On 2016/2/23 1:31, Jaegeuk Kim wrote:
> Hi Junling,
>
> On Thu, Feb 18, 2016 at 04:59:04PM +0800, Junling Zheng wrote:
>> In most cases, sit_bitmap_size is smaller than MAX_SIT_BITMAP_SIZE.
>>
>> However, in some extreme scenarios, such as 16TB, sit_bitmap_size
>> could be larger than MAX_SIT_BITMAP_SIZE.
>>
>> In this case, we should recalculate the sit_segments through
>> max_sit_bitmap_size to prevent sit_ver_bitmap_bytesize got from
>> segment_count_sit in f2fs_write_check_point_pack() being over
>> MAX_SIT_BITMAP_SIZE.
>>
>> Signed-off-by: Junling Zheng <zhengjunl...@huawei.com>
>> ---
>> mkfs/f2fs_format.c | 24 +++++++++++++++++-------
>> 1 file changed, 17 insertions(+), 7 deletions(-)
>>
>> diff --git a/mkfs/f2fs_format.c b/mkfs/f2fs_format.c
>> index 645c2aa..24b2cc1 100644
>> --- a/mkfs/f2fs_format.c
>> +++ b/mkfs/f2fs_format.c
>> @@ -191,6 +191,23 @@ static int f2fs_prepare_super_block(void)
>>
>> sit_segments = SEG_ALIGN(blocks_for_sit);
>>
>> + /*
>> + * In most cases, sit_bitmap_size is smaller than MAX_SIT_BITMAP_SIZE.
>> + * However, in an extreme scenario(16TB), sit_bitmap_size could be
>> larger
>> + * than MAX_SIT_BITMAP_SIZE. Thus, we should recalculate the
>> sit_segments
>> + * to prevent sit_ver_bitmap_bytesize got from segment_count_sit in
>> + * f2fs_write_check_point_pack() being over MAX_SIT_BITMAP_SIZE.
>> + */
>> + sit_bitmap_size = (sit_segments << log_blks_per_seg) / 8;
>> +
>> + if (sit_bitmap_size > MAX_SIT_BITMAP_SIZE) {
>> + max_sit_bitmap_size = MAX_SIT_BITMAP_SIZE;
>> + sit_segments = max_sit_bitmap_size * 8 >> log_blks_per_seg;
>> + blocks_for_sit = sit_segments << log_blks_per_seg;
>> + }
>> + else
>> + max_sit_bitmap_size = sit_bitmap_size;
>
> Codling style.
> if {
> } else {
> }
>
> Anyway, what about just redefining MAX_SIT_BITMAP_SIZE as
> (ALIGN(F2FS_MAX_SEGMENT, SIT_ENTRY_PER_BLOCK) / 8) ?
>
(ALIGN(F2FS_MAX_SEGMENT, SIT_ENTRY_PER_BLOCK) / 8) is also 19065, is the same
with
the original. But the sit_ver_bitmap_bytesize in cp could be 19072. Please see
the
following calculations:
F2FS_MAX_SEGMENT:
16 * 1024 * 1024 / 2 = 8388608
max blocks_for_sit:
ALIGN(F2FS_MAX_SEGMENT, SIT_ENTRY_PER_BLOCK) = 152521 // Here, not
all blocks are used, some are redundant.
max sit_segments:
ALIGN(blocks_for_sit, config.blks_per_seg) = 298 // Here, not all
segments are used, some are redundant.
max segment_count_sit:
sit_segments * 2 = 596
max sit_ver_bitmap_bytesize:
((segment_count_sit / 2) << log_blocks_per_seg) / 8 = 19072 // Here, bitmap
size is too large because of the redundant blocks and segments.
At the beginning, I just doubted why segment_count_sit wasn't recalculated by
max_sit_bitmap_size. And then, I found that the sit_ver_bitmap_bytesize got
from segment_count_sit might exceed MAX_SIT_BITMAP_SIZE.
So, if segment_count_sit is not recalculated by max_sit_bitmap_size, checking
and
limiting sit_ver_bitmap_bytesize in cp during f2fs_write_check_point_pack() is
necessary.
> Thanks,
>
>> +
>> set_sb(segment_count_sit, sit_segments * 2);
>>
>> set_sb(nat_blkaddr, get_sb(sit_blkaddr) + get_sb(segment_count_sit) *
>> @@ -208,13 +225,6 @@ static int f2fs_prepare_super_block(void)
>> * This number resizes NAT bitmap area in a CP page.
>> * So the threshold is determined not to overflow one CP page
>> */
>> - sit_bitmap_size = ((get_sb(segment_count_sit) / 2) <<
>> - log_blks_per_seg) / 8;
>> -
>> - if (sit_bitmap_size > MAX_SIT_BITMAP_SIZE)
>> - max_sit_bitmap_size = MAX_SIT_BITMAP_SIZE;
>> - else
>> - max_sit_bitmap_size = sit_bitmap_size;
>>
>> /*
>> * It should be reserved minimum 1 segment for nat.
>> --
>> 1.9.1
>>
>>
>> ------------------------------------------------------------------------------
>> Site24x7 APM Insight: Get Deep Visibility into Application Performance
>> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
>> Monitor end-to-end web transactions and take corrective actions now
>> Troubleshoot faster and improve end-user experience. Signup Now!
>> http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
>> _______________________________________________
>> Linux-f2fs-devel mailing list
>> Linux-f2fs-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>
> .
>
------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
