md: More sector_t conversions -- intro
cover md: More sector_t conversions -- intro
Hi
this patch set against the for-next tree as of last Friday converts a couple of remaining internal representation of md sizes to sector_t, fixes a bug in the reshape code and adds a couple of trivial cleanups.
Patch #1 converts mddev->chunk_size which was previously stored in bytes. Patch #2 fixes what I believe is a bug in super_1_sync() and might be a candiate for -stable. Patches #3 and #4 are further straight-forward conversions that deal with mddev->new_chunk, conf->chunk_size and conf->prev_chunk. Patches #5 and #6 are pure cleanups.
All patches pass the tests of the mdadm test suite except 07reshape5intr which appears fail also for the vanilla for-next tree.
Please review and consider for inclusion into for-next.
Thanks
Andre
PATCH 1/6 md: Make mddev->chunk_size sector-based.
This patch changes the type of the chunk_size field from int to sector_t and renames it to chunk_sectors with the implied change of semantics. Since
is_power_of_2(chunk_size) = is_power_of_2(chunk_sectors << 9) = is_power_of_2(chunk_sectors)
these bits don't need an adjustment for the shift.
--- drivers/md/linear.c | 2 +- drivers/md/md.c | 49 +++++++++++++++++++++++++------------------------ drivers/md/md.h | 2 +- drivers/md/raid0.c | 29 +++++++++++++++-------------- drivers/md/raid1.c | 4 ++-- drivers/md/raid10.c | 15 ++++++++------- drivers/md/raid5.c | 41 ++++++++++++++++++++++------------------- 7 files changed, 74 insertions(+), 68 deletions(-)
drivers/md/linear.c +1 -1
@@ -305,7 +305,7 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)static void linear_status (struct seq_file *seq,mddev_t *mddev) { -seq_printf (seq," %dk rounding" , mddev->chunk_size/1024 ); +seq_printf (seq," %dk rounding" , (int )mddev->chunk_sectors /2 ); }
drivers/md/md.c +25 -24
172 lines not on this page · raw message
drivers/md/md.h +1 -1
@@ -145,7 +145,7 @@ struct mddev_sint external;/* metadata is * managed externally */ char metadata_type[17 ];/* externally set*/ -int chunk_size; +sector_t chunk_sectors;time_t ctime, utime;int level, layout;char clevel[16 ];
drivers/md/raid0.c +15 -14
@@ -240,10 +240,10 @@ static int create_strip_zones(mddev_t *mddev)* now since we have the hard sector sizes, we can make sure * chunk size is a multiple of that sector size */ -if (mddev->chunk_size % mddev->queue->hardsect_size) { +if ((mddev->chunk_sectors <<9 ) % mddev->queue->hardsect_size) {printk (KERN_ERR "%s chunk_size of %d not valid\n" ,mdname (mddev), - mddev->chunk_size); + (int )mddev->chunk_sectors <<9 ); goto abort; }printk (KERN_INFO "raid0: done.\n" );
@@ -272,10 +272,10 @@ static int raid0_mergeable_bvec(struct request_queue *q,mddev_t *mddev = q->queuedata;sector_t sector = bvm->bi_sector +get_start_sect (bvm->bi_bdev);int max; -unsigned int chunk_sectors = mddev->chunk_size >>9 ; +unsigned int chunk_sectors = mddev->chunk_sectors;unsigned int bio_sectors = bvm->bi_size >>9 ; -if (is_power_of_2 (mddev->chunk_size)) +if (is_power_of_2 (mddev->chunk_sectors)) max = (chunk_sectors - ((sector & (chunk_sectors-1 )) + bio_sectors)) <<9 ;else
@@ -306,12 +306,12 @@ static int raid0_run(mddev_t *mddev) {int ret; -if (mddev->chunk_size ==0 || - !is_power_of_2 (mddev->chunk_size)) { +if (mddev->chunk_sectors ==0 || + !is_power_of_2 (mddev->chunk_sectors)) {printk (KERN_ERR "md/raid0: chunk size must be a power of 2.\n" );return -EINVAL ; } -blk_queue_max_sectors (mddev->queue, mddev->chunk_size >>9 ); +blk_queue_max_sectors (mddev->queue, mddev->chunk_sectors); mddev->queue->queue_lock = &mddev->queue->__queue_lock; ret =create_strip_zones (mddev);
