Message ID | 20210601033319.100737-1-gshan@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [RFC] mm/page_reporting: Adjust threshold according to MAX_ORDER | expand |
On 01.06.21 05:33, Gavin Shan wrote: > The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as > minimal order (threshold) to trigger page reporting. The page reporting > is never triggered with the following configurations and settings on > aarch64. In the particular scenario, the page reporting won't be triggered > until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the > page freeing. The condition is very hard, or even impossible to be met. > > CONFIG_ARM64_PAGE_SHIFT: 16 > CONFIG_HUGETLB_PAGE: Y > CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N > pageblock_order: 13 > CONFIG_FORCE_MAX_ZONEORDER: 14 > MAX_ORDER: 14 > > The issue can be reproduced in VM, running kernel with above configurations > and settings. The 'memhog' is used inside the VM to access 512MB anonymous > area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. > > /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ > -accel kvm -machine virt,gic-version=host \ > -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ > -object memory-backend-ram,id=mem0,size=2048M \ > -object memory-backend-ram,id=mem1,size=2048M \ > -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ > -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ > : \ > -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes > > This tries to fix the issue by adjusting the threshold to the smaller value > of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS > drops after 'memhog' exits. IIRC, we use pageblock_order to a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that.
On Tue, 1 Jun 2021 11:33:19 +0800 Gavin Shan <gshan@redhat.com> wrote: > The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as > minimal order (threshold) to trigger page reporting. The page reporting > is never triggered with the following configurations and settings on > aarch64. In the particular scenario, the page reporting won't be triggered > until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the > page freeing. The condition is very hard, or even impossible to be met. > > CONFIG_ARM64_PAGE_SHIFT: 16 > CONFIG_HUGETLB_PAGE: Y > CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N > pageblock_order: 13 > CONFIG_FORCE_MAX_ZONEORDER: 14 > MAX_ORDER: 14 > > The issue can be reproduced in VM, running kernel with above configurations > and settings. The 'memhog' is used inside the VM to access 512MB anonymous > area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. > > /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ > -accel kvm -machine virt,gic-version=host \ > -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ > -object memory-backend-ram,id=mem0,size=2048M \ > -object memory-backend-ram,id=mem1,size=2048M \ > -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ > -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ > : \ > -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes > > This tries to fix the issue by adjusting the threshold to the smaller value > of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS > drops after 'memhog' exits. > Sounds promising, but... > --- a/mm/page_reporting.h > +++ b/mm/page_reporting.h > @@ -10,9 +10,10 @@ > #include <linux/pgtable.h> > #include <linux/scatterlist.h> > > -#define PAGE_REPORTING_MIN_ORDER pageblock_order > - > #ifdef CONFIG_PAGE_REPORTING > +#define PAGE_REPORTING_MIN_ORDER \ > + min_t(unsigned int, pageblock_order, (MAX_ORDER / 2)) > + > DECLARE_STATIC_KEY_FALSE(page_reporting_enabled); > void __page_reporting_notify(void); Could you please redo this as a regular old static function in page_reporting.c? Bonus points for commenting its design ;)
On 11.06.21 09:44, Gavin Shan wrote: > On 6/1/21 6:01 PM, David Hildenbrand wrote: >> On 01.06.21 05:33, Gavin Shan wrote: >>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>> minimal order (threshold) to trigger page reporting. The page reporting >>> is never triggered with the following configurations and settings on >>> aarch64. In the particular scenario, the page reporting won't be triggered >>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>> page freeing. The condition is very hard, or even impossible to be met. >>> >>> CONFIG_ARM64_PAGE_SHIFT: 16 >>> CONFIG_HUGETLB_PAGE: Y >>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>> pageblock_order: 13 >>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>> MAX_ORDER: 14 >>> >>> The issue can be reproduced in VM, running kernel with above configurations >>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>> >>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>> -accel kvm -machine virt,gic-version=host \ >>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>> -object memory-backend-ram,id=mem0,size=2048M \ >>> -object memory-backend-ram,id=mem1,size=2048M \ >>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>> : \ >>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>> >>> This tries to fix the issue by adjusting the threshold to the smaller value >>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>> drops after 'memhog' exits. >> >> IIRC, we use pageblock_order to >> >> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >> >> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >> >> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >> >> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >> >> >> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >> >> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >> > > [Remove Alexander from the cc list as his mail isn't reachable] > [adding his gmail address which should be the right one] > David, thanks for your time to review and sorry for the delay and late response. > I spent some time to get myself familiar with the code, but there are still some > questions to me, explained as below. > > Yes, @pageblock_order is currently taken as page reporting threshold. It will > incur more overhead if the threshold is decreased as you said in (a). Right. Alex did quite some performance/overhead evaluation when introducing this feature. Changing the reporting granularity on most setups (esp., x86-64) is not desired IMHO. > > This patch tries to decrease the free page reporting threshold. The @pageblock_order > isn't touched. I don't understand how the code changes affecting THP splitting > and the creation of page blocks mentioned in (b) and (c). David, could you please > provide more details? Think of it like this: while reporting to the hypervisor, we temporarily turn free/"movable" pieces part of a pageblock "unmovable" -- see __isolate_free_page()->del_page_from_free_list(). While reporting them to the hypervisor, these pages are not available and not even marked as PageBuddy() anymore. There are at least two scenarios where this could affect creation of free pageblocks I can see: a. Compaction. While compacting, we might identify completely movable/free pageblocks, however, actual compaction on that pageblock can fail because some part is temporarily unmovable. b. Free/alloc sequences. Assume a pageblocks is mostly free, except two pages (x and y). Assume the following sequence: 1. free(x) 2. free(y) 3. alloc Before your change, after 1. and 2. we'll have a free pageblock. 3 won't allocate from that pageblock. With your change, free page reporting might run after 1. After 2, we'll not have a free pageblock (until free page reporting finished), and 3. might just reallocate what we freed in 2 and prevent having a free pageblock. No idea how relevant both points are in practice, however, the fundamental difference to current handling is that we would turn parts of pageblocks temporarily unmovable, instead of complete pageblocks.
On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: > > On 11.06.21 09:44, Gavin Shan wrote: > > On 6/1/21 6:01 PM, David Hildenbrand wrote: > >> On 01.06.21 05:33, Gavin Shan wrote: > >>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as > >>> minimal order (threshold) to trigger page reporting. The page reporting > >>> is never triggered with the following configurations and settings on > >>> aarch64. In the particular scenario, the page reporting won't be triggered > >>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the > >>> page freeing. The condition is very hard, or even impossible to be met. > >>> > >>> CONFIG_ARM64_PAGE_SHIFT: 16 > >>> CONFIG_HUGETLB_PAGE: Y > >>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N > >>> pageblock_order: 13 > >>> CONFIG_FORCE_MAX_ZONEORDER: 14 > >>> MAX_ORDER: 14 > >>> > >>> The issue can be reproduced in VM, running kernel with above configurations > >>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous > >>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. > >>> > >>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ > >>> -accel kvm -machine virt,gic-version=host \ > >>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ > >>> -object memory-backend-ram,id=mem0,size=2048M \ > >>> -object memory-backend-ram,id=mem1,size=2048M \ > >>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ > >>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ > >>> : \ > >>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes > >>> > >>> This tries to fix the issue by adjusting the threshold to the smaller value > >>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS > >>> drops after 'memhog' exits. > >> > >> IIRC, we use pageblock_order to > >> > >> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. > >> > >> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. > >> > >> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. > >> > >> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. > >> > >> > >> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. > >> > >> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. > >> > > > > [Remove Alexander from the cc list as his mail isn't reachable] > > > > [adding his gmail address which should be the right one] > > > David, thanks for your time to review and sorry for the delay and late response. > > I spent some time to get myself familiar with the code, but there are still some > > questions to me, explained as below. > > > > Yes, @pageblock_order is currently taken as page reporting threshold. It will > > incur more overhead if the threshold is decreased as you said in (a). > > Right. Alex did quite some performance/overhead evaluation when > introducing this feature. Changing the reporting granularity on most > setups (esp., x86-64) is not desired IMHO. Yes, generally reporting pages comes at a fairly high cost so it is important to find the right trade-off between the size of the page and the size of the batch of pages being reported. If the size of the pages is reduced it maybe important to increase the batch size in order to avoid paying too much in the way of overhead. The other main reason for holding to pageblock_order on x86 is to avoid THP splitting. Anything smaller than pageblock_order will trigger THP splitting which will significantly hurt the performance of the VM in general as it forces it down to order 0 pages.
