Hi Chao,
On Tue, Dec 08, 2020 at 04:15:59PM +0800, Chao Yu wrote:
> On 2020/12/7 9:23, Gao Xiang wrote:
> > Previously, we played around with magical page->mapping for short-lived
> > temporary pages since we need to identify different types of pages in
> > the same pcluster but both invalidated and short-lived temporary pages
> > can have page->mapping == NULL. It was considered as safe because that
> > temporary pages are all non-LRU / non-movable pages.
> >
> > This patch tends to use specific page->private to identify short-lived
> > pages instead so it won't rely on page->mapping anymore. Details are
> > described in "compress.h" as well.
> >
> > Signed-off-by: Gao Xiang <hsiang...@redhat.com>
> > ---
> > tested with ro_fsstress for a whole night.
> >
> > The old "[PATCH 4/4] erofs: complete a missing case for inplace I/O" is
> > temporarily dropped since ro_fsstress failed with such modification,
> > will look into later.
> >
> > fs/erofs/compress.h | 50 ++++++++++++++++++++++++++++++-----------
> > fs/erofs/decompressor.c | 2 +-
> > fs/erofs/zdata.c | 42 +++++++++++++++++++++-------------
> > fs/erofs/zdata.h | 1 +
> > 4 files changed, 65 insertions(+), 30 deletions(-)
> >
> > diff --git a/fs/erofs/compress.h b/fs/erofs/compress.h
> > index 3d452443c545..2bbf47f353ef 100644
> > --- a/fs/erofs/compress.h
> > +++ b/fs/erofs/compress.h
> > @@ -26,30 +26,54 @@ struct z_erofs_decompress_req {
> > bool inplace_io, partial_decoding;
> > };
> > +#define Z_EROFS_SHORTLIVED_PAGE (-1UL << 2)
> > +
> > /*
> > - * - 0x5A110C8D ('sallocated', Z_EROFS_MAPPING_STAGING) -
> > - * used to mark temporary allocated pages from other
> > - * file/cached pages and NULL mapping pages.
> > + * For all pages in a pcluster, page->private should be one of
> > + * Type Last 2bits page->private
> > + * short-lived page 00 Z_EROFS_SHORTLIVED_PAGE
> > + * cached/managed page 00 pointer to z_erofs_pcluster
> > + * online page (file-backed, 01/10/11 sub-index << 2 | count
> > + * some pages can be used for inplace I/O)
> > + *
> > + * page->mapping should be one of
> > + * Type page->mapping
> > + * short-lived page NULL
> > + * cached/managed page non-NULL or NULL (invalidated/truncated page)
> > + * online page non-NULL
> > + *
> > + * For all managed pages, PG_private should be set with 1 extra refcount,
> > + * which is used for page reclaim / migration.
>
> FYI, there is a generic way to set/clear page_private, it binds the private
> value set and page count operation in one function:
>
> attach_page_private()
> detach_page_private()
>
> If there are use cases, let's try to use them as much as possible.
I discussed this case in the original thread,
https://lore.kernel.org/r/20200519100612.GA3687@hsiangkao-HP-ZHAN-66-Pro-G1
The previous conclusion is that for EROFS case (see Matthew's reply) this
pair won't have too much usage. since EROFS pattern saves extra page
reference count (- and +) by cases.
I could use attach_page_private() and detach_page_private() if possible,
but the problem is I'm not not its such pair internal implementation is
stable (but the PG_Private rule is stable for decades);
>
> > */
> > -#define Z_EROFS_MAPPING_STAGING ((void *)0x5A110C8D)
> > -/* check if a page is marked as staging */
> > -static inline bool z_erofs_page_is_staging(struct page *page)
> > +/*
> > + * short-lived pages are pages directly from buddy system with specific
> > + * page->private (no need to set PagePrivate since these are non-LRU /
> > + * non-movable pages and bypass reclaim / migration code).
