Re: [PATCH 3/3] zram: rework reported to end-user zram statistics

[Jerome Marchand <jmarchan@redhat.com>]

On 01/14/2014 11:57 AM, Sergey Senozhatsky wrote:
> 
> Hello Jerome,
> 
> On (01/14/14 11:38), Jerome Marchand wrote:
>> On 01/14/2014 10:37 AM, Sergey Senozhatsky wrote:
>>> 1) Introduce ZRAM_ATTR_RO macro to generate zram atomic64_t stats
>>> `show' functions and reduce code duplication.
>>>
>>> 2) Account and report back to user numbers of failed READ and WRITE
>>> operations.
>>>
>>> 3) Remove `good' and `bad' compressed sub-requests stats. RW request may
>>> cause a number of RW sub-requests. zram used to account `good' compressed
>>> sub-queries (with compressed size less than 50% of original size), `bad'
>>> compressed sub-queries (with compressed size greater that 75% of original
>>> size), leaving sub-requests with compression size between 50% and 75% of
>>> original size not accounted and not reported.
>>
>> That's weird: good/bad_compress are accounted, but it seems to me that
>> they are to never used in any way. If so, there is indeed no reason to
>> keep them.
>>
>>
>>> Account each sub-request
>>> compression size so we can calculate real device compression ratio.
>>
>> Your patch doesn't change the way pages_stored and compr[essed]_size
>> are accounted. What does your patch change that allow us to calculate
>> the "real" compression ratio?
> 
> we have compressed size, number of stored pages and reported by zs pool
> (as a zram memory_used attr) number of bytes used
> 
> u64 zs_get_total_size_bytes(struct zs_pool *pool)
> {
>         int i;
>         u64 npages = 0;
> 
>         for (i = 0; i < ZS_SIZE_CLASSES; i++)
>                 npages += pool->size_class[i].pages_allocated;
> 
>         return npages << PAGE_SHIFT;
> }
> 
> looks enough to calculate device overall data compression ratio.

Yes. But don't we have all that already without your patch applied?
What does this patch change?

> 
> 	-ss
> 
>>>
>>> 4) reported zram stats:
>>>   - num_writes  -- number of writes
>>>   - num_reads  -- number of reads
>>>   - pages_stored -- number of pages currently stored
>>>   - compressed_size -- compressed size of pages stored
>>
>> Wouldn't it be more practical to report the original and compressed
>> data sizes using the same units as it is currently done?
>>
> 
> sorry, not sure I understand.

Currently users have access to orig_data_size and compr_data_size,
both in bytes. With your patch, they have access to pages_stored in
pages and compressed_size in bytes. I find the current set more
practical.

