erofs: random decompression cleanups
Gao Xiang
maintainercover erofs: random decompression cleanups
From: Gao Xiang <[email protected]>
Hi folks,
Now I'm working on cleanuping decompression code and doing some folio adaption for the next 5.20 cycle, see:
https://git.kernel.org/pub/scm/linux/kernel/git/xiang/linux.git/log/?h=erofs/folios
This cleanup work completely gets rid of PageError usage finally and tries to prepare for introducing rolling hashing for EROFS since EROFS supports compressing variable-sized data instead of fixed-sized clusters.
Therefore, EROFS can support rolling hash easily and our mechanism can make full use of filesystem interfaces (byte-addressed) naturally.
Before that, I'd like to submit some trivial cleanups in advance for the 5.19 cycle. All patches are without any logical change, so I can have a more recent codebase for the next rework cycle.
Thanks,
Gao Xiang
Gao Xiang (3):
erofs: get rid of `struct z_erofs_collection'
erofs: get rid of label `restart_now'
erofs: simplify z_erofs_pcluster_readmore()
fs/erofs/zdata.c | 165 +++++++++++++++++++---------------------------- fs/erofs/zdata.h | 50 +++++++------- 2 files changed, 88 insertions(+), 127 deletions(-)
Signature · 1 line
2.30.2
Gao Xiang
maintainerPATCH 1/3 erofs: get rid of `struct z_erofs_collection'
From: Gao Xiang <[email protected]>
It was incompletely introduced for deduplication between different logical extents backed with the same pcluster.
We will have a better in-memory representation in the next release cycle for this, as well as partial memory folios support. So get rid of it instead.
No logic changes.
--- fs/erofs/zdata.c | 111 ++++++++++++++++++----------------------------- fs/erofs/zdata.h | 50 ++++++++++----------- 2 files changed, 65 insertions(+), 96 deletions(-)
fs/erofs/zdata.c +43 -68
@@ -199,7 +199,6 @@ struct z_erofs_decompress_frontend {struct z_erofs_pagevec_ctor vector;struct z_erofs_pcluster *pcl, *tailpcl; -struct z_erofs_collection *cl;/* a pointer used to pick up inplace I/O pages */ struct page **icpage_ptr;z_erofs_next_pcluster_t owned_head;
@@ -357,7 +356,7 @@ static bool z_erofs_try_inplace_io(struct z_erofs_decompress_frontend *fe,return false; } -/* callers must be with collection lock held */ +/* callers must be with pcluster lock held */ static int z_erofs_attach_page (struct z_erofs_decompress_frontend *fe,struct page *page,enum z_erofs_page_type type,bool pvec_safereuse)
@@ -372,7 +371,7 @@ static int z_erofs_attach_page(struct z_erofs_decompress_frontend *fe, ret =z_erofs_pagevec_enqueue (&fe->vector, page, type, pvec_safereuse); - fe->cl->vcnt += (unsigned int )ret; + fe->pcl->vcnt += (unsigned int )ret;return ret ?0 : -EAGAIN ; }
@@ -405,12 +404,11 @@ static void z_erofs_try_to_claim_pcluster(struct z_erofs_decompress_frontend *f) f->mode =COLLECT_PRIMARY ; } -static int z_erofs_lookup_collection (struct z_erofs_decompress_frontend *fe, -struct inode *inode, -struct erofs_map_blocks *map) +static int z_erofs_lookup_pcluster (struct z_erofs_decompress_frontend *fe, +struct inode *inode, +struct erofs_map_blocks *map) {struct z_erofs_pcluster *pcl = fe->pcl; -struct z_erofs_collection *cl;unsigned int length;/* to avoid unexpected loop formed by corrupted images */