@@ -333,7 +333,8 @@ static int raid0_run(mddev_t *mddev)* chunksize should be used in that case. */ { -int stripe = mddev->raid_disks * mddev->chunk_size /PAGE_SIZE ; +int stripe = mddev->raid_disks * + (mddev->chunk_sectors <<9 ) /PAGE_SIZE ;if (mddev->queue->backing_dev_info.ra_pages <2 * stripe) mddev->queue->backing_dev_info.ra_pages =2 * stripe; }
@@ -384,9 +385,9 @@ static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,unsigned int sect_in_chunk;sector_t chunk;raid0_conf_t *conf = mddev->private; -unsigned int chunk_sects = mddev->chunk_size >>9 ; +unsigned int chunk_sects = mddev->chunk_sectors; -if (is_power_of_2 (mddev->chunk_size)) { +if (is_power_of_2 (mddev->chunk_sectors)) {int chunksect_bits =ffz (~chunk_sects);/* find the sector offset inside the chunk */ sect_in_chunk = sector & (chunk_sects -1 );
@@ -416,7 +417,7 @@ static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,static inline int is_io_in_chunk_boundary (mddev_t *mddev,unsigned int chunk_sects,struct bio *bio) { -if (likely (is_power_of_2 (mddev->chunk_size))) { +if (likely (is_power_of_2 (mddev->chunk_sectors))) {return chunk_sects >= ((bio->bi_sector & (chunk_sects-1 )) + (bio->bi_size >>9 )); }else {
@@ -447,7 +448,7 @@ static int raid0_make_request(struct request_queue *q, struct bio *bio) bio_sectors(bio));part_stat_unlock (); - chunk_sects = mddev->chunk_size >>9 ; + chunk_sects = mddev->chunk_sectors;if (unlikely (!is_io_in_chunk_boundary (mddev, chunk_sects, bio))) {sector_t sector = bio->bi_sector;struct bio_pair *bp;
@@ -458,7 +459,7 @@ static int raid0_make_request(struct request_queue *q, struct bio *bio)/* This is a one page bio that upper layers * refuse to split for us, so we need to split it. */ -if (likely (is_power_of_2 (mddev->chunk_size))) +if (likely (is_power_of_2 (mddev->chunk_sectors))) bp =bio_split (bio, chunk_sects - (sector & (chunk_sects-1 )));else
@@ -522,7 +523,7 @@ static void raid0_status(struct seq_file *seq, mddev_t *mddev) zone_start = conf->strip_zone[j].zone_end; }#endif -seq_printf (seq," %dk chunks" , mddev->chunk_size/1024 ); +seq_printf (seq," %dk chunks" , (int )mddev->chunk_sectors /2 );return ; }
drivers/md/raid1.c +2 -2
@@ -2161,10 +2161,10 @@ static int raid1_reshape(mddev_t *mddev)int d, d2, err;/* Cannot change chunk_size, layout, or level */ -if (mddev->chunk_size != mddev->new_chunk || +if (mddev->chunk_sectors <<9 != mddev->new_chunk || mddev->layout != mddev->new_layout || mddev->level != mddev->new_level) { - mddev->new_chunk = mddev->chunk_size; + mddev->new_chunk = mddev->chunk_sectors <<9 ; mddev->new_layout = mddev->layout; mddev->new_level = mddev->level;return -EINVAL ;
drivers/md/raid10.c +8 -7
@@ -461,7 +461,7 @@ static int raid10_mergeable_bvec(struct request_queue *q,mddev_t *mddev = q->queuedata;sector_t sector = bvm->bi_sector +get_start_sect (bvm->bi_bdev);int max; -unsigned int chunk_sectors = mddev->chunk_size >>9 ; +unsigned int chunk_sectors = mddev->chunk_sectors;unsigned int bio_sectors = bvm->bi_size >>9 ; max = (chunk_sectors - ((sector & (chunk_sectors -1 )) + bio_sectors)) <<9 ;
@@ -985,7 +985,7 @@ static void status(struct seq_file *seq, mddev_t *mddev)int i;if (conf->near_copies < conf->raid_disks) -seq_printf (seq," %dK chunks" , mddev->chunk_size/1024 ); +seq_printf (seq," %dK chunks" , (int )mddev->chunk_sectors /2 );if (conf->near_copies >1 )seq_printf (seq," %d near-copies" , conf->near_copies);if (conf->far_copies >1 ) {
@@ -2050,8 +2050,8 @@ static int run(mddev_t *mddev)int nc, fc, fo;sector_t stride, size; -if (mddev->chunk_size <PAGE_SIZE || - !is_power_of_2 (mddev->chunk_size)) { +if (mddev->chunk_sectors <<9 <PAGE_SIZE || + !is_power_of_2 (mddev->chunk_sectors)) {printk (KERN_ERR "md/raid10: chunk size must be " "at least PAGE_SIZE(%ld) and be a power of 2.\n" ,PAGE_SIZE );return -EINVAL ;