> > + */
> > +static inline bool z_erofs_is_shortlived_page(struct page *page)
> > {
> > - return page->mapping == Z_EROFS_MAPPING_STAGING;
> > + if (page->private != Z_EROFS_SHORTLIVED_PAGE)
> > + return false;
> > +
> > + DBG_BUGON(page->mapping);
> > + return true;
> > }
> > -static inline bool z_erofs_put_stagingpage(struct list_head *pagepool,
> > - struct page *page)
> > +static inline bool z_erofs_put_shortlivedpage(struct list_head *pagepool,
> > + struct page *page)
> > {
> > - if (!z_erofs_page_is_staging(page))
> > + if (!z_erofs_is_shortlived_page(page))
> > return false;
> > - /* staging pages should not be used by others at the same time */
> > - if (page_ref_count(page) > 1)
> > + /* short-lived pages should not be used by others at the same time */
> > + if (page_ref_count(page) > 1) {
>
> Does this be a possible case?
Yes, see decompress.c, if will get_page(page);
>
> > put_page(page);
> > - else
> > + } else {
> > + /* follow the pcluster rule above. */
> > + set_page_private(page, 0);
> > list_add(&page->lru, pagepool);
> > + }
> > return true;
> > }
> > diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
> > index cbadbf55c6c2..1cb1ffd10569 100644
> > --- a/fs/erofs/decompressor.c
> > +++ b/fs/erofs/decompressor.c
> > @@ -76,7 +76,7 @@ static int z_erofs_lz4_prepare_destpages(struct
> > z_erofs_decompress_req *rq,
> > victim = erofs_allocpage(pagepool, GFP_KERNEL);
> > if (!victim)
> > return -ENOMEM;
> > - victim->mapping = Z_EROFS_MAPPING_STAGING;
> > + set_page_private(victim, Z_EROFS_SHORTLIVED_PAGE);
> > }
> > rq->out[i] = victim;
> > }
> > diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
> > index 86fd3bf62af6..afeadf413c2c 100644
> > --- a/fs/erofs/zdata.c
> > +++ b/fs/erofs/zdata.c
> > @@ -255,6 +255,7 @@ int erofs_try_to_free_cached_page(struct address_space
> > *mapping,
> > erofs_workgroup_unfreeze(&pcl->obj, 1);
> > if (ret) {
> > + set_page_private(page, 0);
> > ClearPagePrivate(page);
> > put_page(page);
>
> detach_page_private()?
The same as the above.
Thanks,
Gao Xiang
>
> Thanks,
>
> > }
> > @@ -648,12 +649,12 @@ static int z_erofs_do_read_page(struct
> > z_erofs_decompress_frontend *fe,
> > retry:
> > err = z_erofs_attach_page(clt, page, page_type);
> > - /* should allocate an additional staging page for pagevec */
> > + /* should allocate an additional short-lived page for pagevec */
> > if (err == -EAGAIN) {
> > struct page *const newpage =
> > alloc_page(GFP_NOFS | __GFP_NOFAIL);
> > - newpage->mapping = Z_EROFS_MAPPING_STAGING;
> > + set_page_private(newpage, Z_EROFS_SHORTLIVED_PAGE);
> > err = z_erofs_attach_page(clt, newpage,
> > Z_EROFS_PAGE_TYPE_EXCLUSIVE);
> > if (!err)
> > @@ -710,6 +711,11 @@ static void z_erofs_decompress_kickoff(struct
> > z_erofs_decompressqueue *io,
> > queue_work(z_erofs_workqueue, &io->u.work);
> > }
> > +static bool z_erofs_page_is_invalidated(struct page *page)
> > +{
> > + return !page->mapping && !z_erofs_is_shortlived_page(page);
> > +}
> > +
> > static void z_erofs_decompressqueue_endio(struct bio *bio)
> > {
> > tagptr1_t t = tagptr_init(tagptr1_t, bio->bi_private);
> > @@ -722,7 +728,7 @@ static void z_erofs_decompressqueue_endio(struct bio
> > *bio)
> > struct page *page = bvec->bv_page;
> > DBG_BUGON(PageUptodate(page));
> > - DBG_BUGON(!page->mapping);
> > + DBG_BUGON(z_erofs_page_is_invalidated(page));
> > if (err)
> > SetPageError(page);
> > @@ -795,9 +801,9 @@ static int z_erofs_decompress_pcluster(struct
> > super_block *sb,
