| Message ID | 20221213212053.106058-1-sidhartha.kumar@oracle.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [mm-unstable] mm: move folio_set_compound_order() to mm/internal.h | expand |
On 12/13/22 13:20, Sidhartha Kumar wrote: > folio_set_compound_order() is moved to an mm-internal location so external > folio users cannot misuse this function. Change the name of the function > to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, > handle the case if a non-large folio is passed and add clarifying comments > to the function. > > Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ > Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") > > Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> > Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> > Suggested-by: Muchun Song <songmuchun@bytedance.com> > Suggested-by: Matthew Wilcox <willy@infradead.org> > Suggested-by: John Hubbard <jhubbard@nvidia.com> > --- > include/linux/mm.h | 16 ---------------- > mm/hugetlb.c | 6 +++--- > mm/internal.h | 21 +++++++++++++++++++++ > 3 files changed, 24 insertions(+), 19 deletions(-) I think this looks good. One small question below. > > diff --git a/include/linux/mm.h b/include/linux/mm.h > index 7dc376052d40..300d92d2b49d 100644 > --- a/include/linux/mm.h > +++ b/include/linux/mm.h > @@ -1019,22 +1019,6 @@ static inline void set_compound_order(struct page *page, unsigned int order) > #endif > } > > -/* > - * folio_set_compound_order is generally passed a non-zero order to > - * initialize a large folio. However, hugetlb code abuses this by > - * passing in zero when 'dissolving' a large folio. > - */ > -static inline void folio_set_compound_order(struct folio *folio, > - unsigned int order) > -{ > - VM_BUG_ON_FOLIO(!folio_test_large(folio), folio); > - > - folio->_folio_order = order; > -#ifdef CONFIG_64BIT > - folio->_folio_nr_pages = order ? 1U << order : 0; > -#endif > -} > - > /* Returns the number of pages in this potentially compound page. */ > static inline unsigned long compound_nr(struct page *page) > { > diff --git a/mm/hugetlb.c b/mm/hugetlb.c > index 7cdbcc22587b..810e840bb4f1 100644 > --- a/mm/hugetlb.c > +++ b/mm/hugetlb.c > @@ -1344,7 +1344,7 @@ static void __destroy_compound_gigantic_folio(struct folio *folio, > set_page_refcounted(p); > } > > - folio_set_compound_order(folio, 0); > + folio_set_order(folio, 0); > __folio_clear_head(folio); > } > > @@ -1808,7 +1808,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, > __folio_clear_reserved(folio); > __folio_set_head(folio); > /* we rely on prep_new_hugetlb_folio to set the destructor */ > - folio_set_compound_order(folio, order); > + folio_set_order(folio, order); > for (i = 0; i < nr_pages; i++) { > p = folio_page(folio, i); > > @@ -1872,7 +1872,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, > p = folio_page(folio, j); > __ClearPageReserved(p); > } > - folio_set_compound_order(folio, 0); > + folio_set_order(folio, 0); > __folio_clear_head(folio); > return false; > } > diff --git a/mm/internal.h b/mm/internal.h > index bcf75a8b032d..829b6a60ceb7 100644 > --- a/mm/internal.h > +++ b/mm/internal.h > @@ -378,6 +378,27 @@ extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, > int split_free_page(struct page *free_page, > unsigned int order, unsigned long split_pfn_offset); > > +/* > + * This will have no effect, other than possibly generating a warning, if the > + * caller passes in a non-large folio. > + */ > +static inline void folio_set_order(struct folio *folio, unsigned int order) > +{ > + if (!folio_test_large(folio)) { > + WARN_ON_ONCE(order); > + return; > + } Would it be better to do this (below)? I'm not sure of the value of warning on "order"--it's a little odd and unexplained and doesn't really do anything more helpful than simply warning about what why the code is failing, which is really about !large, rather than order. Unless I'm missing something? if (WARN_ON_ONCE(!folio_test_large(folio))) return; Sorry to drive you crazy over nits. This is the last one from me. :) thanks,
