Re: [Xen-devel] [PATCH 1/2] xen: vnuma support for PV guests runningas domU.

[Stefano Stabellini]

On Wed, 13 Nov 2013, Elena Ufimtseva wrote:
> Issues Xen hypercall subop XENMEM_get_vnumainfo and sets the
> NUMA topology, otherwise sets dummy NUMA node and prevents
> numa_init from calling other numa initializators as they may
> break other guests.
> 
> Signed-off-by: Elena Ufimtseva <ufimtseva@xxxxxxxxx>
> ---
>  arch/x86/include/asm/xen/vnuma.h |   12 ++++
>  arch/x86/mm/numa.c               |    5 ++
>  arch/x86/xen/Makefile            |    2 +-
>  arch/x86/xen/vnuma.c             |  119 ++++++++++++++++++++++++++++++++++++++
>  include/xen/interface/memory.h   |   28 +++++++++
>  5 files changed, 165 insertions(+), 1 deletion(-)
>  create mode 100644 arch/x86/include/asm/xen/vnuma.h
>  create mode 100644 arch/x86/xen/vnuma.c
> 
> diff --git a/arch/x86/include/asm/xen/vnuma.h b/arch/x86/include/asm/xen/vnuma.h
> new file mode 100644
> index 0000000..1ba1e06
> --- /dev/null
> +++ b/arch/x86/include/asm/xen/vnuma.h
> @@ -0,0 +1,12 @@
> +#ifndef _ASM_X86_VNUMA_H
> +#define _ASM_X86_VNUMA_H
> +
> +#ifdef CONFIG_XEN
> +int xen_vnuma_supported(void);
> +int xen_numa_init(void);
> +#else
> +int xen_vnuma_supported(void) { return 0; };
> +int xen_numa_init(void) { return -1; };

static inline?

> +#endif
>
> +#endif /* _ASM_X86_VNUMA_H */
> diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
> index 8bf93ba..c8a61dc 100644
> --- a/arch/x86/mm/numa.c
> +++ b/arch/x86/mm/numa.c
> @@ -19,6 +19,7 @@
>  #include <asm/amd_nb.h>
>  
>  #include "numa_internal.h"
> +#include "asm/xen/vnuma.h"
>  
>  int __initdata numa_off;
>  nodemask_t numa_nodes_parsed __initdata;
> @@ -621,6 +622,10 @@ static int __init dummy_numa_init(void)
>  void __init x86_numa_init(void)
>  {
>  	if (!numa_off) {
> +#ifdef CONFIG_XEN
> +		if (xen_vnuma_supported() && !numa_init(xen_numa_init))
> +			return;
> +#endif

Given the non-Xen function definitions above, you can remove the ifdef
CONFIG_XEN here.


>  #ifdef CONFIG_X86_NUMAQ
>  		if (!numa_init(numaq_numa_init))
>  			return;
> diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile
> index 96ab2c0..de9deab 100644
> --- a/arch/x86/xen/Makefile
> +++ b/arch/x86/xen/Makefile
> @@ -13,7 +13,7 @@ CFLAGS_mmu.o			:= $(nostackp)
>  obj-y		:= enlighten.o setup.o multicalls.o mmu.o irq.o \
>  			time.o xen-asm.o xen-asm_$(BITS).o \
>  			grant-table.o suspend.o platform-pci-unplug.o \
> -			p2m.o
> +			p2m.o vnuma.o
>  
>  obj-$(CONFIG_EVENT_TRACING) += trace.o
>  
> diff --git a/arch/x86/xen/vnuma.c b/arch/x86/xen/vnuma.c
> new file mode 100644
> index 0000000..b4fc667
> --- /dev/null
> +++ b/arch/x86/xen/vnuma.c
> @@ -0,0 +1,119 @@
> +#include <linux/err.h>
> +#include <linux/memblock.h>
> +#include <xen/interface/xen.h>
> +#include <xen/interface/memory.h>
> +#include <asm/xen/interface.h>
> +#include <asm/xen/hypercall.h>
> +#include <asm/xen/vnuma.h>
> +
> +#ifdef CONFIG_NUMA
> +
> +/* Checks if hypercall is suported */
                               ^ supported
> +int xen_vnuma_supported()
> +{
> +	return HYPERVISOR_memory_op(XENMEM_get_vnuma_info, NULL) == -ENOSYS ? 0 : 1;
> +}
> +
> +int __init xen_numa_init(void)
> +{
> +	int rc;
> +	unsigned int i, j, nr_nodes, cpu, idx, pcpus;
> +	u64 physm, physd, physc;
> +	unsigned int *vdistance, *cpu_to_node;
> +	unsigned long mem_size, dist_size, cpu_to_node_size;

physm, physd and physc need to be initialized to 0, otherwise the
vnumaout error path below is erroneous