On 6/14/21 9:03 PM, David Hildenbrand wrote: > On 11.06.21 09:44, Gavin Shan wrote: >> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>> On 01.06.21 05:33, Gavin Shan wrote: >>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>> minimal order (threshold) to trigger page reporting. The page reporting >>>> is never triggered with the following configurations and settings on >>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>> page freeing. The condition is very hard, or even impossible to be met. >>>> >>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>> CONFIG_HUGETLB_PAGE: Y >>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>> pageblock_order: 13 >>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>> MAX_ORDER: 14 >>>> >>>> The issue can be reproduced in VM, running kernel with above configurations >>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>> >>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>> -accel kvm -machine virt,gic-version=host \ >>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>> : \ >>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>> >>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>> drops after 'memhog' exits. >>> >>> IIRC, we use pageblock_order to >>> >>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>> >>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>> >>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>> >>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>> >>> >>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>> >>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>> >> >> [Remove Alexander from the cc list as his mail isn't reachable] >> > > [adding his gmail address which should be the right one] > >> David, thanks for your time to review and sorry for the delay and late response. >> I spent some time to get myself familiar with the code, but there are still some >> questions to me, explained as below. >> >> Yes, @pageblock_order is currently taken as page reporting threshold. It will >> incur more overhead if the threshold is decreased as you said in (a). > > Right. Alex did quite some performance/overhead evaluation when introducing this feature. Changing the reporting granularity on most setups (esp., x86-64) is not desired IMHO. > Thanks for adding Alex's correct mail address, David. >> >> This patch tries to decrease the free page reporting threshold. The @pageblock_order >> isn't touched. I don't understand how the code changes affecting THP splitting >> and the creation of page blocks mentioned in (b) and (c). David, could you please >> provide more details? > > Think of it like this: while reporting to the hypervisor, we temporarily turn free/"movable" pieces part of a pageblock "unmovable" -- see __isolate_free_page()->del_page_from_free_list(). While reporting them to the hypervisor, these pages are not available and not even marked as PageBuddy() anymore. > > There are at least two scenarios where this could affect creation of free pageblocks I can see: > > a. Compaction. While compacting, we might identify completely movable/free pageblocks, however, actual compaction on that pageblock can fail because some part is temporarily unmovable. > > b. Free/alloc sequences. Assume a pageblocks is mostly free, except two pages (x and y). Assume the following sequence: > > 1. free(x) > 2. free(y) > 3. alloc > > Before your change, after 1. and 2. we'll have a free pageblock. 3 won't allocate from that pageblock. > > With your change, free page reporting might run after 1. After 2, we'll not have a free pageblock (until free page reporting finished), and 3. might just reallocate what we freed in 2 and prevent having a free pageblock. > > > No idea how relevant both points are in practice, however, the fundamental difference to current handling is that we would turn parts of pageblocks temporarily unmovable, instead of complete pageblocks. > Thank you for the details. Without my changes and the page reporting threshold is @pageblock_order, the whole page block can become 'movable' from 'unmovable'. I don't think it's what we want, but I need Alex's confirm. If we needn't change page block's migration type in page reporting, I guess I need additional parameter for __isolate_free_page() so that the migration type won't be changed by page reporting. For (a), the 'movable' and 'unmovable' type is maintained with page block granularity. So it seems the pages in one page block can't have different migration types, or I missed something. For (b), the scenario is possible to happen. It means the changed page reporting threshold could affect the page allocator's behaviour, which could introduce more fragmentations. However, it's really depending on how the memory is allocated. Thanks, Gavin
On 16.06.21 03:53, Gavin Shan wrote: > On 6/14/21 9:03 PM, David Hildenbrand wrote: >> On 11.06.21 09:44, Gavin Shan wrote: >>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>> is never triggered with the following configurations and settings on >>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>> >>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>> CONFIG_HUGETLB_PAGE: Y >>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>> pageblock_order: 13 >>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>> MAX_ORDER: 14 >>>>> >>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>> >>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>> -accel kvm -machine virt,gic-version=host \ >>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>> : \ >>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>> >>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>> drops after 'memhog' exits. >>>> >>>> IIRC, we use pageblock_order to >>>> >>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>> >>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>> >>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>> >>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>> >>>> >>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>> >>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>> >>> >>> [Remove Alexander from the cc list as his mail isn't reachable] >>> >> >> [adding his gmail address which should be the right one] >> >>> David, thanks for your time to review and sorry for the delay and late response. >>> I spent some time to get myself familiar with the code, but there are still some >>> questions to me, explained as below. >>> >>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>> incur more overhead if the threshold is decreased as you said in (a). >> >> Right. Alex did quite some performance/overhead evaluation when introducing this feature. Changing the reporting granularity on most setups (esp., x86-64) is not desired IMHO. >> > > Thanks for adding Alex's correct mail address, David. > >>> >>> This patch tries to decrease the free page reporting threshold. The @pageblock_order >>> isn't touched. I don't understand how the code changes affecting THP splitting >>> and the creation of page blocks mentioned in (b) and (c). David, could you please >>> provide more details? >> >> Think of it like this: while reporting to the hypervisor, we temporarily turn free/"movable" pieces part of a pageblock "unmovable" -- see __isolate_free_page()->del_page_from_free_list(). While reporting them to the hypervisor, these pages are not available and not even marked as PageBuddy() anymore. >> >> There are at least two scenarios where this could affect creation of free pageblocks I can see: >> >> a. Compaction. While compacting, we might identify completely movable/free pageblocks, however, actual compaction on that pageblock can fail because some part is temporarily unmovable. >> >> b. Free/alloc sequences. Assume a pageblocks is mostly free, except two pages (x and y). Assume the following sequence: >> >> 1. free(x) >> 2. free(y) >> 3. alloc >> >> Before your change, after 1. and 2. we'll have a free pageblock. 3 won't allocate from that pageblock. >> >> With your change, free page reporting might run after 1. After 2, we'll not have a free pageblock (until free page reporting finished), and 3. might just reallocate what we freed in 2 and prevent having a free pageblock. >> >> >> No idea how relevant both points are in practice, however, the fundamental difference to current handling is that we would turn parts of pageblocks temporarily unmovable, instead of complete pageblocks. >> > > Thank you for the details. Without my changes and the page reporting threshold > is @pageblock_order, the whole page block can become 'movable' from 'unmovable'. > I don't think it's what we want, but I need Alex's confirm. __isolate_free_page() will set the pageblock MIGRATE_MOVABLE in that case. It's only temporarily unmovable, while we're hinting. Note that MOVABLE vs. UNMOVABLE is just grouping for free pages, and even setting it to the wrong migratetype isn't "wrong" as in "correctness". It doesn't make a difference if there are no free pages because the whole block is isolated.
