disk: 500 MB/sec
   - zram disk: 160 MB/sec

Issues:

 - Double caching: We can potentially waste memory by having
two copies of a page -- one in page cache (uncompress) and
second in the device memory (compressed). However, during
reclaim, clean page cache pages are quickly freed, so this
does not seem to be a big problem.

 - Stale data: Not all filesystems support issuing 'discard'
requests to underlying block devices. So, if such filesystems
are used over zram devices, we can accumulate lot of stale
data in memory. Even for filesystems to do support discard
(example, ext4), we need to see how effective it is.

 - Scalability: There is only one (per-device) de/compression
buffer stats. This can lead to significant contention, especially
when used for generic (non-swap) purposes.

Signed-off-by: default avatarNitin Gupta <ngupta@vflare.org>
Acked-by: default avatarPekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: default avatarGreg Kroah-Hartman <gregkh@suse.de>
parent 36e574fe

drivers/staging/ramzswap/ramzswap_drv.c

+147 −179
Original line number Diff line number Diff line
@@ -101,20 +101,6 @@ static void ramzswap_set_disksize(struct ramzswap *rzs, size_t totalram_bytes) 
	rzs->disksize &= PAGE_MASK;

}



/*

 * Swap header (1st page of swap device) contains information

 * about a swap file/partition. Prepare such a header for the

 * given ramzswap device so that swapon can identify it as a

 * swap partition.

 */

static void setup_swap_header(struct ramzswap *rzs, union swap_header *s)

{

	s->info.version = 1;

	s->info.last_page = (rzs->disksize >> PAGE_SHIFT) - 1;

	s->info.nr_badpages = 0;

	memcpy(s->magic.magic, "SWAPSPACE2", 10);

}



static void ramzswap_ioctl_get_stats(struct ramzswap *rzs,

			struct ramzswap_ioctl_stats *s)

{
@@ -202,31 +188,22 @@ out: 
	rzs->table[index].offset = 0;

}



static int handle_zero_page(struct bio *bio)

static void handle_zero_page(struct page *page)

{

	void *user_mem;

	struct page *page = bio->bi_io_vec[0].bv_page;



	user_mem = kmap_atomic(page, KM_USER0);

	memset(user_mem, 0, PAGE_SIZE);

	kunmap_atomic(user_mem, KM_USER0);



	flush_dcache_page(page);



	set_bit(BIO_UPTODATE, &bio->bi_flags);

	bio_endio(bio, 0);

	return 0;

}



static int handle_uncompressed_page(struct ramzswap *rzs, struct bio *bio)

static void handle_uncompressed_page(struct ramzswap *rzs,

				struct page *page, u32 index)

{

	u32 index;

	struct page *page;

	unsigned char *user_mem, *cmem;



	page = bio->bi_io_vec[0].bv_page;

	index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;



	user_mem = kmap_atomic(page, KM_USER0);

	cmem = kmap_atomic(rzs->table[index].page, KM_USER1) +

			rzs->table[index].offset;
@@ -236,55 +213,45 @@ static int handle_uncompressed_page(struct ramzswap *rzs, struct bio *bio) 
	kunmap_atomic(cmem, KM_USER1);



	flush_dcache_page(page);



	set_bit(BIO_UPTODATE, &bio->bi_flags);

	bio_endio(bio, 0);

	return 0;

}



/*

 * Called when request page is not present in ramzswap.

 * This is an attempt to read before any previous write

 * to this location - this happens due to readahead when

 * swap device is read from user-space (e.g. during swapon)

 */

static int handle_ramzswap_fault(struct ramzswap *rzs, struct bio *bio)

static int ramzswap_read(struct ramzswap *rzs, struct bio *bio)

{

	pr_debug("Read before write on swap device: "

		"sector=%lu, size=%u, offset=%u\n",

		(ulong)(bio->bi_sector), bio->bi_size,

		bio->bi_io_vec[0].bv_offset);



	/* Do nothing. Just return success */

	set_bit(BIO_UPTODATE, &bio->bi_flags);

	bio_endio(bio, 0);

	return 0;

}

	int i;

	u32 index;

	struct bio_vec *bvec;



static int ramzswap_read(struct ramzswap *rzs, struct bio *bio)