@@ -2096,8 +2096,8 @@ static int run(mddev_t *mddev) conf->far_copies = fc; conf->copies = nc*fc; conf->far_offset = fo; - conf->chunk_mask = (sector_t)(mddev->chunk_size>>9 )-1 ; - conf->chunk_shift =ffz (~mddev->chunk_size) -9 ; + conf->chunk_mask = mddev->chunk_sectors -1 ; + conf->chunk_shift =ffz (~mddev->chunk_sectors); size = mddev->dev_sectors >> conf->chunk_shift;sector_div (size, fc); size = size * conf->raid_disks;
@@ -2205,7 +2205,8 @@ static int run(mddev_t *mddev)* maybe... */ { -int stripe = conf->raid_disks * (mddev->chunk_size /PAGE_SIZE ); +int stripe = conf->raid_disks * + ((mddev->chunk_sectors <<9 ) /PAGE_SIZE ); stripe /= conf->near_copies;if (mddev->queue->backing_dev_info.ra_pages <2 * stripe) mddev->queue->backing_dev_info.ra_pages =2 * stripe;
drivers/md/raid5.c +22 -19
142 lines not on this page · raw message
Signature · 1 line
1.5.4.3
Re: [PATCH 1/6] md: Make mddev->chunk_size sector-based.
Quoted 8 lines · On Monday May 25, [email protected] wrote:
This patch changes the type of the
chunk_sizefield from int tosector_tand renames it tochunk_sectorswith the implied change of semantics. Sinceis_power_of_2(chunk_size) = is_power_of_2(chunk_sectors << 9) = is_power_of_2(chunk_sectors)these bits don't need an adjustment for the shift.
Thanks.
I don't really want chunk_sectors to be sector_t - it is a waste on 32bit, and would cause
Review
Quoted 1 line
(quoted diff) +1
-0+ if (sector_div(temp, mddev->chunk_sectors))
to fail.
I've left it as 'int' and removed all the (int) casts and (sector_t)
casts. I adjusted the comment too.
I also changed:
if (mddev->chunk_sectors << 9 < PAGE_SIZE || to if (mddev->chunk_sectors < (PAGE_SIZE >> 9) ||
primarily because I need the brackets around the shift, even if the C compiler doesn't.
Thanks,
NeilBrown
PATCH 2/6 md: Fix a bug in super_1_sync().
In reply to Andre NollThe new_chunk fields of both struct mddev_s and struct mdp_superblock_1 are measured in bytes. However, in super_1_sync(), mddev->new_chunk is converted to sectors before it assigned to sb->new_chunk.
This results in a too small number being stored in the in-memory image of the superblock which could confuse the reshape process.
--- drivers/md/md.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/md/md.c +1 -1
@@ -1402,7 +1402,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->new_layout =cpu_to_le32 (mddev->new_layout); sb->delta_disks =cpu_to_le32 (mddev->delta_disks); sb->new_level =cpu_to_le32 (mddev->new_level); - sb->new_chunk =cpu_to_le32 (mddev->new_chunk>>9 ); + sb->new_chunk =cpu_to_le32 (mddev->new_chunk); } max_dev =0 ;
Signature · 1 line
1.5.4.3
Re: [PATCH 2/6] md: Fix a bug in super_1_sync().
Quoted 17 lines · Andre Noll
The
new_chunkfields of both structmddev_sand structmdp_superblock_1are measured in bytes. However, insuper_1_sync(),mddev->new_chunkis converted to sectors before it assigned tosb->new_chunk.This results in a too small number being stored in the in-memory image of the superblock which could confuse the reshape process.
Signed-off-by Andre Noll <[email protected]>--- drivers/md/md.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)drivers/md/md.c +0
-0
drivers/md/md.c static void super_1_sync(mddev_t *mddev, mdk_rdev_t · line 1406
Quoted 6 lines
drivers/md/md.c +1
-1@@ -1402,7 +1402,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->new_layout =cpu_to_le32 (mddev->new_layout); sb->delta_disks =cpu_to_le32 (mddev->delta_disks); sb->new_level =cpu_to_le32 (mddev->new_level); - sb->new_chunk =cpu_to_le32 (mddev->new_chunk>>9 ); + sb->new_chunk =cpu_to_le32 (mddev->new_chunk);
I think the current code is correct.
