On 2020/7/8 2:23, Aravind Ramesh wrote:
> Thanks for review Chao Yu.
> Please find my response inline.
> I will re-send a V2 after incorporating your comments.
>
> Regards,
> Aravind
>
>> -----Original Message-----
>> From: Chao Yu <yuchao0@xxxxxxxxxx>
>> Sent: Tuesday, July 7, 2020 5:49 PM
>> To: Aravind Ramesh <Aravind.Ramesh@xxxxxxx>; jaegeuk@xxxxxxxxxx; linux-
>> fsdevel@xxxxxxxxxxxxxxx; linux-f2fs-devel@xxxxxxxxxxxxxxxxxxxxx; hch@xxxxxx
>> Cc: Damien Le Moal <Damien.LeMoal@xxxxxxx>; Niklas Cassel
>> <Niklas.Cassel@xxxxxxx>; Matias Bjorling <Matias.Bjorling@xxxxxxx>
>> Subject: Re: [PATCH 1/2] f2fs: support zone capacity less than zone size
>>
>> On 2020/7/2 23:54, Aravind Ramesh wrote:
>>> NVMe Zoned Namespace devices can have zone-capacity less than zone-size.
>>> Zone-capacity indicates the maximum number of sectors that are usable
>>> in a zone beginning from the first sector of the zone. This makes the
>>> sectors sectors after the zone-capacity till zone-size to be unusable.
>>> This patch set tracks zone-size and zone-capacity in zoned devices and
>>> calculate the usable blocks per segment and usable segments per section.
>>>
>>> If zone-capacity is less than zone-size mark only those segments which
>>> start before zone-capacity as free segments. All segments at and
>>> beyond zone-capacity are treated as permanently used segments. In
>>> cases where zone-capacity does not align with segment size the last
>>> segment will start before zone-capacity and end beyond the
>>> zone-capacity of the zone. For such spanning segments only sectors within the
>> zone-capacity are used.
>>>
>>> Signed-off-by: Aravind Ramesh <aravind.ramesh@xxxxxxx>
>>> Signed-off-by: Damien Le Moal <damien.lemoal@xxxxxxx>
>>> Signed-off-by: Niklas Cassel <niklas.cassel@xxxxxxx>
>>> ---
>>> fs/f2fs/f2fs.h | 5 ++
>>> fs/f2fs/segment.c | 136
>> ++++++++++++++++++++++++++++++++++++++++++++--
>>> fs/f2fs/segment.h | 6 +-
>>> fs/f2fs/super.c | 41 ++++++++++++--
>>> 4 files changed, 176 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index
>>> e6e47618a357..73219e4e1ba4 100644
>>> --- a/fs/f2fs/f2fs.h
>>> +++ b/fs/f2fs/f2fs.h
>>> @@ -1232,6 +1232,7 @@ struct f2fs_dev_info { #ifdef
>>> CONFIG_BLK_DEV_ZONED
>>> unsigned int nr_blkz; /* Total number of zones */
>>> unsigned long *blkz_seq; /* Bitmap indicating sequential zones */
>>> + block_t *zone_capacity_blocks; /* Array of zone capacity in blks */
>>> #endif
>>> };
>>>
>>> @@ -3395,6 +3396,10 @@ void f2fs_destroy_segment_manager_caches(void);
>>> int f2fs_rw_hint_to_seg_type(enum rw_hint hint); enum rw_hint
>>> f2fs_io_type_to_rw_hint(struct f2fs_sb_info *sbi,
>>> enum page_type type, enum temp_type temp);
>>> +unsigned int f2fs_usable_segs_in_sec(struct f2fs_sb_info *sbi,
>>> + unsigned int segno);
>>> +unsigned int f2fs_usable_blks_in_seg(struct f2fs_sb_info *sbi,
>>> + unsigned int segno);
>>>
>>> /*
>>> * checkpoint.c
>>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index
>>> c35614d255e1..d2156f3f56a5 100644
>>> --- a/fs/f2fs/segment.c
>>> +++ b/fs/f2fs/segment.c
>>> @@ -4294,9 +4294,12 @@ static void init_free_segmap(struct
>>> f2fs_sb_info *sbi) {
>>> unsigned int start;
>>> int type;
>>> + struct seg_entry *sentry;
>>>
>>> for (start = 0; start < MAIN_SEGS(sbi); start++) {
>>> - struct seg_entry *sentry = get_seg_entry(sbi, start);
>>> + if (f2fs_usable_blks_in_seg(sbi, start) == 0)
>>
>> If usable blocks count is zero, shouldn't we update SIT_I(sbi)->written_valid_blocks
>> as we did when there is partial usable block in current segment?
