[PATCH] zram: factor out partial IO routine

[PATCH] zram: factor out partial IO routine

[Minchan Kim]

For architecture(PAGE_SIZE > 4K), zram have supported partial IO.
However, the mixed code for handling normal/partial IO is too mess,
error-prone to modify IO handler functions with upcoming feature
so this patch aims for cleaning up via factoring out partial IO
routines to zram_bvec_partial_[read|write] which will be disabled
for most 4K page architecures.

x86(4K architecure)
add/remove: 0/1 grow/shrink: 0/1 up/down: 0/-664 (-664)
function                                     old     new   delta
zram_bvec_rw                                2301    2039    -262
zram_decompress_page.isra                    402       -    -402

So, we will save 662 bytes.

However, as side effect, it will increase binary size in
non-4K architecure but it's not major for zram so I believe
benefit(maintainance, binary size for most architecture) is bigger.

Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
---
 drivers/block/zram/zram_drv.c | 259 ++++++++++++++++++++++++++++++------------
 1 file changed, 186 insertions(+), 73 deletions(-)

diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index fefdf260503a..0f92dae177d6 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -98,10 +98,17 @@ static void zram_set_obj_size(struct zram_meta *meta,
 	meta->table[index].value = (flags << ZRAM_FLAG_SHIFT) | size;
 }
 
+#if PAGE_SIZE != 4096
 static inline bool is_partial_io(struct bio_vec *bvec)
 {
 	return bvec->bv_len != PAGE_SIZE;
 }
+#else
+static inline bool is_partial_io(struct bio_vec *bvec)
+{
+	return false;
+}
+#endif
 
 static void zram_revalidate_disk(struct zram *zram)
 {
@@ -543,8 +550,8 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
 	return 0;
 }
 
-static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
-			  u32 index, int offset)
+static int zram_bvec_partial_read(struct zram *zram, struct bio_vec *bvec,
+				u32 index, int offset)
 {
 	int ret;
 	struct page *page;
@@ -561,81 +568,87 @@ static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
 	}
 	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
 
-	if (is_partial_io(bvec))
-		/* Use  a temporary buffer to decompress the page */
-		uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
-
-	user_mem = kmap_atomic(page);
-	if (!is_partial_io(bvec))
-		uncmem = user_mem;
-
+	/* Use  a temporary buffer to decompress the page */
+	uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
 	if (!uncmem) {
 		pr_err("Unable to allocate temp memory\n");
-		ret = -ENOMEM;
-		goto out_cleanup;
+		return -ENOMEM;
 	}
 
+	user_mem = kmap_atomic(page);
 	ret = zram_decompress_page(zram, uncmem, index);
-	/* Should NEVER happen. Return bio error if it does. */
 	if (unlikely(ret))
 		goto out_cleanup;
 
-	if (is_partial_io(bvec))
-		memcpy(user_mem + bvec->bv_offset, uncmem + offset,
+	memcpy(user_mem + bvec->bv_offset, uncmem + offset,
 				bvec->bv_len);
-
 	flush_dcache_page(page);
 	ret = 0;
 out_cleanup:
 	kunmap_atomic(user_mem);
-	if (is_partial_io(bvec))
-		kfree(uncmem);
+	kfree(uncmem);
 	return ret;
 }
 
-static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
-			   int offset)
+static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
+			  u32 index, int offset)
 {
-	int ret = 0;
+	int ret;
+	struct page *page;
+	unsigned char *user_mem;
+	struct zram_meta *meta = zram->meta;
+	page = bvec->bv_page;
+
+	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
+	if (unlikely(!meta->table[index].handle) ||
+			zram_test_flag(meta, index, ZRAM_SAME)) {
+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
+		handle_same_page(bvec, meta->table[index].element);
+		return 0;
+	}
+	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
+
+	user_mem = kmap_atomic(page);
+	ret = zram_decompress_page(zram, user_mem, index);
+	kunmap_atomic(user_mem);
+
+	flush_dcache_page(page);
+	return ret;
+}
+
+static int zram_bvec_partial_write(struct zram *zram, struct bio_vec *bvec,
+				u32 index, int offset)
+{
+	int ret = -ENOMEM;
 	unsigned int clen;
 	unsigned long handle = 0;
 	struct page *page;
 	unsigned char *user_mem, *cmem, *src, *uncmem = NULL;
 	struct zram_meta *meta = zram->meta;
-	struct zcomp_strm *zstrm = NULL;
+	struct zcomp_strm *zstrm;
 	unsigned long alloced_pages;
 	unsigned long element;
 