For v1.x metadata, the chunk size (both chunksize and new_chunk) is stored in sectors. in 0.90 it is stored in bytes. (in DDF, it is stored as the base-2 log of sectors, but that is not relevant for the kernel).
So as mddev->new_chunk is (still) in bytes, we need to shift by 9. ??
I'll sort through the rest of the patches tomorrow, but that look fine.
Thanks,
NeilBrown
Re: [PATCH 2/6] md: Fix a bug in super_1_sync().
Quoted 7 lines · NeilBrown
Quoted 7 lines
drivers/md/md.c +0
-0@@ -1402,7 +1402,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->new_layout =cpu_to_le32 (mddev->new_layout);
Quoted 8 lines
drivers/md/md.c static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) · line 1405
Quoted 4 lines
drivers/md/md.c +1
-1sb->delta_disks =cpu_to_le32 (mddev->delta_disks); sb->new_level =cpu_to_le32 (mddev->new_level); - sb->new_chunk =cpu_to_le32 (mddev->new_chunk>>9 ); + sb->new_chunk =cpu_to_le32 (mddev->new_chunk);I think the current code is correct.
For v1.x metadata, the chunk size (both chunksize andnew_chunk) is stored in sectors. in 0.90 it is stored in bytes.
Oops you're right, I got confused by the comment in md_p.h that says new_chunk is in bytes, even for v1.x metadata. So please drop this second patch.
Note that the corresponding hunk in the third patch needs to be changed as well:
- sb->new_chunk = cpu_to_le32(mddev->new_chunk);
+ sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors << 9);
must become
- sb->new_chunk = cpu_to_le32(mddev->new_chunk>>9);
+ sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors);
Thanks
Andre
Signature · 1 line
The only person who always got his work done by Friday was Robinson Crusoe
PATCH 3/6 md: Convert mddev->new_chunk to sectors.
In reply to Andre NollA straight-forward conversion which gets rid of some multiplications/divisions/shifts. The patch also introduces a couple of new ones, most of which are due to conf->chunk_size still being represented in bytes. This will be cleaned up in subsequent patches.
--- drivers/md/md.c | 29 +++++++++++++++-------------- drivers/md/md.h | 3 ++- drivers/md/raid1.c | 4 ++-- drivers/md/raid5.c | 45 ++++++++++++++++++++++++--------------------- 4 files changed, 43 insertions(+), 38 deletions(-)
drivers/md/md.c +15 -14
@@ -886,13 +886,13 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev) mddev->delta_disks = sb->delta_disks; mddev->new_level = sb->new_level; mddev->new_layout = sb->new_layout; - mddev->new_chunk = sb->new_chunk; + mddev->new_chunk_sectors = sb->new_chunk >>9 ; }else { mddev->reshape_position = MaxSector; mddev->delta_disks =0 ; mddev->new_level = mddev->level; mddev->new_layout = mddev->layout; - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors; }if (sb->state & (1 <<MD_SB_CLEAN ))
@@ -1007,7 +1007,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->new_level = mddev->new_level; sb->delta_disks = mddev->delta_disks; sb->new_layout = mddev->new_layout; - sb->new_chunk = mddev->new_chunk; + sb->new_chunk = mddev->new_chunk_sectors <<9 ; } mddev->minor_version = sb->minor_version;if (mddev->in_sync)
@@ -1300,13 +1300,13 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev) mddev->delta_disks =le32_to_cpu (sb->delta_disks); mddev->new_level =le32_to_cpu (sb->new_level); mddev->new_layout =le32_to_cpu (sb->new_layout); - mddev->new_chunk =le32_to_cpu (sb->new_chunk)<<9 ; + mddev->new_chunk_sectors =le32_to_cpu (sb->new_chunk); }else { mddev->reshape_position = MaxSector; mddev->delta_disks =0 ; mddev->new_level = mddev->level; mddev->new_layout = mddev->layout; - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors; } }else if (mddev->pers ==NULL ) {
