Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1572946 > unrolled thread

[PATCH] mm: extend zero pages to same element pages for zram

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH] mm: extend zero pages to same element pages for zram <zhouxianrong@huawei.com> - 2017-02-03 09:50 +0100
    Re: [PATCH] mm: extend zero pages to same element pages for zram Matthew Wilcox <willy@infradead.org> - 2017-02-03 16:50 +0100
      Re: [PATCH] mm: extend zero pages to same element pages for zram zhouxianrong <zhouxianrong@huawei.com> - 2017-02-04 04:40 +0100
    Re: [PATCH] mm: extend zero pages to same element pages for zram Minchan Kim <minchan@kernel.org> - 2017-02-05 15:30 +0100
      Re: [PATCH] mm: extend zero pages to same element pages for zram zhouxianrong <zhouxianrong@huawei.com> - 2017-02-06 02:40 +0100
        Re: [PATCH] mm: extend zero pages to same element pages for zram Matthew Wilcox <willy@infradead.org> - 2017-02-06 15:20 +0100
        Re: [PATCH] mm: extend zero pages to same element pages for zram Minchan Kim <minchan@kernel.org> - 2017-02-07 00:50 +0100
          Re: [PATCH] mm: extend zero pages to same element pages for zram zhouxianrong <zhouxianrong@huawei.com> - 2017-02-07 03:30 +0100
            Re: [PATCH] mm: extend zero pages to same element pages for zram Minchan Kim <minchan@kernel.org> - 2017-02-07 04:00 +0100
              Re: [PATCH] mm: extend zero pages to same element pages for zram zhouxianrong <zhouxianrong@huawei.com> - 2017-02-07 04:30 +0100
                Re: [PATCH] mm: extend zero pages to same element pages for zram Minchan Kim <minchan@kernel.org> - 2017-02-07 06:00 +0100

#1572946 — [PATCH] mm: extend zero pages to same element pages for zram

From: zhouxianrong <zhouxianrong@huawei.com>

test result as listed below:

zero   pattern_char pattern_short pattern_int pattern_long total   (unit)
162989 14454        3534          23516       2769         3294399 (page)

statistics for the result:

        zero  pattern_char  pattern_short  pattern_int  pattern_long
AVERAGE 0.745696298 0.085937175 0.015957701 0.131874915 0.020533911
STDEV   0.035623777 0.016892402 0.004454534 0.021657123 0.019420072
MAX     0.973813421 0.222222222 0.021409518 0.211812245 0.176512625
MIN     0.645431905 0.004634398 0           0           0

Signed-off-by: zhouxianrong <zhouxianrong@huawei.com>
---
 drivers/block/zram/zram_drv.c |  124 +++++++++++++++++++++++++++++++----------
 drivers/block/zram/zram_drv.h |   11 ++--
 2 files changed, 103 insertions(+), 32 deletions(-)

diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index e5ab7d9..6a8c9c5 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -95,6 +95,17 @@ static void zram_clear_flag(struct zram_meta *meta, u32 index,
 	meta->table[index].value &= ~BIT(flag);
 }
 
+static inline void zram_set_element(struct zram_meta *meta, u32 index,
+			unsigned long element)
+{
+	meta->table[index].element = element;
+}
+
+static inline void zram_clear_element(struct zram_meta *meta, u32 index)
+{
+	meta->table[index].element = 0;
+}
+
 static size_t zram_get_obj_size(struct zram_meta *meta, u32 index)
 {
 	return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1);
@@ -167,31 +178,78 @@ static inline void update_used_max(struct zram *zram,
 	} while (old_max != cur_max);
 }
 
-static bool page_zero_filled(void *ptr)
+static inline void zram_fill_page(char *ptr, unsigned long value)
+{
+	int i;
+	unsigned long *page = (unsigned long *)ptr;
+
+	if (likely(value == 0)) {
+		clear_page(ptr);
+	} else {
+		for (i = 0; i < PAGE_SIZE / sizeof(*page); i++)
+			page[i] = value;
+	}
+}
+
+static inline void zram_fill_page_partial(char *ptr, unsigned int size,
+		unsigned long value)
+{
+	int i;
+	unsigned long *page;
+
+	if (likely(value == 0)) {
+		memset(ptr, 0, size);
+		return;
+	}
+
+	i = ((unsigned long)ptr) % sizeof(*page);
+	if (i) {
+		while (i < sizeof(*page)) {
+			*ptr++ = (value >> (i * 8)) & 0xff;
+			--size;
+			++i;
+		}
+	}
+
+	for (i = size / sizeof(*page); i > 0; --i) {
+		page = (unsigned long *)ptr;
+		*page = value;
+		ptr += sizeof(*page);
+		size -= sizeof(*page);
+	}
+
+	for (i = 0; i < size; ++i)
+		*ptr++ = (value >> (i * 8)) & 0xff;
+}
+
+static bool page_same_filled(void *ptr, unsigned long *element)
 {
 	unsigned int pos;
 	unsigned long *page;
 
 	page = (unsigned long *)ptr;
 