@@ -419,8 +417,7 @@ static int z_erofs_lookup_collection(struct z_erofs_decompress_frontend *fe,return -EFSCORRUPTED ; } - cl =z_erofs_primarycollection (pcl); -if (cl->pageofs != (map->m_la & ~PAGE_MASK )) { +if (pcl->pageofs_out != (map->m_la & ~PAGE_MASK )) {DBG_BUGON (1 );return -EFSCORRUPTED ; }
@@ -443,23 +440,21 @@ static int z_erofs_lookup_collection(struct z_erofs_decompress_frontend *fe, length =READ_ONCE (pcl->length); } } -mutex_lock (&cl->lock); +mutex_lock (&pcl->lock);/* used to check tail merging loop due to corrupted images */ if (fe->owned_head ==Z_EROFS_PCLUSTER_TAIL ) fe->tailpcl = pcl;z_erofs_try_to_claim_pcluster (fe); - fe->cl = cl;return 0 ; } -static int z_erofs_register_collection (struct z_erofs_decompress_frontend *fe, -struct inode *inode, -struct erofs_map_blocks *map) +static int z_erofs_register_pcluster (struct z_erofs_decompress_frontend *fe, +struct inode *inode, +struct erofs_map_blocks *map) {bool ztailpacking = map->m_flags &EROFS_MAP_META ;struct z_erofs_pcluster *pcl; -struct z_erofs_collection *cl;struct erofs_workgroup *grp;int err;
@@ -482,17 +477,15 @@ static int z_erofs_register_collection(struct z_erofs_decompress_frontend *fe,/* new pclusters should be claimed as type 1, primary and followed */ pcl->next = fe->owned_head; + pcl->pageofs_out = map->m_la & ~PAGE_MASK ; fe->mode =COLLECT_PRIMARY_FOLLOWED ; - cl =z_erofs_primarycollection (pcl); - cl->pageofs = map->m_la & ~PAGE_MASK ; -/* * lock all primary followed works before visible to others * and mutex_trylock *never* fails for a new pcluster. */ -mutex_init (&cl->lock); -DBG_BUGON (!mutex_trylock (&cl->lock)); +mutex_init (&pcl->lock); +DBG_BUGON (!mutex_trylock (&pcl->lock));if (ztailpacking) { pcl->obj.index =0 ;/* which indicates ztailpacking */
@@ -519,11 +512,10 @@ static int z_erofs_register_collection(struct z_erofs_decompress_frontend *fe, fe->tailpcl = pcl; fe->owned_head = &pcl->next; fe->pcl = pcl; - fe->cl = cl;return 0 ; err_out: -mutex_unlock (&cl->lock); +mutex_unlock (&pcl->lock);z_erofs_free_pcluster (pcl);return err; }
@@ -535,9 +527,9 @@ static int z_erofs_collector_begin(struct z_erofs_decompress_frontend *fe,struct erofs_workgroup *grp;int ret; -DBG_BUGON (fe->cl); +DBG_BUGON (fe->pcl); -/* must be Z_EROFS_PCLUSTER_TAIL or pointed to previous collection */ +/* must be Z_EROFS_PCLUSTER_TAIL or pointed to previous pcluster */ DBG_BUGON (fe->owned_head ==Z_EROFS_PCLUSTER_NIL );DBG_BUGON (fe->owned_head ==Z_EROFS_PCLUSTER_TAIL_CLOSED );
@@ -554,14 +546,14 @@ static int z_erofs_collector_begin(struct z_erofs_decompress_frontend *fe, fe->pcl = container_of(grp, struct z_erofs_pcluster, obj); }else { tailpacking: - ret =z_erofs_register_collection (fe, inode, map); + ret =z_erofs_register_pcluster (fe, inode, map);if (!ret) goto out;if (ret != -EEXIST )return ret; } - ret =z_erofs_lookup_collection (fe, inode, map); + ret =z_erofs_lookup_pcluster (fe, inode, map);if (ret) {erofs_workgroup_put (&fe->pcl->obj);return ret;
@@ -569,7 +561,7 @@ static int z_erofs_collector_begin(struct z_erofs_decompress_frontend *fe, out:z_erofs_pagevec_ctor_init (&fe->vector,Z_EROFS_NR_INLINE_PAGEVECS , - fe->cl->pagevec, fe->cl->vcnt); + fe->pcl->pagevec, fe->pcl->vcnt);/* since file-backed online pages are traversed in reverse order */ fe->icpage_ptr = fe->pcl->compressed_pages +z_erofs_pclusterpages (fe->pcl);