Jerome

> 
>> Jerome
>>
>>>   - pages_zero  -- number of zero filled pages
>>>   - failed_read -- number of failed reads
>>>   - failed_writes -- can happen when memory is too low
>>>   - invalid_io   -- non-page-aligned I/O requests
>>>   - notify_free  -- number of swap slot free notifications
>>>   - memory_used -- zs pool zs_get_total_size_bytes()
>>>
>>> Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
>>> ---
>>>  drivers/block/zram/zram_drv.c | 167 ++++++++++++------------------------------
>>>  drivers/block/zram/zram_drv.h |  17 ++---
>>>  2 files changed, 54 insertions(+), 130 deletions(-)
>>>
>>> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
>>> index 2a7682c..8bddaff 100644
>>> --- a/drivers/block/zram/zram_drv.c
>>> +++ b/drivers/block/zram/zram_drv.c
>>> @@ -42,6 +42,17 @@ static struct zram *zram_devices;
>>>  /* Module params (documentation at end) */
>>>  static unsigned int num_devices = 1;
>>>  
>>> +#define ZRAM_ATTR_RO(name)						\
>>> +static ssize_t zram_attr_##name##_show(struct device *d,		\
>>> +				struct device_attribute *attr, char *b)	\
>>> +{									\
>>> +	struct zram *zram = dev_to_zram(d);				\
>>> +	return sprintf(b, "%llu\n",					\
>>> +		(u64)atomic64_read(&zram->stats.name));			\
>>> +}									\
>>> +static struct device_attribute dev_attr_##name =			\
>>> +	__ATTR(name, S_IRUGO, zram_attr_##name##_show, NULL);
>>> +
>>>  static inline int init_done(struct zram *zram)
>>>  {
>>>  	return zram->meta != NULL;
>>> @@ -52,97 +63,36 @@ static inline struct zram *dev_to_zram(struct device *dev)
>>>  	return (struct zram *)dev_to_disk(dev)->private_data;
>>>  }
>>>  
>>> -static ssize_t disksize_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n", zram->disksize);
>>> -}
>>> -
>>> -static ssize_t initstate_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%u\n", init_done(zram));
>>> -}
>>> -
>>> -static ssize_t num_reads_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n",
>>> -			(u64)atomic64_read(&zram->stats.num_reads));
>>> -}
>>> -
>>> -static ssize_t num_writes_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n",
>>> -			(u64)atomic64_read(&zram->stats.num_writes));
>>> -}
>>> -
>>> -static ssize_t invalid_io_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n",
>>> -			(u64)atomic64_read(&zram->stats.invalid_io));
>>> -}
>>> -
>>> -static ssize_t notify_free_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n",
>>> -			(u64)atomic64_read(&zram->stats.notify_free));
>>> -}
>>> -
>>> -static ssize_t zero_pages_show(struct device *dev,
>>> -		struct device_attribute *attr, char *buf)
>>> -{
>>> -	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%u\n", atomic_read(&zram->stats.pages_zero));
>>> -}
>>> -
>>> -static ssize_t orig_data_size_show(struct device *dev,
>>> +static ssize_t memory_used_show(struct device *dev,
>>>  		struct device_attribute *attr, char *buf)
>>>  {
>>> +	u64 val = 0;
>>>  	struct zram *zram = dev_to_zram(dev);
>>> +	struct zram_meta *meta = zram->meta;
>>>  
>>> -	return sprintf(buf, "%llu\n",
>>> -		(u64)(atomic_read(&zram->stats.pages_stored)) << PAGE_SHIFT);
>>> +	down_read(&zram->init_lock);
>>> +	if (init_done(zram))
>>> +		val = zs_get_total_size_bytes(meta->mem_pool);
>>> +	up_read(&zram->init_lock);
>>> +	return sprintf(buf, "%llu\n", val);
>>>  }
>>>  
>>> -static ssize_t compr_data_size_show(struct device *dev,
>>> +static ssize_t disksize_show(struct device *dev,
>>>  		struct device_attribute *attr, char *buf)
>>>  {
>>>  	struct zram *zram = dev_to_zram(dev);
>>> -
>>> -	return sprintf(buf, "%llu\n",
>>> -			(u64)atomic64_read(&zram->stats.compr_size));
>>> +	return sprintf(buf, "%llu\n", zram->disksize);
>>>  }
>>>  
>>> -static ssize_t mem_used_total_show(struct device *dev,
