| Message ID | 1554178276-10372-1-git-send-email-fanglinxu@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mem-hotplug: fix node spanned pages when we have a node with only zone_movable | expand |
On Tue, Apr 02, 2019 at 12:11:16PM +0800, Linxu Fang wrote: > commit <342332e6a925> ("mm/page_alloc.c: introduce kernelcore=mirror > option") and series patches rewrote the calculation of node spanned > pages. > commit <e506b99696a2> (mem-hotplug: fix node spanned pages when we have a > movable node), but the current code still has problems, > when we have a node with only zone_movable and the node id is not zero, > the size of node spanned pages is double added. > That's because we have an empty normal zone, and zone_start_pfn or > zone_end_pfn is not between arch_zone_lowest_possible_pfn and > arch_zone_highest_possible_pfn, so we need to use clamp to constrain the > range just like the commit <96e907d13602> (bootmem: Reimplement > __absent_pages_in_range() using for_each_mem_pfn_range()). So, let me see if I understood this correctly: When calling zone_spanned_pages_in_node() for any node which is not node 0, > *zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type]; > *zone_end_pfn = arch_zone_highest_possible_pfn[zone_type]; will actually set zone_start_pfn/zone_end_pfn to the values from node0's ZONE_NORMAL? So we use clamp to actually check if such values fall within what node1's memory spans, and ignore them otherwise? Btw, mem-hotplug does not hit this path anymore. > > e.g. > Zone ranges: > DMA [mem 0x0000000000001000-0x0000000000ffffff] > DMA32 [mem 0x0000000001000000-0x00000000ffffffff] > Normal [mem 0x0000000100000000-0x000000023fffffff] > Movable zone start for each node > Node 0: 0x0000000100000000 > Node 1: 0x0000000140000000 > Early memory node ranges > node 0: [mem 0x0000000000001000-0x000000000009efff] > node 0: [mem 0x0000000000100000-0x00000000bffdffff] > node 0: [mem 0x0000000100000000-0x000000013fffffff] > node 1: [mem 0x0000000140000000-0x000000023fffffff] > > node 0 DMA spanned:0xfff present:0xf9e absent:0x61 > node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 > node 0 Normal spanned:0 present:0 absent:0 > node 0 Movable spanned:0x40000 present:0x40000 absent:0 > On node 0 totalpages(node_present_pages): 1048446 > node_spanned_pages:1310719 > node 1 DMA spanned:0 present:0 absent:0 > node 1 DMA32 spanned:0 present:0 absent:0 > node 1 Normal spanned:0x100000 present:0x100000 absent:0 > node 1 Movable spanned:0x100000 present:0x100000 absent:0 > On node 1 totalpages(node_present_pages): 2097152 > node_spanned_pages:2097152 > Memory: 6967796K/12582392K available (16388K kernel code, 3686K rwdata, > 4468K rodata, 2160K init, 10444K bss, 5614596K reserved, 0K > cma-reserved) > > It shows that the current memory of node 1 is double added. > After this patch, the problem is fixed. > > node 0 DMA spanned:0xfff present:0xf9e absent:0x61 > node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 > node 0 Normal spanned:0 present:0 absent:0 > node 0 Movable spanned:0x40000 present:0x40000 absent:0 > On node 0 totalpages(node_present_pages): 1048446 > node_spanned_pages:1310719 > node 1 DMA spanned:0 present:0 absent:0 > node 1 DMA32 spanned:0 present:0 absent:0 > node 1 Normal spanned:0 present:0 absent:0 > node 1 Movable spanned:0x100000 present:0x100000 absent:0 > On node 1 totalpages(node_present_pages): 1048576 > node_spanned_pages:1048576 > memory: 6967796K/8388088K available (16388K kernel code, 3686K rwdata, > 4468K rodata, 2160K init, 10444K bss, 1420292K reserved, 0K > cma-reserved) > > Signed-off-by: Linxu Fang <fanglinxu@huawei.com> > --- > mm/page_alloc.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index 3eb01de..5cd0cb2 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -6233,13 +6233,15 @@ static unsigned long __init zone_spanned_pages_in_node(int nid, > unsigned long *zone_end_pfn, > unsigned long *ignored) > { > + unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type]; > + unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type]; > /* When hotadd a new node from cpu_up(), the node should be empty */ > if (!node_start_pfn && !node_end_pfn) > return 0; > > /* Get the start and end of the zone */ > - *zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type]; > - *zone_end_pfn = arch_zone_highest_possible_pfn[zone_type]; > + *zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high); > + *zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high); > adjust_zone_range_for_zone_movable(nid, zone_type, > node_start_pfn, node_end_pfn, > zone_start_pfn, zone_end_pfn); > -- > 1.8.5.6 > >
On Tue, Apr 02, 2019 at 04:57:11PM +0200, Oscar Salvador wrote: > On Tue, Apr 02, 2019 at 12:11:16PM +0800, Linxu Fang wrote: > > commit <342332e6a925> ("mm/page_alloc.c: introduce kernelcore=mirror > > option") and series patches rewrote the calculation of node spanned > > pages. > > commit <e506b99696a2> (mem-hotplug: fix node spanned pages when we have a > > movable node), but the current code still has problems, > > when we have a node with only zone_movable and the node id is not zero, > > the size of node spanned pages is double added. > > That's because we have an empty normal zone, and zone_start_pfn or > > zone_end_pfn is not between arch_zone_lowest_possible_pfn and > > arch_zone_highest_possible_pfn, so we need to use clamp to constrain the > > range just like the commit <96e907d13602> (bootmem: Reimplement > > __absent_pages_in_range() using for_each_mem_pfn_range()). > > So, let me see if I understood this correctly: > > When calling zone_spanned_pages_in_node() for any node which is not node 0, > > > *zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type]; > > *zone_end_pfn = arch_zone_highest_possible_pfn[zone_type]; > > will actually set zone_start_pfn/zone_end_pfn to the values from node0's > ZONE_NORMAL? Of course, I meant when calling it being zone_type == ZONE_NORMAL.