 	page = bvec->bv_page;
-	if (is_partial_io(bvec)) {
-		/*
-		 * This is a partial IO. We need to read the full page
-		 * before to write the changes.
-		 */
-		uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
-		if (!uncmem) {
-			ret = -ENOMEM;
-			goto out;
-		}
-		ret = zram_decompress_page(zram, uncmem, index);
-		if (ret)
-			goto out;
-	}
+	/*
+	 * This is a partial IO. We need to read the full page
+	 * before to write the changes.
+	 */
+	uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
+	if (!uncmem)
+		return ret;
+
+	ret = zram_decompress_page(zram, uncmem, index);
+	if (ret)
+		goto out;
 
 compress_again:
 	user_mem = kmap_atomic(page);
-	if (is_partial_io(bvec)) {
-		memcpy(uncmem + offset, user_mem + bvec->bv_offset,
-		       bvec->bv_len);
-		kunmap_atomic(user_mem);
-		user_mem = NULL;
-	} else {
-		uncmem = user_mem;
-	}
+	memcpy(uncmem + offset, user_mem + bvec->bv_offset,
+	       bvec->bv_len);
+	kunmap_atomic(user_mem);
 
 	if (page_same_filled(uncmem, &element)) {
-		if (user_mem)
-			kunmap_atomic(user_mem);
 		/* Free memory associated with this sector now. */
 		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 		zram_free_page(zram, index);
@@ -650,13 +663,8 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 
 	zstrm = zcomp_stream_get(zram->comp);
 	ret = zcomp_compress(zstrm, uncmem, &clen);
-	if (!is_partial_io(bvec)) {
-		kunmap_atomic(user_mem);
-		user_mem = NULL;
-		uncmem = NULL;
-	}
-
 	if (unlikely(ret)) {
+		zcomp_stream_put(zram->comp);
 		pr_err("Compression failed! err=%d\n", ret);
 		goto out;
 	}
@@ -664,8 +672,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	src = zstrm->buffer;
 	if (unlikely(clen > max_zpage_size)) {
 		clen = PAGE_SIZE;
-		if (is_partial_io(bvec))
-			src = uncmem;
+		src = uncmem;
 	}
 
 	/*
@@ -689,8 +696,6 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 				__GFP_MOVABLE);
 	if (!handle) {
 		zcomp_stream_put(zram->comp);
-		zstrm = NULL;
-
 		atomic64_inc(&zram->stats.writestall);
 
 		handle = zs_malloc(meta->mem_pool, clen,
@@ -701,7 +706,6 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 
 		pr_err("Error allocating memory for compressed page: %u, size=%u\n",
 			index, clen);
-		ret = -ENOMEM;
 		goto out;
 	}
 
@@ -710,22 +714,131 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 
 	if (zram->limit_pages && alloced_pages > zram->limit_pages) {
 		zs_free(meta->mem_pool, handle);
-		ret = -ENOMEM;
+		zcomp_stream_put(zram->comp);
 		goto out;
 	}
 
 	cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO);
 