>>> +static ssize_t initstate_show(struct device *dev,
>>>  		struct device_attribute *attr, char *buf)
>>>  {
>>> -	u64 val = 0;
>>> +	u32 val = 0;
>>>  	struct zram *zram = dev_to_zram(dev);
>>> -	struct zram_meta *meta = zram->meta;
>>> -
>>>  	down_read(&zram->init_lock);
>>> -	if (init_done(zram))
>>> -		val = zs_get_total_size_bytes(meta->mem_pool);
>>> +	val = init_done(zram);
>>>  	up_read(&zram->init_lock);
>>> -
>>> -	return sprintf(buf, "%llu\n", val);
>>> +	return sprintf(buf, "%u\n", val);
>>>  }
>>>  
>>>  /* flag operations needs meta->tb_lock */
>>> @@ -293,7 +243,6 @@ static void zram_free_page(struct zram *zram, size_t index)
>>>  {
>>>  	struct zram_meta *meta = zram->meta;
>>>  	unsigned long handle = meta->table[index].handle;
>>> -	u16 size = meta->table[index].size;
>>>  
>>>  	if (unlikely(!handle)) {
>>>  		/*
>>> @@ -302,21 +251,15 @@ static void zram_free_page(struct zram *zram, size_t index)
>>>  		 */
>>>  		if (zram_test_flag(meta, index, ZRAM_ZERO)) {
>>>  			zram_clear_flag(meta, index, ZRAM_ZERO);
>>> -			atomic_dec(&zram->stats.pages_zero);
>>> +			atomic64_dec(&zram->stats.pages_zero);
>>>  		}
>>>  		return;
>>>  	}
>>>  
>>> -	if (unlikely(size > max_zpage_size))
>>> -		atomic_dec(&zram->stats.bad_compress);
>>> -
>>>  	zs_free(meta->mem_pool, handle);
>>>  
>>> -	if (size <= PAGE_SIZE / 2)
>>> -		atomic_dec(&zram->stats.good_compress);
>>> -
>>> -	atomic64_sub(meta->table[index].size, &zram->stats.compr_size);
>>> -	atomic_dec(&zram->stats.pages_stored);
>>> +	atomic64_sub(meta->table[index].size, &zram->stats.compressed_size);
>>> +	atomic64_dec(&zram->stats.pages_stored);
>>>  
>>>  	meta->table[index].handle = 0;
>>>  	meta->table[index].size = 0;
>>> @@ -362,7 +305,7 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
>>>  static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
>>>  			  u32 index, int offset, struct bio *bio)
>>>  {
>>> -	int ret;
>>> +	int ret = -EINVAL;
>>>  	struct page *page;
>>>  	unsigned char *user_mem, *uncmem = NULL;
>>>  	struct zram_meta *meta = zram->meta;
>>> @@ -406,6 +349,8 @@ out_cleanup:
>>>  	kunmap_atomic(user_mem);
>>>  	if (is_partial_io(bvec))
>>>  		kfree(uncmem);
>>> +	if (ret)
>>> +		atomic64_inc(&zram->stats.failed_reads);
>>>  	return ret;
>>>  }
>>>  
>>> @@ -459,7 +404,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
>>>  		zram_set_flag(meta, index, ZRAM_ZERO);
>>>  		write_unlock(&zram->meta->tb_lock);
>>>  
>>> -		atomic_inc(&zram->stats.pages_zero);
>>> +		atomic64_inc(&zram->stats.pages_zero);
>>>  		ret = 0;
>>>  		goto out;
>>>  	}
>>> @@ -478,7 +423,6 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
>>>  	}
>>>  
>>>  	if (unlikely(clen > max_zpage_size)) {
>>> -		atomic_inc(&zram->stats.bad_compress);
>>>  		clen = PAGE_SIZE;
>>>  		src = NULL;
>>>  		if (is_partial_io(bvec))
>>> @@ -516,11 +460,8 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
>>>  	write_unlock(&zram->meta->tb_lock);
>>>  
>>>  	/* Update stats */
>>> -	atomic64_add(clen, &zram->stats.compr_size);
>>> -	atomic_inc(&zram->stats.pages_stored);
>>> -	if (clen <= PAGE_SIZE / 2)
>>> -		atomic_inc(&zram->stats.good_compress);
>>> -
>>> +	atomic64_add(clen, &zram->stats.compressed_size);
>>> +	atomic64_inc(&zram->stats.pages_stored);
>>>  out:
>>>  	if (locked)
>>>  		mutex_unlock(&meta->buffer_lock);
>>> @@ -586,23 +527,8 @@ static void zram_reset_device(struct zram *zram, bool reset_capacity)
>>>  
>>>  static void zram_init_device(struct zram *zram, struct zram_meta *meta)
>>>  {
>>> -	if (zram->disksize > 2 * (totalram_pages << PAGE_SHIFT)) {
>>> -		pr_info(
>>> -		"There is little point creating a zram of greater than "
>>> -		"twice the size of memory since we expect a 2:1 compression "