On 12/14/22 12:43 AM, John Hubbard wrote: > On 12/13/22 13:20, Sidhartha Kumar wrote: >> folio_set_compound_order() is moved to an mm-internal location so external >> folio users cannot misuse this function. Change the name of the function >> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, >> handle the case if a non-large folio is passed and add clarifying comments >> to the function. >> >> Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ >> Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") >> >> Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> >> Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> >> Suggested-by: Muchun Song <songmuchun@bytedance.com> >> Suggested-by: Matthew Wilcox <willy@infradead.org> >> Suggested-by: John Hubbard <jhubbard@nvidia.com> >> --- >> include/linux/mm.h | 16 ---------------- >> mm/hugetlb.c | 6 +++--- >> mm/internal.h | 21 +++++++++++++++++++++ >> 3 files changed, 24 insertions(+), 19 deletions(-) > > I think this looks good. One small question below. > >> >> diff --git a/include/linux/mm.h b/include/linux/mm.h >> index 7dc376052d40..300d92d2b49d 100644 >> --- a/include/linux/mm.h >> +++ b/include/linux/mm.h >> @@ -1019,22 +1019,6 @@ static inline void set_compound_order(struct page *page, unsigned int order) >> #endif >> } >> >> -/* >> - * folio_set_compound_order is generally passed a non-zero order to >> - * initialize a large folio. However, hugetlb code abuses this by >> - * passing in zero when 'dissolving' a large folio. >> - */ >> -static inline void folio_set_compound_order(struct folio *folio, >> - unsigned int order) >> -{ >> - VM_BUG_ON_FOLIO(!folio_test_large(folio), folio); >> - >> - folio->_folio_order = order; >> -#ifdef CONFIG_64BIT >> - folio->_folio_nr_pages = order ? 1U << order : 0; >> -#endif >> -} >> - >> /* Returns the number of pages in this potentially compound page. */ >> static inline unsigned long compound_nr(struct page *page) >> { >> diff --git a/mm/hugetlb.c b/mm/hugetlb.c >> index 7cdbcc22587b..810e840bb4f1 100644 >> --- a/mm/hugetlb.c >> +++ b/mm/hugetlb.c >> @@ -1344,7 +1344,7 @@ static void __destroy_compound_gigantic_folio(struct folio *folio, >> set_page_refcounted(p); >> } >> >> - folio_set_compound_order(folio, 0); >> + folio_set_order(folio, 0); >> __folio_clear_head(folio); >> } >> >> @@ -1808,7 +1808,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, >> __folio_clear_reserved(folio); >> __folio_set_head(folio); >> /* we rely on prep_new_hugetlb_folio to set the destructor */ >> - folio_set_compound_order(folio, order); >> + folio_set_order(folio, order); >> for (i = 0; i < nr_pages; i++) { >> p = folio_page(folio, i); >> >> @@ -1872,7 +1872,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, >> p = folio_page(folio, j); >> __ClearPageReserved(p); >> } >> - folio_set_compound_order(folio, 0); >> + folio_set_order(folio, 0); >> __folio_clear_head(folio); >> return false; >> } >> diff --git a/mm/internal.h b/mm/internal.h >> index bcf75a8b032d..829b6a60ceb7 100644 >> --- a/mm/internal.h >> +++ b/mm/internal.h >> @@ -378,6 +378,27 @@ extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, >> int split_free_page(struct page *free_page, >> unsigned int order, unsigned long split_pfn_offset); >> >> +/* >> + * This will have no effect, other than possibly generating a warning, if the >> + * caller passes in a non-large folio. >> + */ >> +static inline void folio_set_order(struct folio *folio, unsigned int order) >> +{ >> + if (!folio_test_large(folio)) { >> + WARN_ON_ONCE(order); >> + return; >> + } > > Would it be better to do this (below)? I'm not sure of the value of > warning on "order"--it's a little odd and unexplained and doesn't really > do anything more helpful than simply warning about what why the code is > failing, which is really about !large, rather than order. Unless I'm > missing something? > > if (WARN_ON_ONCE(!folio_test_large(folio))) > return; >I agree that warning this way is clearer, I will change in a v2. > Sorry to drive you crazy over nits. This is the last one from me. :) > No worries, I appreciate the feedback. Thanks, Sidhartha Kumar > thanks,
On 12/14/22 12:35, Sidhartha Kumar wrote: >> if (WARN_ON_ONCE(!folio_test_large(folio))) >> return; >> I agree that warning this way is clearer, I will change in a v2. With that change, please feel free to add: Reviewed-by: John Hubbard <jhubbard@nvidia.com> thanks,
> On Dec 14, 2022, at 05:20, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: > > folio_set_compound_order() is moved to an mm-internal location so external > folio users cannot misuse this function. Change the name of the function > to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, > handle the case if a non-large folio is passed and add clarifying comments > to the function. > > Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ > Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") > > Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> > Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> > Suggested-by: Muchun Song <songmuchun@bytedance.com> > Suggested-by: Matthew Wilcox <willy@infradead.org> > Suggested-by: John Hubbard <jhubbard@nvidia.com> Reviewed-by: Muchun Song <songmuchun@bytedance.com> Thanks.