On 16.06.21 11:10, Gavin Shan wrote: > On 6/15/21 12:26 PM, Alexander Duyck wrote: >> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>> On 11.06.21 09:44, Gavin Shan wrote: >>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>> is never triggered with the following configurations and settings on >>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>> >>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>> pageblock_order: 13 >>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>> MAX_ORDER: 14 >>>>>> >>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>> >>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>> : \ >>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>> >>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>> drops after 'memhog' exits. >>>>> >>>>> IIRC, we use pageblock_order to >>>>> >>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>> >>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>> >>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>> >>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>> >>>>> >>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>> >>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>> >>>> >>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>> >>> >>> [adding his gmail address which should be the right one] >>> >>>> David, thanks for your time to review and sorry for the delay and late response. >>>> I spent some time to get myself familiar with the code, but there are still some >>>> questions to me, explained as below. >>>> >>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>> incur more overhead if the threshold is decreased as you said in (a). >>> >>> Right. Alex did quite some performance/overhead evaluation when >>> introducing this feature. Changing the reporting granularity on most >>> setups (esp., x86-64) is not desired IMHO. >> >> Yes, generally reporting pages comes at a fairly high cost so it is >> important to find the right trade-off between the size of the page and >> the size of the batch of pages being reported. If the size of the >> pages is reduced it maybe important to increase the batch size in >> order to avoid paying too much in the way of overhead. >> >> The other main reason for holding to pageblock_order on x86 is to >> avoid THP splitting. Anything smaller than pageblock_order will >> trigger THP splitting which will significantly hurt the performance of >> the VM in general as it forces it down to order 0 pages. >> > > Alex, Thanks for your reply and sorry for taking your time to this > discussion. > > Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget > used in page_reporting_cycle() when you're talking about "batch"? > > I don't understand how the THP splitting is triggered. As I understood, > the free page reporting works like this: the free pages are reported > back to QEMU through virtio-balloon, madvise(DONTNEED) is called on the > free pages. For THP related VMA, its memory is guranteed to be physically > 2MB contiguous. The 2MB memory is free'd at once as I understand. I > don't think the 2MB memory can be free'd partially apart from the THP > splitting caused by some reasons. One of the reason is memory reclaim > and swapping. However, I think the THP splitting is caused by memory > reclaim is irrelevant to what you were saying. Anonymous memory: Assume you have a 2 MB THP in the hypervisor. If you madvise(DONTNEED, 4K), you'll split the THP and free the single 4k page back to the buddy. That virtual memory region is no longer backed by a physically contiguous 2 MB page and there is no huge mapping in the page tables. Instead, there is now an "ordinary" 4k mapping with a hole. Bad for performance.
On 6/15/21 12:26 PM, Alexander Duyck wrote: > On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >> On 11.06.21 09:44, Gavin Shan wrote: >>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>> is never triggered with the following configurations and settings on >>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>> >>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>> CONFIG_HUGETLB_PAGE: Y >>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>> pageblock_order: 13 >>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>> MAX_ORDER: 14 >>>>> >>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>> >>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>> -accel kvm -machine virt,gic-version=host \ >>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>> : \ >>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>> >>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>> drops after 'memhog' exits. >>>> >>>> IIRC, we use pageblock_order to >>>> >>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>> >>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>> >>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>> >>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>> >>>> >>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>> >>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>> >>> >>> [Remove Alexander from the cc list as his mail isn't reachable] >>> >> >> [adding his gmail address which should be the right one] >> >>> David, thanks for your time to review and sorry for the delay and late response. >>> I spent some time to get myself familiar with the code, but there are still some >>> questions to me, explained as below. >>> >>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>> incur more overhead if the threshold is decreased as you said in (a). >> >> Right. Alex did quite some performance/overhead evaluation when >> introducing this feature. Changing the reporting granularity on most >> setups (esp., x86-64) is not desired IMHO. > > Yes, generally reporting pages comes at a fairly high cost so it is > important to find the right trade-off between the size of the page and > the size of the batch of pages being reported. If the size of the > pages is reduced it maybe important to increase the batch size in > order to avoid paying too much in the way of overhead. > > The other main reason for holding to pageblock_order on x86 is to > avoid THP splitting. Anything smaller than pageblock_order will > trigger THP splitting which will significantly hurt the performance of > the VM in general as it forces it down to order 0 pages. > Alex, Thanks for your reply and sorry for taking your time to this discussion. Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget used in page_reporting_cycle() when you're talking about "batch"? I don't understand how the THP splitting is triggered. As I understood, the free page reporting works like this: the free pages are reported back to QEMU through virtio-balloon, madvise(DONTNEED) is called on the free pages. For THP related VMA, its memory is guranteed to be physically 2MB contiguous. The 2MB memory is free'd at once as I understand. I don't think the 2MB memory can be free'd partially apart from the THP splitting caused by some reasons. One of the reason is memory reclaim and swapping. However, I think the THP splitting is caused by memory reclaim is irrelevant to what you were saying. Thanks, Gavin
On 16.06.21 14:59, Gavin Shan wrote: > On 6/16/21 5:59 PM, David Hildenbrand wrote: >> On 16.06.21 03:53, Gavin Shan wrote: >>> On 6/14/21 9:03 PM, David Hildenbrand wrote: >>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>>> is never triggered with the following configurations and settings on >>>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>>> >>>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>>> pageblock_order: 13 >>>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>>> MAX_ORDER: 14 >>>>>>> >>>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>>> >>>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>>> : \ >>>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>>> >>>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>>> drops after 'memhog' exits. >>>>>> >>>>>> IIRC, we use pageblock_order to >>>>>> >>>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>>> >>>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>>> >>>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>>> >>>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>>> >>>>>> >>>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>>> >>>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>>> >>>>> >>>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>>> >>>> >>>> [adding his gmail address which should be the right one] >>>> >>>>> David, thanks for your time to review and sorry for the delay and late response. >>>>> I spent some time to get myself familiar with the code, but there are still some >>>>> questions to me, explained as below. >>>>> >>>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>>> incur more overhead if the threshold is decreased as you said in (a). >>>> >>>> Right. Alex did quite some performance/overhead evaluation when introducing this feature. Changing the reporting granularity on most setups (esp., x86-64) is not desired IMHO. >>>> >>> >>> Thanks for adding Alex's correct mail address, David. >>> >>>>> >>>>> This patch tries to decrease the free page reporting threshold. The @pageblock_order >>>>> isn't touched. I don't understand how the code changes affecting THP splitting >>>>> and the creation of page blocks mentioned in (b) and (c). David, could you please >>>>> provide more details? >>>> >>>> Think of it like this: while reporting to the hypervisor, we temporarily turn free/"movable" pieces part of a pageblock "unmovable" -- see __isolate_free_page()->del_page_from_free_list(). While reporting them to the hypervisor, these pages are not available and not even marked as PageBuddy() anymore. >>>> >>>> There are at least two scenarios where this could affect creation of free pageblocks I can see: >>>> >>>> a. Compaction. While compacting, we might identify completely movable/free pageblocks, however, actual compaction on that pageblock can fail because some part is temporarily unmovable. >>>> >>>> b. Free/alloc sequences. Assume a pageblocks is mostly free, except two pages (x and y). Assume the following sequence: >>>> >>>> 1. free(x) >>>> 2. free(y) >>>> 3. alloc >>>> >>>> Before your change, after 1. and 2. we'll have a free pageblock. 3 won't allocate from that pageblock. >>>> >>>> With your change, free page reporting might run after 1. After 2, we'll not have a free pageblock (until free page reporting finished), and 3. might just reallocate what we freed in 2 and prevent having a free pageblock. >>>> >>>> >>>> No idea how relevant both points are in practice, however, the fundamental difference to current handling is that we would turn parts of pageblocks temporarily unmovable, instead of complete pageblocks. >>>> >>> >>> Thank you for the details. Without my changes and the page reporting threshold >>> is @pageblock_order, the whole page block can become 'movable' from 'unmovable'. >>> I don't think it's what we want, but I need Alex's confirm. >> >> __isolate_free_page() will set the pageblock MIGRATE_MOVABLE in that case. It's only temporarily unmovable, while we're hinting. >> >> Note that MOVABLE vs. UNMOVABLE is just grouping for free pages, and even setting it to the wrong migratetype isn't "wrong" as in "correctness". It doesn't make a difference if there are no free pages because the whole block is isolated. >> > > Yes, It doesn't matter since these pages have been isolated. The migration type is changed to MIGRATE_MOVABLE > in __isolated_free_page(). My questions are actually: > > (1) Is it possible the migration type is changed from MIGRATE_UNMOVABLE to MIGRATE_MOVABLE > in __isolated_free_page()? Yes, if the isolated page covers at least half the pageblock. So either if we isolate the complete pageblock (as it's free, there is nothing unmovable) or half the pageblock. The latter seems to be some heuristic that says if it's half-free, make it MIGRATE_MOVABLE -- maybe because that increases the chances that we might get a completely movable pageblock later (would have too look into the details). > (2) After the free page reporting is completed, the migrate type is restored to MIGRATE_UNMOVABLE? No, don't think so. And it also doesn't make too much sense if we decided when isolating that we're better off using MIGRATE_MOVABLE. After all, we're just putting back a free page we previously isolated from the free lists.