> If usable_block_count is zero, then it is like a dead segment, all blocks in the segment lie after the
> zone-capacity in the zone. So there can never be a valid written content on these segments, hence it is not updated.
> In the other case, when a segment start before the zone-capacity and it ends beyond zone-capacity, then there are
> some blocks before zone-capacity which can be used, so they are accounted for.
I'm thinking that for limit_free_user_blocks() function, it assumes all
unwritten blocks as potential reclaimable blocks, however segment after
zone-capacity should never be used or reclaimable, it looks calculation
could be not correct here.
static inline block_t limit_free_user_blocks(struct f2fs_sb_info *sbi)
{
block_t reclaimable_user_blocks = sbi->user_block_count -
written_block_count(sbi);
return (long)(reclaimable_user_blocks * LIMIT_FREE_BLOCK) / 100;
}
static inline bool has_enough_invalid_blocks(struct f2fs_sb_info *sbi)
{
block_t invalid_user_blocks = sbi->user_block_count -
written_block_count(sbi);
/*
* Background GC is triggered with the following conditions.
* 1. There are a number of invalid blocks.
* 2. There is not enough free space.
*/
if (invalid_user_blocks > limit_invalid_user_blocks(sbi) &&
free_user_blocks(sbi) < limit_free_user_blocks(sbi))
-- In this condition, free_user_blocks() doesn't include segments after
zone-capacity, however limit_free_user_blocks() includes them.
return true;
return false;
}
>>
>>> + continue;
>>> + sentry = get_seg_entry(sbi, start);
>>> if (!sentry->valid_blocks)
>>> __set_free(sbi, start);
>>> else
>>> @@ -4316,7 +4319,7 @@ static void init_dirty_segmap(struct f2fs_sb_info *sbi)
>>> struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
>>> struct free_segmap_info *free_i = FREE_I(sbi);
>>> unsigned int segno = 0, offset = 0, secno;
>>> - unsigned short valid_blocks;
>>> + unsigned short valid_blocks, usable_blks_in_seg;
>>> unsigned short blks_per_sec = BLKS_PER_SEC(sbi);
>>>
>>> while (1) {
>>> @@ -4326,9 +4329,10 @@ static void init_dirty_segmap(struct f2fs_sb_info *sbi)
>>> break;
>>> offset = segno + 1;
>>> valid_blocks = get_valid_blocks(sbi, segno, false);
>>> - if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
>>> + usable_blks_in_seg = f2fs_usable_blks_in_seg(sbi, segno);
>>> + if (valid_blocks == usable_blks_in_seg || !valid_blocks)
>>
>> It needs to traverse .cur_valid_map bitmap to check whether blocks in range of [0,
>> usable_blks_in_seg] are all valid or not, if there is at least one usable block in the
>> range, segment should be dirty.
> For the segments which start and end before zone-capacity are just like any normal segments.
> Segments which start after the zone-capacity are fully unusable and are marked as used in the free_seg_bitmap, so these segments are never used.
> Segments which span across the zone-capacity have some unusable blocks. Even when blocks from these segments are allocated/deallocated the valid_blocks counter is incremented/decremented, reflecting the current valid_blocks count.
> Comparing valid_blocks count with usable_blocks count in the segment can indicate if the segment is dirty or fully used.
I thought that if there is one valid block locates in range of
[usable_blks_in_seg, blks_per_seg] (after zone-capacity), the condition
will be incorrect. That should never happen, right?
If so, how about adjusting check_block_count() to do sanity check on bitmap
locates after zone-capacity to make sure there is no free slots there.
> Sorry, but could you please share why cur_valid_map needs to be traversed ?
