MFC r271889, 272799, 272800, 274976

This brings in bus_get_domain() and the related reporting via devinfo,
dmesg, and sysctl.

Obtained from:	adrian, jhb
Sponsored by:	Netflix, Inc.
This commit is contained in:
scottl 2015-03-12 07:07:41 +00:00
parent 27912a446f
commit 565f091b88
8 changed files with 158 additions and 15 deletions

View File

@ -207,6 +207,7 @@ static device_method_t acpi_methods[] = {
DEVMETHOD(bus_setup_intr, bus_generic_setup_intr),
DEVMETHOD(bus_teardown_intr, bus_generic_teardown_intr),
DEVMETHOD(bus_hint_device_unit, acpi_hint_device_unit),
DEVMETHOD(bus_get_domain, acpi_get_domain),
/* ACPI bus */
DEVMETHOD(acpi_id_probe, acpi_device_id_probe),
@ -797,6 +798,7 @@ acpi_print_child(device_t bus, device_t child)
retval += resource_list_print_type(rl, "drq", SYS_RES_DRQ, "%ld");
if (device_get_flags(child))
retval += printf(" flags %#x", device_get_flags(child));
retval += bus_print_child_domain(bus, child);
retval += bus_print_child_footer(bus, child);
return (retval);
@ -853,11 +855,18 @@ acpi_child_location_str_method(device_t cbdev, device_t child, char *buf,
size_t buflen)
{
struct acpi_device *dinfo = device_get_ivars(child);
char buf2[32];
int pxm;
if (dinfo->ad_handle)
snprintf(buf, buflen, "handle=%s", acpi_name(dinfo->ad_handle));
else
snprintf(buf, buflen, "unknown");
if (dinfo->ad_handle) {
snprintf(buf, buflen, "handle=%s", acpi_name(dinfo->ad_handle));
if (ACPI_SUCCESS(acpi_GetInteger(dinfo->ad_handle, "_PXM", &pxm))) {
snprintf(buf2, 32, " _PXM=%d", pxm);
strlcat(buf, buf2, buflen);
}
} else {
snprintf(buf, buflen, "unknown");
}
return (0);
}
@ -1062,6 +1071,35 @@ acpi_hint_device_unit(device_t acdev, device_t child, const char *name,
}
}
/*
* Fech the NUMA domain for the given device.
*
* If a device has a _PXM method, map that to a NUMA domain.
*
* If none is found, then it'll call the parent method.
* If there's no domain, return ENOENT.
*/
int
acpi_get_domain(device_t dev, device_t child, int *domain)
{
#if MAXMEMDOM > 1
ACPI_HANDLE h;
int d, pxm;
h = acpi_get_handle(child);
if ((h != NULL) &&
ACPI_SUCCESS(acpi_GetInteger(h, "_PXM", &pxm))) {
d = acpi_map_pxm_to_vm_domainid(pxm);
if (d < 0)
return (ENOENT);
*domain = d;
return (0);
}
#endif
/* No _PXM node; go up a level */
return (bus_generic_get_domain(dev, child, domain));
}
/*
* Pre-allocate/manage all memory and IO resources. Since rman can't handle
* duplicates, we merge any in the sysresource attach routine.

View File

@ -94,6 +94,7 @@ static device_method_t acpi_pci_methods[] = {
DEVMETHOD(bus_write_ivar, acpi_pci_write_ivar),
DEVMETHOD(bus_child_location_str, acpi_pci_child_location_str_method),
DEVMETHOD(bus_get_dma_tag, acpi_pci_get_dma_tag),
DEVMETHOD(bus_get_domain, acpi_get_domain),
/* PCI interface */
DEVMETHOD(pci_set_powerstate, acpi_pci_set_powerstate_method),
@ -149,12 +150,19 @@ acpi_pci_child_location_str_method(device_t cbdev, device_t child, char *buf,
size_t buflen)
{
struct acpi_pci_devinfo *dinfo = device_get_ivars(child);
int pxm;
char buf2[32];
pci_child_location_str_method(cbdev, child, buf, buflen);
if (dinfo->ap_handle) {
strlcat(buf, " handle=", buflen);
strlcat(buf, acpi_name(dinfo->ap_handle), buflen);
strlcat(buf, " handle=", buflen);
strlcat(buf, acpi_name(dinfo->ap_handle), buflen);
if (ACPI_SUCCESS(acpi_GetInteger(dinfo->ap_handle, "_PXM", &pxm))) {
snprintf(buf2, 32, " _PXM=%d", pxm);
strlcat(buf, buf2, buflen);
}
}
return (0);
}

