[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH V9 08/12] NUMA: set guest numa nodes memory poli
From: |
Andrew Jones |
Subject: |
Re: [Qemu-devel] [PATCH V9 08/12] NUMA: set guest numa nodes memory policy |
Date: |
Fri, 23 Aug 2013 04:44:44 -0400 (EDT) |
----- Original Message -----
> Set the guest numa nodes memory policies using the mbind(2)
> system call node by node.
> After this patch, we are able to set guest nodes memory policies
> through the QEMU options, this arms to solve the guest cross
> nodes memory access performance issue.
> And as you all know, if PCI-passthrough is used,
> direct-attached-device uses DMA transfer between device and qemu process.
> All pages of the guest will be pinned by get_user_pages().
>
> KVM_ASSIGN_PCI_DEVICE ioctl
> kvm_vm_ioctl_assign_device()
> =>kvm_assign_device()
> => kvm_iommu_map_memslots()
> => kvm_iommu_map_pages()
> => kvm_pin_pages()
>
> So, with direct-attached-device, all guest page's page count will be +1 and
> any page migration will not work. AutoNUMA won't too.
>
> So, we should set the guest nodes memory allocation policies before
> the pages are really mapped.
>
> Signed-off-by: Andre Przywara <address@hidden>
> Signed-off-by: Wanlong Gao <address@hidden>
> ---
> numa.c | 90
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 90 insertions(+)
>
> diff --git a/numa.c b/numa.c
> index 4ccc6cb..4a9c368 100644
> --- a/numa.c
> +++ b/numa.c
> @@ -28,6 +28,16 @@
> #include "qapi-visit.h"
> #include "qapi/opts-visitor.h"
> #include "qapi/dealloc-visitor.h"
> +#include "exec/memory.h"
> +
> +#ifdef CONFIG_NUMA
> +#include <numa.h>
> +#include <numaif.h>
> +#ifndef MPOL_F_RELATIVE_NODES
> +#define MPOL_F_RELATIVE_NODES (1 << 14)
> +#define MPOL_F_STATIC_NODES (1 << 15)
> +#endif
> +#endif
>
> QemuOptsList qemu_numa_opts = {
> .name = "numa",
> @@ -219,6 +229,79 @@ void set_numa_nodes(void)
> }
> }
>
> +#ifdef CONFIG_NUMA
> +static int node_parse_bind_mode(unsigned int nodeid)
> +{
> + int bind_mode;
> +
> + switch (numa_info[nodeid].policy) {
> + case NUMA_NODE_POLICY_MEMBIND:
> + bind_mode = MPOL_BIND;
> + break;
> + case NUMA_NODE_POLICY_INTERLEAVE:
> + bind_mode = MPOL_INTERLEAVE;
> + break;
> + case NUMA_NODE_POLICY_PREFERRED:
> + bind_mode = MPOL_PREFERRED;
> + break;
> + case NUMA_NODE_POLICY_DEFAULT:
> + default:
> + bind_mode = MPOL_DEFAULT;
> + return bind_mode;
> + }
> +
> + bind_mode |= numa_info[nodeid].relative ?
> + MPOL_F_RELATIVE_NODES : MPOL_F_STATIC_NODES;
> +
> + return bind_mode;
> +}
> +#endif
> +
> +static int set_node_mem_policy(int nodeid)
> +{
> +#ifdef CONFIG_NUMA
> + void *ram_ptr;
> + RAMBlock *block;
> + ram_addr_t len, ram_offset = 0;
> + int bind_mode;
> + int i;
> +
> + QTAILQ_FOREACH(block, &ram_list.blocks, next) {
> + if (!strcmp(block->mr->name, "pc.ram")) {
> + break;
> + }
> + }
> +
> + if (block->host == NULL) {
> + return -1;
> + }
> +
> + ram_ptr = block->host;
> + for (i = 0; i < nodeid; i++) {
> + len = numa_info[i].node_mem;
> + ram_offset += len;
> + }
> +
> + len = numa_info[nodeid].node_mem;
> + bind_mode = node_parse_bind_mode(nodeid);
> + unsigned long *nodes = numa_info[nodeid].host_mem;
> +
> + /* This is a workaround for a long standing bug in Linux'
> + * mbind implementation, which cuts off the last specified
> + * node. To stay compatible should this bug be fixed, we
> + * specify one more node and zero this one out.
> + */
> + unsigned long maxnode = find_last_bit(nodes, MAX_CPUMASK_BITS);
> + clear_bit(maxnode + 1, nodes);
This clear_bit() isn't necessary. We know that maxnode+1 is certainly
already clear, because find_last_bit() just returned maxnode as the last
"non-clear" bit.
> + if (mbind(ram_ptr + ram_offset, len, bind_mode, nodes, maxnode + 1, 0))
> {
> + perror("mbind");
> + return -1;
> + }
> +#endif
> +
> + return 0;
> +}
> +
> void set_numa_modes(void)
> {
> CPUState *cpu;
> @@ -231,4 +314,11 @@ void set_numa_modes(void)
> }
> }
> }
> +
> + for (i = 0; i < nb_numa_nodes; i++) {
> + if (set_node_mem_policy(i) == -1) {
> + fprintf(stderr,
> + "qemu: can not set host memory policy for node%d\n", i);
> + }
> + }
> }
> --
> 1.8.4.rc4
>
>
>
- [Qemu-devel] [PATCH V9 00/12] Add support for binding guest numa nodes to host numa nodes, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 01/12] NUMA: add NumaOptions, NumaNodeOptions and NumaMemOptions, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 04/12] NUMA: move numa related code to numa.c, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 05/12] NUMA: Add numa_info structure to contain numa nodes info, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 06/12] NUMA: Add Linux libnuma detection, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 08/12] NUMA: set guest numa nodes memory policy, Wanlong Gao, 2013/08/23
- Re: [Qemu-devel] [PATCH V9 08/12] NUMA: set guest numa nodes memory policy,
Andrew Jones <=
- [Qemu-devel] [PATCH V9 07/12] NUMA: parse guest numa nodes memory policy, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 10/12] NUMA: add hmp command set-mem-policy, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 03/12] NUMA: check if the total numa memory size is equal to ram_size, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 11/12] NUMA: add qmp command query-numa, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 12/12] NUMA: convert hmp command info_numa to use qmp command query_numa, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 02/12] NUMA: split -numa option, Wanlong Gao, 2013/08/23
- [Qemu-devel] [PATCH V9 09/12] NUMA: add qmp command set-mem-policy to set memory policy for NUMA node, Wanlong Gao, 2013/08/23