Hi Yunlei, On Fri, May 29, 2015 at 01:05:06PM +0800, He YunLei wrote: > Hi Jaegeuk, > > I found that when files(size > MAX_INLINE_DATA) shrink(size < MAX_INLINE_DATA), > inline_data option is not used in the file after narrow > > I mount f2fs with inline_data as follow: > > /dev/block/mmcblk0p40 /data f2fs rw,nosuid,nodev,noatime,background_gc=on,discard, > user_xattr,inline_xattr,acl,inline_data,active_logs=6 0 0 > > First,I create a small file by command echo for testing inline_data. > > root@hyl:/data # echo 123 > tmp1 > root@hyl:/data # busybox stat tmp1 > File: tmp1 > Size: 4 Blocks: 8 IO Block: 4096 regular file > Device: 10300h/66304d Inode: 2173 Links: 1 > Access: (0666/-rw-rw-rw-) Uid: ( 0/ UNKNOWN) Gid: ( 0/ UNKNOWN) > Access: 2015-05-29 02:59:53.000000000 > Modify: 2015-05-29 02:59:53.000000000 > Change: 2015-05-29 02:59:53.000000000 > > It works well, then I create a file with size 4096(>MAX_INLINE_DATA). > > 127|root@hyl:/data # busybox stat tmp > File: tmp > Size: 4096 Blocks: 16 IO Block: 4096 regular file > Device: 10300h/66304d Inode: 109 Links: 1 > Access: (0666/-rw-rw-rw-) Uid: ( 0/ UNKNOWN) Gid: ( 0/ UNKNOWN) > Access: 2015-05-28 08:48:50.000000000 > Modify: 2015-05-28 08:48:50.000000000 > Change: 2015-05-28 08:53:18.000000000 > > It doesn't use inline_data because file size 4096 > MAX_INLINE_DATA, > So I shrink the file by I/O redirection > > root@hyl:/data # echo 123 > tmp > root@hyl:/data # busybox stat tmp > File: tmp > Size: 4 Blocks: 16 IO Block: 4096 regular file > Device: 10300h/66304d Inode: 109 Links: 1 > Access: (0666/-rw-rw-rw-) Uid: ( 0/ UNKNOWN) Gid: ( 0/ UNKNOWN) > Access: 2015-05-28 08:48:50.000000000 > Modify: 2015-05-29 02:58:31.000000000 > Change: 2015-05-29 02:58:31.000000000 > > We can see that file still uses 16 Blocks > > How do we deal with such situation? it's meaningful to reuse inline_data? > I wish you and other developers in this list could help me in a correct way. As Chao mentioned, inline_data is not recoverable; once it is unset, there is no way to set again. The reason is to avoid live/deadlock, data loss after power-cuts, and another heavy lock contention overhead. Historically, Huajun Li [1] implemented recoverable inline_data, but I decided to removed that due to lots of subtle bugs. [1] http://lwn.net/Articles/573408/ > > Thanks, > He -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