View File

@ -489,5 +489,16 @@ ACPI_HANDLE acpi_GetReference(ACPI_HANDLE scope, ACPI_OBJECT *obj);
SYSCTL_DECL(_debug_acpi);
/*
* Map a PXM to a VM domain.
*
* Returns the VM domain ID if found, or -1 if not found / invalid.
*/
#if MAXMEMDOM > 1
extern int acpi_map_pxm_to_vm_domainid(int pxm);
#endif
extern int acpi_get_domain(device_t dev, device_t child, int *domain);
#endif /* _KERNEL */
#endif /* !_ACPIVAR_H_ */

View File

@ -3781,6 +3781,7 @@ pci_print_child(device_t dev, device_t child)
retval += printf(" at device %d.%d", pci_get_slot(child),
pci_get_function(child));
retval += bus_print_child_domain(dev, child);
retval += bus_print_child_footer(dev, child);
return (retval);

View File

@ -670,3 +670,16 @@ METHOD int remap_intr {
device_t _child;
u_int _irq;
} DEFAULT null_remap_intr;
/**
* @brief Get the VM domain handle for the given bus and child.
*
* @param _dev the bus device
* @param _child the child device
* @param _domain a pointer to the bus's domain handle identifier
*/
METHOD int get_domain {
device_t _dev;
device_t _child;
int *_domain;
} DEFAULT bus_generic_get_domain;

View File

@ -54,6 +54,7 @@ __FBSDID("$FreeBSD$");
#include <sys/uio.h>
#include <sys/bus.h>
#include <sys/interrupt.h>
#include <sys/cpuset.h>
#include <net/vnet.h>
@ -284,6 +285,7 @@ static void
device_sysctl_init(device_t dev)
{
devclass_t dc = dev->devclass;
int domain;
if (dev->sysctl_tree != NULL)
return;
@ -313,6 +315,10 @@ device_sysctl_init(device_t dev)
OID_AUTO, "%parent", CTLTYPE_STRING | CTLFLAG_RD,
dev, DEVICE_SYSCTL_PARENT, device_sysctl_handler, "A",
"parent device");
if (bus_get_domain(dev, &domain) == 0)
SYSCTL_ADD_INT(&dev->sysctl_ctx,
SYSCTL_CHILDREN(dev->sysctl_tree), OID_AUTO, "%domain",
CTLFLAG_RD, NULL, domain, "NUMA domain");
}
static void
@ -3716,6 +3722,25 @@ bus_print_child_footer(device_t dev, device_t child)
return (printf(" on %s\n", device_get_nameunit(dev)));
}
/**
* @brief Helper function for implementing BUS_PRINT_CHILD().
*
* This function prints out the VM domain for the given device.
*
* @returns the number of characters printed
*/
int
bus_print_child_domain(device_t dev, device_t child)
{
int domain;
/* No domain? Don't print anything */
if (BUS_GET_DOMAIN(dev, child, &domain) != 0)
return (0);
return (printf(" numa-domain %d", domain));
}
/**
* @brief Helper function for implementing BUS_PRINT_CHILD().
*
@ -3730,6 +3755,7 @@ bus_generic_print_child(device_t dev, device_t child)
int retval = 0;
retval += bus_print_child_header(dev, child);
retval += bus_print_child_domain(dev, child);
retval += bus_print_child_footer(dev, child);
return (retval);
@ -4144,6 +4170,16 @@ bus_generic_child_present(device_t dev, device_t child)
return (BUS_CHILD_PRESENT(device_get_parent(dev), dev));
}
int
bus_generic_get_domain(device_t dev, device_t child, int *domain)
{
if (dev->parent)
return (BUS_GET_DOMAIN(dev->parent, dev, domain));
return (ENOENT);
}
/*
* Some convenience functions to make it easier for drivers to use the
* resource-management functions. All these really do is hide the
@ -4476,6 +4512,18 @@ bus_get_dma_tag(device_t dev)
return (BUS_GET_DMA_TAG(parent, dev));
}
/**
* @brief Wrapper function for BUS_GET_DOMAIN().
*
* This function simply calls the BUS_GET_DOMAIN() method of the
* parent of @p dev.
*/
int
bus_get_domain(device_t dev, int *domain)
{
return (BUS_GET_DOMAIN(device_get_parent(dev), dev, domain));
}
/* Resume all devices and then notify userland that we're up again. */
static int
root_resume(device_t dev)