@@ -582,48 +574,36 @@ static int z_erofs_collector_begin(struct z_erofs_decompress_frontend *fe,*/ static void z_erofs_rcu_callback (struct rcu_head *head) { -struct z_erofs_collection *const cl = - container_of(head, struct z_erofs_collection, rcu); - -z_erofs_free_pcluster (container_of(cl ,struct z_erofs_pcluster , -primary_collection )); +z_erofs_free_pcluster (container_of(head , +struct z_erofs_pcluster ,rcu )); }void erofs_workgroup_free_rcu (struct erofs_workgroup *grp) {struct z_erofs_pcluster *const pcl = container_of(grp, struct z_erofs_pcluster, obj); -struct z_erofs_collection *const cl =z_erofs_primarycollection (pcl); -call_rcu (&cl->rcu, z_erofs_rcu_callback); -} - -static void z_erofs_collection_put (struct z_erofs_collection *cl) -{ -struct z_erofs_pcluster *const pcl = - container_of(cl, struct z_erofs_pcluster, primary_collection); - -erofs_workgroup_put (&pcl->obj); +call_rcu (&pcl->rcu, z_erofs_rcu_callback); }static bool z_erofs_collector_end (struct z_erofs_decompress_frontend *fe) { -struct z_erofs_collection *cl = fe->cl; +struct z_erofs_pcluster *pcl = fe->pcl; -if (!cl) +if (!pcl)return false;z_erofs_pagevec_ctor_exit (&fe->vector, false); -mutex_unlock (&cl->lock); +mutex_unlock (&pcl->lock);/* * if all pending pages are added, don't hold its reference * any longer if the pcluster isn't hosted by ourselves. */ if (fe->mode <COLLECT_PRIMARY_FOLLOWED_NOINPLACE ) -z_erofs_collection_put (cl); +erofs_workgroup_put (&pcl->obj); - fe->cl =NULL ; + fe->pcl =NULL ;return true; }
@@ -666,8 +646,8 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,/* lucky, within the range of the current map_blocks */ if (offset + cur >= map->m_la && offset + cur < map->m_la + map->m_llen) { -/* didn't get a valid collection previously (very rare) */ -if (!fe->cl) +/* didn't get a valid pcluster previously (very rare) */ +if (!fe->pcl) goto restart_now; goto hitted; }
@@ -766,7 +746,7 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,/* bump up the number of spiltted parts of a page */ ++spiltted;/* also update nr_pages */ - fe->cl->nr_pages =max_t (pgoff_t, fe->cl->nr_pages, index +1 ); + fe->pcl->nr_pages =max_t (pgoff_t, fe->pcl->nr_pages, index +1 ); next_part:/* can be used for verification */ map->m_llen = offset + cur - map->m_la;
@@ -821,15 +801,13 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,enum z_erofs_page_type page_type;bool overlapped, partial; -struct z_erofs_collection *cl;int err;might_sleep (); - cl =z_erofs_primarycollection (pcl); -DBG_BUGON (!READ_ONCE (cl->nr_pages)); +DBG_BUGON (!READ_ONCE (pcl->nr_pages)); -mutex_lock (&cl->lock); - nr_pages = cl->nr_pages; +mutex_lock (&pcl->lock); + nr_pages = pcl->nr_pages;if (nr_pages <=Z_EROFS_VMAP_ONSTACK_PAGES ) { pages = pages_onstack;
@@ -857,9 +835,9 @@ static int z_erofs_decompress_pcluster(struct super_block *sb, err =0 ;z_erofs_pagevec_ctor_init (&ctor,Z_EROFS_NR_INLINE_PAGEVECS , - cl->pagevec,0 ); + pcl->pagevec,0 ); -for (i =0 ; i < cl->vcnt; ++i) { +for (i =0 ; i < pcl->vcnt; ++i) {unsigned int pagenr; page =z_erofs_pagevec_dequeue (&ctor, &page_type);