>
>>
>> One question, if we select dirty segment which across zone-capacity as opened
>> segment (in curseg), how can we avoid allocating usable block beyong zone-capacity
>> in such segment via .cur_valid_map?
> For zoned devices, we have to allocate blocks sequentially, so it's always in LFS manner it is allocated.
> The __has_curseg_space() checks for the usable blocks and stops allocating blocks after zone-capacity.
Oh, that was implemented in patch 2, I haven't checked that patch...sorry,
however, IMO, patch should be made to apply independently, what if do allocation
only after applying patch 1..., do we need to merge them into one?
>>
>>> continue;
>>> - if (valid_blocks > sbi->blocks_per_seg) {
>>> + if (valid_blocks > usable_blks_in_seg) {
>>> f2fs_bug_on(sbi, 1);
>>> continue;
>>> }
>>> @@ -4678,6 +4682,101 @@ int f2fs_check_write_pointer(struct
>>> f2fs_sb_info *sbi)
>>>
>>> return 0;
>>> }
>>> +
>>> +static bool is_conv_zone(struct f2fs_sb_info *sbi, unsigned int zone_idx,
>>> + unsigned int dev_idx)
>>> +{
>>> + if (!bdev_is_zoned(FDEV(dev_idx).bdev))
>>> + return true;
>>> + return !test_bit(zone_idx, FDEV(dev_idx).blkz_seq); }
>>> +
>>> +/* Return the zone index in the given device */ static unsigned int
>>> +get_zone_idx(struct f2fs_sb_info *sbi, unsigned int secno,
>>> + int dev_idx)
>>> +{
>>> + block_t sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi,
>>> +secno));
>>> +
>>> + return (sec_start_blkaddr - FDEV(dev_idx).start_blk) >>
>>> + sbi->log_blocks_per_blkz;
>>> +}
>>> +
>>> +/*
>>> + * Return the usable segments in a section based on the zone's
>>> + * corresponding zone capacity. Zone is equal to a section.
>>> + */
>>> +static inline unsigned int f2fs_usable_zone_segs_in_sec(
>>> + struct f2fs_sb_info *sbi, unsigned int segno) {
>>> + unsigned int dev_idx, zone_idx, unusable_segs_in_sec;
>>> +
>>> + dev_idx = f2fs_target_device_index(sbi, START_BLOCK(sbi, segno));
>>> + zone_idx = get_zone_idx(sbi, GET_SEC_FROM_SEG(sbi, segno), dev_idx);
>>> +
>>> + /* Conventional zone's capacity is always equal to zone size */
>>> + if (is_conv_zone(sbi, zone_idx, dev_idx))
>>> + return sbi->segs_per_sec;
>>> +
>>> + /*
>>> + * If the zone_capacity_blocks array is NULL, then zone capacity
>>> + * is equal to the zone size for all zones
>>> + */
>>> + if (!FDEV(dev_idx).zone_capacity_blocks)
>>> + return sbi->segs_per_sec;
>>> +
>>> + /* Get the segment count beyond zone capacity block */
>>> + unusable_segs_in_sec = (sbi->blocks_per_blkz -
>>> + FDEV(dev_idx).zone_capacity_blocks[zone_idx])
>>>>
>>> + sbi->log_blocks_per_seg;
>>> + return sbi->segs_per_sec - unusable_segs_in_sec; }
>>> +
>>> +/*
>>> + * Return the number of usable blocks in a segment. The number of
>>> +blocks
>>> + * returned is always equal to the number of blocks in a segment for
>>> + * segments fully contained within a sequential zone capacity or a
>>> + * conventional zone. For segments partially contained in a
>>> +sequential
>>> + * zone capacity, the number of usable blocks up to the zone capacity
>>> + * is returned. 0 is returned in all other cases.
>>> + */
>>> +static inline unsigned int f2fs_usable_zone_blks_in_seg(
>>> + struct f2fs_sb_info *sbi, unsigned int segno) {
>>> + block_t seg_start, sec_start_blkaddr, sec_cap_blkaddr;
>>> + unsigned int zone_idx, dev_idx, secno;
>>> +
>>> + secno = GET_SEC_FROM_SEG(sbi, segno);
>>> + seg_start = START_BLOCK(sbi, segno);
>>> + dev_idx = f2fs_target_device_index(sbi, seg_start);
>>> + zone_idx = get_zone_idx(sbi, secno, dev_idx);
>>> +
>>> + /*
>>> + * Conventional zone's capacity is always equal to zone size,
>>> + * so, blocks per segment is unchanged.
