mm/page_io: use sio->len for PSWPIN accounting in sio_read_complete()
Subsystems (2)
PATCH v4 1/2 mm/page_io: use sio->len for PSWPIN accounting in sio_read_complete()
sio_read_complete() uses sio->pages to account global PSWPIN vm events, but sio->pages tracks the number of bvec entries (folios), not base pages.
While large folios cannot currently reach this path (SWP_FS_OPS and SWP_SYNCHRONOUS_IO are mutually exclusive, and mTHP swap-in allocation is gated on SWP_SYNCHRONOUS_IO), the accounting is semantically inconsistent with the per-memcg path which correctly uses folio_nr_pages().
Use sio->len >> PAGE_SHIFT instead, which gives the correct base page count since sio->len is accumulated via folio_size(folio).
--- mm/page_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
mm/page_io.c +1 -1
@@ -497,7 +497,7 @@ static void sio_read_complete(struct kiocb *iocb, long ret)folio_mark_uptodate (folio);folio_unlock (folio); } -count_vm_events (PSWPIN , sio->pages); +count_vm_events (PSWPIN , sio->len >>PAGE_SHIFT ); }else {for (p =0 ; p < sio->pages; p++) {struct folio *folio =page_folio (sio->bvec[p].bv_page);
Signature · 1 line
2.53.0
PATCH v4 2/2 mm/page_io: rename swap_iocb->pages to swap_iocb->bvecs
swap_iocb->pages tracks the number of bvec entries (folios), not base pages. Rename it to bvecs to accurately reflect its purpose.
--- mm/page_io.c | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 deletions(-)
mm/page_io.c +15 -15
@@ -327,7 +327,7 @@ static void bio_associate_blkg_from_page(struct bio *bio, struct folio *folio)struct swap_iocb {struct kiocb iocb;struct bio_vec bvec[SWAP_CLUSTER_MAX ]; -int pages; +int bvecs;int len; };static mempool_t *sio_pool;
@@ -362,14 +362,14 @@ static void sio_write_complete(struct kiocb *iocb, long ret)*/ pr_err_ratelimited ("Write error %ld on dio swapfile (%llu)\n" , ret,swap_dev_pos (page_swap_entry (page))); -for (p =0 ; p < sio->pages; p++) { +for (p =0 ; p < sio->bvecs; p++) { page = sio->bvec[p].bv_page;set_page_dirty (page);ClearPageReclaim (page); } } -for (p =0 ; p < sio->pages; p++) +for (p =0 ; p < sio->bvecs; p++)end_page_writeback (sio->bvec[p].bv_page);mempool_free (sio, sio_pool);
@@ -397,13 +397,13 @@ static void swap_writepage_fs(struct folio *folio, struct swap_iocb **swap_plug)init_sync_kiocb (&sio->iocb, swap_file); sio->iocb.ki_complete = sio_write_complete; sio->iocb.ki_pos = pos; - sio->pages =0 ; + sio->bvecs =0 ; sio->len =0 ; } -bvec_set_folio (&sio->bvec[sio->pages], folio,folio_size (folio),0 ); +bvec_set_folio (&sio->bvec[sio->bvecs], folio,folio_size (folio),0 ); sio->len +=folio_size (folio); - sio->pages +=1 ; -if (sio->pages ==ARRAY_SIZE (sio->bvec) || !swap_plug) { + sio->bvecs +=1 ; +if (sio->bvecs ==ARRAY_SIZE (sio->bvec) || !swap_plug) {swap_write_unplug (sio); sio =NULL ; }
@@ -477,7 +477,7 @@ void swap_write_unplug(struct swap_iocb *sio)struct address_space *mapping = sio->iocb.ki_filp->f_mapping;int ret; -iov_iter_bvec (&from,ITER_SOURCE , sio->bvec, sio->pages, sio->len); +iov_iter_bvec (&from,ITER_SOURCE , sio->bvec, sio->bvecs, sio->len); ret = mapping->a_ops->swap_rw (&sio->iocb, &from);if (ret != -EIOCBQUEUED )sio_write_complete (&sio->iocb, ret);
@@ -489,7 +489,7 @@ static void sio_read_complete(struct kiocb *iocb, long ret)int p;if (ret == sio->len) { -for (p =0 ; p < sio->pages; p++) { +for (p =0 ; p < sio->bvecs; p++) {struct folio *folio =page_folio (sio->bvec[p].bv_page);count_mthp_stat (folio_order (folio),MTHP_STAT_SWPIN );
@@ -499,7 +499,7 @@ static void sio_read_complete(struct kiocb *iocb, long ret) }count_vm_events (PSWPIN , sio->len >>PAGE_SHIFT ); }else { -for (p =0 ; p < sio->pages; p++) { +for (p =0 ; p < sio->bvecs; p++) {struct folio *folio =page_folio (sio->bvec[p].bv_page);folio_unlock (folio);