-	for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) {
-		if (page[pos])
+	for (pos = 0; pos < PAGE_SIZE / sizeof(*page) - 1; pos++) {
+		if (page[pos] != page[pos + 1])
 			return false;
 	}
 
+	*element = page[pos];
+
 	return true;
 }
 
-static void handle_zero_page(struct bio_vec *bvec)
+static void handle_same_page(struct bio_vec *bvec, unsigned long element)
 {
 	struct page *page = bvec->bv_page;
 	void *user_mem;
 
 	user_mem = kmap_atomic(page);
 	if (is_partial_io(bvec))
-		memset(user_mem + bvec->bv_offset, 0, bvec->bv_len);
+		zram_fill_page_partial(user_mem + bvec->bv_offset, bvec->bv_len,
+			element);
 	else
-		clear_page(user_mem);
+		zram_fill_page(user_mem, element);
 	kunmap_atomic(user_mem);
 
 	flush_dcache_page(page);
@@ -440,7 +498,7 @@ static ssize_t mm_stat_show(struct device *dev,
 			mem_used << PAGE_SHIFT,
 			zram->limit_pages << PAGE_SHIFT,
 			max_used << PAGE_SHIFT,
-			(u64)atomic64_read(&zram->stats.zero_pages),
+			(u64)atomic64_read(&zram->stats.same_pages),
 			pool_stats.pages_compacted);
 	up_read(&zram->init_lock);
 
@@ -473,7 +531,7 @@ static ssize_t debug_stat_show(struct device *dev,
 ZRAM_ATTR_RO(failed_writes);
 ZRAM_ATTR_RO(invalid_io);
 ZRAM_ATTR_RO(notify_free);
-ZRAM_ATTR_RO(zero_pages);
+ZRAM_ATTR_RO(same_pages);
 ZRAM_ATTR_RO(compr_data_size);
 
 static inline bool zram_meta_get(struct zram *zram)
@@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
 
 	/* Free all pages that are still in this zram device */
 	for (index = 0; index < num_pages; index++) {
-		unsigned long handle = meta->table[index].handle;
+		unsigned long handle;
+
+		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
+		handle = meta->table[index].handle;
 
-		if (!handle)
+		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
+			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
 			continue;
+		}
 
+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
 		zs_free(meta->mem_pool, handle);
 	}
 
@@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
 static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
 {
 	size_t num_pages;
-	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
+	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
 
 	if (!meta)
 		return NULL;
@@ -547,18 +611,20 @@ static void zram_free_page(struct zram *zram, size_t index)
 	struct zram_meta *meta = zram->meta;
 	unsigned long handle = meta->table[index].handle;
 
-	if (unlikely(!handle)) {
-		/*
-		 * No memory is allocated for zero filled pages.
-		 * Simply clear zero page flag.
-		 */
-		if (zram_test_flag(meta, index, ZRAM_ZERO)) {
-			zram_clear_flag(meta, index, ZRAM_ZERO);
-			atomic64_dec(&zram->stats.zero_pages);
-		}
+	/*
+	 * No memory is allocated for same element filled pages.
+	 * Simply clear same page flag.
+	 */
+	if (zram_test_flag(meta, index, ZRAM_SAME)) {
+		zram_clear_flag(meta, index, ZRAM_SAME);
+		zram_clear_element(meta, index);
+		atomic64_dec(&zram->stats.same_pages);
 		return;
 	}
 
+	if (unlikely(!handle))
+		return;
+
 	zs_free(meta->mem_pool, handle);
 
 	atomic64_sub(zram_get_obj_size(meta, index),
@@ -581,9 +647,9 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
 	handle = meta->table[index].handle;
 	size = zram_get_obj_size(meta, index);
 
-	if (!handle || zram_test_flag(meta, index, ZRAM_ZERO)) {
+	if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
 		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
-		clear_page(mem);
+		zram_fill_page(mem, meta->table[index].element);
 		return 0;
 	}
 
@@ -619,9 +685,9 @@ static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
 
 	bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
 	if (unlikely(!meta->table[index].handle) ||
-			zram_test_flag(meta, index, ZRAM_ZERO)) {
+			zram_test_flag(meta, index, ZRAM_SAME)) {
 		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
-		handle_zero_page(bvec);
+		handle_same_page(bvec, meta->table[index].element);
 		return 0;
 	}
 	bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
@@ -669,6 +735,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 	struct zram_meta *meta = zram->meta;
 	struct zcomp_strm *zstrm = NULL;
 	unsigned long alloced_pages;
+	unsigned long element;
 
 	page = bvec->bv_page;
 	if (is_partial_io(bvec)) {
@@ -697,16 +764,17 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
 		uncmem = user_mem;
 	}
 