-	if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) {
-		src = kmap_atomic(page);
-		copy_page(cmem, src);
-		kunmap_atomic(src);
+	memcpy(cmem, src, clen);
+	zcomp_stream_put(zram->comp);
+	zs_unmap_object(meta->mem_pool, handle);
+
+	/*
+	 * Free memory associated with this sector
+	 * before overwriting unused sectors.
+	 */
+	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
+	zram_free_page(zram, index);
+	meta->table[index].handle = handle;
+	zram_set_obj_size(meta, index, clen);
+	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
+
+	/* Update stats */
+	atomic64_add(clen, &zram->stats.compr_data_size);
+	atomic64_inc(&zram->stats.pages_stored);
+	ret = 0;
+out:
+	kfree(uncmem);
+	return ret;
+}
+
+static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
+			   int offset)
+{
+	int ret;
+	unsigned int clen;
+	unsigned long handle = 0;
+	struct page *page;
+	void *user_mem, *cmem;
+	struct zram_meta *meta = zram->meta;
+	struct zcomp_strm *zstrm;
+	unsigned long alloced_pages;
+	unsigned long element;
+
+	page = bvec->bv_page;
+compress_again:
+	user_mem = kmap_atomic(page);
+	if (page_same_filled(user_mem, &element)) {
+		kunmap_atomic(user_mem);
+
+		/* Free memory associated with this sector now. */
+		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
+		zram_free_page(zram, index);
+		zram_set_flag(meta, index, ZRAM_SAME);
+		zram_set_element(meta, index, element);
+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
+
+		atomic64_inc(&zram->stats.same_pages);
+		return 0;
+	}
+
+	zstrm = zcomp_stream_get(zram->comp);
+	ret = zcomp_compress(zstrm, user_mem, &clen);
+	kunmap_atomic(user_mem);
+
+	if (unlikely(ret)) {
+		pr_err("Compression failed! err=%d\n", ret);
+		zcomp_stream_put(zram->comp);
+		return ret;
+	}
+
+	if (unlikely(clen > max_zpage_size))
+		clen = PAGE_SIZE;
+
+	/*
+	 * handle allocation has 2 paths:
+	 * a) fast path is executed with preemption disabled (for
+	 *  per-cpu streams) and has __GFP_DIRECT_RECLAIM bit clear,
+	 *  since we can't sleep;
+	 * b) slow path enables preemption and attempts to allocate
+	 *  the page with __GFP_DIRECT_RECLAIM bit set. we have to
+	 *  put per-cpu compression stream and, thus, to re-do
+	 *  the compression once handle is allocated.
+	 *
+	 * if we have a 'non-null' handle here then we are coming
+	 * from the slow path and handle has already been allocated.
+	 */
+	if (!handle)
+		handle = zs_malloc(meta->mem_pool, clen,
+				__GFP_KSWAPD_RECLAIM |
+				__GFP_NOWARN |
+				__GFP_HIGHMEM |
+				__GFP_MOVABLE);
+	if (!handle) {
+		zcomp_stream_put(zram->comp);
+		atomic64_inc(&zram->stats.writestall);
+		handle = zs_malloc(meta->mem_pool, clen,
+				GFP_NOIO | __GFP_HIGHMEM |
+				__GFP_MOVABLE);
+		if (handle)
+			goto compress_again;
+
+		pr_err("Error allocating memory for compressed page: %u, size=%u\n",
+			index, clen);
+		return -ENOMEM;
+	}
+
+	alloced_pages = zs_get_total_pages(meta->mem_pool);
+	update_used_max(zram, alloced_pages);
+
+	if (zram->limit_pages && alloced_pages > zram->limit_pages) {
+		zs_free(meta->mem_pool, handle);
+		zcomp_stream_put(zram->comp);
+		return -ENOMEM;
+	}
+
+	cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO);
+
+	if (clen == PAGE_SIZE) {
+		user_mem = kmap_atomic(page);
+		copy_page(cmem, user_mem);
+		kunmap_atomic(user_mem);
 	} else {
-		memcpy(cmem, src, clen);
+		memcpy(cmem, zstrm->buffer, clen);
 	}
 
 	zcomp_stream_put(zram->comp);
-	zstrm = NULL;
 	zs_unmap_object(meta->mem_pool, handle);
 
 	/*
@@ -734,7 +847,6 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	 */
 	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 	zram_free_page(zram, index);
-
 	meta->table[index].handle = handle;
 	zram_set_obj_size(meta, index, clen);
 	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
@@ -742,12 +854,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	/* Update stats */
 	atomic64_add(clen, &zram->stats.compr_data_size);
 	atomic64_inc(&zram->stats.pages_stored);
-out:
-	if (zstrm)
-		zcomp_stream_put(zram->comp);
-	if (is_partial_io(bvec))
-		kfree(uncmem);
-	return ret;
+	return 0;
 }
 