{

	rzs_stat64_inc(rzs, &rzs->stats.num_reads);



	index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;

	bio_for_each_segment(bvec, bio, i) {

		int ret;

	u32 index;

		size_t clen;

		struct page *page;

		struct zobj_header *zheader;

		unsigned char *user_mem, *cmem;



	rzs_stat64_inc(rzs, &rzs->stats.num_reads);

		page = bvec->bv_page;



	page = bio->bi_io_vec[0].bv_page;

	index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;



	if (rzs_test_flag(rzs, index, RZS_ZERO))

		return handle_zero_page(bio);

		if (rzs_test_flag(rzs, index, RZS_ZERO)) {

			handle_zero_page(page);

			continue;

		}



		/* Requested page is not present in compressed area */

	if (!rzs->table[index].page)

		return handle_ramzswap_fault(rzs, bio);

		if (unlikely(!rzs->table[index].page)) {

			pr_debug("Read before write: sector=%lu, size=%u",

				(ulong)(bio->bi_sector), bio->bi_size);

			/* Do nothing */

			continue;

		}



		/* Page is stored uncompressed since it's incompressible */

	if (unlikely(rzs_test_flag(rzs, index, RZS_UNCOMPRESSED)))

		return handle_uncompressed_page(rzs, bio);

		if (unlikely(rzs_test_flag(rzs, index, RZS_UNCOMPRESSED))) {

			handle_uncompressed_page(rzs, page, index);

			continue;

		}



		user_mem = kmap_atomic(page, KM_USER0);

		clen = PAGE_SIZE;
@@ -300,7 +267,7 @@ static int ramzswap_read(struct ramzswap *rzs, struct bio *bio) 
		kunmap_atomic(user_mem, KM_USER0);

		kunmap_atomic(cmem, KM_USER1);



	/* should NEVER happen */

		/* Should NEVER happen. Return bio error if it does. */

		if (unlikely(ret != LZO_E_OK)) {

			pr_err("Decompression failed! err=%d, page=%u\n",

				ret, index);
@@ -309,6 +276,8 @@ static int ramzswap_read(struct ramzswap *rzs, struct bio *bio) 
		}



		flush_dcache_page(page);

		index++;

	}



	set_bit(BIO_UPTODATE, &bio->bi_flags);

	bio_endio(bio, 0);
@@ -321,20 +290,33 @@ out: 


static int ramzswap_write(struct ramzswap *rzs, struct bio *bio)

{

	int i;

	u32 index;

	struct bio_vec *bvec;



	rzs_stat64_inc(rzs, &rzs->stats.num_writes);



	index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;



	bio_for_each_segment(bvec, bio, i) {

		int ret;

	u32 offset, index;

		u32 offset;

		size_t clen;

		struct zobj_header *zheader;

		struct page *page, *page_store;

		unsigned char *user_mem, *cmem, *src;



	rzs_stat64_inc(rzs, &rzs->stats.num_writes);



	page = bio->bi_io_vec[0].bv_page;

	index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;



		page = bvec->bv_page;

		src = rzs->compress_buffer;



		/*

		 * System overwrites unused sectors. Free memory associated

		 * with this sector now.

		 */

		if (rzs->table[index].page ||

				rzs_test_flag(rzs, index, RZS_ZERO))

			ramzswap_free_page(rzs, index);



		mutex_lock(&rzs->lock);



		user_mem = kmap_atomic(page, KM_USER0);
@@ -343,10 +325,7 @@ static int ramzswap_write(struct ramzswap *rzs, struct bio *bio) 
			mutex_unlock(&rzs->lock);

			rzs_stat_inc(&rzs->stats.pages_zero);

			rzs_set_flag(rzs, index, RZS_ZERO);



		set_bit(BIO_UPTODATE, &bio->bi_flags);

		bio_endio(bio, 0);

		return 0;

			continue;

		}



		ret = lzo1x_1_compress(user_mem, PAGE_SIZE, src, &clen,
@@ -371,8 +350,8 @@ static int ramzswap_write(struct ramzswap *rzs, struct bio *bio) 
			page_store = alloc_page(GFP_NOIO | __GFP_HIGHMEM);

			if (unlikely(!page_store)) {

				mutex_unlock(&rzs->lock);

			pr_info("Error allocating memory for incompressible "

				"page: %u\n", index);

				pr_info("Error allocating memory for "

					"incompressible page: %u\n", index);

				rzs_stat64_inc(rzs, &rzs->stats.failed_writes);

				goto out;

			}
@@ -423,6 +402,8 @@ memstore: 
			rzs_stat_inc(&rzs->stats.good_compress);



		mutex_unlock(&rzs->lock);

		index++;

	}



	set_bit(BIO_UPTODATE, &bio->bi_flags);

	bio_endio(bio, 0);
@@ -436,19 +417,17 @@ out: 
/*

 * Check if request is within bounds and page aligned.