> > /* all pages in pagevec ought to be valid */
> > DBG_BUGON(!page);
> > - DBG_BUGON(!page->mapping);
> > + DBG_BUGON(z_erofs_page_is_invalidated(page));
> > - if (z_erofs_put_stagingpage(pagepool, page))
> > + if (z_erofs_put_shortlivedpage(pagepool, page))
> > continue;
> > if (page_type == Z_EROFS_VLE_PAGE_TYPE_HEAD)
> > @@ -831,9 +837,9 @@ static int z_erofs_decompress_pcluster(struct
> > super_block *sb,
> > /* all compressed pages ought to be valid */
> > DBG_BUGON(!page);
> > - DBG_BUGON(!page->mapping);
> > + DBG_BUGON(z_erofs_page_is_invalidated(page));
> > - if (!z_erofs_page_is_staging(page)) {
> > + if (!z_erofs_is_shortlived_page(page)) {
> > if (erofs_page_is_managed(sbi, page)) {
> > if (!PageUptodate(page))
> > err = -EIO;
> > @@ -858,7 +864,7 @@ static int z_erofs_decompress_pcluster(struct
> > super_block *sb,
> > overlapped = true;
> > }
> > - /* PG_error needs checking for inplaced and staging pages */
> > + /* PG_error needs checking for all non-managed pages */
> > if (PageError(page)) {
> > DBG_BUGON(PageUptodate(page));
> > err = -EIO;
> > @@ -897,8 +903,8 @@ static int z_erofs_decompress_pcluster(struct
> > super_block *sb,
> > if (erofs_page_is_managed(sbi, page))
> > continue;
> > - /* recycle all individual staging pages */
> > - (void)z_erofs_put_stagingpage(pagepool, page);
> > + /* recycle all individual short-lived pages */
> > + (void)z_erofs_put_shortlivedpage(pagepool, page);
> > WRITE_ONCE(compressed_pages[i], NULL);
> > }
> > @@ -908,10 +914,10 @@ static int z_erofs_decompress_pcluster(struct
> > super_block *sb,
> > if (!page)
> > continue;
> > - DBG_BUGON(!page->mapping);
> > + DBG_BUGON(z_erofs_page_is_invalidated(page));
> > - /* recycle all individual staging pages */
> > - if (z_erofs_put_stagingpage(pagepool, page))
> > + /* recycle all individual short-lived pages */
> > + if (z_erofs_put_shortlivedpage(pagepool, page))
> > continue;
> > if (err < 0)
> > @@ -1011,13 +1017,17 @@ static struct page
> > *pickup_page_for_submission(struct z_erofs_pcluster *pcl,
> > mapping = READ_ONCE(page->mapping);
> > /*
> > - * unmanaged (file) pages are all locked solidly,
> > + * file-backed online pages in plcuster are all locked steady,
> > * therefore it is impossible for `mapping' to be NULL.
> > */
> > if (mapping && mapping != mc)
> > /* ought to be unmanaged pages */
> > goto out;
> > + /* directly return for shortlived page as well */
> > + if (z_erofs_is_shortlived_page(page))
> > + goto out;
> > +
> > lock_page(page);
> > /* only true if page reclaim goes wrong, should never happen */
> > @@ -1062,8 +1072,8 @@ static struct page *pickup_page_for_submission(struct
> > z_erofs_pcluster *pcl,
> > out_allocpage:
> > page = erofs_allocpage(pagepool, gfp | __GFP_NOFAIL);
> > if (!tocache || add_to_page_cache_lru(page, mc, index + nr, gfp)) {
> > - /* non-LRU / non-movable temporary page is needed */
> > - page->mapping = Z_EROFS_MAPPING_STAGING;
> > + /* turn into temporary page if fails */
> > + set_page_private(page, Z_EROFS_SHORTLIVED_PAGE);
> > tocache = false;
> > }
> > diff --git a/fs/erofs/zdata.h b/fs/erofs/zdata.h
> > index 68c9b29fc0ca..b503b353d4ab 100644
> > --- a/fs/erofs/zdata.h
> > +++ b/fs/erofs/zdata.h
> > @@ -173,6 +173,7 @@ static inline void z_erofs_onlinepage_endio(struct page
> > *page)
> > v = atomic_dec_return(u.o);
> > if (!(v & Z_EROFS_ONLINEPAGE_COUNT_MASK)) {
> > + set_page_private(page, 0);
> > ClearPagePrivate(page);
> > if (!PageError(page))
> > SetPageUptodate(page);
> >
>