 /*
@@ -801,10 +908,16 @@ static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index,
 
 	if (!is_write) {
 		atomic64_inc(&zram->stats.num_reads);
-		ret = zram_bvec_read(zram, bvec, index, offset);
+		if (!is_partial_io(bvec))
+			ret = zram_bvec_read(zram, bvec, index, offset);
+		else
+			ret = zram_bvec_partial_read(zram, bvec, index, offset);
 	} else {
 		atomic64_inc(&zram->stats.num_writes);
-		ret = zram_bvec_write(zram, bvec, index, offset);
+		if (!is_partial_io(bvec))
+			ret = zram_bvec_write(zram, bvec, index, offset);
+		else
+			ret = zram_bvec_partial_write(zram, bvec, index, offset);
 	}
 
 	generic_end_io_acct(rw_acct, &zram->disk->part0, start_time);
-- 
2.7.4

Re: [PATCH] zram: factor out partial IO routine

[Sergey Senozhatsky]

On (03/29/17 16:48), Minchan Kim wrote:
> For architecture(PAGE_SIZE > 4K), zram have supported partial IO.
> However, the mixed code for handling normal/partial IO is too mess,
> error-prone to modify IO handler functions with upcoming feature
> so this patch aims for cleaning up via factoring out partial IO
> routines to zram_bvec_partial_[read|write] which will be disabled
> for most 4K page architecures.
> 
> x86(4K architecure)
> add/remove: 0/1 grow/shrink: 0/1 up/down: 0/-664 (-664)
> function                                     old     new   delta
> zram_bvec_rw                                2301    2039    -262
> zram_decompress_page.isra                    402       -    -402
> 
> So, we will save 662 bytes.
> 
> However, as side effect, it will increase binary size in
> non-4K architecure but it's not major for zram so I believe
> benefit(maintainance, binary size for most architecture) is bigger.

a bigger side effect is that now we double the amount of lines we need
to change in certain patches and, thus, the amount of work - when we
add new functionality/fix something in zram_bvec_{write, read} we also
would need to touch zram_bvec_partial_{write, read}.

still probably worth it.

Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>

	-ss

Re: [PATCH] zram: factor out partial IO routine

[Minchan Kim]

Hi Sergey,

On Thu, Mar 30, 2017 at 01:12:38PM +0900, Sergey Senozhatsky wrote:
> On (03/29/17 16:48), Minchan Kim wrote:
> > For architecture(PAGE_SIZE > 4K), zram have supported partial IO.
> > However, the mixed code for handling normal/partial IO is too mess,
> > error-prone to modify IO handler functions with upcoming feature
> > so this patch aims for cleaning up via factoring out partial IO
> > routines to zram_bvec_partial_[read|write] which will be disabled
> > for most 4K page architecures.
> > 
> > x86(4K architecure)
> > add/remove: 0/1 grow/shrink: 0/1 up/down: 0/-664 (-664)
> > function                                     old     new   delta
> > zram_bvec_rw                                2301    2039    -262
> > zram_decompress_page.isra                    402       -    -402
> > 
> > So, we will save 662 bytes.
> > 
> > However, as side effect, it will increase binary size in
> > non-4K architecure but it's not major for zram so I believe
> > benefit(maintainance, binary size for most architecture) is bigger.
> 
> a bigger side effect is that now we double the amount of lines we need
> to change in certain patches and, thus, the amount of work - when we
> add new functionality/fix something in zram_bvec_{write, read} we also
> would need to touch zram_bvec_partial_{write, read}.

Yes, that is a pain, too. However, I thought it would be more easier
because as-is partial IO routine is more error-prone to me. :)
> 
> still probably worth it.
> 
> Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
 
Thanks for the review.
so I tried clean-up further to make you happy. :)

How about this?
It's totally untested and I have no time until Monday next week.
So, please review with having enough time.

diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index fefdf260503a..68eee48c5a9d 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -98,10 +98,17 @@ static void zram_set_obj_size(struct zram_meta *meta,
 	meta->table[index].value = (flags << ZRAM_FLAG_SHIFT) | size;
 }
 
+#if PAGE_SIZE != 4096
 static inline bool is_partial_io(struct bio_vec *bvec)
 {
 	return bvec->bv_len != PAGE_SIZE;
 }
+#else
+static inline bool is_partial_io(struct bio_vec *bvec)
+{
+	return false;
+}
+#endif
 
 static void zram_revalidate_disk(struct zram *zram)
 {
@@ -191,7 +198,7 @@ static bool page_same_filled(void *ptr, unsigned long *element)
 	return true;
 }
 
