| Message ID | 20220930141931.174362-7-david@redhat.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | mm/ksm: break_ksm() cleanups and fixes | expand |
On Fri, Sep 30, 2022 at 04:19:30PM +0200, David Hildenbrand wrote: > FOLL_MIGRATION exists only for the purpose of break_ksm(), and > actually, there is not even the need to wait for the migration to > finish, we only want to know if we're dealing with a KSM page. > > Using follow_page() just to identify a KSM page overcomplicates GUP > code. Let's use walk_page_range_vma() instead, because we don't actually > care about the page itself, we only need to know a single property -- > no need to even grab a reference on the page. > > In my setup (AMD Ryzen 9 3900X), running the KSM selftest to test unmerge > performance on 2 GiB (taskset 0x8 ./ksm_tests -D -s 2048), this results in > a performance degradation of ~4% (old: ~5010 MiB/s, new: ~4800 MiB/s). > I don't think we particularly care for now. > > Signed-off-by: David Hildenbrand <david@redhat.com> [...] > +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, > + struct mm_walk *walk) > +{ > + /* We only care about page tables to walk to a single base page. */ > + if (pud_leaf(*pud) || !pud_present(*pud)) > + return 1; > + return 0; > +} Is this needed? I thought the pgtable walker handlers this already. [...] > static int break_ksm(struct vm_area_struct *vma, unsigned long addr) > { > - struct page *page; > vm_fault_t ret = 0; > > + if (WARN_ON_ONCE(!IS_ALIGNED(addr, PAGE_SIZE))) > + return -EINVAL; > + > do { > bool ksm_page = false; > > cond_resched(); > - page = follow_page(vma, addr, > - FOLL_GET | FOLL_MIGRATION | FOLL_REMOTE); > - if (IS_ERR_OR_NULL(page)) > - break; > - if (PageKsm(page)) > - ksm_page = true; > - put_page(page); > + ret = walk_page_range_vma(vma, addr, addr + PAGE_SIZE, > + &break_ksm_ops, &ksm_page); > + if (WARN_ON_ONCE(ret < 0)) > + return ret; I'm not sure this would be worth it, especially with a 4% degrade. The next patch will be able to bring 50- LOC, but this patch does 60+ anyway, based on another new helper just introduced... I just don't see whether there's strong enough reason to do so to drop FOLL_MIGRATE. It's different to the previous VM_FAULT_WRITE refactor because of the unshare approach was much of a good reasoning to me. Perhaps I missed something?
>> +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, >> + struct mm_walk *walk) >> +{ >> + /* We only care about page tables to walk to a single base page. */ >> + if (pud_leaf(*pud) || !pud_present(*pud)) >> + return 1; >> + return 0; >> +} > > Is this needed? I thought the pgtable walker handlers this already. > > [...] > Most probably yes. I was trying to avoid about PUD splits, but I guess we simply should not care in VMAs that are considered by KSM (MERGABLE). Most probably never ever happens. >> static int break_ksm(struct vm_area_struct *vma, unsigned long addr) >> { >> - struct page *page; >> vm_fault_t ret = 0; >> >> + if (WARN_ON_ONCE(!IS_ALIGNED(addr, PAGE_SIZE))) >> + return -EINVAL; >> + >> do { >> bool ksm_page = false; >> >> cond_resched(); >> - page = follow_page(vma, addr, >> - FOLL_GET | FOLL_MIGRATION | FOLL_REMOTE); >> - if (IS_ERR_OR_NULL(page)) >> - break; >> - if (PageKsm(page)) >> - ksm_page = true; >> - put_page(page); >> + ret = walk_page_range_vma(vma, addr, addr + PAGE_SIZE, >> + &break_ksm_ops, &ksm_page); >> + if (WARN_ON_ONCE(ret < 0)) >> + return ret; > > I'm not sure this would be worth it, especially with a 4% degrade. The > next patch will be able to bring 50- LOC, but this patch does 60+ anyway, > based on another new helper just introduced... > > I just don't see whether there's strong enough reason to do so to drop > FOLL_MIGRATE. It's different to the previous VM_FAULT_WRITE refactor > because of the unshare approach was much of a good reasoning to me. > > Perhaps I missed something? My main motivation is to remove most of that GUP hackery here, which is 1) Getting a reference on a page and waiting for migration to finish even though both is unnecessary. 2) As we don't have sufficient control, we added FOLL_MIGRATION hacks to MM core to work around limitations in the GUP-based approacj. 3) We rely on legacy follow_page() interface that we should really get rid of in the long term. All we want to do is walk the page tables and make a decision if something we care about is mapped. Instead of leaking these details via hacks into GUP code and making that code harder to grasp/maintain, this patch moves that logic to the actual user, while reusing generic page walking code. Yes, we have to extend page walking code, but it's just the natural, non-hacky way of doing it. Regarding the 4% performance degradation (if I wouldn't have added the benchmarks, nobody would know and probably care ;) ), I am not quite sure why that is the case. We're just walking page tables after all in both cases. Maybe the callback-based implementation of pagewalk code is less efficient, but we might be able to improve that implementation if we really care about performance here. Maybe removing break_ksm_pud_entry() already improves the numbers slightly. Thanks!
