Sneed-Group/gems-kernel
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
gems-kernel/source/THIRDPARTY/xnu/bsd/skywalk/nexus/nexus_kern.c
Sam Sneed f5727805f2 add darwin/xnu functionality
2024-06-03 11:29:39 -05:00

2347 lines
61 KiB
C
Raw Blame History

/*
* Copyright (c) 2015-2021 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* compliance with the License. The rights granted to you under the License
* may not be used to create, or enable the creation or redistribution of,
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
*
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
*
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
#include <skywalk/os_skywalk_private.h>
#include <skywalk/nexus/upipe/nx_user_pipe.h>
#include <skywalk/nexus/kpipe/nx_kernel_pipe.h>
#include <skywalk/nexus/flowswitch/nx_flowswitch.h>
#include <skywalk/nexus/netif/nx_netif.h>
#include <skywalk/nexus/monitor/nx_monitor.h>
static STAILQ_HEAD(, nxdom) nexus_domains =
STAILQ_HEAD_INITIALIZER(nexus_domains);
static void nxdom_attach(struct nxdom *);
static void nxdom_detach(struct nxdom *);
static void nxdom_init(struct nxdom *);
static void nxdom_terminate(struct nxdom *);
static void nxdom_fini(struct nxdom *);
static void nxdom_del_provider_final(struct kern_nexus_domain_provider *);
static int nxdom_prov_ext_init(struct kern_nexus_domain_provider *);
static void nxdom_prov_ext_fini(struct kern_nexus_domain_provider *);
static struct kern_nexus_domain_provider *nxdom_prov_alloc(zalloc_flags_t);
static void nxdom_prov_free(struct kern_nexus_domain_provider *);
static uint32_t nxprov_bound_var(uint32_t *, uint32_t, uint32_t, uint32_t,
const char *);
static void nxprov_detaching_enqueue(struct kern_nexus_domain_provider *);
static struct kern_nexus_domain_provider *nxprov_detaching_dequeue(void);
static void nxprov_detacher(void *, wait_result_t);
static int nxprov_detacher_cont(int);
static struct nexus_controller *ncd_alloc(zalloc_flags_t);
static void ncd_free(struct nexus_controller *);
static struct nexus_attr *nxa_alloc(zalloc_flags_t);
static void nxa_free(struct nexus_attr *);
static int _kern_nexus_ifattach(struct nxctl *nxctl, const uuid_t nx_uuid,
struct ifnet *ifp, const uuid_t nx_uuid_attachee, boolean_t host,
uuid_t *nx_if_uuid);
static SKMEM_TYPE_DEFINE(ncd_zone, struct nexus_controller);
static SKMEM_TYPE_DEFINE(nxdom_prov_zone, struct kern_nexus_domain_provider);
static SKMEM_TYPE_DEFINE(nxa_zone, struct nexus_attr);
static int __nxdom_inited = 0;
static STAILQ_HEAD(, kern_nexus_domain_provider) nxprov_detaching_head =
STAILQ_HEAD_INITIALIZER(nxprov_detaching_head);
static uint32_t nxprov_detaching_cnt;
static void *nxprov_detach_wchan; /* wait channel for detacher */
/*
* Array of default nexus domain providers. Initialized once during
* domain attach time; no lock is needed to read as they can be treated
* as immutables, since default providers imply built-in ones and they
* never detach in practice.
*/
struct kern_nexus_domain_provider *nxdom_prov_default[NEXUS_TYPE_MAX];
void
nxdom_attach_all(void)
{
struct nxdom *nxdom;
thread_t tp = THREAD_NULL;
SK_LOCK_ASSERT_HELD();
ASSERT(!__nxdom_inited);
ASSERT(STAILQ_EMPTY(&nexus_domains));
#if CONFIG_NEXUS_FLOWSWITCH
nxdom_attach(&nx_flowswitch_dom_s);
#endif /* CONFIG_NEXUS_FLOWSWITCH */
#if CONFIG_NEXUS_USER_PIPE
nxdom_attach(&nx_upipe_dom_s);
#endif /* CONFIG_NEXUS_USER_PIPE */
#if CONFIG_NEXUS_KERNEL_PIPE
nxdom_attach(&nx_kpipe_dom_s);
#endif /* CONFIG_NEXUS_KERNEL_PIPE */
#if CONFIG_NEXUS_NETIF
nxdom_attach(&nx_netif_dom_s);
#endif /* CONFIG_NEXUS_NETIF */
#if CONFIG_NEXUS_MONITOR
nxdom_attach(&nx_monitor_dom_s);
#endif /* CONFIG_NEXUS_MONITOR */
/* ask domains to initialize */
STAILQ_FOREACH(nxdom, &nexus_domains, nxdom_link)
nxdom_init(nxdom);
if (kernel_thread_start(nxprov_detacher, NULL, &tp) != KERN_SUCCESS) {
panic_plain("%s: couldn't create detacher thread", __func__);
/* NOTREACHED */
__builtin_unreachable();
}
thread_deallocate(tp);
__nxdom_inited = 1;
}
void
nxdom_detach_all(void)
{
struct nxdom *nxdom, *tnxdom;
SK_LOCK_ASSERT_HELD();
if (__nxdom_inited) {
STAILQ_FOREACH_SAFE(nxdom, &nexus_domains, nxdom_link, tnxdom) {
nxdom_terminate(nxdom);
nxdom_fini(nxdom);
nxdom_detach(nxdom);
}
/*
* TODO: adi@apple.com -- terminate detacher thread.