-	if (page_zero_filled(uncmem)) {
+	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);
-		zram_set_flag(meta, index, ZRAM_ZERO);
+		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.zero_pages);
+		atomic64_inc(&zram->stats.same_pages);
 		ret = 0;
 		goto out;
 	}
@@ -1206,7 +1274,7 @@ static int zram_open(struct block_device *bdev, fmode_t mode)
 	&dev_attr_compact.attr,
 	&dev_attr_invalid_io.attr,
 	&dev_attr_notify_free.attr,
-	&dev_attr_zero_pages.attr,
+	&dev_attr_same_pages.attr,
 	&dev_attr_orig_data_size.attr,
 	&dev_attr_compr_data_size.attr,
 	&dev_attr_mem_used_total.attr,
diff --git a/drivers/block/zram/zram_drv.h b/drivers/block/zram/zram_drv.h
index 74fcf10..4bb92e1 100644
--- a/drivers/block/zram/zram_drv.h
+++ b/drivers/block/zram/zram_drv.h
@@ -60,8 +60,8 @@
 
 /* Flags for zram pages (table[page_no].value) */
 enum zram_pageflags {
-	/* Page consists entirely of zeros */
-	ZRAM_ZERO = ZRAM_FLAG_SHIFT,
+	/* Page consists entirely of same elements */
+	ZRAM_SAME = ZRAM_FLAG_SHIFT,
 	ZRAM_ACCESS,	/* page is now accessed */
 
 	__NR_ZRAM_PAGEFLAGS,
@@ -71,7 +71,10 @@ enum zram_pageflags {
 
 /* Allocated for each disk page */
 struct zram_table_entry {
-	unsigned long handle;
+	union {
+		unsigned long handle;
+		unsigned long element;
+	};
 	unsigned long value;
 };
 
@@ -83,7 +86,7 @@ struct zram_stats {
 	atomic64_t failed_writes;	/* can happen when memory is too low */
 	atomic64_t invalid_io;	/* non-page-aligned I/O requests */
 	atomic64_t notify_free;	/* no. of swap slot free notifications */
-	atomic64_t zero_pages;		/* no. of zero filled pages */
+	atomic64_t same_pages;		/* no. of same element filled pages */
 	atomic64_t pages_stored;	/* no. of pages currently stored */
 	atomic_long_t max_used_pages;	/* no. of maximum pages stored */
 	atomic64_t writestall;		/* no. of write slow paths */
-- 
1.7.9.5

[toc] | [next] | [standalone]

#1573184

On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
> +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
> +		unsigned long value)
> +{
> +	int i;
> +	unsigned long *page;
> +
> +	if (likely(value == 0)) {
> +		memset(ptr, 0, size);
> +		return;
> +	}
> +
> +	i = ((unsigned long)ptr) % sizeof(*page);
> +	if (i) {
> +		while (i < sizeof(*page)) {
> +			*ptr++ = (value >> (i * 8)) & 0xff;
> +			--size;
> +			++i;
> +		}
> +	}
> +
> +	for (i = size / sizeof(*page); i > 0; --i) {
> +		page = (unsigned long *)ptr;
> +		*page = value;
> +		ptr += sizeof(*page);
> +		size -= sizeof(*page);
> +	}
> +
> +	for (i = 0; i < size; ++i)
> +		*ptr++ = (value >> (i * 8)) & 0xff;
> +}

You're assuming little-endian here.  I think you need to do a
cpu_to_le() here, but I don't think we have a cpu_to_leul, only
cpu_to_le64/cpu_to_le32.  So you may have some work to do ...

[toc] | [prev] | [next] | [standalone]

#1573588

right, thanks.

On 2017/2/3 23:33, Matthew Wilcox wrote:
> On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
>> +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
>> +		unsigned long value)
>> +{
>> +	int i;
>> +	unsigned long *page;
>> +
>> +	if (likely(value == 0)) {
>> +		memset(ptr, 0, size);
>> +		return;
>> +	}
>> +
>> +	i = ((unsigned long)ptr) % sizeof(*page);
>> +	if (i) {
>> +		while (i < sizeof(*page)) {
>> +			*ptr++ = (value >> (i * 8)) & 0xff;
>> +			--size;
>> +			++i;
>> +		}
>> +	}
>> +
>> +	for (i = size / sizeof(*page); i > 0; --i) {
>> +		page = (unsigned long *)ptr;
>> +		*page = value;
>> +		ptr += sizeof(*page);
>> +		size -= sizeof(*page);
>> +	}
>> +
>> +	for (i = 0; i < size; ++i)
>> +		*ptr++ = (value >> (i * 8)) & 0xff;
>> +}
>
> You're assuming little-endian here.  I think you need to do a
> cpu_to_le() here, but I don't think we have a cpu_to_leul, only
> cpu_to_le64/cpu_to_le32.  So you may have some work to do ...
>
>
> .
>

[toc] | [prev] | [next] | [standalone]

#1573881

Hi zhouxianrong,

On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
> From: zhouxianrong <zhouxianrong@huawei.com>
> 
> test result as listed below:
> 
> zero   pattern_char pattern_short pattern_int pattern_long total   (unit)
> 162989 14454        3534          23516       2769         3294399 (page)
> 
> statistics for the result:
> 
>         zero  pattern_char  pattern_short  pattern_int  pattern_long
> AVERAGE 0.745696298 0.085937175 0.015957701 0.131874915 0.020533911
> STDEV   0.035623777 0.016892402 0.004454534 0.021657123 0.019420072
> MAX     0.973813421 0.222222222 0.021409518 0.211812245 0.176512625
> MIN     0.645431905 0.004634398 0           0           0

The description in old version was better for justifying same page merging
feature.

