nilfs2: insert checkpoint number in segment summary header
PATCH 1/1 nilfs2: insert checkpoint number in segment summary header
This patch extends the format of segment summary so that it stores a checkpoint number in the header. This is normally not used, but I think this would help recovery from critcal situation in which the filesystem has lost its pointer to the latest log.
I'm planning to send this to the next release unless problem occurs.
Thanks,
Ryusuke Konishi
--
From: Ryusuke Konishi <[email protected]>
This adds a field to record the latest checkpoint number in the nilfs_segment_summary structure. This will help to recover the latest checkpoint number from logs on disk. This field is intended for crucial cases in which super blocks have lost pointer to the latest log.
Even though this will change the disk format, both backward and forward compatibility is preserved by a size field prepared in the segment summary header.
--- fs/nilfs2/recovery.c | 2 ++ fs/nilfs2/segbuf.c | 4 +++- fs/nilfs2/segbuf.h | 4 +++- fs/nilfs2/segment.c | 3 ++- include/linux/nilfs2_fs.h | 2 ++ 5 files changed, 12 insertions(+), 3 deletions(-)
fs/nilfs2/recovery.c +2 -0
@@ -104,6 +104,8 @@ static void store_segsum_info(struct nilfs_segsum_info *ssi, ssi->nsumblk =DIV_ROUND_UP (ssi->sumbytes, blocksize); ssi->nfileblk = ssi->nblocks - ssi->nsumblk - !!NILFS_SEG_HAS_SR (ssi); + +/* need to verify ->ss_bytes field if read ->ss_cno */ }/**
fs/nilfs2/segbuf.c +3 -1
@@ -133,7 +133,7 @@ int nilfs_segbuf_extend_payload(struct nilfs_segment_buffer *segbuf, }int nilfs_segbuf_reset (struct nilfs_segment_buffer *segbuf,unsigned flags, -time_t ctime) +time_t ctime,__u64 cno) {int err;
@@ -146,6 +146,7 @@ int nilfs_segbuf_reset(struct nilfs_segment_buffer *segbuf, unsigned flags, segbuf->sb_sum.sumbytes =sizeof (struct nilfs_segment_summary ); segbuf->sb_sum.nfinfo = segbuf->sb_sum.nfileblk =0 ; segbuf->sb_sum.ctime = ctime; + segbuf->sb_sum.cno = cno;return 0 ; }
@@ -171,6 +172,7 @@ void nilfs_segbuf_fill_in_segsum(struct nilfs_segment_buffer *segbuf) raw_sum->ss_nfinfo =cpu_to_le32 (segbuf->sb_sum.nfinfo); raw_sum->ss_sumbytes =cpu_to_le32 (segbuf->sb_sum.sumbytes); raw_sum->ss_pad =0 ; + raw_sum->ss_cno =cpu_to_le64 (segbuf->sb_sum.cno); }/*
fs/nilfs2/segbuf.h +3 -1
@@ -37,6 +37,7 @@* @sumbytes: Byte count of segment summary * @nfileblk: Total number of file blocks * @seg_seq: Segment sequence number +* @cno: Checkpoint number * @ctime: Creation time * @next: Block number of the next full segment */
@@ -48,6 +49,7 @@ struct nilfs_segsum_info {unsigned long sumbytes;unsigned long nfileblk;u64 seg_seq; +__u64 cno;time_t ctime;sector_t next; };
@@ -135,7 +137,7 @@ void nilfs_segbuf_map_cont(struct nilfs_segment_buffer *segbuf, struct nilfs_segment_buffer *prev);void nilfs_segbuf_set_next_segnum (struct nilfs_segment_buffer *,__u64 ,struct the_nilfs *); -int nilfs_segbuf_reset (struct nilfs_segment_buffer *,unsigned ,time_t ); +int nilfs_segbuf_reset (struct nilfs_segment_buffer *,unsigned ,time_t ,__u64 );int nilfs_segbuf_extend_segsum (struct nilfs_segment_buffer *);int nilfs_segbuf_extend_payload (struct nilfs_segment_buffer *,struct buffer_head **);
fs/nilfs2/segment.c +2 -1
@@ -365,7 +365,8 @@ static int nilfs_segctor_reset_segment_buffer(struct nilfs_sc_info *sci)if (nilfs_doing_gc ()) flags =NILFS_SS_GC ; - err =nilfs_segbuf_reset (segbuf, flags, sci->sc_seg_ctime); + err =nilfs_segbuf_reset (segbuf, flags, sci->sc_seg_ctime, + sci->sc_sbi->s_nilfs->ns_cno);if (unlikely (err))return err;
include/linux/nilfs2_fs.h +2 -0
@@ -377,6 +377,7 @@ union nilfs_binfo {* @ss_nfinfo: number of finfo structures * @ss_sumbytes: total size of segment summary in bytes * @ss_pad: padding +* @ss_cno: checkpoint number */ struct nilfs_segment_summary {__le32 ss_datasum;
@@ -391,6 +392,7 @@ struct nilfs_segment_summary {__le32 ss_nfinfo;__le32 ss_sumbytes;__le32 ss_pad; +__le64 ss_cno;/* array of finfo structures */ };
Signature · 6 lines
1.6.3.4 -- To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
A message that is not in the archive (sent off-list or not yet received): [email protected]
Hi Ryusuke,
Quoted 4 lines · Ryusuke Konishi
This patch extends the format of segment summary so that it stores a checkpoint number in the header. This is normally not used, but I think this would help recovery from critcal situation in which the filesystem has lost its pointer to the latest log.
