Track device's NUMA domain in ifnet & alloc ifnet from NUMA local memory

This commit adds new if_alloc_domain() and if_alloc_dev() methods to
allocate ifnets.  When called with a domain on a NUMA machine,
ifalloc_domain() will record the NUMA domain in the ifnet, and it will
allocate the ifnet struct from memory which is local to that NUMA
node.  Similarly, if_alloc_dev() is a wrapper for if_alloc_domain
which uses a driver supplied device_t to call ifalloc_domain() with
the appropriate domain.

Note that the new if_numa_domain field fits in an alignment pad in
struct ifnet, and so does not alter the size of the structure.

Reviewed by:	glebius, kib, markj
Sponsored by:	Netflix
Differential Revision:	https://reviews.freebsd.org/D19930
This commit is contained in:
Andrew Gallatin 2019-04-22 19:24:21 +00:00
parent 613d28127b
commit 7687707dd4
6 changed files with 73 additions and 7 deletions

View File

@ -1175,6 +1175,8 @@ MLINKS+=iflibtxrx.9 isc_rxd_available.9 \
iflibtxrx.9 isc_txd_flush.9
MLINKS+=ifnet.9 if_addmulti.9 \
ifnet.9 if_alloc.9 \
ifnet.9 if_alloc_dev.9 \
ifnet.9 if_alloc_domain.9 \
ifnet.9 if_allmulti.9 \
ifnet.9 if_attach.9 \
ifnet.9 if_data.9 \

View File

@ -48,6 +48,10 @@
.Ss "Interface Manipulation Functions"
.Ft "struct ifnet *"
.Fn if_alloc "u_char type"
.Ft "struct ifnet *"
.Fn if_alloc_dev "u_char type" "device_t dev"
.Ft "struct ifnet *"
.Fn if_alloc_domain "u_char type" "int numa_domain"
.Ft void
.Fn if_attach "struct ifnet *ifp"
.Ft void
@ -440,6 +444,15 @@ It is used to cache the type passed to
but unlike
.Va if_type ,
it would not be changed by drivers.
.It Va if_numa_domain
.Pq Vt uint8_t
The NUMA domain of the hardware device associated with the interface.
This is filled in with a wildcard value unless the kernel is NUMA
aware, the system is a NUMA system, and the ifnet is allocated
using
.Fn if_alloc_dev
or
.Fn if_alloc_domain .
.El
.Pp
References to
@ -1151,6 +1164,24 @@ include the allocation of a
.Fa type
specific structure in
.Va if_l2com .
.It Fn if_alloc_dev
Allocate and initialize
.Vt "struct ifnet"
as
.Fn if_alloc
does, with the addition that the ifnet can be tagged with the
appropriate NUMA domain derived from the
.Fa dev
argument passed by the caller.
.It Fn if_alloc_domain
Allocate and initialize
.Vt "struct ifnet"
as
.Fn if_alloc
does, with the addition that the ifnet will be tagged with the NUMA
domain via the
.Fa numa_domain
argument passed by the caller.
.It Fn if_attach
Link the specified interface
.Fa ifp
@ -1168,7 +1199,10 @@ function.)
The
.Fa ifp
must have been allocated by
.Fn if_alloc .
.Fn if_alloc ,
.Fn if_alloc_dev
or
.Fn if_alloc_domain .
.It Fn if_detach
Shut down and unlink the specified
.Fa ifp

View File

@ -1636,7 +1636,7 @@ cxgbe_vi_attach(device_t dev, struct vi_info *vi)
callout_init(&vi->tick, 1);
/* Allocate an ifnet and set it up */
ifp = if_alloc(IFT_ETHER);
ifp = if_alloc_dev(IFT_ETHER, dev);
if (ifp == NULL) {
device_printf(dev, "Cannot allocate ifnet\n");
return (ENOMEM);

View File

@ -3682,7 +3682,7 @@ mlx5e_create_ifp(struct mlx5_core_dev *mdev)
M_MLX5EN, M_WAITOK | M_ZERO);
mlx5e_priv_mtx_init(priv);
ifp = priv->ifp = if_alloc(IFT_ETHER);
ifp = priv->ifp = if_alloc_dev(IFT_ETHER, mdev->pdev->dev.bsddev);
if (ifp == NULL) {
mlx5_core_err(mdev, "if_alloc() failed\n");
goto err_free_priv;

View File

@ -39,6 +39,7 @@
#include <sys/types.h>
#include <sys/conf.h>
#include <sys/malloc.h>
#include <sys/domainset.h>
#include <sys/sbuf.h>
#include <sys/bus.h>
#include <sys/epoch.h>
@ -526,13 +527,19 @@ if_grow(void)
* registered for the passed type.
*/
struct ifnet *
if_alloc(u_char type)
if_alloc_domain(u_char type, int numa_domain)
{
struct ifnet *ifp;
u_short idx;
void *old;
ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
if (numa_domain == IF_NODOM)
ifp = malloc(sizeof(struct ifnet), M_IFNET,
M_WAITOK | M_ZERO);
else
ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
restart:
IFNET_WLOCK();
idx = ifindex_alloc(&old);
@ -547,6 +554,7 @@ if_alloc(u_char type)
ifp->if_index = idx;
ifp->if_type = type;
ifp->if_alloctype = type;
ifp->if_numa_domain = numa_domain;
#ifdef VIMAGE
ifp->if_vnet = curvnet;
#endif
@ -580,6 +588,22 @@ if_alloc(u_char type)
return (ifp);
}
struct ifnet *
if_alloc_dev(u_char type, device_t dev)
{
int numa_domain;
if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
return (if_alloc_domain(type, IF_NODOM));
return (if_alloc_domain(type, numa_domain));
}
struct ifnet *
if_alloc(u_char type)
{
return (if_alloc_domain(type, IF_NODOM));
}
/*
* Do the actual work of freeing a struct ifnet, and layer 2 common
* structure. This call is made when the last reference to an
@ -608,7 +632,10 @@ if_free_internal(struct ifnet *ifp)
free(ifp->if_description, M_IFDESCR);
free(ifp->if_hw_addr, M_IFADDR);
free(ifp, M_IFNET);
if (ifp->if_numa_domain == IF_NODOM)
free(ifp, M_IFNET);
else
free_domain(ifp, M_IFNET);
}
static void

View File

@ -244,7 +244,7 @@ struct ifnet {
CK_STAILQ_HEAD(, ifg_list) if_groups; /* linked list of groups per if (CK_) */
/* protected by if_addr_lock */
u_char if_alloctype; /* if_type at time of allocation */
uint8_t if_numa_domain; /* NUMA domain of device */
/* Driver and protocol specific information that remains stable. */
void *if_softc; /* pointer to driver state */
void *if_llsoftc; /* link layer softc */
@ -394,6 +394,7 @@ struct ifnet {
/* for compatibility with other BSDs */
#define if_name(ifp) ((ifp)->if_xname)
#define IF_NODOM 255
/*
* Locks for address lists on the network interface.
*/
@ -622,6 +623,8 @@ int if_delgroup(struct ifnet *, const char *);
int if_addmulti(struct ifnet *, struct sockaddr *, struct ifmultiaddr **);
int if_allmulti(struct ifnet *, int);
struct ifnet* if_alloc(u_char);
struct ifnet* if_alloc_dev(u_char, device_t dev);
struct ifnet* if_alloc_domain(u_char, int numa_domain);
void if_attach(struct ifnet *);
void if_dead(struct ifnet *);
int if_delmulti(struct ifnet *, struct sockaddr *);