-static void handle_same_page(struct bio_vec *bvec, unsigned long element)
+static void __handle_same_page(struct bio_vec *bvec, unsigned long element)
 {
 	struct page *page = bvec->bv_page;
 	void *user_mem;
@@ -199,8 +206,6 @@ static void handle_same_page(struct bio_vec *bvec, unsigned long element)
 	user_mem = kmap_atomic(page);
 	zram_fill_page(user_mem + bvec->bv_offset, bvec->bv_len, element);
 	kunmap_atomic(user_mem);
-
-	flush_dcache_page(page);
 }
 
 static ssize_t initstate_show(struct device *dev,
@@ -504,13 +509,14 @@ static void zram_free_page(struct zram *zram, size_t index)
 	zram_set_obj_size(meta, index, 0);
 }
 
-static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
+static int zram_decompress_page(struct zram *zram, struct page *page, u32 index)
 {
 	int ret = 0;
 	unsigned char *cmem;
 	struct zram_meta *meta = zram->meta;
 	unsigned long handle;
 	unsigned int size;
+	void *mem;
 
 	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 	handle = meta->table[index].handle;
@@ -518,17 +524,23 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
 
 	if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
 		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
+		mem = kmap_atomic(page);
 		zram_fill_page(mem, PAGE_SIZE, meta->table[index].element);
+		kunmap_atomic(mem);
 		return 0;
 	}
 
 	cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_RO);
 	if (size == PAGE_SIZE) {
+		mem = kmap_atomic(page);
 		copy_page(mem, cmem);
+		kunmap_atomic(mem);
 	} else {
 		struct zcomp_strm *zstrm = zcomp_stream_get(zram->comp);
 
+		mem = kmap_atomic(page);
 		ret = zcomp_decompress(zstrm, cmem, size, mem);
+		kunmap_atomic(mem);
 		zcomp_stream_put(zram->comp);
 	}
 	zs_unmap_object(meta->mem_pool, handle);
@@ -543,99 +555,70 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
 	return 0;
 }
 
-static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
-			  u32 index, int offset)
+static bool handle_special_page(struct zram *zram, struct bio_vec *bvec,
+				u32 index)
 {
-	int ret;
-	struct page *page;
-	unsigned char *user_mem, *uncmem = NULL;
 	struct zram_meta *meta = zram->meta;
-	page = bvec->bv_page;
 
 	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 	if (unlikely(!meta->table[index].handle) ||
 			zram_test_flag(meta, index, ZRAM_SAME)) {
 		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
-		handle_same_page(bvec, meta->table[index].element);
-		return 0;
+		__handle_same_page(bvec, meta->table[index].element);
+		return true;
 	}
 	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
 
-	if (is_partial_io(bvec))
-		/* Use  a temporary buffer to decompress the page */
-		uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
+	return false;
+}
 
-	user_mem = kmap_atomic(page);
-	if (!is_partial_io(bvec))
-		uncmem = user_mem;
+static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
+				u32 index, int offset)
+{
+	int ret;
+	struct page *page;
 
-	if (!uncmem) {
-		pr_err("Unable to allocate temp memory\n");
-		ret = -ENOMEM;
-		goto out_cleanup;
+	if (handle_special_page(zram, bvec, index))
+		return 0;
+
+	page = bvec->bv_page;
+	if (is_partial_io(bvec)) {
+		/* Use a temporary buffer to decompress the page */
+		page = alloc_page(GFP_NOIO|__GFP_HIGHMEM);
+		if (!page)
+			return -ENOMEM;
 	}
 
-	ret = zram_decompress_page(zram, uncmem, index);
-	/* Should NEVER happen. Return bio error if it does. */
+	ret = zram_decompress_page(zram, page, index);
 	if (unlikely(ret))
-		goto out_cleanup;
+		goto out;
 
-	if (is_partial_io(bvec))
-		memcpy(user_mem + bvec->bv_offset, uncmem + offset,
-				bvec->bv_len);
+	if (is_partial_io(bvec)) {
+		void *user_mem = kmap_atomic(bvec->bv_page);
+		void *uncmem = kmap_atomic(page);
 
-	flush_dcache_page(page);
-	ret = 0;
-out_cleanup:
-	kunmap_atomic(user_mem);
+		memcpy(user_mem + bvec->bv_offset, uncmem + offset,
+					bvec->bv_len);
+		kunmap_atomic(uncmem);
+		kunmap_atomic(user_mem);
+	}
+out:
 	if (is_partial_io(bvec))
-		kfree(uncmem);
+		__free_page(page);
+
+	flush_dcache_page(bvec->bv_page);
 	return ret;
 }
 