> 
> Signed-off-by: zhouxianrong <zhouxianrong@huawei.com>
> ---
>  drivers/block/zram/zram_drv.c |  124 +++++++++++++++++++++++++++++++----------
>  drivers/block/zram/zram_drv.h |   11 ++--
>  2 files changed, 103 insertions(+), 32 deletions(-)
> 
> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
> index e5ab7d9..6a8c9c5 100644
> --- a/drivers/block/zram/zram_drv.c
> +++ b/drivers/block/zram/zram_drv.c
> @@ -95,6 +95,17 @@ static void zram_clear_flag(struct zram_meta *meta, u32 index,
>  	meta->table[index].value &= ~BIT(flag);
>  }
>  
> +static inline void zram_set_element(struct zram_meta *meta, u32 index,
> +			unsigned long element)
> +{
> +	meta->table[index].element = element;
> +}
> +
> +static inline void zram_clear_element(struct zram_meta *meta, u32 index)
> +{
> +	meta->table[index].element = 0;
> +}
> +
>  static size_t zram_get_obj_size(struct zram_meta *meta, u32 index)
>  {
>  	return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1);
> @@ -167,31 +178,78 @@ static inline void update_used_max(struct zram *zram,
>  	} while (old_max != cur_max);
>  }
>  
> -static bool page_zero_filled(void *ptr)
> +static inline void zram_fill_page(char *ptr, unsigned long value)
> +{
> +	int i;
> +	unsigned long *page = (unsigned long *)ptr;
> +
> +	if (likely(value == 0)) {
> +		clear_page(ptr);
> +	} else {
> +		for (i = 0; i < PAGE_SIZE / sizeof(*page); i++)
> +			page[i] = value;
> +	}
> +}
> +
> +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
> +		unsigned long value)
> +{
> +	int i;
> +	unsigned long *page;
> +
> +	if (likely(value == 0)) {
> +		memset(ptr, 0, size);
> +		return;
> +	}
> +
> +	i = ((unsigned long)ptr) % sizeof(*page);
> +	if (i) {
> +		while (i < sizeof(*page)) {
> +			*ptr++ = (value >> (i * 8)) & 0xff;
> +			--size;
> +			++i;
> +		}
> +	}
> +

I don't think we need this part because block layer works with sector
size or multiple times of it so it must be aligned unsigned long.

[toc] | [prev] | [next] | [standalone]

#1574318

On 2017/2/5 22:21, Minchan Kim wrote:
> Hi zhouxianrong,
>
> On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
>> From: zhouxianrong <zhouxianrong@huawei.com>
>>
>> test result as listed below:
>>
>> zero   pattern_char pattern_short pattern_int pattern_long total   (unit)
>> 162989 14454        3534          23516       2769         3294399 (page)
>>
>> statistics for the result:
>>
>>         zero  pattern_char  pattern_short  pattern_int  pattern_long
>> AVERAGE 0.745696298 0.085937175 0.015957701 0.131874915 0.020533911
>> STDEV   0.035623777 0.016892402 0.004454534 0.021657123 0.019420072
>> MAX     0.973813421 0.222222222 0.021409518 0.211812245 0.176512625
>> MIN     0.645431905 0.004634398 0           0           0
>
> The description in old version was better for justifying same page merging
> feature.
>
>>
>> Signed-off-by: zhouxianrong <zhouxianrong@huawei.com>
>> ---
>>  drivers/block/zram/zram_drv.c |  124 +++++++++++++++++++++++++++++++----------
>>  drivers/block/zram/zram_drv.h |   11 ++--
>>  2 files changed, 103 insertions(+), 32 deletions(-)
>>
>> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
>> index e5ab7d9..6a8c9c5 100644
>> --- a/drivers/block/zram/zram_drv.c
>> +++ b/drivers/block/zram/zram_drv.c
>> @@ -95,6 +95,17 @@ static void zram_clear_flag(struct zram_meta *meta, u32 index,
>>  	meta->table[index].value &= ~BIT(flag);
>>  }
>>
>> +static inline void zram_set_element(struct zram_meta *meta, u32 index,
>> +			unsigned long element)
>> +{
>> +	meta->table[index].element = element;
>> +}
>> +
>> +static inline void zram_clear_element(struct zram_meta *meta, u32 index)
>> +{
>> +	meta->table[index].element = 0;
>> +}
>> +
>>  static size_t zram_get_obj_size(struct zram_meta *meta, u32 index)
>>  {
>>  	return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1);
>> @@ -167,31 +178,78 @@ static inline void update_used_max(struct zram *zram,
>>  	} while (old_max != cur_max);
>>  }
>>
>> -static bool page_zero_filled(void *ptr)
>> +static inline void zram_fill_page(char *ptr, unsigned long value)
>> +{
>> +	int i;
>> +	unsigned long *page = (unsigned long *)ptr;
>> +
>> +	if (likely(value == 0)) {
>> +		clear_page(ptr);
>> +	} else {
>> +		for (i = 0; i < PAGE_SIZE / sizeof(*page); i++)
>> +			page[i] = value;
>> +	}
>> +}
>> +
>> +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
>> +		unsigned long value)
>> +{
>> +	int i;
>> +	unsigned long *page;
>> +
>> +	if (likely(value == 0)) {
>> +		memset(ptr, 0, size);
>> +		return;
>> +	}
>> +
>> +	i = ((unsigned long)ptr) % sizeof(*page);
>> +	if (i) {
>> +		while (i < sizeof(*page)) {
>> +			*ptr++ = (value >> (i * 8)) & 0xff;
>> +			--size;
>> +			++i;
>> +		}
>> +	}
>> +
>
> I don't think we need this part because block layer works with sector
> size or multiple times of it so it must be aligned unsigned long.
>
>
>
>
> .
>