On 16.06.21 15:16, Gavin Shan wrote: > On 6/16/21 6:03 PM, David Hildenbrand wrote: >> On 16.06.21 11:10, Gavin Shan wrote: >>> On 6/15/21 12:26 PM, Alexander Duyck wrote: >>>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>>>> is never triggered with the following configurations and settings on >>>>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>>>> >>>>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>>>> pageblock_order: 13 >>>>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>>>> MAX_ORDER: 14 >>>>>>>> >>>>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>>>> >>>>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>>>> : \ >>>>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>>>> >>>>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>>>> drops after 'memhog' exits. >>>>>>> >>>>>>> IIRC, we use pageblock_order to >>>>>>> >>>>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>>>> >>>>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>>>> >>>>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>>>> >>>>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>>>> >>>>>>> >>>>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>>>> >>>>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>>>> >>>>>> >>>>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>>>> >>>>> >>>>> [adding his gmail address which should be the right one] >>>>> >>>>>> David, thanks for your time to review and sorry for the delay and late response. >>>>>> I spent some time to get myself familiar with the code, but there are still some >>>>>> questions to me, explained as below. >>>>>> >>>>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>>>> incur more overhead if the threshold is decreased as you said in (a). >>>>> >>>>> Right. Alex did quite some performance/overhead evaluation when >>>>> introducing this feature. Changing the reporting granularity on most >>>>> setups (esp., x86-64) is not desired IMHO. >>>> >>>> Yes, generally reporting pages comes at a fairly high cost so it is >>>> important to find the right trade-off between the size of the page and >>>> the size of the batch of pages being reported. If the size of the >>>> pages is reduced it maybe important to increase the batch size in >>>> order to avoid paying too much in the way of overhead. >>>> >>>> The other main reason for holding to pageblock_order on x86 is to >>>> avoid THP splitting. Anything smaller than pageblock_order will >>>> trigger THP splitting which will significantly hurt the performance of >>>> the VM in general as it forces it down to order 0 pages. >>>> >>> >>> Alex, Thanks for your reply and sorry for taking your time to this >>> discussion. >>> >>> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget >>> used in page_reporting_cycle() when you're talking about "batch"? >>> >>> I don't understand how the THP splitting is triggered. As I understood, >>> the free page reporting works like this: the free pages are reported >>> back to QEMU through virtio-balloon, madvise(DONTNEED) is called on the >>> free pages. For THP related VMA, its memory is guranteed to be physically >>> 2MB contiguous. The 2MB memory is free'd at once as I understand. I >>> don't think the 2MB memory can be free'd partially apart from the THP >>> splitting caused by some reasons. One of the reason is memory reclaim >>> and swapping. However, I think the THP splitting is caused by memory >>> reclaim is irrelevant to what you were saying. >> >> >> Anonymous memory: Assume you have a 2 MB THP in the hypervisor. If you madvise(DONTNEED, 4K), you'll split the THP and free the single 4k page back to the buddy. That virtual memory region is no longer backed by a physically contiguous 2 MB page and there is no huge mapping in the page tables. Instead, there is now an "ordinary" 4k mapping with a hole. Bad for performance. >> > > Ok, now I see and thanks for your explanation. In this regard, we need gurantee > the page reporting threshold is larger or equal to THP size. The THP size is 2MB > or 512MB if base page size is 4KB or 64KB. > > It makes the issue hard to be fixed as we have 512MB THP size with 64KB base page > size on arm64. The following configurations are used in this case. > > CONFIG_FORCE_MAX_ZONEORDER 14 > MAX_ORDER 14 > pageblock_order 13 > > The free page reporting won't be started until the page freeing comes up with 512MB > free area. On system, which has limited memory (e.g. 4GB), 512MB free area is hard > to have due to memory fragmentation. > FWIW, in an ideal world the hypervisor would tell us (guest) which granularity it prefers. Could be that the hypervisor is using a different page size / thp size ... but that's a different story :) From a guest POV, it usually makes sense to report only whole pageblocks. But as identified, this is an issue with abnormally large pageblocks. 512MB pageblocks / THP is simply far from ideal for VMs, kindly phrased. :)
> Indeed. 512MB pageblocks are rare, especially on systems which have been > up and running for long time. > > The free page reporting starts from guest. Taking an extreme case: guest has > 512MB memory and it's backed by one THP on host. The free page reporting won't > work at all. > > Besides, it seems free page reporting isn't guranteed to work all the time. > For example, on system where we have 4KB base page size. Freeing individual > 4KB pages can't come up with a free 2MB pageblock due to fragmentation. > In this case, the free'd page won't be reported immediately, but might be > reported after swapping or compaction due to memory pressure. The free page > isn't reported immediately at least. Exactly, it's a pure optimization that won't work, especially when guest memory is heavily fragmented. There has to be a balance between reclaiming free memory in the hypervisor, degrading VM performance, and overhead of the feature. Further, there are no guarantees when a VM will reuse the memory again. In the worst case, all VMs that reported free pages reuse memory at the same time. In that case, one definitely needs sufficient backend memory in the hypervisor (-> swap) to not run out of memory, and performance will be degraded. As MST once phrased it, if the feature has a higher overhead than swapping in the hypervisor, it's of little use. > > David, how about taking your suggestion to have different threshold size only > for arm64 (64KB base page size). The threshold will be smaller than pageblock_order > for sure. There are two ways to do so and please let me know which is the preferred > way to go if you (and Alex) agree to do it. > > (a) Introduce CONFIG_PAGE_REPORTING_ORDER for individual archs to choose the > value. The threshold falls back to pageblock_order if isn't configurated. > (b) Rename PAGE_REPORTING_MIN_ORDER to PAGE_REPORTING_ORDER. archs can decide > its value. If it's not provided by arch, it falls back to pageblock_order. > I wonder if we could further define it as a (module/cmdline) parameter and make it configurable when booting. The default could then be set based on CONFIG_PAGE_REPORTING_ORDER. CONFIG_PAGE_REPORTING_ORDER would default to pageblock_order (if easily possible) and could be special-cases to arm64 with 64k. > By the way, I recently had some performance testing on different page sizes. > We get much more performance gain from 64KB (vs 4KB) page size in guest than > 512MB (vs 2MB) THP on host. It means the performance won't be affected too > much even the 512MB THP is splitted on arm64 host. Yes, if one is even able to get 512MB THP populated in the hypervisor -- because once again, 512MB THP are just a bad fit for many workloads.