On Thu, Oct 06, 2022 at 11:20:42AM +0200, David Hildenbrand wrote: > > > +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, > > > + struct mm_walk *walk) > > > +{ > > > + /* We only care about page tables to walk to a single base page. */ > > > + if (pud_leaf(*pud) || !pud_present(*pud)) > > > + return 1; > > > + return 0; > > > +} > > > > Is this needed? I thought the pgtable walker handlers this already. > > > > [...] > > > > Most probably yes. I was trying to avoid about PUD splits, but I guess we > simply should not care in VMAs that are considered by KSM (MERGABLE). Most > probably never ever happens. I was surprised the split is the default approach; didn't really notice that before. Yeah maybe better to keep it. > > > > static int break_ksm(struct vm_area_struct *vma, unsigned long addr) > > > { > > > - struct page *page; > > > vm_fault_t ret = 0; > > > + if (WARN_ON_ONCE(!IS_ALIGNED(addr, PAGE_SIZE))) > > > + return -EINVAL; > > > + > > > do { > > > bool ksm_page = false; > > > cond_resched(); > > > - page = follow_page(vma, addr, > > > - FOLL_GET | FOLL_MIGRATION | FOLL_REMOTE); > > > - if (IS_ERR_OR_NULL(page)) > > > - break; > > > - if (PageKsm(page)) > > > - ksm_page = true; > > > - put_page(page); > > > + ret = walk_page_range_vma(vma, addr, addr + PAGE_SIZE, > > > + &break_ksm_ops, &ksm_page); > > > + if (WARN_ON_ONCE(ret < 0)) > > > + return ret; > > > > I'm not sure this would be worth it, especially with a 4% degrade. The > > next patch will be able to bring 50- LOC, but this patch does 60+ anyway, > > based on another new helper just introduced... > > > > I just don't see whether there's strong enough reason to do so to drop > > FOLL_MIGRATE. It's different to the previous VM_FAULT_WRITE refactor > > because of the unshare approach was much of a good reasoning to me. > > > > Perhaps I missed something? > > My main motivation is to remove most of that GUP hackery here, which is > 1) Getting a reference on a page and waiting for migration to finish > even though both is unnecessary. > 2) As we don't have sufficient control, we added FOLL_MIGRATION hacks to > MM core to work around limitations in the GUP-based approacj. I saw one thing of adding FOLL_MIGRATION from Hugh was to have a hint for follow page users: I'd have preferred to avoid another flag, and do it every time, in case someone else makes the same easy mistake.. Though.. > 3) We rely on legacy follow_page() interface that we should really get > rid of in the long term. ..this is part of effort to remove follow_page()? More context will be helpful in that case. > > All we want to do is walk the page tables and make a decision if something > we care about is mapped. Instead of leaking these details via hacks into GUP > code and making that code harder to grasp/maintain, this patch moves that > logic to the actual user, while reusing generic page walking code. Indeed there's only one ksm user, at least proving that the flag was not widely used. > > Yes, we have to extend page walking code, but it's just the natural, > non-hacky way of doing it. > > Regarding the 4% performance degradation (if I wouldn't have added the > benchmarks, nobody would know and probably care ;) ), I am not quite sure > why that is the case. We're just walking page tables after all in both > cases. Maybe the callback-based implementation of pagewalk code is less > efficient, but we might be able to improve that implementation if we really > care about performance here. Maybe removing break_ksm_pud_entry() already > improves the numbers slightly. Yeah it could be the walker is just slower. And for !ksm walking your code should be faster when hit migration entries, but that should really be rare anyway.