@@ -1402,7 +1402,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->new_layout =cpu_to_le32 (mddev->new_layout); sb->delta_disks =cpu_to_le32 (mddev->delta_disks); sb->new_level =cpu_to_le32 (mddev->new_level); - sb->new_chunk =cpu_to_le32 (mddev->new_chunk); + sb->new_chunk =cpu_to_le32 (mddev->new_chunk_sectors <<9 ); } max_dev =0 ;
@@ -2746,7 +2746,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)if (IS_ERR (priv)) { mddev->new_level = mddev->level; mddev->new_layout = mddev->layout; - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors; mddev->raid_disks -= mddev->delta_disks; mddev->delta_disks =0 ;module_put (pers->owner);
@@ -2764,7 +2764,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)strlcpy (mddev->clevel, pers->name,sizeof (mddev->clevel)); mddev->level = mddev->new_level; mddev->layout = mddev->new_layout; - mddev->chunk_sectors = mddev->new_chunk >>9 ; + mddev->chunk_sectors = mddev->new_chunk_sectors; mddev->delta_disks =0 ; pers->run (mddev);mddev_resume (mddev);
@@ -2857,8 +2857,9 @@ static ssize_t chunk_size_show(mddev_t *mddev, char *page) {if (mddev->reshape_position != MaxSector && - mddev->chunk_sectors <<9 != mddev->new_chunk) -return sprintf (page,"%d (%d)\n" , mddev->new_chunk, + mddev->chunk_sectors != mddev->new_chunk_sectors) +return sprintf (page,"%d (%d)\n" , + (int )mddev->new_chunk_sectors <<9 , (int )mddev->chunk_sectors <<9 );return sprintf (page,"%d\n" , (int )mddev->chunk_sectors <<9 ); }
@@ -2880,7 +2881,7 @@ chunk_size_store(mddev_t *mddev, const char *buf, size_t len)if (err)return err; }else { - mddev->new_chunk = n; + mddev->new_chunk_sectors = n >>9 ;if (mddev->reshape_position == MaxSector) mddev->chunk_sectors = n >>9 ; }
@@ -3651,7 +3652,7 @@ reshape_position_store(mddev_t *mddev, const char *buf, size_t len) mddev->delta_disks =0 ; mddev->new_level = mddev->level; mddev->new_layout = mddev->layout; - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors;return len; }
@@ -4400,7 +4401,7 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open) mddev->delta_disks =0 ; mddev->new_level =LEVEL_NONE ; mddev->new_layout =0 ; - mddev->new_chunk =0 ; + mddev->new_chunk_sectors =0 ; mddev->curr_resync =0 ; mddev->resync_mismatches =0 ; mddev->suspend_lo = mddev->suspend_hi =0 ;
@@ -5069,7 +5070,7 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)get_random_bytes (mddev->uuid,16 ); mddev->new_level = mddev->level; - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors; mddev->new_layout = mddev->layout; mddev->delta_disks =0 ;
drivers/md/md.h +2 -1
@@ -166,7 +166,8 @@ struct mddev_s* If reshape_position is MaxSector, then no reshape is happening (yet). */ sector_t reshape_position; -int delta_disks, new_level, new_layout, new_chunk; +int delta_disks, new_level, new_layout; +sector_t new_chunk_sectors;struct mdk_thread_s *thread;/* management thread */ struct mdk_thread_s *sync_thread;/* doing resync or reconstruct */
drivers/md/raid1.c +2 -2
@@ -2161,10 +2161,10 @@ static int raid1_reshape(mddev_t *mddev)int d, d2, err;/* Cannot change chunk_size, layout, or level */ -if (mddev->chunk_sectors <<9 != mddev->new_chunk || +if (mddev->chunk_sectors != mddev->new_chunk_sectors || mddev->layout != mddev->new_layout || mddev->level != mddev->new_level) { - mddev->new_chunk = mddev->chunk_sectors <<9 ; + mddev->new_chunk_sectors = mddev->chunk_sectors; mddev->new_layout = mddev->layout; mddev->new_level = mddev->level;return -EINVAL ;
drivers/md/raid5.c +24 -21
137 lines not on this page · raw message
Signature · 1 line
1.5.4.3
PATCH 4/6 md: convert conf->chunk_size and conf->prev_chunk to sectors.
In reply to Andre NollThis kills some more shifts.