On 6/16/21 5:59 PM, David Hildenbrand wrote: > On 16.06.21 03:53, Gavin Shan wrote: >> On 6/14/21 9:03 PM, David Hildenbrand wrote: >>> On 11.06.21 09:44, Gavin Shan wrote: >>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>> is never triggered with the following configurations and settings on >>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>> >>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>> pageblock_order: 13 >>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>> MAX_ORDER: 14 >>>>>> >>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>> >>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>> : \ >>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>> >>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>> drops after 'memhog' exits. >>>>> >>>>> IIRC, we use pageblock_order to >>>>> >>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>> >>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>> >>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>> >>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>> >>>>> >>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>> >>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>> >>>> >>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>> >>> >>> [adding his gmail address which should be the right one] >>> >>>> David, thanks for your time to review and sorry for the delay and late response. >>>> I spent some time to get myself familiar with the code, but there are still some >>>> questions to me, explained as below. >>>> >>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>> incur more overhead if the threshold is decreased as you said in (a). >>> >>> Right. Alex did quite some performance/overhead evaluation when introducing this feature. Changing the reporting granularity on most setups (esp., x86-64) is not desired IMHO. >>> >> >> Thanks for adding Alex's correct mail address, David. >> >>>> >>>> This patch tries to decrease the free page reporting threshold. The @pageblock_order >>>> isn't touched. I don't understand how the code changes affecting THP splitting >>>> and the creation of page blocks mentioned in (b) and (c). David, could you please >>>> provide more details? >>> >>> Think of it like this: while reporting to the hypervisor, we temporarily turn free/"movable" pieces part of a pageblock "unmovable" -- see __isolate_free_page()->del_page_from_free_list(). While reporting them to the hypervisor, these pages are not available and not even marked as PageBuddy() anymore. >>> >>> There are at least two scenarios where this could affect creation of free pageblocks I can see: >>> >>> a. Compaction. While compacting, we might identify completely movable/free pageblocks, however, actual compaction on that pageblock can fail because some part is temporarily unmovable. >>> >>> b. Free/alloc sequences. Assume a pageblocks is mostly free, except two pages (x and y). Assume the following sequence: >>> >>> 1. free(x) >>> 2. free(y) >>> 3. alloc >>> >>> Before your change, after 1. and 2. we'll have a free pageblock. 3 won't allocate from that pageblock. >>> >>> With your change, free page reporting might run after 1. After 2, we'll not have a free pageblock (until free page reporting finished), and 3. might just reallocate what we freed in 2 and prevent having a free pageblock. >>> >>> >>> No idea how relevant both points are in practice, however, the fundamental difference to current handling is that we would turn parts of pageblocks temporarily unmovable, instead of complete pageblocks. >>> >> >> Thank you for the details. Without my changes and the page reporting threshold >> is @pageblock_order, the whole page block can become 'movable' from 'unmovable'. >> I don't think it's what we want, but I need Alex's confirm. > > __isolate_free_page() will set the pageblock MIGRATE_MOVABLE in that case. It's only temporarily unmovable, while we're hinting. > > Note that MOVABLE vs. UNMOVABLE is just grouping for free pages, and even setting it to the wrong migratetype isn't "wrong" as in "correctness". It doesn't make a difference if there are no free pages because the whole block is isolated. > Yes, It doesn't matter since these pages have been isolated. The migration type is changed to MIGRATE_MOVABLE in __isolated_free_page(). My questions are actually: (1) Is it possible the migration type is changed from MIGRATE_UNMOVABLE to MIGRATE_MOVABLE in __isolated_free_page()? (2) After the free page reporting is completed, the migrate type is restored to MIGRATE_UNMOVABLE? Thanks, Gavin
On 6/16/21 6:03 PM, David Hildenbrand wrote: > On 16.06.21 11:10, Gavin Shan wrote: >> On 6/15/21 12:26 PM, Alexander Duyck wrote: >>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>>> is never triggered with the following configurations and settings on >>>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>>> >>>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>>> pageblock_order: 13 >>>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>>> MAX_ORDER: 14 >>>>>>> >>>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>>> >>>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>>> : \ >>>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>>> >>>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>>> drops after 'memhog' exits. >>>>>> >>>>>> IIRC, we use pageblock_order to >>>>>> >>>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>>> >>>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>>> >>>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>>> >>>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>>> >>>>>> >>>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>>> >>>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>>> >>>>> >>>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>>> >>>> >>>> [adding his gmail address which should be the right one] >>>> >>>>> David, thanks for your time to review and sorry for the delay and late response. >>>>> I spent some time to get myself familiar with the code, but there are still some >>>>> questions to me, explained as below. >>>>> >>>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>>> incur more overhead if the threshold is decreased as you said in (a). >>>> >>>> Right. Alex did quite some performance/overhead evaluation when >>>> introducing this feature. Changing the reporting granularity on most >>>> setups (esp., x86-64) is not desired IMHO. >>> >>> Yes, generally reporting pages comes at a fairly high cost so it is >>> important to find the right trade-off between the size of the page and >>> the size of the batch of pages being reported. If the size of the >>> pages is reduced it maybe important to increase the batch size in >>> order to avoid paying too much in the way of overhead. >>> >>> The other main reason for holding to pageblock_order on x86 is to >>> avoid THP splitting. Anything smaller than pageblock_order will >>> trigger THP splitting which will significantly hurt the performance of >>> the VM in general as it forces it down to order 0 pages. >>> >> >> Alex, Thanks for your reply and sorry for taking your time to this >> discussion. >> >> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget >> used in page_reporting_cycle() when you're talking about "batch"? >> >> I don't understand how the THP splitting is triggered. As I understood, >> the free page reporting works like this: the free pages are reported >> back to QEMU through virtio-balloon, madvise(DONTNEED) is called on the >> free pages. For THP related VMA, its memory is guranteed to be physically >> 2MB contiguous. The 2MB memory is free'd at once as I understand. I >> don't think the 2MB memory can be free'd partially apart from the THP >> splitting caused by some reasons. One of the reason is memory reclaim >> and swapping. However, I think the THP splitting is caused by memory >> reclaim is irrelevant to what you were saying. > > > Anonymous memory: Assume you have a 2 MB THP in the hypervisor. If you madvise(DONTNEED, 4K), you'll split the THP and free the single 4k page back to the buddy. That virtual memory region is no longer backed by a physically contiguous 2 MB page and there is no huge mapping in the page tables. Instead, there is now an "ordinary" 4k mapping with a hole. Bad for performance. > Ok, now I see and thanks for your explanation. In this regard, we need gurantee the page reporting threshold is larger or equal to THP size. The THP size is 2MB or 512MB if base page size is 4KB or 64KB. It makes the issue hard to be fixed as we have 512MB THP size with 64KB base page size on arm64. The following configurations are used in this case. CONFIG_FORCE_MAX_ZONEORDER 14 MAX_ORDER 14 pageblock_order 13 The free page reporting won't be started until the page freeing comes up with 512MB free area. On system, which has limited memory (e.g. 4GB), 512MB free area is hard to have due to memory fragmentation. Thanks, Gavin