On 12/14/22 7:44 PM, Muchun Song wrote: > > >> On Dec 14, 2022, at 05:20, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: >> >> folio_set_compound_order() is moved to an mm-internal location so external >> folio users cannot misuse this function. Change the name of the function >> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, >> handle the case if a non-large folio is passed and add clarifying comments >> to the function. >> >> Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ >> Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") >> >> Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> >> Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> >> Suggested-by: Muchun Song <songmuchun@bytedance.com> >> Suggested-by: Matthew Wilcox <willy@infradead.org> >> Suggested-by: John Hubbard <jhubbard@nvidia.com> > > Reviewed-by: Muchun Song <songmuchun@bytedance.com> Hi Muchun, Does this review include the change from + if (!folio_test_large(folio)) { + WARN_ON_ONCE(order); + return; + } to if (WARN_ON_ONCE(!folio_test_large(folio))) return; discussed in in this thread? Thanks, Sidhartha Kumar > > Thanks.
> On Dec 15, 2022, at 13:09, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: > > On 12/14/22 7:44 PM, Muchun Song wrote: >>> On Dec 14, 2022, at 05:20, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: >>> >>> folio_set_compound_order() is moved to an mm-internal location so external >>> folio users cannot misuse this function. Change the name of the function >>> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, >>> handle the case if a non-large folio is passed and add clarifying comments >>> to the function. >>> >>> Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ >>> Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") >>> >>> Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> >>> Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> >>> Suggested-by: Muchun Song <songmuchun@bytedance.com> >>> Suggested-by: Matthew Wilcox <willy@infradead.org> >>> Suggested-by: John Hubbard <jhubbard@nvidia.com> >> Reviewed-by: Muchun Song <songmuchun@bytedance.com> > Hi Muchun, > > Does this review include the change from Yes. Thanks. > > + if (!folio_test_large(folio)) { > + WARN_ON_ONCE(order); > + return; > + } > > to > > if (WARN_ON_ONCE(!folio_test_large(folio))) > return; > > discussed in in this thread? > > Thanks, > Sidhartha Kumar >> Thanks.
On Tue, 13 Dec 2022 13:20:53 -0800 Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: > folio_set_compound_order() is moved to an mm-internal location so external > folio users cannot misuse this function. Change the name of the function > to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, > handle the case if a non-large folio is passed and add clarifying comments > to the function. > This differs from the version I previously merged: --- a/mm/internal.h~mm-move-folio_set_compound_order-to-mm-internalh-update +++ a/mm/internal.h @@ -384,8 +384,10 @@ int split_free_page(struct page *free_pa */ static inline void folio_set_order(struct folio *folio, unsigned int order) { - if (WARN_ON_ONCE(!folio_test_large(folio))) + if (!folio_test_large(folio)) { + WARN_ON_ONCE(order); return; + } folio->_folio_order = order; #ifdef CONFIG_64BIT Makes sense. But wouldn't if (WARN_ON_ONCE(order && !folio_test_large(folio))) be clearer?
