| Message ID | 6bd71ff55e48686fc917736e686143ca7d5d2c64.1652711187.git.johannes.thumshirn@wdc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: introduce raid-stripe-tree | expand |
On 2022/5/16 22:31, Johannes Thumshirn wrote: > In preparation for upcoming changes, move 'struct btrfs_io_context' to > volumes.h, so we can use it outside of volumes.c In fact I don't think the naming itself (from myself) is that good. It maybe a good idea to also do a rename here. I have some bad alternatives, but doesn't seem better than the current generic naming either: - btrfs_io_mapping - btrfs_mapping_context Thus I guess the current name is chosen mostly due to lack of better ones. Thanks, Qu > > Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> > --- > fs/btrfs/volumes.h | 90 +++++++++++++++++++++++----------------------- > 1 file changed, 45 insertions(+), 45 deletions(-) > > diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h > index bd297f23d19e..894d289a3b50 100644 > --- a/fs/btrfs/volumes.h > +++ b/fs/btrfs/volumes.h > @@ -32,6 +32,51 @@ struct btrfs_io_geometry { > u64 raid56_stripe_offset; > }; > > +struct btrfs_io_stripe { > + struct btrfs_device *dev; > + u64 physical; > + u64 length; /* only used for discard mappings */ > +}; > + > +/* > + * Context for IO subsmission for device stripe. > + * > + * - Track the unfinished mirrors for mirror based profiles > + * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. > + * > + * - Contain the logical -> physical mapping info > + * Used by submit_stripe_bio() for mapping logical bio > + * into physical device address. > + * > + * - Contain device replace info > + * Used by handle_ops_on_dev_replace() to copy logical bios > + * into the new device. > + * > + * - Contain RAID56 full stripe logical bytenrs > + */ > +struct btrfs_io_context { > + refcount_t refs; > + atomic_t stripes_pending; > + struct btrfs_fs_info *fs_info; > + u64 map_type; /* get from map_lookup->type */ > + bio_end_io_t *end_io; > + struct bio *orig_bio; > + void *private; > + atomic_t error; > + int max_errors; > + int num_stripes; > + int mirror_num; > + int num_tgtdevs; > + int *tgtdev_map; > + /* > + * logical block numbers for the start of each stripe > + * The last one or two are p/q. These are sorted, > + * so raid_map[0] is the start of our full stripe > + */ > + u64 *raid_map; > + struct btrfs_io_stripe stripes[]; > +}; > + > /* > * Use sequence counter to get consistent device stat data on > * 32-bit processors. > @@ -354,51 +399,6 @@ static inline void btrfs_bio_free_csum(struct btrfs_bio *bbio) > } > } > > -struct btrfs_io_stripe { > - struct btrfs_device *dev; > - u64 physical; > - u64 length; /* only used for discard mappings */ > -}; > - > -/* > - * Context for IO subsmission for device stripe. > - * > - * - Track the unfinished mirrors for mirror based profiles > - * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. > - * > - * - Contain the logical -> physical mapping info > - * Used by submit_stripe_bio() for mapping logical bio > - * into physical device address. > - * > - * - Contain device replace info > - * Used by handle_ops_on_dev_replace() to copy logical bios > - * into the new device. > - * > - * - Contain RAID56 full stripe logical bytenrs > - */ > -struct btrfs_io_context { > - refcount_t refs; > - atomic_t stripes_pending; > - struct btrfs_fs_info *fs_info; > - u64 map_type; /* get from map_lookup->type */ > - bio_end_io_t *end_io; > - struct bio *orig_bio; > - void *private; > - atomic_t error; > - int max_errors; > - int num_stripes; > - int mirror_num; > - int num_tgtdevs; > - int *tgtdev_map; > - /* > - * logical block numbers for the start of each stripe > - * The last one or two are p/q. These are sorted, > - * so raid_map[0] is the start of our full stripe > - */ > - u64 *raid_map; > - struct btrfs_io_stripe stripes[]; > -}; > - > struct btrfs_device_info { > struct btrfs_device *dev; > u64 dev_offset;
On 17/05/2022 09:42, Qu Wenruo wrote: > > On 2022/5/16 22:31, Johannes Thumshirn wrote: >> In preparation for upcoming changes, move 'struct btrfs_io_context' to >> volumes.h, so we can use it outside of volumes.c > In fact I don't think the naming itself (from myself) is that good. > > It maybe a good idea to also do a rename here. > > I have some bad alternatives, but doesn't seem better than the current > generic naming either: > > - btrfs_io_mapping > - btrfs_mapping_context > > Thus I guess the current name is chosen mostly due to lack of better ones. Yep but I'm not any better in naming *cough* btrfs_dp_stripe *cough*. Maybe someone else has an idea.