that sounds like i good plan. I'm not that happy about the position in the structure you added the cno to... Would have like to see `next' as one of the first fields and all extra appended but thats not the issue i am writing about.
Quoted 13 lines
fs/nilfs2/segbuf.h +1
-0@@ -37,6 +37,7 @@* @sumbytes: Byte count of segment summary * @nfileblk: Total number of file blocks * @seg_seq: Segment sequence number +* @cno: Checkpoint number * @ctime: Creation time * @next: Block number of the next full segment */
fs/nilfs2/segbuf.h struct nilfs_segsum_info { · line 55
Quoted 7 lines
fs/nilfs2/segbuf.h +1
-0@@ -48,6 +49,7 @@ struct nilfs_segsum_info {unsigned long sumbytes;unsigned long nfileblk;u64 seg_seq; +__u64 cno;time_t ctime;sector_t next; };
What i miss in the patch is the roll-forward support for this. The field sector_t next' has moved but i dont see any reference in the patch that takes into account the size of the structure to see if the next' field has shifted position or not. Am i missing/overseeing a thing?
With regards,
Reinoud
P.S. the write support of the NetBSD NiLFS implementation has been delayed quite some time due to simple lack of time (AKA too many projects running simultaniously) but after some time dormant i've picked it up again.
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
A message that is not in the archive (sent off-list or not yet received): 20100502105058.GA812-bVHBekiX4bNgoMqBc1r0ESegHCQxtGRMHZ5vskTnxNA@public.gmane.org
Hi Reinoud,
Quoted 8 lines · Reinoud Zandijk
Hi Ryusuke,
Quoted 4 lines · Ryusuke Konishi
This patch extends the format of segment summary so that it stores a checkpoint number in the header. This is normally not used, but I think this would help recovery from critcal situation in which the filesystem has lost its pointer to the latest log.
Quoted 7 lines
that sounds like i good plan. I'm not that happy about the position in the structure you added the cno to...
What i miss in the patch is the roll-forward support for this. The field
sector_t next' has moved but i dont see any reference in the patch that takes into account the size of the structure to see if thenext' field has shifted position or not. Am i missing/overseeing a thing?
There seems a misunderstanding on use of the structure in question.
nilfs_segsum_info is an on-memory structure and is not for the on-disk format. The on-disk summary header is defined by nilfs_segment_summary as below:
Quoted 13 lines
include/linux/nilfs2_fs.h +1
-0@@ -377,6 +377,7 @@ union nilfs_binfo {* @ss_nfinfo: number of finfo structures * @ss_sumbytes: total size of segment summary in bytes * @ss_pad: padding +* @ss_cno: checkpoint number */ struct nilfs_segment_summary {__le32 ss_datasum;
include/linux/nilfs2_fs.h struct nilfs_segment_summary { · line 397
Quoted 6 lines
include/linux/nilfs2_fs.h +1
-0@@ -391,6 +392,7 @@ struct nilfs_segment_summary {__le32 ss_nfinfo;__le32 ss_sumbytes;__le32 ss_pad; +__le64 ss_cno;/* array of finfo structures */ };
The checkpoint number field is appended to the tail in order to keep compatibility; the ``next'' and other fields are not moved in the header.
nilfs_segsum_info is volatile, so we don't have to care about the order except for the viewpoint of optimization or simplicity.
Quoted 3 lines
P.S. the write support of the NetBSD NiLFS implementation has been delayed quite some time due to simple lack of time (AKA too many projects running simultaniously) but after some time dormant i've picked it up again.
Me neither for lack of time, thanks anyway :)
With regards,
Ryusuke Konishi
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html