On 12/16/22 14:27, Andrew Morton wrote: > On Tue, 13 Dec 2022 13:20:53 -0800 Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: > >> folio_set_compound_order() is moved to an mm-internal location so external >> folio users cannot misuse this function. Change the name of the function >> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, >> handle the case if a non-large folio is passed and add clarifying comments >> to the function. >> > > This differs from the version I previously merged: > > --- a/mm/internal.h~mm-move-folio_set_compound_order-to-mm-internalh-update > +++ a/mm/internal.h > @@ -384,8 +384,10 @@ int split_free_page(struct page *free_pa > */ > static inline void folio_set_order(struct folio *folio, unsigned int order) > { > - if (WARN_ON_ONCE(!folio_test_large(folio))) > + if (!folio_test_large(folio)) { > + WARN_ON_ONCE(order); > return; > + } I think that's out of date? We eventually settled on the version that is (as of this a few minutes ago) already in mm-unstable (commit fdea060a130d: "mm: move folio_set_compound_order() to mm/internal.h"), which has it like this: if (WARN_ON_ONCE(!folio_test_large(folio))) return; > > folio->_folio_order = order; > #ifdef CONFIG_64BIT > > Makes sense. But wouldn't > > if (WARN_ON_ONCE(order && !folio_test_large(folio))) > > be clearer? That's a little narrower of a check. But maybe that's desirable. Could someone (Mike, Muchun, Sidhartha) comment on which behavior is preferable, please? I think I'm a little dizzy at this point. :) thanks,
On Fri, 16 Dec 2022 14:56:47 -0800 John Hubbard <jhubbard@nvidia.com> wrote: > On 12/16/22 14:27, Andrew Morton wrote: > > On Tue, 13 Dec 2022 13:20:53 -0800 Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote: > > > >> folio_set_compound_order() is moved to an mm-internal location so external > >> folio users cannot misuse this function. Change the name of the function > >> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, > >> handle the case if a non-large folio is passed and add clarifying comments > >> to the function. > >> > > > > This differs from the version I previously merged: > > > > --- a/mm/internal.h~mm-move-folio_set_compound_order-to-mm-internalh-update > > +++ a/mm/internal.h > > @@ -384,8 +384,10 @@ int split_free_page(struct page *free_pa > > */ > > static inline void folio_set_order(struct folio *folio, unsigned int order) > > { > > - if (WARN_ON_ONCE(!folio_test_large(folio))) > > + if (!folio_test_large(folio)) { > > + WARN_ON_ONCE(order); > > return; > > + } > > I think that's out of date? Yes thanks, I'm looking at patches in the wrong order. I'll drop the above delta.
On 12/16/22 2:56 PM, John Hubbard wrote: > On 12/16/22 14:27, Andrew Morton wrote: >> On Tue, 13 Dec 2022 13:20:53 -0800 Sidhartha Kumar >> <sidhartha.kumar@oracle.com> wrote: >> >>> folio_set_compound_order() is moved to an mm-internal location so >>> external >>> folio users cannot misuse this function. Change the name of the function >>> to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, >>> handle the case if a non-large folio is passed and add clarifying >>> comments >>> to the function. >>> >> >> This differs from the version I previously merged: >> >> --- >> a/mm/internal.h~mm-move-folio_set_compound_order-to-mm-internalh-update >> +++ a/mm/internal.h >> @@ -384,8 +384,10 @@ int split_free_page(struct page *free_pa >> */ >> static inline void folio_set_order(struct folio *folio, unsigned int >> order) >> { >> - if (WARN_ON_ONCE(!folio_test_large(folio))) >> + if (!folio_test_large(folio)) { >> + WARN_ON_ONCE(order); >> return; >> + } > > I think that's out of date? > > We eventually settled on the version that is (as of this a few minutes > ago) already in mm-unstable (commit fdea060a130d: "mm: move > folio_set_compound_order() to mm/internal.h"), which has it like this: > Hi Andrew, yes this version that is already in mm-unstable represents the v2 of this patch which is what we agreed on. I think the patch mm-move-folio_set_compound_order-to-mm-internalh-update with description "alter the folio_set_order() warning" which was just added to mm-unstable should be removed as our discussion lead us away from that version. > if (WARN_ON_ONCE(!folio_test_large(folio))) > return; > >> folio->_folio_order = order; >> #ifdef CONFIG_64BIT >> >> Makes sense. But wouldn't >> >> if (WARN_ON_ONCE(order && !folio_test_large(folio))) >> >> be clearer? > That's a little narrower of a check. But maybe that's desirable. Could Ya I think it would helpful to have a wider catch for the warn as any user calling folio_set_order() with a non-large folio should be aware as they could misuse the folio later on even if they passed in a 0 order because order itself would be an OOB access. Thanks, Sidhartha Kumar > someone (Mike, Muchun, Sidhartha) comment on which behavior is > preferable, please? I think I'm a little dizzy at this point. :) > > > thanks,
diff --git a/include/linux/mm.h b/include/linux/mm.h index 7dc376052d40..300d92d2b49d 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -1019,22 +1019,6 @@ static inline void set_compound_order(struct page *page, unsigned int order) #endif } -/* - * folio_set_compound_order is generally passed a non-zero order to - * initialize a large folio. However, hugetlb code abuses this by - * passing in zero when 'dissolving' a large folio. - */ -static inline void folio_set_compound_order(struct folio *folio, - unsigned int order) -{ - VM_BUG_ON_FOLIO(!folio_test_large(folio), folio); - - folio->_folio_order = order; -#ifdef CONFIG_64BIT - folio->_folio_nr_pages = order ? 1U << order : 0; -#endif -} - /* Returns the number of pages in this potentially compound page. */ static inline unsigned long compound_nr(struct page *page) { diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 7cdbcc22587b..810e840bb4f1 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -1344,7 +1344,7 @@ static void __destroy_compound_gigantic_folio(struct folio *folio, set_page_refcounted(p); } - folio_set_compound_order(folio, 0); + folio_set_order(folio, 0); __folio_clear_head(folio); } @@ -1808,7 +1808,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, __folio_clear_reserved(folio); __folio_set_head(folio); /* we rely on prep_new_hugetlb_folio to set the destructor */ - folio_set_compound_order(folio, order); + folio_set_order(folio, order); for (i = 0; i < nr_pages; i++) { p = folio_page(folio, i); @@ -1872,7 +1872,7 @@ static bool __prep_compound_gigantic_folio(struct folio *folio, p = folio_page(folio, j); __ClearPageReserved(p); } - folio_set_compound_order(folio, 0); + folio_set_order(folio, 0); __folio_clear_head(folio); return false; } diff --git a/mm/internal.h b/mm/internal.h index bcf75a8b032d..829b6a60ceb7 100644 --- a/mm/internal.h +++ b/mm/internal.h @@ -378,6 +378,27 @@ extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, int split_free_page(struct page *free_page, unsigned int order, unsigned long split_pfn_offset); +/* + * This will have no effect, other than possibly generating a warning, if the + * caller passes in a non-large folio. + */ +static inline void folio_set_order(struct folio *folio, unsigned int order) +{ + if (!folio_test_large(folio)) { + WARN_ON_ONCE(order); + return; + } + + folio->_folio_order = order; +#ifdef CONFIG_64BIT + /* + * When hugetlb dissolves a folio, we need to clear the tail + * page, rather than setting nr_pages to 1. + */ + folio->_folio_nr_pages = order ? 1U << order : 0; +#endif +} + #if defined CONFIG_COMPACTION || defined CONFIG_CMA /*
folio_set_compound_order() is moved to an mm-internal location so external folio users cannot misuse this function. Change the name of the function to folio_set_order() and use WARN_ON_ONCE() rather than BUG_ON. Also, handle the case if a non-large folio is passed and add clarifying comments to the function. Link: https://lore.kernel.org/lkml/20221207223731.32784-1-sidhartha.kumar@oracle.com/T/ Fixes: 9fd330582b2f ("mm: add folio dtor and order setter functions") Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> Suggested-by: Mike Kravetz <mike.kravetz@oracle.com> Suggested-by: Muchun Song <songmuchun@bytedance.com> Suggested-by: Matthew Wilcox <willy@infradead.org> Suggested-by: John Hubbard <jhubbard@nvidia.com> --- include/linux/mm.h | 16 ---------------- mm/hugetlb.c | 6 +++--- mm/internal.h | 21 +++++++++++++++++++++ 3 files changed, 24 insertions(+), 19 deletions(-)