dm: bio split bvec fix
Alasdair G Kergon
maintainerThe code that handles bios that span table target boundaries by breaking them up into smaller bios will not split an individual struct bio_vec into more than two pieces. Sometimes more than that are required.
This patch adds a loop to break the second piece up into as many pieces as are necessary.
drivers/md/dm.c +25 -20
@@ -573,30 +573,35 @@ static void __clone_and_map(struct clone }else {/* -* Create two copy bios to deal with io that has -* been split across a target. +* Handle a bvec that must be split between two or more targets. */ struct bio_vec *bv = bio->bi_io_vec + ci->idx; +sector_t remaining =to_sector (bv->bv_len); +unsigned int offset =0 ; - clone =split_bvec (bio, ci->sector, ci->idx, - bv->bv_offset, max); -__map_bio (ti, clone, tio); - - ci->sector += max; - ci->sector_count -= max; - ti =dm_table_find_target (ci->map, ci->sector); - - len =to_sector (bv->bv_len) - max; - clone =split_bvec (bio, ci->sector, ci->idx, - bv->bv_offset +to_bytes (max), len); - tio =alloc_tio (ci->md); - tio->io = ci->io; - tio->ti = ti; -memset (&tio->info,0 ,sizeof (tio->info)); -__map_bio (ti, clone, tio); +do { +if (offset) { + ti =dm_table_find_target (ci->map, ci->sector); + max =max_io_len (ci->md, ci->sector, ti); + + tio =alloc_tio (ci->md); + tio->io = ci->io; + tio->ti = ti; +memset (&tio->info,0 ,sizeof (tio->info)); + } + + len = (max > remaining) ? remaining : max; + + clone =split_bvec (bio, ci->sector, ci->idx, + bv->bv_offset + offset, len); + +__map_bio (ti, clone, tio); + + ci->sector += len; + ci->sector_count -= len; + offset +=to_bytes (len); + }while (remaining -= len); - ci->sector += len; - ci->sector_count -= len; ci->idx++; } }
Quoted 7 lines · Alasdair G Kergon
The code that handles bios that span table target boundaries by breaking
them up into smaller bios will not split an individual structbio_vecinto more than two pieces. Sometimes more than that are required.This patch adds a loop to break the second piece up into as many pieces as are necessary.
Should this go into 2.6.16.1?
Quoted 6 lines · Alasdair G Kergon
The code that handles bios that span table target boundaries by breaking them up into smaller bios will not split an individual struct
bio_vecinto more than two pieces. Sometimes more than that are required.This patch adds a loop to break the second piece up into as many pieces as are necessary.
Why isn't this just handled in the merge callback? Can a single page bio span > 2 targets?
Signature · 1 line
Jens Axboe
Alasdair G Kergon
maintainerQuoted 2 lines · Jens Axboe
Why isn't this just handled in the merge callback? Can a single page bio span > 2 targets?
Yes. (Unit of size if the sector - and things don't have to be aligned nicely, just aligned to sector.)
IIRC the merge function assumes the number of bytes that can be added is only a function of the offset: but in our case it's also a function of time. To make this work it should reserve those bytes with device-mapper, and guarantee either to supply them to us subsequently (and preferably quickly) or to cancel that reservation. Device-mapper for its part would guarantee to accept the bio without needing to split it. Or dm could have a rejection mechanism that refuses bios that are too big (because the max number of bytes we accept got reduced between the initial call and the bio actually being presented) and they go back and get processed again.
Alasdair
Signature · 1 line
[email protected]
Quoted 7 lines · Alasdair G Kergon
Quoted 2 lines · Jens Axboe
Why isn't this just handled in the merge callback? Can a single page bio span > 2 targets?
Yes. (Unit of size if the sector - and things don't have to be aligned nicely, just aligned to sector.)
Quoted 11 lines
IIRC the merge function assumes the number of bytes that can be added is only a function of the offset: but in our case it's also a function of time. To make this work it should reserve those bytes with device-mapper, and guarantee either to supply them to us subsequently (and preferably quickly) or to cancel that reservation. Device-mapper for its part would guarantee to accept the bio without needing to split it. Or dm could have a rejection mechanism that refuses bios that are too big (because the max number of bytes we accept got reduced between the initial call and the bio actually being presented) and they go back and get processed again.
Ok, thanks for the followup explanation, I guess there's no way around that then.
Signature · 1 line
Jens Axboe
Quoted 28 lines · Alasdair G Kergon
Quoted 2 lines · Jens Axboe
Why isn't this just handled in the merge callback? Can a single page bio span > 2 targets?
Yes. (Unit of size if the sector - and things don't have to be aligned nicely, just aligned to sector.)
IIRC the merge function assumes the number of bytes that can be added is only a function of the offset: but in our case it's also a function of time. To make this work it should reserve those bytes with device-mapper, and guarantee either to supply them to us subsequently (and preferably quickly) or to cancel that reservation. Device-mapper for its part would guarantee to accept the bio without needing to split it. Or dm could have a rejection mechanism that refuses bios that are too big (because the max number of bytes we accept got reduced between the initial call and the bio actually being presented) and they go back and get processed again.
Alasdair
Signature · 6 lines
[email protected] - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Signature · 1 line
Jens Axboe