--- drivers/md/raid5.c | 31 +++++++++++++++---------------- drivers/md/raid5.h | 6 ++++-- 2 files changed, 19 insertions(+), 18 deletions(-)
drivers/md/raid5.c +15 -16
@@ -1274,8 +1274,8 @@ static sector_t raid5_compute_sector(raid5_conf_t *conf, sector_t r_sector,sector_t new_sector;int algorithm = previous ? conf->prev_algo : conf->algorithm; -int sectors_per_chunk = previous ? (conf->prev_chunk >>9 ) - : (conf->chunk_size >>9 ); +int sectors_per_chunk = previous ? conf->prev_chunk_sectors + : conf->chunk_sectors;int raid_disks = previous ? conf->previous_raid_disks : conf->raid_disks;int data_disks = raid_disks - conf->max_degraded;
@@ -1480,8 +1480,8 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)int raid_disks = sh->disks;int data_disks = raid_disks - conf->max_degraded;sector_t new_sector = sh->sector, check; -int sectors_per_chunk = previous ? (conf->prev_chunk >>9 ) - : (conf->chunk_size >>9 ); +int sectors_per_chunk = previous ? conf->prev_chunk_sectors + : conf->chunk_sectors;int algorithm = previous ? conf->prev_algo : conf->algorithm;sector_t stripe;
@@ -1997,8 +1997,7 @@ static void stripe_set_idx(sector_t stripe, raid5_conf_t *conf, int previous, struct stripe_head *sh) {int sectors_per_chunk = - previous ? (conf->prev_chunk >>9 ) - : (conf->chunk_size >>9 ); + previous ? conf->prev_chunk_sectors : conf->chunk_sectors;int dd_idx;int chunk_offset =sector_div (stripe, sectors_per_chunk);int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
@@ -3916,9 +3915,9 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skippedraid5_compute_sector (conf, stripe_addr*(new_data_disks),1 , &dd_idx,NULL ); last_sector = -raid5_compute_sector (conf, ((stripe_addr+conf->chunk_size/512 ) - *(new_data_disks) -1 ), -1 , &dd_idx,NULL ); +raid5_compute_sector (conf, (stripe_addr + conf->chunk_sectors) + * new_data_disks -1 , +1 , &dd_idx,NULL );if (last_sector >= mddev->dev_sectors) last_sector = mddev->dev_sectors -1 ;while (first_sector <= last_sector) {
@@ -4403,7 +4402,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev) conf->fullsync =1 ; } - conf->chunk_size = mddev->new_chunk_sectors <<9 ; + conf->chunk_sectors = mddev->new_chunk_sectors; conf->level = mddev->new_level;if (conf->level ==6 ) conf->max_degraded =2 ;
@@ -4413,7 +4412,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev) conf->max_nr_stripes =NR_STRIPES ; conf->reshape_progress = mddev->reshape_position;if (conf->reshape_progress != MaxSector) { - conf->prev_chunk = mddev->chunk_sectors <<9 ; + conf->prev_chunk_sectors = mddev->chunk_sectors; conf->prev_algo = mddev->layout; }
@@ -4931,8 +4930,8 @@ static int raid5_start_reshape(mddev_t *mddev)spin_lock_irq (&conf->device_lock); conf->previous_raid_disks = conf->raid_disks; conf->raid_disks += mddev->delta_disks; - conf->prev_chunk = conf->chunk_size; - conf->chunk_size = mddev->new_chunk_sectors <<9 ; + conf->prev_chunk_sectors = conf->chunk_sectors; + conf->chunk_sectors = mddev->new_chunk_sectors; conf->prev_algo = conf->algorithm; conf->algorithm = mddev->new_layout;if (mddev->delta_disks <0 )
@@ -5014,7 +5013,7 @@ static void end_reshape(raid5_conf_t *conf)*/ {int data_disks = conf->raid_disks - conf->max_degraded; -int stripe = data_disks * (conf->chunk_size +int stripe = data_disks * ((conf->chunk_sectors <<9 ) /PAGE_SIZE );if (conf->mddev->queue->backing_dev_info.ra_pages <2 * stripe) conf->mddev->queue->backing_dev_info.ra_pages =2 * stripe;
@@ -5059,7 +5058,7 @@ static void raid5_finish_reshape(mddev_t *mddev)raid5_remove_disk (mddev, d); } mddev->layout = conf->algorithm; - mddev->chunk_sectors = conf->chunk_size >>9 ; + mddev->chunk_sectors = conf->chunk_sectors; mddev->reshape_position = MaxSector; mddev->delta_disks =0 ; }
@@ -5187,7 +5186,7 @@ static int raid5_reconfig(mddev_t *mddev, int new_layout, int new_chunk) mddev->layout = mddev->new_layout = new_layout; }if (new_chunk >0 ) { - conf->chunk_size = new_chunk; + conf->chunk_sectors = new_chunk >>9 ; mddev->new_chunk_sectors = new_chunk >>9 ; mddev->chunk_sectors = new_chunk >>9 ; }
drivers/md/raid5.h +4 -2
@@ -334,7 +334,8 @@ struct raid5_private_data {struct hlist_head *stripe_hashtbl;mddev_t *mddev;struct disk_info *spare; -int chunk_size, level, algorithm; +sector_t chunk_sectors; +int level, algorithm;int max_degraded;int raid_disks;int max_nr_stripes;
@@ -350,7 +351,8 @@ struct raid5_private_data {*/ sector_t reshape_safe;int previous_raid_disks; -int prev_chunk, prev_algo; +sector_t prev_chunk_sectors; +int prev_algo;short generation;/* increments with every reshape */ unsigned long reshape_checkpoint;/* Time we last updated * metadata */
Signature · 1 line
1.5.4.3
Re: [PATCH 4/6] md: convert conf->chunk_size and conf->prev_chunk to sectors.