 */

static inline int valid_swap_request(struct ramzswap *rzs, struct bio *bio)

static inline int valid_io_request(struct ramzswap *rzs, struct bio *bio)

{

	if (unlikely(

		(bio->bi_sector >= (rzs->disksize >> SECTOR_SHIFT)) ||

		(bio->bi_sector & (SECTORS_PER_PAGE - 1)) ||

		(bio->bi_vcnt != 1) ||

		(bio->bi_size != PAGE_SIZE) ||

		(bio->bi_io_vec[0].bv_offset != 0))) {

		(bio->bi_size & (PAGE_SIZE - 1)))) {



		return 0;

	}



	/* swap request is valid */

	/* I/O request is valid */

	return 1;

}
@@ -465,7 +444,7 @@ static int ramzswap_make_request(struct request_queue *queue, struct bio *bio) 
		return 0;

	}



	if (!valid_swap_request(rzs, bio)) {

	if (!valid_io_request(rzs, bio)) {

		rzs_stat64_inc(rzs, &rzs->stats.invalid_io);

		bio_io_error(bio);

		return 0;
@@ -531,8 +510,6 @@ static int ramzswap_ioctl_init_device(struct ramzswap *rzs) 
{

	int ret;

	size_t num_pages;

	struct page *page;

	union swap_header *swap_header;



	if (rzs->init_done) {

		pr_info("Device already initialized!\n");
@@ -566,19 +543,6 @@ static int ramzswap_ioctl_init_device(struct ramzswap *rzs) 
	}

	memset(rzs->table, 0, num_pages * sizeof(*rzs->table));



	page = alloc_page(__GFP_ZERO);

	if (!page) {

		pr_err("Error allocating swap header page\n");

		ret = -ENOMEM;

		goto fail;

	}

	rzs->table[0].page = page;

	rzs_set_flag(rzs, 0, RZS_UNCOMPRESSED);



	swap_header = kmap(page);

	setup_swap_header(rzs, swap_header);

	kunmap(page);



	set_capacity(rzs->disk, rzs->disksize >> SECTOR_SHIFT);



	/* ramzswap devices sort of resembles non-rotational disks */
@@ -689,11 +653,9 @@ void ramzswap_slot_free_notify(struct block_device *bdev, unsigned long index) 
	rzs = bdev->bd_disk->private_data;

	ramzswap_free_page(rzs, index);

	rzs_stat64_inc(rzs, &rzs->stats.notify_free);



	return;

}



static struct block_device_operations ramzswap_devops = {

static const struct block_device_operations ramzswap_devops = {

	.ioctl = ramzswap_ioctl,

	.swap_slot_free_notify = ramzswap_slot_free_notify,

	.owner = THIS_MODULE
@@ -737,8 +699,14 @@ static int create_device(struct ramzswap *rzs, int device_id) 
	/* Actual capacity set using RZSIO_SET_DISKSIZE_KB ioctl */

	set_capacity(rzs->disk, 0);



	/*

	 * To ensure that we always get PAGE_SIZE aligned

	 * and n*PAGE_SIZED sized I/O requests.

	 */

	blk_queue_physical_block_size(rzs->disk->queue, PAGE_SIZE);

	blk_queue_logical_block_size(rzs->disk->queue, PAGE_SIZE);

	blk_queue_io_min(rzs->disk->queue, PAGE_SIZE);

	blk_queue_io_opt(rzs->disk->queue, PAGE_SIZE);



	add_disk(rzs->disk);

drivers/staging/ramzswap/ramzswap_drv.h

+2 −1
Original line number Diff line number Diff line
@@ -112,7 +112,8 @@ struct ramzswap { 
	void *compress_buffer;

	struct table *table;

	spinlock_t stat64_lock;	/* protect 64-bit stats */

	struct mutex lock;

	struct mutex lock;	/* protect compression buffers against

				 * concurrent writes */

	struct request_queue *queue;

	struct gendisk *disk;

	int init_done;

CRA Git | Maintained and supported by SUSTech CRA and CCSE