"There is a HTML error in the previous email, so I send this one.If you already
received this before, please ignore it.Sorry for the inconvenience"
This patch change the method of calculating the number of summary blocks in
function npages_for_summary_flush.
npages_for_summary_flush uses (SUMMARY_SIZE + 1) as the size of a f2fs_summary
while the actual size is just SUMMARY_SIZE.
As a result, occasionally the return value of npages_for_summary_flush will be
bigger than the actual number by one, and the checkpoint won't be written
contiguously into disk.
Besides, struct f2fs_summary can't be split to two pages, so it could take more
space than the sum of its size, the current version seems not to take it into
account.
Signed-off-by: Fan Li <fanofcode...@samsung.com>
---
fs/f2fs/segment.c | 14 ++++++--------
1 file changed, 6 insertions(+), 8 deletions(-)
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 487af61..ba2930f 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -279,9 +279,8 @@ static void __add_sum_entry(struct f2fs_sb_info *sbi, int
type,
*/
int npages_for_summary_flush(struct f2fs_sb_info *sbi) {
- int total_size_bytes = 0;
int valid_sum_count = 0;
- int i, sum_space;
+ int i, sum_in_page;
for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
if (sbi->ckpt->alloc_type[i] == SSR)
@@ -290,13 +289,12 @@ int npages_for_summary_flush(struct f2fs_sb_info *sbi)
valid_sum_count += curseg_blkoff(sbi, i);
}
- total_size_bytes = valid_sum_count * (SUMMARY_SIZE + 1)
- + sizeof(struct nat_journal) + 2
- + sizeof(struct sit_journal) + 2;
- sum_space = PAGE_CACHE_SIZE - SUM_FOOTER_SIZE;
- if (total_size_bytes < sum_space)
+ sum_in_page = (PAGE_CACHE_SIZE - 2*SUM_JOURNAL_SIZE - SUM_FOOTER_SIZE)/
+ SUMMARY_SIZE;
+ if (valid_sum_count <= sum_in_page)
return 1;
- else if (total_size_bytes < 2 * sum_space)
+ else if (valid_sum_count-sum_in_page <=
+ (PAGE_CACHE_SIZE-SUM_FOOTER_SIZE)/SUMMARY_SIZE)
return 2;
return 3;
}
--
1.7.9.5
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