Minchan and Matthew Wilcox:

1. right, but users could open /dev/block/zram0 file and do any read operations.

2. about endian operation for long, the modification is trivial and low efficient.
    i have not better method. do you have any good idea for this?

3. the below should be modified.

static inline bool zram_meta_get(struct zram *zram)
@@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)

  	/* Free all pages that are still in this zram device */
  	for (index = 0; index < num_pages; index++) {
-		unsigned long handle = meta->table[index].handle;
+		unsigned long handle;
+
+		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
+		handle = meta->table[index].handle;

-		if (!handle)
+		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
+			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
  			continue;
+		}

+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
  		zs_free(meta->mem_pool, handle);
  	}

@@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
  static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
  {
  	size_t num_pages;
-	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
+	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);

[toc] | [prev] | [next] | [standalone]

#1574817

On Mon, Feb 06, 2017 at 09:28:18AM +0800, zhouxianrong wrote:
> > > +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
> > > +		unsigned long value)
> > > +{
> > > +	int i;
> > > +	unsigned long *page;
> > > +
> > > +	if (likely(value == 0)) {
> > > +		memset(ptr, 0, size);
> > > +		return;
> > > +	}
> > > +
> > > +	i = ((unsigned long)ptr) % sizeof(*page);
> > > +	if (i) {
> > > +		while (i < sizeof(*page)) {
> > > +			*ptr++ = (value >> (i * 8)) & 0xff;
> > > +			--size;
> > > +			++i;
> > > +		}
> > > +	}
> > > +
> > 
> > I don't think we need this part because block layer works with sector
> > size or multiple times of it so it must be aligned unsigned long.
> 
> Minchan and Matthew Wilcox:
> 
> 1. right, but users could open /dev/block/zram0 file and do any read operations.

But any such read operation would go through the page cache, so will
be page aligned.  Unless they do an O_DIRECT operation, in which case
it must be aligned to block size.  Please, try it.  I/Os which are not
aligned should be failed long before they reach your driver.

[toc] | [prev] | [next] | [standalone]