> will actually set zone_start_pfn/zone_end_pfn to the values from node0's > ZONE_NORMAL? > So we use clamp to actually check if such values fall within what node1's > memory spans, and ignore them otherwise? That's right. Normally, zone_start_pfn/zone_end_pfn has the same value for all nodes. Let's look at another example, which is obtained by adding some debugging information. e.g. Zone ranges: DMA [mem 0x0000000000001000-0x0000000000ffffff] DMA32 [mem 0x0000000001000000-0x00000000ffffffff] Normal [mem 0x0000000100000000-0x0000000792ffffff] Movable zone start for each node Node 0: 0x0000000100000000 Node 1: 0x00000002b1000000 Node 2: 0x0000000522000000 Early memory node ranges node 0: [mem 0x0000000000001000-0x000000000009efff] node 0: [mem 0x0000000000100000-0x00000000bffdefff] node 0: [mem 0x0000000100000000-0x00000002b0ffffff] node 1: [mem 0x00000002b1000000-0x0000000521ffffff] node 2: [mem 0x0000000522000000-0x0000000792ffffff] Node 0: node_start_pfn=1 node_end_pfn=2822144 DMA zone_low=1 zone_high=4096 DMA32 zone_low=4096 zone_high=1048576 Normal zone_low=1048576 zone_high=7942144 Movable zone_low=0 zone_high=0 Node 1: node_start_pfn=2822144 node_end_pfn=5382144 DMA zone_low=1 zone_high=4096 DMA32 zone_low=4096 zone_high=1048576 Normal zone_low=1048576 zone_high=7942144 Movable zone_low=0 zone_high=0 Node 2: node_start_pfn=5382144 node_end_pfn=7942144 DMA zone_low=1 zone_high=4096 DMA32 zone_low=4096 zone_high=1048576 Normal zone_low=1048576 zone_high=7942144 Movable zone_low=0 zone_high=0 Before this patch, zone_start_pfn/zone_end_pfn in node 0,1,2 is the same: DMA zone_start_pfn:1 zone_end_pfn:4096 DMA32 zone_start_pfn:4096 zone_end_pfn:1048576 Normal zone_start_pfn:1048576 zone_end_pfn:7942144 Movable zone_start_pfn:0 zone_end_pfn:0 spaned pages resuelt: node 0: DMA spanned:4095 DMA32 spanned:1044480 Normal spanned:0 Movable spanned:1773568 totalpages:2559869 node 1: DMA spanned:0 DMA32 spanned:0 Normal spanned:2560000 Movable spanned:2560000 totalpages:5120000 node 2: DMA spanned:0 DMA32 spanned:0 Normal spanned:2560000 Movable spanned:2560000 totalpages:5120000 After this patch: node 0: DMA zone_start_pfn:1 zone_end_pfn:4096 spanned:4095 DMA32 zone_start_pfn:4096 zone_end_pfn:1048576 spanned:1044480 Normal zone_start_pfn:1048576 zone_end_pfn:2822144 spanned:0 Movable zone_start_pfn:0 zone_end_pfn:0 spanned:1773568 totalpages:2559869 node 1: DMA zone_start_pfn:4096 zone_end_pfn:4096 spanned:0 DMA32 zone_start_pfn:1048576 zone_end_pfn:1048576 spanned:0 Normal zone_start_pfn:2822144 zone_end_pfn:5382144 spanned:0 Movable zone_start_pfn:0 zone_end_pfn:0 spanned:2560000 totalpages:2560000 node 2: DMA zone_start_pfn:4096 zone_end_pfn:4096 spanned:0 DMA32 zone_start_pfn:1048576 zone_end_pfn:1048576 spanned:0 Normal zone_start_pfn:5382144 zone_end_pfn:7942144 spanned:0 Movable zone_start_pfn:0 zone_end_pfn:0 spanned:2560000 totalpages:2560000 It is easy to construct such a scenario by configuring kernelcore=mirror in a multi-NUMA machine without full mirrored memory. Of course, it can be a machine without any mirrored memory. A great difference can be observed by startup information and viewing /proc/pagetypeinfo. On earlier kernel versions, such BUGs will directly double the memory of some nodes. Although these redundant memory exists in the form of reserved memory, this should not be expected.