On 30.09.22 16:19, David Hildenbrand wrote: > FOLL_MIGRATION exists only for the purpose of break_ksm(), and > actually, there is not even the need to wait for the migration to > finish, we only want to know if we're dealing with a KSM page. > > Using follow_page() just to identify a KSM page overcomplicates GUP > code. Let's use walk_page_range_vma() instead, because we don't actually > care about the page itself, we only need to know a single property -- > no need to even grab a reference on the page. > > In my setup (AMD Ryzen 9 3900X), running the KSM selftest to test unmerge > performance on 2 GiB (taskset 0x8 ./ksm_tests -D -s 2048), this results in > a performance degradation of ~4% (old: ~5010 MiB/s, new: ~4800 MiB/s). > I don't think we particularly care for now. > > Signed-off-by: David Hildenbrand <david@redhat.com> > --- > mm/ksm.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++------- > 1 file changed, 62 insertions(+), 8 deletions(-) > > diff --git a/mm/ksm.c b/mm/ksm.c > index 4d7bcf7da7c3..814c1a37c323 100644 > --- a/mm/ksm.c > +++ b/mm/ksm.c > @@ -39,6 +39,7 @@ > #include <linux/freezer.h> > #include <linux/oom.h> > #include <linux/numa.h> > +#include <linux/pagewalk.h> > > #include <asm/tlbflush.h> > #include "internal.h" > @@ -452,6 +453,60 @@ static inline bool ksm_test_exit(struct mm_struct *mm) > return atomic_read(&mm->mm_users) == 0; > } > > +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, > + struct mm_walk *walk) > +{ > + /* We only care about page tables to walk to a single base page. */ > + if (pud_leaf(*pud) || !pud_present(*pud)) > + return 1; > + return 0; > +} > + > +int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next, > + struct mm_walk *walk) > +{ > + bool *ksm_page = walk->private; > + struct page *page = NULL; > + pte_t *pte, ptent; > + spinlock_t *ptl; > + > + /* We only care about page tables to walk to a single base page. */ > + if (pmd_leaf(*pmd) || !pmd_present(*pmd)) > + return 1; > + > + /* > + * We only lookup a single page (a) no need to iterate; and (b) > + * always return 1 to exit immediately and not iterate in the caller. > + */ > + pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); > + ptent = *pte; > + > + if (pte_none(ptent)) > + return 1; As reported by Janosch, we fail to drop the lock here. t480s: ~/git/linux ksm_unshare $ git diff diff --git a/mm/ksm.c b/mm/ksm.c index 26aec41b127c..94f8e114c89f 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -435,7 +435,7 @@ int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next, pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); if (pte_none(*pte)) - return 1; + goto out_unlock; if (!pte_present(*pte)) { swp_entry_t entry = pte_to_swp_entry(*pte); @@ -451,6 +451,7 @@ int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next, } if (page && PageKsm(page)) *ksm_page = true; +out_unlock: pte_unmap_unlock(pte, ptl); return 1; }
On 06.10.22 21:28, Peter Xu wrote: > On Thu, Oct 06, 2022 at 11:20:42AM +0200, David Hildenbrand wrote: >>>> +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, >>>> + struct mm_walk *walk) >>>> +{ >>>> + /* We only care about page tables to walk to a single base page. */ >>>> + if (pud_leaf(*pud) || !pud_present(*pud)) >>>> + return 1; >>>> + return 0; >>>> +} >>> >>> Is this needed? I thought the pgtable walker handlers this already. >>> >>> [...] >>> >> >> Most probably yes. I was trying to avoid about PUD splits, but I guess we >> simply should not care in VMAs that are considered by KSM (MERGABLE). Most >> probably never ever happens. > > I was surprised the split is the default approach; didn't really notice > that before. Yeah maybe better to keep it. Interestingly, one callback reduces the benchmark result by 100-200 MiB. With only break_ksm_pmd_entry(), I get ~4900 MiB/s instead of ~5010 MiB/s (~2%). I came to the conclusion that we really shouldn't have to worry about pud THPs here: it could only be a file PUD and splitting that only zaps the entry, but doesn't PMD- or PTE-map it. Also, I think we will barely see large pud THP in a mergable mapping ... :) [...] >> My main motivation is to remove most of that GUP hackery here, which is >> 1) Getting a reference on a page and waiting for migration to finish >> even though both is unnecessary. >> 2) As we don't have sufficient