>>> -		"ratio. Note that zram uses about 0.1%% of the size of "
>>> -		"the disk when not in use so a huge zram is "
>>> -		"wasteful.\n"
>>> -		"\tMemory Size: %lu kB\n"
>>> -		"\tSize you selected: %llu kB\n"
>>> -		"Continuing anyway ...\n",
>>> -		(totalram_pages << PAGE_SHIFT) >> 10, zram->disksize >> 10
>>> -		);
>>> -	}
>>> -
>>>  	/* zram devices sort of resembles non-rotational disks */
>>>  	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue);
>>> -
>>>  	zram->meta = meta;
>>>  	pr_debug("Initialization done!\n");
>>>  }
>>> @@ -774,14 +700,15 @@ static DEVICE_ATTR(disksize, S_IRUGO | S_IWUSR,
>>>  		disksize_show, disksize_store);
>>>  static DEVICE_ATTR(initstate, S_IRUGO, initstate_show, NULL);
>>>  static DEVICE_ATTR(reset, S_IWUSR, NULL, reset_store);
>>> -static DEVICE_ATTR(num_reads, S_IRUGO, num_reads_show, NULL);
>>> -static DEVICE_ATTR(num_writes, S_IRUGO, num_writes_show, NULL);
>>> -static DEVICE_ATTR(invalid_io, S_IRUGO, invalid_io_show, NULL);
>>> -static DEVICE_ATTR(notify_free, S_IRUGO, notify_free_show, NULL);
>>> -static DEVICE_ATTR(zero_pages, S_IRUGO, zero_pages_show, NULL);
>>> -static DEVICE_ATTR(orig_data_size, S_IRUGO, orig_data_size_show, NULL);
>>> -static DEVICE_ATTR(compr_data_size, S_IRUGO, compr_data_size_show, NULL);
>>> -static DEVICE_ATTR(mem_used_total, S_IRUGO, mem_used_total_show, NULL);
>>> +static DEVICE_ATTR(memory_used, S_IRUGO, memory_used_show, NULL);
>>> +
>>> +ZRAM_ATTR_RO(num_reads);
>>> +ZRAM_ATTR_RO(num_writes);
>>> +ZRAM_ATTR_RO(pages_stored);
>>> +ZRAM_ATTR_RO(invalid_io);
>>> +ZRAM_ATTR_RO(notify_free);
>>> +ZRAM_ATTR_RO(pages_zero);
>>> +ZRAM_ATTR_RO(compressed_size);
>>>  
>>>  static struct attribute *zram_disk_attrs[] = {
>>>  	&dev_attr_disksize.attr,
>>> @@ -789,12 +716,12 @@ static struct attribute *zram_disk_attrs[] = {
>>>  	&dev_attr_reset.attr,
>>>  	&dev_attr_num_reads.attr,
>>>  	&dev_attr_num_writes.attr,
>>> +	&dev_attr_pages_stored.attr,
>>>  	&dev_attr_invalid_io.attr,
>>>  	&dev_attr_notify_free.attr,
>>> -	&dev_attr_zero_pages.attr,
>>> -	&dev_attr_orig_data_size.attr,
>>> -	&dev_attr_compr_data_size.attr,
>>> -	&dev_attr_mem_used_total.attr,
>>> +	&dev_attr_pages_zero.attr,
>>> +	&dev_attr_compressed_size.attr,
>>> +	&dev_attr_memory_used.attr,
>>>  	NULL,
>>>  };
>>>  
>>> diff --git a/drivers/block/zram/zram_drv.h b/drivers/block/zram/zram_drv.h
>>> index e81e9cd..277023d 100644
>>> --- a/drivers/block/zram/zram_drv.h
>>> +++ b/drivers/block/zram/zram_drv.h
>>> @@ -64,22 +64,19 @@ enum zram_pageflags {
>>>  struct table {
>>>  	unsigned long handle;
>>>  	u16 size;	/* object size (excluding header) */
>>> -	u8 count;	/* object ref count (not yet used) */
>>> -	u8 flags;
>>> +	u16 flags;
>>>  } __aligned(4);
>>>  
>>>  struct zram_stats {
>>> -	atomic64_t compr_size;	/* compressed size of pages stored */
>>> -	atomic64_t num_reads;	/* failed + successful */
>>> -	atomic64_t num_writes;	/* --do-- */
>>> -	atomic64_t failed_reads;	/* should NEVER! happen */
>>> +	atomic64_t num_writes;	/* number of writes */
>>> +	atomic64_t num_reads;	/* number of reads */
>>> +	atomic64_t pages_stored;	/* no. of pages currently stored */
>>> +	atomic64_t compressed_size;	/* compressed size of pages stored */
>>> +	atomic64_t pages_zero;		/* no. of zero filled pages */
>>> +	atomic64_t failed_reads;	/* no. of failed reads */
>>>  	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 */
>>> -	atomic_t pages_zero;		/* no. of zero filled pages */
>>> -	atomic_t pages_stored;	/* no. of pages currently stored */
>>> -	atomic_t good_compress;	/* % of pages with compression ratio<=50% */
>>> -	atomic_t bad_compress;	/* % of pages with compression ratio>=75% */
>>>  };
>>>  
>>>  struct zram_meta {
>>>
>>