@@ -559,13 +559,13 @@ static void swap_read_folio_fs(struct folio *folio, struct swap_iocb **plug)init_sync_kiocb (&sio->iocb, sis->swap_file); sio->iocb.ki_pos = pos; sio->iocb.ki_complete = sio_read_complete; - sio->pages =0 ; + sio->bvecs =0 ; sio->len =0 ; } -bvec_set_folio (&sio->bvec[sio->pages], folio,folio_size (folio),0 ); +bvec_set_folio (&sio->bvec[sio->bvecs], folio,folio_size (folio),0 ); sio->len +=folio_size (folio); - sio->pages +=1 ; -if (sio->pages ==ARRAY_SIZE (sio->bvec) || !plug) { + sio->bvecs +=1 ; +if (sio->bvecs ==ARRAY_SIZE (sio->bvec) || !plug) {swap_read_unplug (sio); sio =NULL ; }
@@ -666,7 +666,7 @@ void __swap_read_unplug(struct swap_iocb *sio)struct address_space *mapping = sio->iocb.ki_filp->f_mapping;int ret; -iov_iter_bvec (&from,ITER_DEST , sio->bvec, sio->pages, sio->len); +iov_iter_bvec (&from,ITER_DEST , sio->bvec, sio->bvecs, sio->len); ret = mapping->a_ops->swap_rw (&sio->iocb, &from);if (ret != -EIOCBQUEUED )sio_read_complete (&sio->iocb, ret);
Signature · 1 line
2.53.0
Re: [PATCH v4 2/2] mm/page_io: rename swap_iocb->pages to swap_iocb->bvecs
Quoted 14 lines · David Carlier
swap_iocb->pagestracks the number of bvec entries (folios), not base pages. Rename it to bvecs to accurately reflect its purpose.Suggested-by Matthew Wilcox (Oracle) <[email protected]>Signed-off-by David Carlier <[email protected]>--- mm/page_io.c | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 deletions(-)mm/page_io.c +0
-0
mm/page_io.c static void bio_associate_blkg_from_page(struct bio *bio, struct folio *folio) · line 330
Quoted 5 lines
mm/page_io.c +1
-1@@ -327,7 +327,7 @@ static void bio_associate_blkg_from_page(struct bio *bio, struct folio *folio)struct swap_iocb {struct kiocb iocb;struct bio_vec bvec[SWAP_CLUSTER_MAX ]; -int pages; +int bvecs;
bvec (an array of bvecs) vs bvecs. Hm.
Wouldn't it be cleaner to have
struct bio_vec bvecs[SWAP_CLUSTER_MAX]; int nr_bvecs;
?
Signature · 3 lines
Cheers, David
Re: [PATCH v4 2/2] mm/page_io: rename swap_iocb->pages to swap_iocb->bvecs
I went for it initially but I thought I saw bvecs in a message :) but I do not mind it is just naming after all, let s go for it.
Quoted 35 lines · David Hildenbrand (Arm)
Quoted 14 lines · David Carlier
swap_iocb->pagestracks the number of bvec entries (folios), not base pages. Rename it to bvecs to accurately reflect its purpose.Suggested-by Matthew Wilcox (Oracle) <[email protected]>Signed-off-by David Carlier <[email protected]>--- mm/page_io.c | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 deletions(-)mm/page_io.c +0
-0mm/page_io.c static void bio_associate_blkg_from_page(struct bio *bio, struct folio *folio) · line 330
Quoted 5 lines
mm/page_io.c +1
-1@@ -327,7 +327,7 @@ static void bio_associate_blkg_from_page(struct bio *bio, struct folio *folio)struct swap_iocb {struct kiocb iocb;struct bio_vec bvec[SWAP_CLUSTER_MAX ]; -int pages; +int bvecs;bvec (an array of bvecs) vs bvecs. Hm.
Wouldn't it be cleaner to have
struct bio_vec bvecs[SWAP_CLUSTER_MAX]; int nr_bvecs;?
--
Cheers,David
Quoted 10 lines · David Carlier
sio_read_complete()usessio->pagesto account global PSWPIN vm events, butsio->pagestracks the number of bvec entries (folios), not base pages.While large folios cannot currently reach this path (
SWP_FS_OPSandSWP_SYNCHRONOUS_IOare mutually exclusive, and mTHP swap-in allocation is gated onSWP_SYNCHRONOUS_IO), the accounting is semantically inconsistent with the per-memcg path which correctly usesfolio_nr_pages().
Quoted 5 lines
Use
sio->len>>PAGE_SHIFTinstead, which gives the correct base page count sincesio->lenis accumulated viafolio_size(folio).Signed-off-by David Carlier <[email protected]>---
Signature · 3 lines
Cheers, David