>>> + */
>>> + if (is_conv_zone(sbi, zone_idx, dev_idx))
>>> + return sbi->blocks_per_seg;
>>> +
>>> + if (!FDEV(dev_idx).zone_capacity_blocks)
>>> + return sbi->blocks_per_seg;
>>> +
>>> + sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, secno));
>>> + sec_cap_blkaddr = sec_start_blkaddr +
>>> + FDEV(dev_idx).zone_capacity_blocks[zone_idx];
>>> +
>>> + /*
>>> + * If segment starts before zone capacity and spans beyond
>>> + * zone capacity, then usable blocks are from seg start to
>>> + * zone capacity. If the segment starts after the zone capacity,
>>> + * then there are no usable blocks.
>>> + */
>>> + if (seg_start >= sec_cap_blkaddr)
>>> + return 0;
>>> + if (seg_start + sbi->blocks_per_seg > sec_cap_blkaddr)
>>> + return sec_cap_blkaddr - seg_start;
>>> +
>>> + return sbi->blocks_per_seg;
>>> +}
>>> #else
>>> int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi) { @@
>>> -4688,7 +4787,36 @@ int f2fs_check_write_pointer(struct f2fs_sb_info
>>> *sbi) {
>>> return 0;
>>> }
>>> +
>>> +static inline unsigned int f2fs_usable_zone_blks_in_seg(struct f2fs_sb_info *sbi,
>>> + unsigned int segno)
>>> +{
>>> + return 0;
>>> +}
>>> +
>>> +static inline unsigned int f2fs_usable_zone_segs_in_sec(struct f2fs_sb_info *sbi,
>>> + unsigned int segno)
>>> +{
>>> + return 0;
>>> +}
>>> #endif
>>> +unsigned int f2fs_usable_blks_in_seg(struct f2fs_sb_info *sbi,
>>> + unsigned int segno)
>>> +{
>>> + if (f2fs_sb_has_blkzoned(sbi))
>>> + return f2fs_usable_zone_blks_in_seg(sbi, segno);
>>> +
>>> + return sbi->blocks_per_seg;
>>> +}
>>> +
>>> +unsigned int f2fs_usable_segs_in_sec(struct f2fs_sb_info *sbi,
>>> + unsigned int segno)
>>> +{
>>> + if (f2fs_sb_has_blkzoned(sbi))
>>> + return f2fs_usable_zone_segs_in_sec(sbi, segno);
>>> +
>>> + return sbi->segs_per_sec;
>>> +}
>>>
>>> /*
>>> * Update min, max modified time for cost-benefit GC algorithm diff
>>> --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h index
>>> f261e3e6a69b..79b0dc33feaf 100644
>>> --- a/fs/f2fs/segment.h
>>> +++ b/fs/f2fs/segment.h
>>> @@ -411,6 +411,7 @@ static inline void __set_free(struct f2fs_sb_info *sbi,
>> unsigned int segno)
>>> unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
>>> unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
>>> unsigned int next;
>>> + unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
>>>
>>> spin_lock(&free_i->segmap_lock);
>>> clear_bit(segno, free_i->free_segmap); @@ -418,7 +419,7 @@ static
>>> inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
>>>
>>> next = find_next_bit(free_i->free_segmap,
>>> start_segno + sbi->segs_per_sec, start_segno);
>>> - if (next >= start_segno + sbi->segs_per_sec) {
>>> + if (next >= start_segno + usable_segs) {
>>> clear_bit(secno, free_i->free_secmap);
>>> free_i->free_sections++;
>>> }
>>> @@ -444,6 +445,7 @@ static inline void __set_test_and_free(struct f2fs_sb_info
>> *sbi,
>>> unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
>>> unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
>>> unsigned int next;
>>> + unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
>>>
>>> spin_lock(&free_i->segmap_lock);
>>> if (test_and_clear_bit(segno, free_i->free_segmap)) { @@ -453,7
>>> +455,7 @@ static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
>>> goto skip_free;
>>> next = find_next_bit(free_i->free_segmap,
>>> start_segno + sbi->segs_per_sec, start_segno);
>>> - if (next >= start_segno + sbi->segs_per_sec) {
>>> + if (next >= start_segno + usable_segs) {
>>> if (test_and_clear_bit(secno, free_i->free_secmap))
>>> free_i->free_sections++;
>>> }
>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index
>>> 80cb7cd358f8..2686b07ae7eb 100644
>>> --- a/fs/f2fs/super.c
>>> +++ b/fs/f2fs/super.c
>>> @@ -1164,6 +1164,7 @@ static void destroy_device_list(struct f2fs_sb_info *sbi)
>>> blkdev_put(FDEV(i).bdev, FMODE_EXCL); #ifdef
>> CONFIG_BLK_DEV_ZONED
>>> kvfree(FDEV(i).blkz_seq);
>>> + kvfree(FDEV(i).zone_capacity_blocks);
>>
>> Now, f2fs_kzalloc won't allocate vmalloc's memory, so it's safe to use kfree().