-static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
-			   int offset)
+static bool handle_same_page(struct zram *zram, u32 index, struct page *page)
 {
-	int ret = 0;
-	unsigned int clen;
-	unsigned long handle = 0;
-	struct page *page;
-	unsigned char *user_mem, *cmem, *src, *uncmem = NULL;
-	struct zram_meta *meta = zram->meta;
-	struct zcomp_strm *zstrm = NULL;
-	unsigned long alloced_pages;
 	unsigned long element;
+	void *user_mem = kmap_atomic(page);
 
-	page = bvec->bv_page;
-	if (is_partial_io(bvec)) {
-		/*
-		 * This is a partial IO. We need to read the full page
-		 * before to write the changes.
-		 */
-		uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
-		if (!uncmem) {
-			ret = -ENOMEM;
-			goto out;
-		}
-		ret = zram_decompress_page(zram, uncmem, index);
-		if (ret)
-			goto out;
-	}
+	if (page_same_filled(user_mem, &element)) {
+		struct zram_meta *meta = zram->meta;
 
-compress_again:
-	user_mem = kmap_atomic(page);
-	if (is_partial_io(bvec)) {
-		memcpy(uncmem + offset, user_mem + bvec->bv_offset,
-		       bvec->bv_len);
 		kunmap_atomic(user_mem);
-		user_mem = NULL;
-	} else {
-		uncmem = user_mem;
-	}
-
-	if (page_same_filled(uncmem, &element)) {
-		if (user_mem)
-			kunmap_atomic(user_mem);
 		/* Free memory associated with this sector now. */
 		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 		zram_free_page(zram, index);
@@ -644,29 +627,35 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
 
 		atomic64_inc(&zram->stats.same_pages);
-		ret = 0;
-		goto out;
+		return true;
 	}
+	kunmap_atomic(user_mem);
 
-	zstrm = zcomp_stream_get(zram->comp);
-	ret = zcomp_compress(zstrm, uncmem, &clen);
-	if (!is_partial_io(bvec)) {
-		kunmap_atomic(user_mem);
-		user_mem = NULL;
-		uncmem = NULL;
-	}
+	return false;
+}
+
+static int zram_compress(struct zram *zram, struct zcomp_strm **zstrm,
+			struct page *page, unsigned long *p_handle,
+			unsigned int *p_clen)
+{
+	int ret;
+	unsigned long handle = 0;
+	unsigned int clen;
+	void *user_mem;
+	struct zram_meta *meta = zram->meta;
+
+compress_again:
+	user_mem = kmap_atomic(page);
+	ret = zcomp_compress(*zstrm, user_mem, &clen);
+	kunmap_atomic(user_mem);
 
 	if (unlikely(ret)) {
 		pr_err("Compression failed! err=%d\n", ret);
-		goto out;
+		return ret;
 	}
 
-	src = zstrm->buffer;
-	if (unlikely(clen > max_zpage_size)) {
+	if (unlikely(clen > max_zpage_size))
 		clen = PAGE_SIZE;
-		if (is_partial_io(bvec))
-			src = uncmem;
-	}
 
 	/*
 	 * handle allocation has 2 paths:
@@ -689,20 +678,41 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 				__GFP_MOVABLE);
 	if (!handle) {
 		zcomp_stream_put(zram->comp);
-		zstrm = NULL;
-
 		atomic64_inc(&zram->stats.writestall);
-
 		handle = zs_malloc(meta->mem_pool, clen,
 				GFP_NOIO | __GFP_HIGHMEM |
 				__GFP_MOVABLE);
+		*zstrm = zcomp_stream_get(zram->comp);
 		if (handle)
 			goto compress_again;
+		return -ENOMEM;
+	}
 