control, we added FOLL_MIGRATION hacks to >> MM core to work around limitations in the GUP-based approacj. > > I saw one thing of adding FOLL_MIGRATION from Hugh was to have a hint for > follow page users: > > I'd have preferred to avoid another flag, and do it every time, in case > someone else makes the same easy mistake.. > > Though.. The important thing I think is that FOLL_MIGRATION really only applies to follow_page(). In case of "modern" GUP we will just wait for migration entries, handle swap entries ... when triggering a page fault. > >> 3) We rely on legacy follow_page() interface that we should really get >> rid of in the long term. > > ..this is part of effort to remove follow_page()? More context will be > helpful in that case. The comment from Hugh is another example why follow_page() adds complexity. One might wonder, how pages in the swapcache, device coherent pages, ... would have to be handled. Short-term, I want to cleanup GUP. Long-term we might want to consider removing follow_page() completely. [...] >> >> Yes, we have to extend page walking code, but it's just the natural, >> non-hacky way of doing it. >> >> Regarding the 4% performance degradation (if I wouldn't have added the >> benchmarks, nobody would know and probably care ;) ), I am not quite sure >> why that is the case. We're just walking page tables after all in both >> cases. Maybe the callback-based implementation of pagewalk code is less >> efficient, but we might be able to improve that implementation if we really >> care about performance here. Maybe removing break_ksm_pud_entry() already >> improves the numbers slightly. > > Yeah it could be the walker is just slower. And for !ksm walking your code > should be faster when hit migration entries, but that should really be rare > anyway. I have the following right now: From 7f767f9e9e673a29793cd35f1c3d66ed593b67cb Mon Sep 17 00:00:00 2001 From: David Hildenbrand <david@redhat.com> Date: Mon, 25 Jul 2022 10:36:20 +0200 Subject: [PATCH] mm/ksm: convert break_ksm() to use walk_page_range_vma() FOLL_MIGRATION exists only for the purpose of break_ksm(), and actually, there is not even the need to wait for the migration to finish, we only want to know if we're dealing with a KSM page. Using follow_page() just to identify a KSM page overcomplicates GUP code. Let's use walk_page_range_vma() instead, because we don't actually care about the page itself, we only need to know a single property -- no need to even grab a reference. So, get rid of follow_page() usage such that we can get rid of FOLL_MIGRATION now and eventually be able to get rid of follow_page() in the future. In my setup (AMD Ryzen 9 3900X), running the KSM selftest to test unmerge performance on 2 GiB (taskset 0x8 ./ksm_tests -D -s 2048), this results in a performance degradation of ~2% (old: ~5010 MiB/s, new: ~4900 MiB/s). I don't think we particularly care for now. Interestingly, the benchmark reduction is due to the single callback. Adding a second callback (e.g., pud_entry()) reduces the benchmark by another 100-200 MiB/s. Signed-off-by: David Hildenbrand <david@redhat.com> --- mm/ksm.c | 49 +++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 39 insertions(+), 10 deletions(-) diff --git a/mm/ksm.c b/mm/ksm.c index c6f58aa6e731..5cdb852ff132 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -39,6 +39,7 @@ #include <linux/freezer.h> #include <linux/oom.h> #include <linux/numa.h> +#include <linux/pagewalk.h> #include <asm/tlbflush.h> #include "internal.h" @@ -419,6 +420,39 @@ static inline bool ksm_test_exit(struct mm_struct *mm) return atomic_read(&mm->mm_users) == 0; } +static int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next, + struct mm_walk *walk) +{ + struct page *page = NULL; + spinlock_t *ptl; + pte_t *pte; + int ret; + + if (pmd_leaf(*pmd) || !pmd_present(*pmd)) + return 0; + + pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); + if (pte_present(*pte)) { + page = vm_normal_page(walk->vma, addr, *pte); + } else if (!pte_none(*pte)) { + swp_entry_t entry = pte_to_swp_entry(*pte); + + /* + * As KSM pages remain KSM pages until freed, no need to wait + * here for migration to end. + */ + if (is_migration_entry(entry)) + page = pfn_swap_entry_to_page(entry); + } + ret = page && PageKsm(page); + pte_unmap_unlock(pte, ptl); + return