#1575253

Hi

On Mon, Feb 06, 2017 at 09:28:18AM +0800, zhouxianrong wrote:
> 
> 
> On 2017/2/5 22:21, Minchan Kim wrote:
> >Hi zhouxianrong,
> >
> >On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
> >>From: zhouxianrong <zhouxianrong@huawei.com>
> >>
> >>test result as listed below:
> >>
> >>zero   pattern_char pattern_short pattern_int pattern_long total   (unit)
> >>162989 14454        3534          23516       2769         3294399 (page)
> >>
> >>statistics for the result:
> >>
> >>        zero  pattern_char  pattern_short  pattern_int  pattern_long
> >>AVERAGE 0.745696298 0.085937175 0.015957701 0.131874915 0.020533911
> >>STDEV   0.035623777 0.016892402 0.004454534 0.021657123 0.019420072
> >>MAX     0.973813421 0.222222222 0.021409518 0.211812245 0.176512625
> >>MIN     0.645431905 0.004634398 0           0           0
> >
> >The description in old version was better for justifying same page merging
> >feature.
> >
> >>
> >>Signed-off-by: zhouxianrong <zhouxianrong@huawei.com>
> >>---
> >> drivers/block/zram/zram_drv.c |  124 +++++++++++++++++++++++++++++++----------
> >> drivers/block/zram/zram_drv.h |   11 ++--
> >> 2 files changed, 103 insertions(+), 32 deletions(-)
> >>
> >>diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
> >>index e5ab7d9..6a8c9c5 100644
> >>--- a/drivers/block/zram/zram_drv.c
> >>+++ b/drivers/block/zram/zram_drv.c
> >>@@ -95,6 +95,17 @@ static void zram_clear_flag(struct zram_meta *meta, u32 index,
> >> 	meta->table[index].value &= ~BIT(flag);
> >> }
> >>
> >>+static inline void zram_set_element(struct zram_meta *meta, u32 index,
> >>+			unsigned long element)
> >>+{
> >>+	meta->table[index].element = element;
> >>+}
> >>+
> >>+static inline void zram_clear_element(struct zram_meta *meta, u32 index)
> >>+{
> >>+	meta->table[index].element = 0;
> >>+}
> >>+
> >> static size_t zram_get_obj_size(struct zram_meta *meta, u32 index)
> >> {
> >> 	return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1);
> >>@@ -167,31 +178,78 @@ static inline void update_used_max(struct zram *zram,
> >> 	} while (old_max != cur_max);
> >> }
> >>
> >>-static bool page_zero_filled(void *ptr)
> >>+static inline void zram_fill_page(char *ptr, unsigned long value)
> >>+{
> >>+	int i;
> >>+	unsigned long *page = (unsigned long *)ptr;
> >>+
> >>+	if (likely(value == 0)) {
> >>+		clear_page(ptr);
> >>+	} else {
> >>+		for (i = 0; i < PAGE_SIZE / sizeof(*page); i++)
> >>+			page[i] = value;
> >>+	}
> >>+}
> >>+
> >>+static inline void zram_fill_page_partial(char *ptr, unsigned int size,
> >>+		unsigned long value)
> >>+{
> >>+	int i;
> >>+	unsigned long *page;
> >>+
> >>+	if (likely(value == 0)) {
> >>+		memset(ptr, 0, size);
> >>+		return;
> >>+	}
> >>+
> >>+	i = ((unsigned long)ptr) % sizeof(*page);
> >>+	if (i) {
> >>+		while (i < sizeof(*page)) {
> >>+			*ptr++ = (value >> (i * 8)) & 0xff;
> >>+			--size;
> >>+			++i;
> >>+		}
> >>+	}
> >>+
> >
> >I don't think we need this part because block layer works with sector
> >size or multiple times of it so it must be aligned unsigned long.
> >
> >
> >
> >
> >.
> >
> 
> Minchan and Matthew Wilcox:
> 
> 1. right, but users could open /dev/block/zram0 file and do any read operations.

Could you make that happen?
I don't think it's possible as Matthew already pointed out, too.

> 
> 2. about endian operation for long, the modification is trivial and low efficient.
>    i have not better method. do you have any good idea for this?

So if assumption 1 is wrong, we don't need 2, either.

> 
> 3. the below should be modified.
> 
> static inline bool zram_meta_get(struct zram *zram)
> @@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
> 
>  	/* Free all pages that are still in this zram device */
>  	for (index = 0; index < num_pages; index++) {
> -		unsigned long handle = meta->table[index].handle;
> +		unsigned long handle;
> +
> +		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
> +		handle = meta->table[index].handle;
> 
> -		if (!handle)
> +		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
> +			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>  			continue;
> +		}
> 
> +		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>  		zs_free(meta->mem_pool, handle);

Could you explain why we need this modification?

>  	}
> 
> @@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>  static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
>  {
>  	size_t num_pages;
> -	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
> +	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);

Ditto

> 
> 

[toc] | [prev] | [next] | [standalone]