@@ -945,11 +923,11 @@ static int z_erofs_decompress_pcluster(struct super_block *sb, goto out; llen = pcl->length >>Z_EROFS_PCLUSTER_LENGTH_BIT ; -if (nr_pages <<PAGE_SHIFT >= cl->pageofs + llen) { +if (nr_pages <<PAGE_SHIFT >= pcl->pageofs_out + llen) { outputsize = llen; partial = !(pcl->length &Z_EROFS_PCLUSTER_FULL_LENGTH ); }else { - outputsize = (nr_pages <<PAGE_SHIFT ) - cl->pageofs; + outputsize = (nr_pages <<PAGE_SHIFT ) - pcl->pageofs_out; partial = true; }
@@ -963,7 +941,7 @@ static int z_erofs_decompress_pcluster(struct super_block *sb, .in = compressed_pages, .out = pages, .pageofs_in = pcl->pageofs_in, - .pageofs_out = cl->pageofs, + .pageofs_out = pcl->pageofs_out, .inputsize = inputsize, .outputsize = outputsize, .alg = pcl->algorithmformat,
@@ -1012,16 +990,12 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,else if (pages != pages_onstack)kvfree (pages); - cl->nr_pages =0 ; - cl->vcnt =0 ; + pcl->nr_pages =0 ; + pcl->vcnt =0 ; -/* all cl locks MUST be taken before the following line */ +/* pcluster lock MUST be taken before the following line */ WRITE_ONCE (pcl->next,Z_EROFS_PCLUSTER_NIL ); - -/* all cl locks SHOULD be released right now */ -mutex_unlock (&cl->lock); - -z_erofs_collection_put (cl); +mutex_unlock (&pcl->lock);return err; }
@@ -1043,6 +1017,7 @@ static void z_erofs_decompress_queue(const struct z_erofs_decompressqueue *io, owned =READ_ONCE (pcl->next);z_erofs_decompress_pcluster (io->sb, pcl, pagepool); +erofs_workgroup_put (&pcl->obj); } }
fs/erofs/zdata.h +22 -28
80 lines not on this page · raw message
Signature · 1 line
2.30.2
Gao Xiang
maintainerPATCH 2/3 erofs: get rid of label `restart_now'
In reply to Gao XiangFrom: Gao Xiang <[email protected]>
Simplify this part of code. No logic changes.
--- fs/erofs/zdata.c | 33 ++++++++++++++------------------- 1 file changed, 14 insertions(+), 19 deletions(-)
fs/erofs/zdata.c +14 -19
@@ -643,28 +643,23 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe, repeat: cur = end -1 ; -/* lucky, within the range of the current map_blocks */ -if (offset + cur >= map->m_la && - offset + cur < map->m_la + map->m_llen) { +if (offset + cur < map->m_la || + offset + cur >= map->m_la + map->m_llen) { +erofs_dbg ("out-of-range map @ pos %llu" , offset + cur); + +if (z_erofs_collector_end (fe)) + fe->backmost = false; + map->m_la = offset + cur; + map->m_llen =0 ; + err =z_erofs_map_blocks_iter (inode, map,0 ); +if (err) + goto err_out; + }else { +if (fe->pcl) + goto hitted;/* didn't get a valid pcluster previously (very rare) */ -if (!fe->pcl) - goto restart_now; - goto hitted; } -/* go ahead the next map_blocks */ -erofs_dbg ("%s: [out-of-range] pos %llu" , __func__, offset + cur); - -if (z_erofs_collector_end (fe)) - fe->backmost = false; - - map->m_la = offset + cur; - map->m_llen =0 ; - err =z_erofs_map_blocks_iter (inode, map,0 ); -if (err) - goto err_out; - -restart_now:if (!(map->m_flags &EROFS_MAP_MAPPED )) goto hitted;
Signature · 1 line
2.30.2
Gao Xiang
maintainerPATCH 3/3 erofs: simplify z_erofs_pcluster_readmore()
In reply to Gao XiangFrom: Gao Xiang <[email protected]>
Get rid of unnecessary label `skip'. No logic changes.