On 2022/5/17 15:51, Johannes Thumshirn wrote: > On 17/05/2022 09:42, Qu Wenruo wrote: >> >> On 2022/5/16 22:31, Johannes Thumshirn wrote: >>> In preparation for upcoming changes, move 'struct btrfs_io_context' to >>> volumes.h, so we can use it outside of volumes.c >> In fact I don't think the naming itself (from myself) is that good. >> >> It maybe a good idea to also do a rename here. >> >> I have some bad alternatives, but doesn't seem better than the current >> generic naming either: >> >> - btrfs_io_mapping >> - btrfs_mapping_context >> >> Thus I guess the current name is chosen mostly due to lack of better ones. > > Yep but I'm not any better in naming *cough* btrfs_dp_stripe *cough*. Maybe > someone else has an idea. Forgot to ask in that thread, what does the "dp" naming mean? Thanks, Qu
On 17/05/2022 09:58, Qu Wenruo wrote: > > > On 2022/5/17 15:51, Johannes Thumshirn wrote: >> On 17/05/2022 09:42, Qu Wenruo wrote: >>> >>> On 2022/5/16 22:31, Johannes Thumshirn wrote: >>>> In preparation for upcoming changes, move 'struct btrfs_io_context' to >>>> volumes.h, so we can use it outside of volumes.c >>> In fact I don't think the naming itself (from myself) is that good. >>> >>> It maybe a good idea to also do a rename here. >>> >>> I have some bad alternatives, but doesn't seem better than the current >>> generic naming either: >>> >>> - btrfs_io_mapping >>> - btrfs_mapping_context >>> >>> Thus I guess the current name is chosen mostly due to lack of better ones. >> >> Yep but I'm not any better in naming *cough* btrfs_dp_stripe *cough*. Maybe >> someone else has an idea. > > Forgot to ask in that thread, what does the "dp" naming mean? Declustered Parity, but that's misleading actually, as RAID1 doesn't do parity at all (although BTRFS RAID1 is a declustered RAID already).
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index bd297f23d19e..894d289a3b50 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -32,6 +32,51 @@ struct btrfs_io_geometry { u64 raid56_stripe_offset; }; +struct btrfs_io_stripe { + struct btrfs_device *dev; + u64 physical; + u64 length; /* only used for discard mappings */ +}; + +/* + * Context for IO subsmission for device stripe. + * + * - Track the unfinished mirrors for mirror based profiles + * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. + * + * - Contain the logical -> physical mapping info + * Used by submit_stripe_bio() for mapping logical bio + * into physical device address. + * + * - Contain device replace info + * Used by handle_ops_on_dev_replace() to copy logical bios + * into the new device. + * + * - Contain RAID56 full stripe logical bytenrs + */ +struct btrfs_io_context { + refcount_t refs; + atomic_t stripes_pending; + struct btrfs_fs_info *fs_info; + u64 map_type; /* get from map_lookup->type */ + bio_end_io_t *end_io; + struct bio *orig_bio; + void *private; + atomic_t error; + int max_errors; + int num_stripes; + int mirror_num; + int num_tgtdevs; + int *tgtdev_map; + /* + * logical block numbers for the start of each stripe + * The last one or two are p/q. These are sorted, + * so raid_map[0] is the start of our full stripe + */ + u64 *raid_map; + struct btrfs_io_stripe stripes[]; +}; + /* * Use sequence counter to get consistent device stat data on * 32-bit processors. @@ -354,51 +399,6 @@ static inline void btrfs_bio_free_csum(struct btrfs_bio *bbio) } } -struct btrfs_io_stripe { - struct btrfs_device *dev; - u64 physical; - u64 length; /* only used for discard mappings */ -}; - -/* - * Context for IO subsmission for device stripe. - * - * - Track the unfinished mirrors for mirror based profiles - * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. - * - * - Contain the logical -> physical mapping info - * Used by submit_stripe_bio() for mapping logical bio - * into physical device address. - * - * - Contain device replace info - * Used by handle_ops_on_dev_replace() to copy logical bios - * into the new device. - * - * - Contain RAID56 full stripe logical bytenrs - */ -struct btrfs_io_context { - refcount_t refs; - atomic_t stripes_pending; - struct btrfs_fs_info *fs_info; - u64 map_type; /* get from map_lookup->type */ - bio_end_io_t *end_io; - struct bio *orig_bio; - void *private; - atomic_t error; - int max_errors; - int num_stripes; - int mirror_num; - int num_tgtdevs; - int *tgtdev_map; - /* - * logical block numbers for the start of each stripe - * The last one or two are p/q. These are sorted, - * so raid_map[0] is the start of our full stripe - */ - u64 *raid_map; - struct btrfs_io_stripe stripes[]; -}; - struct btrfs_device_info { struct btrfs_device *dev; u64 dev_offset;
In preparation for upcoming changes, move 'struct btrfs_io_context' to volumes.h, so we can use it outside of volumes.c Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> --- fs/btrfs/volumes.h | 90 +++++++++++++++++++++++----------------------- 1 file changed, 45 insertions(+), 45 deletions(-)