#1575326

On 2017/2/7 7:48, Minchan Kim wrote:
> Hi
>
> On Mon, Feb 06, 2017 at 09:28:18AM +0800, zhouxianrong wrote:
>>
>>
>> On 2017/2/5 22:21, Minchan Kim wrote:
>>> Hi zhouxianrong,
>>>
>>> On Fri, Feb 03, 2017 at 04:42:27PM +0800, zhouxianrong@huawei.com wrote:
>>>> From: zhouxianrong <zhouxianrong@huawei.com>
>>>>
>>>> test result as listed below:
>>>>
>>>> zero   pattern_char pattern_short pattern_int pattern_long total   (unit)
>>>> 162989 14454        3534          23516       2769         3294399 (page)
>>>>
>>>> statistics for the result:
>>>>
>>>>        zero  pattern_char  pattern_short  pattern_int  pattern_long
>>>> AVERAGE 0.745696298 0.085937175 0.015957701 0.131874915 0.020533911
>>>> STDEV   0.035623777 0.016892402 0.004454534 0.021657123 0.019420072
>>>> MAX     0.973813421 0.222222222 0.021409518 0.211812245 0.176512625
>>>> MIN     0.645431905 0.004634398 0           0           0
>>>
>>> The description in old version was better for justifying same page merging
>>> feature.
>>>
>>>>
>>>> Signed-off-by: zhouxianrong <zhouxianrong@huawei.com>
>>>> ---
>>>> drivers/block/zram/zram_drv.c |  124 +++++++++++++++++++++++++++++++----------
>>>> drivers/block/zram/zram_drv.h |   11 ++--
>>>> 2 files changed, 103 insertions(+), 32 deletions(-)
>>>>
>>>> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
>>>> index e5ab7d9..6a8c9c5 100644
>>>> --- a/drivers/block/zram/zram_drv.c
>>>> +++ b/drivers/block/zram/zram_drv.c
>>>> @@ -95,6 +95,17 @@ static void zram_clear_flag(struct zram_meta *meta, u32 index,
>>>> 	meta->table[index].value &= ~BIT(flag);
>>>> }
>>>>
>>>> +static inline void zram_set_element(struct zram_meta *meta, u32 index,
>>>> +			unsigned long element)
>>>> +{
>>>> +	meta->table[index].element = element;
>>>> +}
>>>> +
>>>> +static inline void zram_clear_element(struct zram_meta *meta, u32 index)
>>>> +{
>>>> +	meta->table[index].element = 0;
>>>> +}
>>>> +
>>>> static size_t zram_get_obj_size(struct zram_meta *meta, u32 index)
>>>> {
>>>> 	return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1);
>>>> @@ -167,31 +178,78 @@ static inline void update_used_max(struct zram *zram,
>>>> 	} while (old_max != cur_max);
>>>> }
>>>>
>>>> -static bool page_zero_filled(void *ptr)
>>>> +static inline void zram_fill_page(char *ptr, unsigned long value)
>>>> +{
>>>> +	int i;
>>>> +	unsigned long *page = (unsigned long *)ptr;
>>>> +
>>>> +	if (likely(value == 0)) {
>>>> +		clear_page(ptr);
>>>> +	} else {
>>>> +		for (i = 0; i < PAGE_SIZE / sizeof(*page); i++)
>>>> +			page[i] = value;
>>>> +	}
>>>> +}
>>>> +
>>>> +static inline void zram_fill_page_partial(char *ptr, unsigned int size,
>>>> +		unsigned long value)
>>>> +{
>>>> +	int i;
>>>> +	unsigned long *page;
>>>> +
>>>> +	if (likely(value == 0)) {
>>>> +		memset(ptr, 0, size);
>>>> +		return;
>>>> +	}
>>>> +
>>>> +	i = ((unsigned long)ptr) % sizeof(*page);
>>>> +	if (i) {
>>>> +		while (i < sizeof(*page)) {
>>>> +			*ptr++ = (value >> (i * 8)) & 0xff;
>>>> +			--size;
>>>> +			++i;
>>>> +		}
>>>> +	}
>>>> +
>>>
>>> I don't think we need this part because block layer works with sector
>>> size or multiple times of it so it must be aligned unsigned long.
>>>
>>>
>>>
>>>
>>> .
>>>
>>
>> Minchan and Matthew Wilcox:
>>
>> 1. right, but users could open /dev/block/zram0 file and do any read operations.
>
> Could you make that happen?
> I don't think it's possible as Matthew already pointed out, too.

yes, Matthew is right, thanks

>
>>
>> 2. about endian operation for long, the modification is trivial and low efficient.
>>    i have not better method. do you have any good idea for this?
>
> So if assumption 1 is wrong, we don't need 2, either.

yes

>
>>
>> 3. the below should be modified.
>>
>> static inline bool zram_meta_get(struct zram *zram)
>> @@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>>
>>  	/* Free all pages that are still in this zram device */
>>  	for (index = 0; index < num_pages; index++) {
>> -		unsigned long handle = meta->table[index].handle;
>> +		unsigned long handle;
>> +
>> +		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
>> +		handle = meta->table[index].handle;
>>
>> -		if (!handle)
>> +		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
>> +			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>>  			continue;
>> +		}
>>
>> +		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>>  		zs_free(meta->mem_pool, handle);
>
> Could you explain why we need this modification?
>
>>  	}
>>
>> @@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>>  static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
>>  {
>>  	size_t num_pages;
>> -	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
>> +	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
>
> Ditto
>
>>
>>
>
> .
>

because of union of handle and element, i think a non-zero element (other than handle) is prevented from freeing.
if zram_meta_get was modified, zram_meta_alloc did so.

[toc] | [prev] | [next] | [standalone]

#1575342

On Tue, Feb 07, 2017 at 10:20:57AM +0800, zhouxianrong wrote:

< snip >

> >>3. the below should be modified.
> >>
> >>static inline bool zram_meta_get(struct zram *zram)
> >>@@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
> >>
> >> 	/* Free all pages that are still in this zram device */
> >> 	for (index = 0; index < num_pages; index++) {
> >>-		unsigned long handle = meta->table[index].handle;
> >>+		unsigned long handle;
> >>+
> >>+		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
> >>+		handle = meta->table[index].handle;
> >>
> >>-		if (!handle)
> >>+		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
> >>+			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
> >> 			continue;
> >>+		}
> >>
> >>+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
> >> 		zs_free(meta->mem_pool, handle);
> >
> >Could you explain why we need this modification?
> >
> >> 	}
> >>
> >>@@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
> >> static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
> >> {
> >> 	size_t num_pages;
> >>-	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
> >>+	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
> >
> >Ditto
> >
> >>
> >>
> >
> >.
> >
> 
> because of union of handle and element, i think a non-zero element (other than handle) is prevented from freeing.
> if zram_meta_get was modified, zram_meta_alloc did so.

Right. Thanks but I don't see why we need the locking in there and modification of
zram_meta_alloc.