drivers/md/raid5.c static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped · line 3920
Quoted 13 lines · On Monday May 25, [email protected] wrote:
This kills some more shifts.
drivers/md/raid5.c +3
-3@@ -3916,9 +3915,9 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skippedraid5_compute_sector (conf, stripe_addr*(new_data_disks),1 , &dd_idx,NULL ); last_sector = -raid5_compute_sector (conf, ((stripe_addr+conf->chunk_size/512 ) - *(new_data_disks) -1 ), -1 , &dd_idx,NULL ); +raid5_compute_sector (conf, (stripe_addr + conf->chunk_sectors) + * new_data_disks -1 , +1 , &dd_idx,NULL );
FYI, the () that you removed there were present for a reason. They help emacs C-mode to get the indenting right.
Without them, the indent function make it look like
drivers/md/raid5.c static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped · line 3920
Quoted 3 lines
drivers/md/raid5.c +3
-0+raid5_compute_sector (conf, (stripe_addr + conf->chunk_sectors) + * new_data_disks -1 , +1 , &dd_idx,NULL );
With them I get
raid5_compute_sector(conf, ((stripe_addr + conf->chunk_sectors) * new_data_disks - 1), 1, &dd_idx, NULL);
which is "correct".
Quoted 15 lines
drivers/md/raid5.h +2
-1@@ -334,7 +334,8 @@ struct raid5_private_data {struct hlist_head *stripe_hashtbl;mddev_t *mddev;struct disk_info *spare; -int chunk_size, level, algorithm; +sector_t chunk_sectors; +int level, algorithm;int max_degraded;int raid_disks;int max_nr_stripes;
drivers/md/raid5.h struct raid5_private_data { · line 358
Quoted 9 lines
drivers/md/raid5.h +2
-1@@ -350,7 +351,8 @@ struct raid5_private_data {*/ sector_t reshape_safe;int previous_raid_disks; -int prev_chunk, prev_algo; +sector_t prev_chunk_sectors; +int prev_algo;short generation;/* increments with every reshape */ unsigned long reshape_checkpoint;/* Time we last updated * metadata */
I restored these to 'int' too.
Thanks!
NeilBrown
PATCH 5/6 md/raid5: Use is_power_of_2() in raid5_reconfig()/raid6_reconfig().
In reply to Andre Noll--- drivers/md/raid5.c | 6 ++---- 1 files changed, 2 insertions(+), 4 deletions(-)
drivers/md/raid5.c +2 -4
@@ -5167,8 +5167,7 @@ static int raid5_reconfig(mddev_t *mddev, int new_layout, int new_chunk)if (new_layout >=0 && !algorithm_valid_raid5 (new_layout))return -EINVAL ;if (new_chunk >0 ) { -if (new_chunk & (new_chunk-1 )) -/* not a power of 2 */ +if (!is_power_of_2 (new_chunk))return -EINVAL ;if (new_chunk <PAGE_SIZE )return -EINVAL ;
@@ -5206,8 +5205,7 @@ static int raid6_reconfig(mddev_t *mddev, int new_layout, int new_chunk)if (new_layout >=0 && !algorithm_valid_raid6 (new_layout))return -EINVAL ;if (new_chunk >0 ) { -if (new_chunk & (new_chunk-1 )) -/* not a power of 2 */ +if (!is_power_of_2 (new_chunk))return -EINVAL ;if (new_chunk <PAGE_SIZE )return -EINVAL ;
Signature · 1 line
1.5.4.3
PATCH 6/6 md/raid5: Kill outdated comment.
In reply to Andre NollRaid5 has learned to take over also raid4 and raid6 arrays.