> Ok
>>
>>> #endif
>>> }
>>> kvfree(sbi->devs);
>>> @@ -3039,13 +3040,26 @@ static int init_percpu_info(struct
>>> f2fs_sb_info *sbi) }
>>>
>>> #ifdef CONFIG_BLK_DEV_ZONED
>>> +
>>> +struct f2fs_report_zones_args {
>>> + struct f2fs_dev_info *dev;
>>> + bool zone_cap_mismatch;
>>> +};
>>> +
>>> static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
>>> - void *data)
>>> + void *data)
>>> {
>>> - struct f2fs_dev_info *dev = data;
>>> + struct f2fs_report_zones_args *rz_args = data;
>>> +
>>> + if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
>>> + return 0;
>>> +
>>> + set_bit(idx, rz_args->dev->blkz_seq);
>>> + rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
>>> + F2FS_LOG_SECTORS_PER_BLOCK;
>>> + if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
>>> + rz_args->zone_cap_mismatch = true;
>>>
>>> - if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
>>> - set_bit(idx, dev->blkz_seq);
>>> return 0;
>>> }
>>>
>>> @@ -3053,6 +3067,7 @@ static int init_blkz_info(struct f2fs_sb_info
>>> *sbi, int devi) {
>>> struct block_device *bdev = FDEV(devi).bdev;
>>> sector_t nr_sectors = bdev->bd_part->nr_sects;
>>> + struct f2fs_report_zones_args rep_zone_arg;
>>> int ret;
>>>
>>> if (!f2fs_sb_has_blkzoned(sbi))
>>> @@ -3078,12 +3093,26 @@ static int init_blkz_info(struct f2fs_sb_info *sbi, int
>> devi)
>>> if (!FDEV(devi).blkz_seq)
>>> return -ENOMEM;
>>>
>>> - /* Get block zones type */
>>> + /* Get block zones type and zone-capacity */
>>> + FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
>>> + FDEV(devi).nr_blkz * sizeof(block_t),
>>> + GFP_KERNEL);
>>> + if (!FDEV(devi).zone_capacity_blocks)
>>> + return -ENOMEM;
>>> +
>>> + rep_zone_arg.dev = &FDEV(devi);
>>> + rep_zone_arg.zone_cap_mismatch = false;
>>> +
>>> ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
>>> - &FDEV(devi));
>>> + &rep_zone_arg);
>>> if (ret < 0)
>>> return ret;
>>
>> Missed to call kfree(FDEV(devi).zone_capacity_blocks)?
> Thanks for catching it. Will free it here also.
>>
>>>
>>> + if (!rep_zone_arg.zone_cap_mismatch) {
>>> + kvfree(FDEV(devi).zone_capacity_blocks);
>>
>> Ditto, kfree().
> Ok.
>>
>> Thanks,
>>
>>> + FDEV(devi).zone_capacity_blocks = NULL;
>>> + }
>>> +
>>> return 0;
>>> }
>>> #endif
>>>
> .
>