-		pr_err("Error allocating memory for compressed page: %u, size=%u\n",
-			index, clen);
-		ret = -ENOMEM;
-		goto out;
+	*p_handle = handle;
+	*p_clen = clen;
+	return 0;
+}
+
+static int __zram_bvec_write(struct zram *zram, struct bio_vec *bvec,
+				u32 index, int offset)
+{
+	int ret;
+	unsigned int clen;
+	unsigned long handle;
+	void *user_mem, *cmem;
+	struct zcomp_strm *zstrm;
+	unsigned long alloced_pages;
+	struct zram_meta *meta = zram->meta;
+	struct page *page = bvec->bv_page;
+
+	if (handle_same_page(zram, index, page))
+		return 0;
+
+	zstrm = zcomp_stream_get(zram->comp);
+	ret = zram_compress(zram, &zstrm, page, &handle, &clen);
+	if (ret) {
+		zcomp_stream_put(zram->comp);
+		return ret;
 	}
 
 	alloced_pages = zs_get_total_pages(meta->mem_pool);
@@ -710,22 +720,21 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 
 	if (zram->limit_pages && alloced_pages > zram->limit_pages) {
 		zs_free(meta->mem_pool, handle);
-		ret = -ENOMEM;
-		goto out;
+		zcomp_stream_put(zram->comp);
+		return -ENOMEM;
 	}
 
 	cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO);
 
-	if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) {
-		src = kmap_atomic(page);
-		copy_page(cmem, src);
-		kunmap_atomic(src);
+	if (clen == PAGE_SIZE) {
+		user_mem = kmap_atomic(page);
+		copy_page(cmem, user_mem);
+		kunmap_atomic(user_mem);
 	} else {
-		memcpy(cmem, src, clen);
+		memcpy(cmem, zstrm->buffer, clen);
 	}
 
 	zcomp_stream_put(zram->comp);
-	zstrm = NULL;
 	zs_unmap_object(meta->mem_pool, handle);
 
 	/*
@@ -734,7 +743,6 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	 */
 	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 	zram_free_page(zram, index);
-
 	meta->table[index].handle = handle;
 	zram_set_obj_size(meta, index, clen);
 	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
@@ -742,11 +750,48 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	/* Update stats */
 	atomic64_add(clen, &zram->stats.compr_data_size);
 	atomic64_inc(&zram->stats.pages_stored);
+	return 0;
+}
+
+static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec,
+				u32 index, int offset)
+{
+	int ret = -ENOMEM;
+	struct page *page = NULL;
+	unsigned char *user_mem;
+	struct bio_vec vec;
+
+	vec = *bvec;
+	if (is_partial_io(bvec)) {
+		void *uncmem;
+		/*
+		 * This is a partial IO. We need to read the full page
+		 * before to write the changes.
+		 */
+		page = alloc_page(GFP_NOIO|__GFP_HIGHMEM);
+		if (!page)
+			return ret;
+
+		ret = zram_decompress_page(zram, page, index);
+		if (ret)
+			goto out;
+
+		user_mem = kmap_atomic(bvec->bv_page);
+		uncmem = kmap_atomic(page);
+		memcpy(uncmem + offset, user_mem + bvec->bv_offset,
+		       bvec->bv_len);
+		kunmap_atomic(uncmem);
+		kunmap_atomic(user_mem);
+
+		vec.bv_page = page;
+		vec.bv_len = PAGE_SIZE;
+		vec.bv_offset = 0;
+	}
+
+	ret = __zram_bvec_write(zram, &vec, index, offset);
 out:
-	if (zstrm)
-		zcomp_stream_put(zram->comp);
 	if (is_partial_io(bvec))
-		kfree(uncmem);
+		__free_page(page);
 	return ret;
 }
 

Re: [PATCH] zram: factor out partial IO routine

[Sergey Senozhatsky]

Hello,

On (03/31/17 08:33), Minchan Kim wrote:
[..]
> > a bigger side effect is that now we double the amount of lines we need
> > to change in certain patches and, thus, the amount of work - when we
> > add new functionality/fix something in zram_bvec_{write, read} we also
> > would need to touch zram_bvec_partial_{write, read}.
> 
> Yes, that is a pain, too. However, I thought it would be more easier
> because as-is partial IO routine is more error-prone to me. :)

yeah, it is quite tricky and a bit difficult. agree.

> > still probably worth it.
> > 
> > Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
>  
> Thanks for the review.
> so I tried clean-up further to make you happy. :)
> 
> How about this?
> It's totally untested and I have no time until Monday next week.
> So, please review with having enough time.

will take a look :)
thanks!

	-ss