> +	struct vmemrange *vblock;
> +
> +	struct vnuma_topology_info numa_topo = {
> +		.domid = DOMID_SELF,
> +		.__pad = 0
> +	};
> +	rc = -EINVAL;
> +
> +	/* For now only PV guests are supported */
> +	if (!xen_pv_domain())
> +		return rc;
> +
> +	pcpus = num_possible_cpus();
> +
> +	mem_size =  pcpus * sizeof(struct vmemrange);
> +	dist_size = pcpus * pcpus * sizeof(*numa_topo.vdistance);
> +	cpu_to_node_size = pcpus * sizeof(*numa_topo.cpu_to_node);
> +
> +	physm = memblock_alloc(mem_size, PAGE_SIZE);
> +	vblock = __va(physm);
> +
> +	physd = memblock_alloc(dist_size, PAGE_SIZE);
> +	vdistance  = __va(physd);
> +
> +	physc = memblock_alloc(cpu_to_node_size, PAGE_SIZE);
> +	cpu_to_node  = __va(physc);
> +
> +	if (!physm || !physc || !physd)
> +		goto vnumaout;
> +
> +	set_xen_guest_handle(numa_topo.nr_nodes, &nr_nodes);
> +	set_xen_guest_handle(numa_topo.vmemblks, vblock);
> +	set_xen_guest_handle(numa_topo.vdistance, vdistance);
> +	set_xen_guest_handle(numa_topo.cpu_to_node, cpu_to_node);
> +
> +	rc = HYPERVISOR_memory_op(XENMEM_get_vnuma_info, &numa_topo);
> +
> +	if (rc < 0)
> +		goto vnumaout;
> +	if (*numa_topo.nr_nodes == 0) {
> +		/* will pass to dummy_numa_init */
> +		goto vnumaout;
> +	}
> +	if (*numa_topo.nr_nodes > num_possible_cpus()) {
> +		pr_debug("vNUMA: Node without cpu is not supported in this version.\n");
> +		goto vnumaout;
> +	}
> +	/*
> +	 * NUMA nodes memory ranges are in pfns, constructed and
> +	 * aligned based on e820 ram domain map.
> +	 */
> +	for (i = 0; i < *numa_topo.nr_nodes; i++) {
> +		if (numa_add_memblk(i, vblock[i].start, vblock[i].end))
> +			/* pass to numa_dummy_init */
> +			goto vnumaout;
> +		node_set(i, numa_nodes_parsed);
> +	}
> +	setup_nr_node_ids();
> +	/* Setting the cpu, apicid to node */
> +	for_each_cpu(cpu, cpu_possible_mask) {
> +		set_apicid_to_node(cpu, cpu_to_node[cpu]);
> +		numa_set_node(cpu, cpu_to_node[cpu]);
> +		cpumask_set_cpu(cpu, node_to_cpumask_map[cpu_to_node[cpu]]);
> +	}
> +	for (i = 0; i < *numa_topo.nr_nodes; i++) {
> +		for (j = 0; j < *numa_topo.nr_nodes; j++) {
> +			idx = (j * *numa_topo.nr_nodes) + i;
> +			numa_set_distance(i, j, *(vdistance + idx));
> +		}
> +	}
> +	rc = 0;
> +vnumaout:
> +	if (physm)
> +		memblock_free(__pa(physm), mem_size);
> +	if (physd)
> +		memblock_free(__pa(physd), dist_size);
> +	if (physc)
> +		memblock_free(__pa(physc), cpu_to_node_size);
> +	/*
> +	 * Set the "dummy" node and exit without error so Linux
> +	 * will not try any NUMA init functions which might break
> +	 * guests in the future. This will discard all previous
> +	 * settings.
> +	 */
> +	if (rc != 0) {
> +		for (i = 0; i < MAX_LOCAL_APIC; i++)
> +			set_apicid_to_node(i, NUMA_NO_NODE);
> +		nodes_clear(numa_nodes_parsed);
> +		nodes_clear(node_possible_map);
> +		nodes_clear(node_online_map);
> +		node_set(0, numa_nodes_parsed);
> +		numa_add_memblk(0, 0, PFN_PHYS(max_pfn));
> +	}
> +	return 0;
> +}
> +#endif
> diff --git a/include/xen/interface/memory.h b/include/xen/interface/memory.h
> index 2ecfe4f..3974e9a 100644
> --- a/include/xen/interface/memory.h
> +++ b/include/xen/interface/memory.h
> @@ -263,4 +263,32 @@ struct xen_remove_from_physmap {
>  };
>  DEFINE_GUEST_HANDLE_STRUCT(xen_remove_from_physmap);
>  
> +/* vNUMA structures */
> +struct vmemrange {
> +	uint64_t start, end;
> +	struct vmemrange *next;
> +};
> +DEFINE_GUEST_HANDLE_STRUCT(vmemrange);
> +
> +struct vnuma_topology_info {
> +	/* OUT */
> +	domid_t domid;
> +	uint32_t __pad;
> +	/* IN */
> +	GUEST_HANDLE(uint) nr_nodes; /* number of virtual numa nodes */
> +	/* distance table */
> +	GUEST_HANDLE(uint) vdistance;
> +	/* cpu mapping to vnodes */
> +	GUEST_HANDLE(uint) cpu_to_node;
> +	/*
> +	* array of numa memory areas constructed by Xen
> +	* where start and end are pfn numbers of the area
> +	* Xen takes into account domains e820 map
> +	*/
> +	GUEST_HANDLE(vmemrange) vmemblks;
> +};
> +DEFINE_GUEST_HANDLE_STRUCT(vnuma_topology_info);
> +
> +#define XENMEM_get_vnuma_info	25
> +
>  #endif /* __XEN_PUBLIC_MEMORY_H__ */
> -- 
> 1.7.10.4
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxx
> http://lists.xen.org/xen-devel
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/