| Message ID | 20200331084613.1258555-1-ying.huang@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm, trivial: Simplify swap related code in try_to_unmap_one() | expand |
On Tue 31-03-20 16:46:13, Huang, Ying wrote: > From: Huang Ying <ying.huang@intel.com> > > Because PageSwapCache() will always return false if PageSwapBacked() returns > false, and PageSwapBacked() will be check for MADV_FREE pages in > try_to_unmap_one(). The swap related code in try_to_unmap_one() can be > simplified to improve the readability. My understanding is that this is a sanity check to let us know if something breaks. Do we really want to get rid of it? Maybe it is not really useful but if that is the case then the changelog should reflect this fact. > Signed-off-by: "Huang, Ying" <ying.huang@intel.com> > Cc: Michal Hocko <mhocko@kernel.org> > Cc: Minchan Kim <minchan@kernel.org> > Cc: Hugh Dickins <hughd@google.com> > Cc: Rik van Riel <riel@surriel.com> > --- > mm/rmap.c | 27 ++++++++++++++------------- > 1 file changed, 14 insertions(+), 13 deletions(-) > > diff --git a/mm/rmap.c b/mm/rmap.c > index 2126fd4a254b..cd3c406aeac7 100644 > --- a/mm/rmap.c > +++ b/mm/rmap.c > @@ -1613,19 +1613,6 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, > } else if (PageAnon(page)) { > swp_entry_t entry = { .val = page_private(subpage) }; > pte_t swp_pte; > - /* > - * Store the swap location in the pte. > - * See handle_pte_fault() ... > - */ > - if (unlikely(PageSwapBacked(page) != PageSwapCache(page))) { > - WARN_ON_ONCE(1); > - ret = false; > - /* We have to invalidate as we cleared the pte */ > - mmu_notifier_invalidate_range(mm, address, > - address + PAGE_SIZE); > - page_vma_mapped_walk_done(&pvmw); > - break; > - } > > /* MADV_FREE page check */ > if (!PageSwapBacked(page)) { > @@ -1648,6 +1635,20 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, > break; > } > > + /* > + * Store the swap location in the pte. > + * See handle_pte_fault() ... > + */ > + if (unlikely(!PageSwapCache(page))) { > + WARN_ON_ONCE(1); > + ret = false; > + /* We have to invalidate as we cleared the pte */ > + mmu_notifier_invalidate_range(mm, address, > + address + PAGE_SIZE); > + page_vma_mapped_walk_done(&pvmw); > + break; > + } > + > if (swap_duplicate(entry) < 0) { > set_pte_at(mm, address, pvmw.pte, pteval); > ret = false; > -- > 2.25.0
Michal Hocko <mhocko@kernel.org> writes: > On Tue 31-03-20 16:46:13, Huang, Ying wrote: >> From: Huang Ying <ying.huang@intel.com> >> >> Because PageSwapCache() will always return false if PageSwapBacked() returns >> false, and PageSwapBacked() will be check for MADV_FREE pages in >> try_to_unmap_one(). The swap related code in try_to_unmap_one() can be >> simplified to improve the readability. > > My understanding is that this is a sanity check to let us know if > something breaks. Do we really want to get rid of it? Maybe it is not > really useful but if that is the case then the changelog should reflect > this fact. Now the definition of PageSwapCache() is, static __always_inline int PageSwapCache(struct page *page) { #ifdef CONFIG_THP_SWAP page = compound_head(page); #endif return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags); } So, if PageSwapBacked() returns false, PageSwapCache() will always return false. The original checking, - if (unlikely(PageSwapBacked(page) != PageSwapCache(page))) { is equivalent to - if (unlikely(PageSwapBacked(page) && !PageSwapCache(page))) { Then what is the check !PageSwapBacked() && PageSwapCache() for? To prevent someone to change the definition of PageSwapCache() in the future to break this? Best Regards, Huang, Ying
On Wed 01-04-20 09:11:13, Huang, Ying wrote: [...] > Then what is the check !PageSwapBacked() && PageSwapCache() for? To > prevent someone to change the definition of PageSwapCache() in the > future to break this? Yes this is my understading. It is essentially an assert that enforces the assumption about swap cache vs. swap backed being coupled.
Michal Hocko <mhocko@kernel.org> writes: > On Wed 01-04-20 09:11:13, Huang, Ying wrote: > [...] >> Then what is the check !PageSwapBacked() && PageSwapCache() for? To >> prevent someone to change the definition of PageSwapCache() in the >> future to break this? > > Yes this is my understading. It is essentially an assert that enforces > the assumption about swap cache vs. swap backed being coupled. OK. Then we can just keep the current code. Best Regards, Huang, Ying
diff --git a/mm/rmap.c b/mm/rmap.c index 2126fd4a254b..cd3c406aeac7 100644 --- a/mm/rmap.c +++ b/mm/rmap.c @@ -1613,19 +1613,6 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, } else if (PageAnon(page)) { swp_entry_t entry = { .val = page_private(subpage) }; pte_t swp_pte; - /* - * Store the swap location in the pte. - * See handle_pte_fault() ... - */ - if (unlikely(PageSwapBacked(page) != PageSwapCache(page))) { - WARN_ON_ONCE(1); - ret = false; - /* We have to invalidate as we cleared the pte */ - mmu_notifier_invalidate_range(mm, address, - address + PAGE_SIZE); - page_vma_mapped_walk_done(&pvmw); - break; - } /* MADV_FREE page check */ if (!PageSwapBacked(page)) { @@ -1648,6 +1635,20 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, break; } + /* + * Store the swap location in the pte. + * See handle_pte_fault() ... + */ + if (unlikely(!PageSwapCache(page))) { + WARN_ON_ONCE(1); + ret = false; + /* We have to invalidate as we cleared the pte */ + mmu_notifier_invalidate_range(mm, address, + address + PAGE_SIZE); + page_vma_mapped_walk_done(&pvmw); + break; + } + if (swap_duplicate(entry) < 0) { set_pte_at(mm, address, pvmw.pte, pteval); ret = false;