On 6/16/21 9:20 PM, David Hildenbrand wrote: > On 16.06.21 15:16, Gavin Shan wrote: >> On 6/16/21 6:03 PM, David Hildenbrand wrote: >>> On 16.06.21 11:10, Gavin Shan wrote: >>>> On 6/15/21 12:26 PM, Alexander Duyck wrote: >>>>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>>>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>>>>> is never triggered with the following configurations and settings on >>>>>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>>>>> >>>>>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>>>>> pageblock_order: 13 >>>>>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>>>>> MAX_ORDER: 14 >>>>>>>>> >>>>>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>>>>> >>>>>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>>>>> : \ >>>>>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>>>>> >>>>>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>>>>> drops after 'memhog' exits. >>>>>>>> >>>>>>>> IIRC, we use pageblock_order to >>>>>>>> >>>>>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>>>>> >>>>>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>>>>> >>>>>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>>>>> >>>>>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>>>>> >>>>>>>> >>>>>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>>>>> >>>>>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>>>>> >>>>>>> >>>>>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>>>>> >>>>>> >>>>>> [adding his gmail address which should be the right one] >>>>>> >>>>>>> David, thanks for your time to review and sorry for the delay and late response. >>>>>>> I spent some time to get myself familiar with the code, but there are still some >>>>>>> questions to me, explained as below. >>>>>>> >>>>>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>>>>> incur more overhead if the threshold is decreased as you said in (a). >>>>>> >>>>>> Right. Alex did quite some performance/overhead evaluation when >>>>>> introducing this feature. Changing the reporting granularity on most >>>>>> setups (esp., x86-64) is not desired IMHO. >>>>> >>>>> Yes, generally reporting pages comes at a fairly high cost so it is >>>>> important to find the right trade-off between the size of the page and >>>>> the size of the batch of pages being reported. If the size of the >>>>> pages is reduced it maybe important to increase the batch size in >>>>> order to avoid paying too much in the way of overhead. >>>>> >>>>> The other main reason for holding to pageblock_order on x86 is to >>>>> avoid THP splitting. Anything smaller than pageblock_order will >>>>> trigger THP splitting which will significantly hurt the performance of >>>>> the VM in general as it forces it down to order 0 pages. >>>>> >>>> >>>> Alex, Thanks for your reply and sorry for taking your time to this >>>> discussion. >>>> >>>> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget >>>> used in page_reporting_cycle() when you're talking about "batch"? >>>> >>>> I don't understand how the THP splitting is triggered. As I understood, >>>> the free page reporting works like this: the free pages are reported >>>> back to QEMU through virtio-balloon, madvise(DONTNEED) is called on the >>>> free pages. For THP related VMA, its memory is guranteed to be physically >>>> 2MB contiguous. The 2MB memory is free'd at once as I understand. I >>>> don't think the 2MB memory can be free'd partially apart from the THP >>>> splitting caused by some reasons. One of the reason is memory reclaim >>>> and swapping. However, I think the THP splitting is caused by memory >>>> reclaim is irrelevant to what you were saying. >>> >>> >>> Anonymous memory: Assume you have a 2 MB THP in the hypervisor. If you madvise(DONTNEED, 4K), you'll split the THP and free the single 4k page back to the buddy. That virtual memory region is no longer backed by a physically contiguous 2 MB page and there is no huge mapping in the page tables. Instead, there is now an "ordinary" 4k mapping with a hole. Bad for performance. >>> >> >> Ok, now I see and thanks for your explanation. In this regard, we need gurantee >> the page reporting threshold is larger or equal to THP size. The THP size is 2MB >> or 512MB if base page size is 4KB or 64KB. >> >> It makes the issue hard to be fixed as we have 512MB THP size with 64KB base page >> size on arm64. The following configurations are used in this case. >> >> CONFIG_FORCE_MAX_ZONEORDER 14 >> MAX_ORDER 14 >> pageblock_order 13 >> >> The free page reporting won't be started until the page freeing comes up with 512MB >> free area. On system, which has limited memory (e.g. 4GB), 512MB free area is hard >> to have due to memory fragmentation. >> > > FWIW, in an ideal world the hypervisor would tell us (guest) which granularity it prefers. Could be that the hypervisor is using a different page size / thp size ... but that's a different story :) > > From a guest POV, it usually makes sense to report only whole pageblocks. But as identified, this is an issue with abnormally large pageblocks. > > 512MB pageblocks / THP is simply far from ideal for VMs, kindly phrased. :) > Indeed. 512MB pageblocks are rare, especially on systems which have been up and running for long time. The free page reporting starts from guest. Taking an extreme case: guest has 512MB memory and it's backed by one THP on host. The free page reporting won't work at all. Besides, it seems free page reporting isn't guranteed to work all the time. For example, on system where we have 4KB base page size. Freeing individual 4KB pages can't come up with a free 2MB pageblock due to fragmentation. In this case, the free'd page won't be reported immediately, but might be reported after swapping or compaction due to memory pressure. The free page isn't reported immediately at least. David, how about taking your suggestion to have different threshold size only for arm64 (64KB base page size). The threshold will be smaller than pageblock_order for sure. There are two ways to do so and please let me know which is the preferred way to go if you (and Alex) agree to do it. (a) Introduce CONFIG_PAGE_REPORTING_ORDER for individual archs to choose the value. The threshold falls back to pageblock_order if isn't configurated. (b) Rename PAGE_REPORTING_MIN_ORDER to PAGE_REPORTING_ORDER. archs can decide its value. If it's not provided by arch, it falls back to pageblock_order. By the way, I recently had some performance testing on different page sizes. We get much more performance gain from 64KB (vs 4KB) page size in guest than 512MB (vs 2MB) THP on host. It means the performance won't be affected too much even the 512MB THP is splitted on arm64 host. Thanks, Gavin
On Wed, Jun 16, 2021 at 12:10 AM Gavin Shan <gshan@redhat.com> wrote: > > On 6/15/21 12:26 PM, Alexander Duyck wrote: > > On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: > >> On 11.06.21 09:44, Gavin Shan wrote: > >>> On 6/1/21 6:01 PM, David Hildenbrand wrote: > >>>> On 01.06.21 05:33, Gavin Shan wrote: > >>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as > >>>>> minimal order (threshold) to trigger page reporting. The page reporting > >>>>> is never triggered with the following configurations and settings on > >>>>> aarch64. In the particular scenario, the page reporting won't be triggered > >>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the > >>>>> page freeing. The condition is very hard, or even impossible to be met. > >>>>> > >>>>> CONFIG_ARM64_PAGE_SHIFT: 16 > >>>>> CONFIG_HUGETLB_PAGE: Y > >>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N > >>>>> pageblock_order: 13 > >>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 > >>>>> MAX_ORDER: 14 > >>>>> > >>>>> The issue can be reproduced in VM, running kernel with above configurations > >>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous > >>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. > >>>>> > >>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ > >>>>> -accel kvm -machine virt,gic-version=host \ > >>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ > >>>>> -object memory-backend-ram,id=mem0,size=2048M \ > >>>>> -object memory-backend-ram,id=mem1,size=2048M \ > >>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ > >>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ > >>>>> : \ > >>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes > >>>>> > >>>>> This tries to fix the issue by adjusting the threshold to the smaller value > >>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS > >>>>> drops after 'memhog' exits. > >>>> > >>>> IIRC, we use pageblock_order to > >>>> > >>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. > >>>> > >>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. > >>>> > >>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. > >>>> > >>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. > >>>> > >>>> > >>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. > >>>> > >>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. > >>>> > >>> > >>> [Remove Alexander from the cc list as his mail isn't reachable] > >>> > >> > >> [adding his gmail address which should be the right one] > >> > >>> David, thanks for your time to review and sorry for the delay and late response. > >>> I spent some time to get myself familiar with the code, but there are still some > >>> questions to me, explained as below. > >>> > >>> Yes, @pageblock_order is currently taken as page reporting threshold. It will > >>> incur more overhead if the threshold is decreased as you said in (a). > >> > >> Right. Alex did quite some performance/overhead evaluation when > >> introducing this feature. Changing the reporting granularity on most > >> setups (esp., x86-64) is not desired IMHO. > > > > Yes, generally reporting pages comes at a fairly high cost so it is > > important to find the right trade-off between the size of the page and > > the size of the batch of pages being reported. If the size of the > > pages is reduced it maybe important to increase the batch size in > > order to avoid paying too much in the way of overhead. > > > > The other main reason for holding to pageblock_order on x86 is to > > avoid THP splitting. Anything smaller than pageblock_order will > > trigger THP splitting which will significantly hurt the performance of > > the VM in general as it forces it down to order 0 pages. > > > > Alex, Thanks for your reply and sorry for taking your time to this > discussion. > > Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget > used in page_reporting_cycle() when you're talking about "batch"? Yes, when I refer to batch it is how many pages we are processing in a single call. That is limited by PAGE_REPORTING_CAPACITY.