ret; +} + +static const struct mm_walk_ops break_ksm_ops = { + .pmd_entry = break_ksm_pmd_entry, +}; + /* * We use break_ksm to break COW on a ksm page by triggering unsharing, * such that the ksm page will get replaced by an exclusive anonymous page. @@ -434,21 +468,16 @@ static inline bool ksm_test_exit(struct mm_struct *mm) */ static int break_ksm(struct vm_area_struct *vma, unsigned long addr) { - struct page *page; vm_fault_t ret = 0; do { - bool ksm_page = false; + int ksm_page; cond_resched(); - page = follow_page(vma, addr, - FOLL_GET | FOLL_MIGRATION | FOLL_REMOTE); - if (IS_ERR_OR_NULL(page)) - break; - if (PageKsm(page)) - ksm_page = true; - put_page(page); - + ksm_page = walk_page_range_vma(vma, addr, addr + 1, + &break_ksm_ops, NULL); + if (WARN_ON_ONCE(ksm_page < 0)) + return ksm_page; if (!ksm_page) return 0; ret = handle_mm_fault(vma, addr,
diff --git a/mm/ksm.c b/mm/ksm.c index 4d7bcf7da7c3..814c1a37c323 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -39,6 +39,7 @@ #include <linux/freezer.h> #include <linux/oom.h> #include <linux/numa.h> +#include <linux/pagewalk.h> #include <asm/tlbflush.h> #include "internal.h" @@ -452,6 +453,60 @@ static inline bool ksm_test_exit(struct mm_struct *mm) return atomic_read(&mm->mm_users) == 0; } +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next, + struct mm_walk *walk) +{ + /* We only care about page tables to walk to a single base page. */ + if (pud_leaf(*pud) || !pud_present(*pud)) + return 1; + return 0; +} + +int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next, + struct mm_walk *walk) +{ + bool *ksm_page = walk->private; + struct page *page = NULL; + pte_t *pte, ptent; + spinlock_t *ptl; + + /* We only care about page tables to walk to a single base page. */ + if (pmd_leaf(*pmd) || !pmd_present(*pmd)) + return 1; + + /* + * We only lookup a single page (a) no need to iterate; and (b) + * always return 1 to exit immediately and not iterate in the caller. + */ + pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); + ptent = *pte; + + if (pte_none(ptent)) + return 1; + if (!pte_present(ptent)) { + swp_entry_t entry = pte_to_swp_entry(ptent); + + /* + * We only care about migration of KSM pages. As KSM pages + * remain KSM pages until freed, no need to wait here for + * migration to end to identify such. + */ + if (is_migration_entry(entry)) + page = pfn_swap_entry_to_page(entry); + } else { + page = vm_normal_page(walk->vma, addr, ptent); + } + if (page && PageKsm(page)) + *ksm_page = true; + pte_unmap_unlock(pte, ptl); + return 1; +} + +static const struct mm_walk_ops break_ksm_ops = { + .pud_entry = break_ksm_pud_entry, + .pmd_entry = break_ksm_pmd_entry, +}; + /* * We use break_ksm to break COW on a ksm page by triggering unsharing, * such that the ksm page will get replaced by an exclusive anonymous page. @@ -467,20 +522,19 @@ static inline bool ksm_test_exit(struct mm_struct *mm) */ static int break_ksm(struct vm_area_struct *vma, unsigned long addr) { - struct page *page; vm_fault_t ret = 0; + if (WARN_ON_ONCE(!IS_ALIGNED(addr, PAGE_SIZE))) + return -EINVAL; + do { bool ksm_page = false; cond_resched(); - page = follow_page(vma, addr, - FOLL_GET | FOLL_MIGRATION | FOLL_REMOTE); - if (IS_ERR_OR_NULL(page)) - break; - if (PageKsm(page)) - ksm_page = true; - put_page(page); + ret = walk_page_range_vma(vma, addr, addr + PAGE_SIZE, + &break_ksm_ops, &ksm_page); + if (WARN_ON_ONCE(ret < 0)) + return ret; if (!ksm_page) return 0;
FOLL_MIGRATION exists only for the purpose of break_ksm(), and actually, there is not even the need to wait for the migration to finish, we only want to know if we're dealing with a KSM page. Using follow_page() just to identify a KSM page overcomplicates GUP code. Let's use walk_page_range_vma() instead, because we don't actually care about the page itself, we only need to know a single property -- no need to even grab a reference on the page. In my setup (AMD Ryzen 9 3900X), running the KSM selftest to test unmerge performance on 2 GiB (taskset 0x8 ./ksm_tests -D -s 2048), this results in a performance degradation of ~4% (old: ~5010 MiB/s, new: ~4800 MiB/s). I don't think we particularly care for now. Signed-off-by: David Hildenbrand <david@redhat.com> --- mm/ksm.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 62 insertions(+), 8 deletions(-)