--- fs/erofs/zdata.c | 23 ++++++++++------------- 1 file changed, 10 insertions(+), 13 deletions(-)
fs/erofs/zdata.c +10 -13
@@ -1436,22 +1436,19 @@ static void z_erofs_pcluster_readmore(struct z_erofs_decompress_frontend *f,struct page *page; page =erofs_grab_cache_page_nowait (inode->i_mapping, index); -if (!page) - goto skip; - -if (PageUptodate (page)) { -unlock_page (page); +if (page) { +if (PageUptodate (page)) { +unlock_page (page); + }else { + err =z_erofs_do_read_page (f, page, pagepool); +if (err) +erofs_err (inode->i_sb, +"readmore error at page %lu @ nid %llu" , + index,EROFS_I (inode)->nid); + }put_page (page); - goto skip; } - err =z_erofs_do_read_page (f, page, pagepool); -if (err) -erofs_err (inode->i_sb, -"readmore error at page %lu @ nid %llu" , - index,EROFS_I (inode)->nid); -put_page (page); -skip:if (cur <PAGE_SIZE )break ; cur = (index <<PAGE_SHIFT ) -1 ;
Signature · 1 line
2.30.2
Thanks,
Quoted 33 lines · Gao Xiang
From: Gao Xiang <[email protected]>
Hi folks,
Now I'm working on cleanuping decompression code and doing some folio adaption for the next 5.20 cycle, see:
https://git.kernel.org/pub/scm/linux/kernel/git/xiang/linux.git/log/?h=erofs/folios
This cleanup work completely gets rid of PageError usage finally and tries to prepare for introducing rolling hashing for EROFS since EROFS supports compressing variable-sized data instead of fixed-sized clusters.
Therefore, EROFS can support rolling hash easily and our mechanism can make full use of filesystem interfaces (byte-addressed) naturally.
Before that, I'd like to submit some trivial cleanups in advance for the 5.19 cycle. All patches are without any logical change, so I can have a more recent codebase for the next rework cycle.
Thanks,
Gao XiangGao Xiang (3):
erofs: get rid of `structz_erofs_collection'
erofs: get rid of label `restart_now'
erofs: simplifyz_erofs_pcluster_readmore()fs/erofs/zdata.c | 165 +++++++++++++++++++---------------------------- fs/erofs/zdata.h | 50 +++++++------- 2 files changed, 88 insertions(+), 127 deletions(-)
Quoted 1 line · Chao Yu
Acked-by Chao Yu <[email protected]>
Thanks Chao! It'd be much helpful for me if we apply these cleanups so I can have a new folio rework+cleanup+rolling hash 5.20 cycle.
Thanks,
Gao Xiang
Quoted 37 lines
Thanks,
Quoted 33 lines · Gao Xiang
From: Gao Xiang <[email protected]>
Hi folks,
Now I'm working on cleanuping decompression code and doing some folio adaption for the next 5.20 cycle, see:
https://git.kernel.org/pub/scm/linux/kernel/git/xiang/linux.git/log/?h=erofs/folios
This cleanup work completely gets rid of PageError usage finally and tries to prepare for introducing rolling hashing for EROFS since EROFS supports compressing variable-sized data instead of fixed-sized clusters.
Therefore, EROFS can support rolling hash easily and our mechanism can make full use of filesystem interfaces (byte-addressed) naturally.
Before that, I'd like to submit some trivial cleanups in advance for the 5.19 cycle. All patches are without any logical change, so I can have a more recent codebase for the next rework cycle.
Thanks,
Gao XiangGao Xiang (3):
erofs: get rid of `structz_erofs_collection'
erofs: get rid of label `restart_now'
erofs: simplifyz_erofs_pcluster_readmore()fs/erofs/zdata.c | 165 +++++++++++++++++++---------------------------- fs/erofs/zdata.h | 50 +++++++------- 2 files changed, 88 insertions(+), 127 deletions(-)