On 6/16/21 10:07 PM, David Hildenbrand wrote: >> Indeed. 512MB pageblocks are rare, especially on systems which have been >> up and running for long time. >> >> The free page reporting starts from guest. Taking an extreme case: guest has >> 512MB memory and it's backed by one THP on host. The free page reporting won't >> work at all. >> >> Besides, it seems free page reporting isn't guranteed to work all the time. >> For example, on system where we have 4KB base page size. Freeing individual >> 4KB pages can't come up with a free 2MB pageblock due to fragmentation. >> In this case, the free'd page won't be reported immediately, but might be >> reported after swapping or compaction due to memory pressure. The free page >> isn't reported immediately at least. > > Exactly, it's a pure optimization that won't work, especially when guest memory is heavily fragmented. There has to be a balance between reclaiming free memory in the hypervisor, degrading VM performance, and overhead of the feature. > > Further, there are no guarantees when a VM will reuse the memory again. In the worst case, all VMs that reported free pages reuse memory at the same time. In that case, one definitely needs sufficient backend memory in the hypervisor (-> swap) to not run out of memory, and performance will be degraded. > > As MST once phrased it, if the feature has a higher overhead than swapping in the hypervisor, it's of little use. > Thanks for the explanation and sorry again for late response, David. I took last week as holiday and didn't work too much. However, it's nice to have unused pages returned back to the host. These pages can be used by other VMs or applications running on the host. >> >> David, how about taking your suggestion to have different threshold size only >> for arm64 (64KB base page size). The threshold will be smaller than pageblock_order >> for sure. There are two ways to do so and please let me know which is the preferred >> way to go if you (and Alex) agree to do it. >> >> (a) Introduce CONFIG_PAGE_REPORTING_ORDER for individual archs to choose the >> value. The threshold falls back to pageblock_order if isn't configurated. >> (b) Rename PAGE_REPORTING_MIN_ORDER to PAGE_REPORTING_ORDER. archs can decide >> its value. If it's not provided by arch, it falls back to pageblock_order. >> > > I wonder if we could further define it as a (module/cmdline) parameter and make it configurable when booting. The default could then be set based on CONFIG_PAGE_REPORTING_ORDER. CONFIG_PAGE_REPORTING_ORDER would default to pageblock_order (if easily possible) and could be special-cases to arm64 with 64k. > The formal patches are posted for review. I used macro PAGE_REPORTING_ORDER instead of CONFIG_PAGE_REPORTING_ORDER. The page reporting order (threshold) is also exported as a module parameter, as you suggested. >> By the way, I recently had some performance testing on different page sizes. >> We get much more performance gain from 64KB (vs 4KB) page size in guest than >> 512MB (vs 2MB) THP on host. It means the performance won't be affected too >> much even the 512MB THP is splitted on arm64 host. > > Yes, if one is even able to get 512MB THP populated in the hypervisor -- because once again, 512MB THP are just a bad fit for many workloads. > Yeah, indeed :) Thanks, Gavin
On 6/17/21 12:15 AM, Alexander Duyck wrote: > On Wed, Jun 16, 2021 at 12:10 AM Gavin Shan <gshan@redhat.com> wrote: >> >> On 6/15/21 12:26 PM, Alexander Duyck wrote: >>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>> On 01.06.21 05:33, Gavin Shan wrote: >>>>>>> The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as >>>>>>> minimal order (threshold) to trigger page reporting. The page reporting >>>>>>> is never triggered with the following configurations and settings on >>>>>>> aarch64. In the particular scenario, the page reporting won't be triggered >>>>>>> until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the >>>>>>> page freeing. The condition is very hard, or even impossible to be met. >>>>>>> >>>>>>> CONFIG_ARM64_PAGE_SHIFT: 16 >>>>>>> CONFIG_HUGETLB_PAGE: Y >>>>>>> CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N >>>>>>> pageblock_order: 13 >>>>>>> CONFIG_FORCE_MAX_ZONEORDER: 14 >>>>>>> MAX_ORDER: 14 >>>>>>> >>>>>>> The issue can be reproduced in VM, running kernel with above configurations >>>>>>> and settings. The 'memhog' is used inside the VM to access 512MB anonymous >>>>>>> area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. >>>>>>> >>>>>>> /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ >>>>>>> -accel kvm -machine virt,gic-version=host \ >>>>>>> -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ >>>>>>> -object memory-backend-ram,id=mem0,size=2048M \ >>>>>>> -object memory-backend-ram,id=mem1,size=2048M \ >>>>>>> -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ >>>>>>> -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ >>>>>>> : \ >>>>>>> -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes >>>>>>> >>>>>>> This tries to fix the issue by adjusting the threshold to the smaller value >>>>>>> of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS >>>>>>> drops after 'memhog' exits. >>>>>> >>>>>> IIRC, we use pageblock_order to >>>>>> >>>>>> a) Reduce the free page reporting overhead. Reporting on small chunks can make us report constantly with little system activity. >>>>>> >>>>>> b) Avoid splitting THP in the hypervisor, avoiding downgraded VM performance. >>>>>> >>>>>> c) Avoid affecting creation of pageblock_order pages while hinting is active. I think there are cases where "temporary pulling sub-pageblock pages" can negatively affect creation of pageblock_order pages. Concurrent compaction would be one of these cases. >>>>>> >>>>>> The monstrosity called aarch64 64k is really special in that sense, because a) does not apply because pageblocks are just very big, b) does sometimes not apply because either our VM isn't backed by (rare) 512MB THP or uses 4k with 2MB THP and c) similarly doesn't apply in smallish VMs because we don't really happen to create 512MB THP either way. >>>>>> >>>>>> >>>>>> For example, going on x86-64 from reporting 2MB to something like 32KB is absolutely undesired. >>>>>> >>>>>> I think if we want to go down that path (and I am not 100% sure yet if we want to), we really want to treat only the special case in a special way. Note that even when doing it only for aarch64 with 64k, you will still end up splitting THP in a hypervisor if it uses 64k base pages (b)) and can affect creation of THP, for example, when compacting (c), so there is a negative side to that. >>>>>> >>>>> >>>>> [Remove Alexander from the cc list as his mail isn't reachable] >>>>> >>>> >>>> [adding his gmail address which should be the right one] >>>> >>>>> David, thanks for your time to review and sorry for the delay and late response. >>>>> I spent some time to get myself familiar with the code, but there are still some >>>>> questions to me, explained as below. >>>>> >>>>> Yes, @pageblock_order is currently taken as page reporting threshold. It will >>>>> incur more overhead if the threshold is decreased as you said in (a). >>>> >>>> Right. Alex did quite some performance/overhead evaluation when >>>> introducing this feature. Changing the reporting granularity on most >>>> setups (esp., x86-64) is not desired IMHO. >>> >>> Yes, generally reporting pages comes at a fairly high cost so it is >>> important to find the right trade-off between the size of the page and >>> the size of the batch of pages being reported. If the size of the >>> pages is reduced it maybe important to increase the batch size in >>> order to avoid paying too much in the way of overhead. >>> >>> The other main reason for holding to pageblock_order on x86 is to >>> avoid THP splitting. Anything smaller than pageblock_order will >>> trigger THP splitting which will significantly hurt the performance of >>> the VM in general as it forces it down to order 0 pages. >>> >> >> Alex, Thanks for your reply and sorry for taking your time to this >> discussion. >> >> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget >> used in page_reporting_cycle() when you're talking about "batch"? > > Yes, when I refer to batch it is how many pages we are processing in a > single call. That is limited by PAGE_REPORTING_CAPACITY. > Thanks for your confirmation. I will do some investigation on how to improve it. Lets resolve the issue we had at first: The page reporting doesn't work on arm64 with 64KB base size. Thanks, Gavin