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 3eb01de..5cd0cb2 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -6233,13 +6233,15 @@ static unsigned long __init zone_spanned_pages_in_node(int nid, unsigned long *zone_end_pfn, unsigned long *ignored) { + unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type]; + unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type]; /* When hotadd a new node from cpu_up(), the node should be empty */ if (!node_start_pfn && !node_end_pfn) return 0; /* Get the start and end of the zone */ - *zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type]; - *zone_end_pfn = arch_zone_highest_possible_pfn[zone_type]; + *zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high); + *zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high); adjust_zone_range_for_zone_movable(nid, zone_type, node_start_pfn, node_end_pfn, zone_start_pfn, zone_end_pfn);
commit <342332e6a925> ("mm/page_alloc.c: introduce kernelcore=mirror option") and series patches rewrote the calculation of node spanned pages. commit <e506b99696a2> (mem-hotplug: fix node spanned pages when we have a movable node), but the current code still has problems, when we have a node with only zone_movable and the node id is not zero, the size of node spanned pages is double added. That's because we have an empty normal zone, and zone_start_pfn or zone_end_pfn is not between arch_zone_lowest_possible_pfn and arch_zone_highest_possible_pfn, so we need to use clamp to constrain the range just like the commit <96e907d13602> (bootmem: Reimplement __absent_pages_in_range() using for_each_mem_pfn_range()). e.g. Zone ranges: DMA [mem 0x0000000000001000-0x0000000000ffffff] DMA32 [mem 0x0000000001000000-0x00000000ffffffff] Normal [mem 0x0000000100000000-0x000000023fffffff] Movable zone start for each node Node 0: 0x0000000100000000 Node 1: 0x0000000140000000 Early memory node ranges node 0: [mem 0x0000000000001000-0x000000000009efff] node 0: [mem 0x0000000000100000-0x00000000bffdffff] node 0: [mem 0x0000000100000000-0x000000013fffffff] node 1: [mem 0x0000000140000000-0x000000023fffffff] node 0 DMA spanned:0xfff present:0xf9e absent:0x61 node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 node 0 Normal spanned:0 present:0 absent:0 node 0 Movable spanned:0x40000 present:0x40000 absent:0 On node 0 totalpages(node_present_pages): 1048446 node_spanned_pages:1310719 node 1 DMA spanned:0 present:0 absent:0 node 1 DMA32 spanned:0 present:0 absent:0 node 1 Normal spanned:0x100000 present:0x100000 absent:0 node 1 Movable spanned:0x100000 present:0x100000 absent:0 On node 1 totalpages(node_present_pages): 2097152 node_spanned_pages:2097152 Memory: 6967796K/12582392K available (16388K kernel code, 3686K rwdata, 4468K rodata, 2160K init, 10444K bss, 5614596K reserved, 0K cma-reserved) It shows that the current memory of node 1 is double added. After this patch, the problem is fixed. node 0 DMA spanned:0xfff present:0xf9e absent:0x61 node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 node 0 Normal spanned:0 present:0 absent:0 node 0 Movable spanned:0x40000 present:0x40000 absent:0 On node 0 totalpages(node_present_pages): 1048446 node_spanned_pages:1310719 node 1 DMA spanned:0 present:0 absent:0 node 1 DMA32 spanned:0 present:0 absent:0 node 1 Normal spanned:0 present:0 absent:0 node 1 Movable spanned:0x100000 present:0x100000 absent:0 On node 1 totalpages(node_present_pages): 1048576 node_spanned_pages:1048576 memory: 6967796K/8388088K available (16388K kernel code, 3686K rwdata, 4468K rodata, 2160K init, 10444K bss, 1420292K reserved, 0K cma-reserved) Signed-off-by: Linxu Fang <fanglinxu@huawei.com> --- mm/page_alloc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)