View File

@ -332,6 +332,7 @@ struct resource_list *
bus_generic_get_resource_list (device_t, device_t);
void bus_generic_new_pass(device_t dev);
int bus_print_child_header(device_t dev, device_t child);
int bus_print_child_domain(device_t dev, device_t child);
int bus_print_child_footer(device_t dev, device_t child);
int bus_generic_print_child(device_t dev, device_t child);
int bus_generic_probe(device_t dev);
@ -363,6 +364,8 @@ int bus_generic_teardown_intr(device_t dev, device_t child,
int bus_generic_write_ivar(device_t dev, device_t child, int which,
uintptr_t value);
int bus_generic_get_domain(device_t dev, device_t child, int *domain);
/*
* Wrapper functions for the BUS_*_RESOURCE methods to make client code
* a little simpler.
@ -389,6 +392,7 @@ int bus_activate_resource(device_t dev, int type, int rid,
int bus_deactivate_resource(device_t dev, int type, int rid,
struct resource *r);
bus_dma_tag_t bus_get_dma_tag(device_t dev);
int bus_get_domain(device_t dev, int *domain);
int bus_release_resource(device_t dev, int type, int rid,
struct resource *r);
int bus_free_resource(device_t dev, int type, struct resource *r);

View File

@ -62,6 +62,8 @@ int num_mem;
static ACPI_TABLE_SRAT *srat;
static vm_paddr_t srat_physaddr;
static int vm_domains[VM_PHYSSEG_MAX];
static void srat_walk_table(acpi_subtable_handler *handler, void *arg);
/*
@ -247,7 +249,6 @@ check_phys_avail(void)
static int
renumber_domains(void)
{
int domains[VM_PHYSSEG_MAX];
int i, j, slot;
/* Enumerate all the domains. */
@ -255,17 +256,17 @@ renumber_domains(void)
for (i = 0; i < num_mem; i++) {
/* See if this domain is already known. */
for (j = 0; j < vm_ndomains; j++) {
if (domains[j] >= mem_info[i].domain)
if (vm_domains[j] >= mem_info[i].domain)
break;
}
if (j < vm_ndomains && domains[j] == mem_info[i].domain)
if (j < vm_ndomains && vm_domains[j] == mem_info[i].domain)
continue;
/* Insert the new domain at slot 'j'. */
slot = j;
for (j = vm_ndomains; j > slot; j--)
domains[j] = domains[j - 1];
domains[slot] = mem_info[i].domain;
vm_domains[j] = vm_domains[j - 1];
vm_domains[slot] = mem_info[i].domain;
vm_ndomains++;
if (vm_ndomains > MAXMEMDOM) {
vm_ndomains = 1;
@ -280,15 +281,15 @@ renumber_domains(void)
* If the domain is already the right value, no need
* to renumber.
*/
if (domains[i] == i)
if (vm_domains[i] == i)
continue;
/* Walk the cpu[] and mem_info[] arrays to renumber. */
for (j = 0; j < num_mem; j++)
if (mem_info[j].domain == domains[i])
if (mem_info[j].domain == vm_domains[i])
mem_info[j].domain = i;
for (j = 0; j <= MAX_APIC_ID; j++)
if (cpus[j].enabled && cpus[j].domain == domains[i])
if (cpus[j].enabled && cpus[j].domain == vm_domains[i])
cpus[j].domain = i;
}
KASSERT(vm_ndomains > 0,
@ -368,4 +369,23 @@ srat_set_cpus(void *dummy)
}
}
SYSINIT(srat_set_cpus, SI_SUB_CPU, SI_ORDER_ANY, srat_set_cpus, NULL);
/*
* Map a _PXM value to a VM domain ID.
*
* Returns the domain ID, or -1 if no domain ID was found.
*/
int
acpi_map_pxm_to_vm_domainid(int pxm)
{
int i;
for (i = 0; i < vm_ndomains; i++) {
if (vm_domains[i] == pxm)
return (i);
}
return (-1);
}
#endif /* MAXMEMDOM > 1 */