On 6/17/21 12:15 AM, Alexander Duyck wrote: > On Wed, Jun 16, 2021 at 12:10 AM Gavin Shan <gshan@redhat.com> wrote: >> On 6/15/21 12:26 PM, Alexander Duyck wrote: >>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: >>>> On 11.06.21 09:44, Gavin Shan wrote: >>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: >>>>>> On 01.06.21 05:33, Gavin Shan wrote: [...] >>> >>> Yes, generally reporting pages comes at a fairly high cost so it is >>> important to find the right trade-off between the size of the page and >>> the size of the batch of pages being reported. If the size of the >>> pages is reduced it maybe important to increase the batch size in >>> order to avoid paying too much in the way of overhead. >>> >>> The other main reason for holding to pageblock_order on x86 is to >>> avoid THP splitting. Anything smaller than pageblock_order will >>> trigger THP splitting which will significantly hurt the performance of >>> the VM in general as it forces it down to order 0 pages. >>> >> >> Alex, Thanks for your reply and sorry for taking your time to this >> discussion. >> >> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget >> used in page_reporting_cycle() when you're talking about "batch"? > > Yes, when I refer to batch it is how many pages we are processing in a > single call. That is limited by PAGE_REPORTING_CAPACITY. > Alex, It seems the batch mechanism is to avoid heavy contention on zone's lock if I'm correct? The current design is to report all pages in the corresponding free list within 17 calls to page_reporting_cycle(). Could you please explain why 17 was chosen? :) budget = DIV_ROUND_UP(area->nr_free, PAGE_REPORTING_CAPACITY * 16); It's related to the magic number ("16"). With the threshold is decreased, for example from 512MB to 2MB on arm64 with 64KB base page size, more page reporting activities will be introduced. From this regard, it's reasonable to increase the magic number as well, so that more calls to page_reporting_cycle() to avoid the contention to zone's lock. If you agree, I will come up with something, similar to what we do for the threshold. However, I'm not sure if 64 is reasonable cycles to have for this particular case. in arch/arm64/include/asm/page.h #ifdef CONFIG_ARM64_64K_PAGES #define PAGE_REPORTING_ORDER 5 #define PAGE_REPORTING_CYCLES 64 #endif in mm/page_reporting.h #ifndef PAGE_REPORTING_CYCLES #define PAGE_REPORTING_CYCLES 16 #endif in mm/page_reporting.c::page_reporting_cycle() budget = DIV_ROUND_UP(area->nr_free, PAGE_REPORTING_CAPACITY * PAGE_REPORTING_CYCLES); Thanks, Gavin
On Sun, Jun 20, 2021 at 10:51 PM Gavin Shan <gshan@redhat.com> wrote: > > On 6/17/21 12:15 AM, Alexander Duyck wrote: > > On Wed, Jun 16, 2021 at 12:10 AM Gavin Shan <gshan@redhat.com> wrote: > >> On 6/15/21 12:26 PM, Alexander Duyck wrote: > >>> On Mon, Jun 14, 2021 at 4:03 AM David Hildenbrand <david@redhat.com> wrote: > >>>> On 11.06.21 09:44, Gavin Shan wrote: > >>>>> On 6/1/21 6:01 PM, David Hildenbrand wrote: > >>>>>> On 01.06.21 05:33, Gavin Shan wrote: > > [...] > > >>> > >>> Yes, generally reporting pages comes at a fairly high cost so it is > >>> important to find the right trade-off between the size of the page and > >>> the size of the batch of pages being reported. If the size of the > >>> pages is reduced it maybe important to increase the batch size in > >>> order to avoid paying too much in the way of overhead. > >>> > >>> The other main reason for holding to pageblock_order on x86 is to > >>> avoid THP splitting. Anything smaller than pageblock_order will > >>> trigger THP splitting which will significantly hurt the performance of > >>> the VM in general as it forces it down to order 0 pages. > >>> > >> > >> Alex, Thanks for your reply and sorry for taking your time to this > >> discussion. > >> > >> Could you please confirm it's PAGE_REPORTING_CAPACITY or the budget > >> used in page_reporting_cycle() when you're talking about "batch"? > > > > Yes, when I refer to batch it is how many pages we are processing in a > > single call. That is limited by PAGE_REPORTING_CAPACITY. > > > > Alex, It seems the batch mechanism is to avoid heavy contention on > zone's lock if I'm correct? The current design is to report all pages > in the corresponding free list within 17 calls to page_reporting_cycle(). > Could you please explain why 17 was chosen? :) > > budget = DIV_ROUND_UP(area->nr_free, PAGE_REPORTING_CAPACITY * 16); It isn't that 17 was chosen. The idea was to only process 1/16th of the free list at a time. The general idea is that by doing that and limiting the page reporting to an interval of once every 2 seconds we should have the entire guest reported out after about 30 seconds assuming it is idle. If it isn't idle then the overhead for reporting only 1/16th of the guest memory should be fairly low. > It's related to the magic number ("16"). With the threshold is decreased, > for example from 512MB to 2MB on arm64 with 64KB base page size, more > page reporting activities will be introduced. From this regard, it's > reasonable to increase the magic number as well, so that more calls > to page_reporting_cycle() to avoid the contention to zone's lock. > > If you agree, I will come up with something, similar to what we do for > the threshold. However, I'm not sure if 64 is reasonable cycles to have > for this particular case. > > in arch/arm64/include/asm/page.h > #ifdef CONFIG_ARM64_64K_PAGES > #define PAGE_REPORTING_ORDER 5 > #define PAGE_REPORTING_CYCLES 64 > #endif You mentioned going from 512MB to 2MB pages. What is the MAX_ORDER for the arm architecture you are working with? One concern I have is that order 5 pages may not be high enough order to keep the page reporting from interfering with the guest memory allocations since you are having to cover so many free areas. Ideally with page reporting we were aiming for MAX_ORDER and MAX_ORDER - 1 as being the main targets for page reporting. The advantage there is that on x86 that also allowed us to avoid splitting THP pages. The other advantage is that when combined with the 16 and the fact that we were rounding up the budget it should come out to about one minute to fully flush out all the memory on an idle guest. If anything we would want to take the (MAX_ORDER - PAGE_REPORTING_ORDER)/2 and use that as a multiple for the 16 value as that would give us the upper limit on how long it should take to report all of the pages in a given block. It gets us to the same value, but does a better job of explaining why. > in mm/page_reporting.h > #ifndef PAGE_REPORTING_CYCLES > #define PAGE_REPORTING_CYCLES 16 > #endif > in mm/page_reporting.c::page_reporting_cycle() > budget = DIV_ROUND_UP(area->nr_free, > PAGE_REPORTING_CAPACITY * PAGE_REPORTING_CYCLES); > > Thanks, > Gavin The 16 isn't about cycles, it is about how fast we want to leak the memory out of the guest. You don't want this to go too fast otherwise you are going to be fighting with anything that is trying to allocate memory. In theory you should only be reporting pages in the top tiers of the memory hierarchy so that it is not very active. One way to think about page reporting is as a leaky bucket approach. After a minute or so you want the bucket to drain assuming the memory allocations/frees for a guest have become inactive. However it the VM is active you want it to do very little in way of page reporting so that you are not having to fault back in a ton of memory.
diff --git a/mm/page_reporting.h b/mm/page_reporting.h index 2c385dd4ddbd..5dae3d171004 100644 --- a/mm/page_reporting.h +++ b/mm/page_reporting.h @@ -10,9 +10,10 @@ #include <linux/pgtable.h> #include <linux/scatterlist.h> -#define PAGE_REPORTING_MIN_ORDER pageblock_order - #ifdef CONFIG_PAGE_REPORTING +#define PAGE_REPORTING_MIN_ORDER \ + min_t(unsigned int, pageblock_order, (MAX_ORDER / 2)) + DECLARE_STATIC_KEY_FALSE(page_reporting_enabled); void __page_reporting_notify(void);
The PAGE_REPORTING_MIN_ORDER is equal to @pageblock_order, taken as minimal order (threshold) to trigger page reporting. The page reporting is never triggered with the following configurations and settings on aarch64. In the particular scenario, the page reporting won't be triggered until the largest (2 ^ (MAX_ORDER-1)) free area is achieved from the page freeing. The condition is very hard, or even impossible to be met. CONFIG_ARM64_PAGE_SHIFT: 16 CONFIG_HUGETLB_PAGE: Y CONFIG_HUGETLB_PAGE_SIZE_VARIABLE: N pageblock_order: 13 CONFIG_FORCE_MAX_ZONEORDER: 14 MAX_ORDER: 14 The issue can be reproduced in VM, running kernel with above configurations and settings. The 'memhog' is used inside the VM to access 512MB anonymous area. The QEMU's RSS doesn't drop accordingly after 'memhog' exits. /home/gavin/sandbox/qemu.main/build/qemu-system-aarch64 \ -accel kvm -machine virt,gic-version=host \ -cpu host -smp 8,sockets=2,cores=4,threads=1 -m 4096M,maxmem=64G \ -object memory-backend-ram,id=mem0,size=2048M \ -object memory-backend-ram,id=mem1,size=2048M \ -numa node,nodeid=0,cpus=0-3,memdev=mem0 \ -numa node,nodeid=1,cpus=4-7,memdev=mem1 \ : \ -device virtio-balloon-pci,id=balloon0,free-page-reporting=yes This tries to fix the issue by adjusting the threshold to the smaller value of @pageblock_order and (MAX_ORDER/2). With this applied, the QEMU's RSS drops after 'memhog' exits. Signed-off-by: Gavin Shan <gshan@redhat.com> --- mm/page_reporting.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)