Isn't it enough with this?

diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index c20b05a84f21..a25d34a8af19 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -425,8 +425,11 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
 	/* Free all pages that are still in this zram device */
 	for (index = 0; index < num_pages; index++) {
 		unsigned long handle = meta->table[index].handle;
-
-		if (!handle)
+		/*
+		 * No memory is allocated for same element filled pages.
+		 * Simply clear same page flag.
+		 */
+		if (!handle || zram_test_flag(meta, index, ZRAM_SAME))
 			continue;
 
 		zs_free(meta->mem_pool, handle);

[toc] | [prev] | [next] | [standalone]

#1575356

On 2017/2/7 10:54, Minchan Kim wrote:
> On Tue, Feb 07, 2017 at 10:20:57AM +0800, zhouxianrong wrote:
>
> < snip >
>
>>>> 3. the below should be modified.
>>>>
>>>> static inline bool zram_meta_get(struct zram *zram)
>>>> @@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>>>>
>>>> 	/* Free all pages that are still in this zram device */
>>>> 	for (index = 0; index < num_pages; index++) {
>>>> -		unsigned long handle = meta->table[index].handle;
>>>> +		unsigned long handle;
>>>> +
>>>> +		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
>>>> +		handle = meta->table[index].handle;
>>>>
>>>> -		if (!handle)
>>>> +		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
>>>> +			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>>>> 			continue;
>>>> +		}
>>>>
>>>> +		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
>>>> 		zs_free(meta->mem_pool, handle);
>>>
>>> Could you explain why we need this modification?
>>>
>>>> 	}
>>>>
>>>> @@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>>>> static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
>>>> {
>>>> 	size_t num_pages;
>>>> -	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
>>>> +	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
>>>
>>> Ditto
>>>
>>>>
>>>>
>>>
>>> .
>>>
>>
>> because of union of handle and element, i think a non-zero element (other than handle) is prevented from freeing.
>> if zram_meta_get was modified, zram_meta_alloc did so.
>
> Right. Thanks but I don't see why we need the locking in there and modification of
> zram_meta_alloc.
>
> Isn't it enough with this?

i am afraid someone do reset_store, so did lock.

yes, i am wrong, zram_meta_alloc should not be modified here. because meta->table has already cleared

	meta->table = vzalloc(num_pages * sizeof(*meta->table));




>
> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
> index c20b05a84f21..a25d34a8af19 100644
> --- a/drivers/block/zram/zram_drv.c
> +++ b/drivers/block/zram/zram_drv.c
> @@ -425,8 +425,11 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
>  	/* Free all pages that are still in this zram device */
>  	for (index = 0; index < num_pages; index++) {
>  		unsigned long handle = meta->table[index].handle;
> -
> -		if (!handle)
> +		/*
> +		 * No memory is allocated for same element filled pages.
> +		 * Simply clear same page flag.
> +		 */
> +		if (!handle || zram_test_flag(meta, index, ZRAM_SAME))
>  			continue;
>
>  		zs_free(meta->mem_pool, handle);
>
> .
>

[toc] | [prev] | [next] | [standalone]

#1575381

On Tue, Feb 07, 2017 at 11:24:40AM +0800, zhouxianrong wrote:
> 
> 
> On 2017/2/7 10:54, Minchan Kim wrote:
> >On Tue, Feb 07, 2017 at 10:20:57AM +0800, zhouxianrong wrote:
> >
> >< snip >
> >
> >>>>3. the below should be modified.
> >>>>
> >>>>static inline bool zram_meta_get(struct zram *zram)
> >>>>@@ -495,11 +553,17 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
> >>>>
> >>>>	/* Free all pages that are still in this zram device */
> >>>>	for (index = 0; index < num_pages; index++) {
> >>>>-		unsigned long handle = meta->table[index].handle;
> >>>>+		unsigned long handle;
> >>>>+
> >>>>+		bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value);
> >>>>+		handle = meta->table[index].handle;
> >>>>
> >>>>-		if (!handle)
> >>>>+		if (!handle || zram_test_flag(meta, index, ZRAM_SAME)) {
> >>>>+			bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
> >>>>			continue;
> >>>>+		}
> >>>>
> >>>>+		bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
> >>>>		zs_free(meta->mem_pool, handle);
> >>>
> >>>Could you explain why we need this modification?
> >>>
> >>>>	}
> >>>>
> >>>>@@ -511,7 +575,7 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
> >>>>static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
> >>>>{
> >>>>	size_t num_pages;
> >>>>-	struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
> >>>>+	struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
> >>>
> >>>Ditto
> >>>
> >>>>
> >>>>
> >>>
> >>>.
> >>>
> >>
> >>because of union of handle and element, i think a non-zero element (other than handle) is prevented from freeing.
> >>if zram_meta_get was modified, zram_meta_alloc did so.
> >
> >Right. Thanks but I don't see why we need the locking in there and modification of
> >zram_meta_alloc.
> >
> >Isn't it enough with this?
> 
> i am afraid someone do reset_store, so did lock.

reset_store is protected by zram->claim and and init_done so I don't
think so.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web