--- drivers/md/raid5.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-)
drivers/md/raid5.c +0 -2
@@ -5231,8 +5231,6 @@ static void *raid5_takeover(mddev_t *mddev)* raid1 - if there are two drives. We need to know the chunk size * raid4 - trivial - just use a raid4 layout. * raid6 - Providing it is a *_6 layout -* -* For now, just do raid1 */ if (mddev->level ==1 )
Signature · 1 line
1.5.4.3
Quoted 6 lines · On Monday May 25, [email protected] wrote:
Hi
this patch set against the for-next tree as of last Friday converts a couple of remaining internal representation of md sizes to
sector_t, fixes a bug in the reshape code and adds a couple of trivial cleanups.
Quoted 6 lines
Patch #1 converts
mddev->chunk_sizewhich was previously stored in bytes. Patch #2 fixes what I believe is a bug insuper_1_sync()and might be a candiate for -stable. Patches #3 and #4 are further straight-forward conversions that deal withmddev->new_chunk,conf->chunk_sizeandconf->prev_chunk. Patches #5 and #6 are pure cleanups.
Thanks.
With the exception of #2, and with the other changes I mentioned, there are all i my for-next now.
Even #2 was useful. It helped me to realise that super_1_sync wasn't writing out any changed to level/layout/chunk_size. Until just recently these couldn't change so there was no point. But now they can. So I added another patch (below) to my for-linus set and modified your patches to match.
Quoted 3 lines
All patches pass the tests of the mdadm test suite except 07reshape5intr which appears fail also for the vanilla for-next tree.
Hmm... I guess I should look into that!
Thanks,
NeilBrown
commit 62e1e389f87a8839ad83b08c44691d1df8320846
Author: NeilBrown <[email protected]> Date: Tue May 26 09:40:59 2009 +1000
md: always update level / chunk_size / layout when writing v1.x metadata.
We previously didn't update these fields when writing the metadata because they could never change. They can now, so we better write them.
v0.90 metadata always updated these fields.
drivers/md/md.c +3 -0
@@ -1375,6 +1375,9 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) sb->raid_disks =cpu_to_le32 (mddev->raid_disks); sb->size =cpu_to_le64 (mddev->dev_sectors); + sb->chunksize =cpu_to_le32 (mddev->chunk_size >>9 ); + sb->level =cpu_to_le32 (mddev->level); + sb->layout =cpu_to_le32 (mddev->layout);if (mddev->bitmap && mddev->bitmap_file ==NULL ) { sb->bitmap_offset =cpu_to_le32 ((__u32 )mddev->bitmap_offset);
Quoted 5 lines · On Tuesday May 26, [email protected] wrote:
Quoted 3 lines
All patches pass the tests of the mdadm test suite except 07reshape5intr which appears fail also for the vanilla for-next tree.
Hmm... I guess I should look into that!
Yep. There is a big there. Thanks for helping find it.
Here is the patch.
Thanks,
NeilBrown
Author: NeilBrown <[email protected]> Date: Tue May 26 12:39:27 2009 +1000
md: raid5: avoid sector values going negative when testing reshape progress.
As sector_t in unsigned, we cannot afford to let 'safepos' etc go negative.
So replace
a -= b;
by
a -= min(b,a);
drivers/md/raid5.c +3 -3
@@ -3810,13 +3810,13 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped safepos = conf->reshape_safe;sector_div (safepos, data_disks);if (mddev->delta_disks <0 ) { - writepos -= reshape_sectors; + writepos -=min (reshape_sectors, writepos); readpos += reshape_sectors; safepos += reshape_sectors; }else { writepos += reshape_sectors; - readpos -= reshape_sectors; - safepos -= reshape_sectors; + readpos -=min (reshape_sectors, readpos); + safepos -=min (reshape_sectors, safepos); }/* 'writepos' is the most advanced device address we might write.
Quoted 10 lines · Neil Brown
Quoted 6 lines
Patch #1 converts
mddev->chunk_sizewhich was previously stored in bytes. Patch #2 fixes what I believe is a bug insuper_1_sync()and might be a candiate for -stable. Patches #3 and #4 are further straight-forward conversions that deal withmddev->new_chunk,conf->chunk_sizeandconf->prev_chunk. Patches #5 and #6 are pure cleanups.Thanks.
With the exception of #2, and with the other changes I mentioned, there are all i my for-next now.
Thanks. BTW: I have a couple of other patches ready that push down code from md.c to the personalities as you suggested. These need to be rebased to current for-next though. I'll send them later this week or at the beginning of next week.
Andre
Signature · 1 line
The only person who always got his work done by Friday was Robinson Crusoe