*/
__nxdom_inited = 0;
}
ASSERT(STAILQ_EMPTY(&nexus_domains));
}
#define ASSERT_NXDOM_PARAMS(_dom, _var) do { \
ASSERT(NXDOM_MIN(_dom, _var) <= NXDOM_MAX(_dom, _var)); \
ASSERT(NXDOM_DEF(_dom, _var) >= NXDOM_MIN(_dom, _var)); \
ASSERT(NXDOM_DEF(_dom, _var) <= NXDOM_MAX(_dom, _var)); \
} while (0)
static void
nxdom_attach(struct nxdom *nxdom)
{
struct nxdom *nxdom1;
SK_LOCK_ASSERT_HELD();
ASSERT(!(nxdom->nxdom_flags & NEXUSDOMF_ATTACHED));
STAILQ_FOREACH(nxdom1, &nexus_domains, nxdom_link) {
if (nxdom1->nxdom_type == nxdom->nxdom_type) {
/* type must be unique; this is a programming error */
VERIFY(0);
/* NOTREACHED */
__builtin_unreachable();
}
}
/* verify this is a valid type */
switch (nxdom->nxdom_type) {
case NEXUS_TYPE_USER_PIPE:
case NEXUS_TYPE_KERNEL_PIPE:
case NEXUS_TYPE_NET_IF:
case NEXUS_TYPE_FLOW_SWITCH:
case NEXUS_TYPE_MONITOR:
break;
default:
VERIFY(0);
/* NOTREACHED */
__builtin_unreachable();
}
/* verify this is a valid metadata type */
switch (nxdom->nxdom_md_type) {
case NEXUS_META_TYPE_QUANTUM:
case NEXUS_META_TYPE_PACKET:
break;
default:
VERIFY(0);
/* NOTREACHED */
__builtin_unreachable();
}
/* verify this is a valid metadata subtype */
switch (nxdom->nxdom_md_subtype) {
case NEXUS_META_SUBTYPE_PAYLOAD:
case NEXUS_META_SUBTYPE_RAW:
break;
default:
VERIFY(0);
/* NOTREACHED */
__builtin_unreachable();
}
#if (DEVELOPMENT || DEBUG)
/*
* Override the default ring sizes for flowswitch if configured
* via boot-args. Each nexus provider instance can still change
* the values if so desired.
*/
if (nxdom->nxdom_type == NEXUS_TYPE_FLOW_SWITCH) {
if (sk_txring_sz != 0) {
if (sk_txring_sz < NXDOM_MIN(nxdom, tx_slots)) {
sk_txring_sz = NXDOM_MIN(nxdom, tx_slots);
} else if (sk_txring_sz > NXDOM_MAX(nxdom, tx_slots)) {
sk_txring_sz = NXDOM_MAX(nxdom, tx_slots);
}
NXDOM_DEF(nxdom, tx_slots) = sk_txring_sz;
}
if (sk_rxring_sz != 0) {
if (sk_rxring_sz < NXDOM_MIN(nxdom, rx_slots)) {
sk_rxring_sz = NXDOM_MIN(nxdom, rx_slots);
} else if (sk_rxring_sz > NXDOM_MAX(nxdom, rx_slots)) {
sk_rxring_sz = NXDOM_MAX(nxdom, rx_slots);
}
NXDOM_DEF(nxdom, rx_slots) = sk_rxring_sz;
}
}
/*
* Override the default ring sizes for netif if configured
* via boot-args. Each nexus provider instance can still change
* the values if so desired.
*/
if (nxdom->nxdom_type == NEXUS_TYPE_NET_IF) {
if (sk_net_txring_sz != 0) {
if (sk_net_txring_sz < NXDOM_MIN(nxdom, tx_slots)) {
sk_net_txring_sz = NXDOM_MIN(nxdom, tx_slots);
} else if (sk_net_txring_sz > NXDOM_MAX(nxdom, tx_slots)) {
sk_net_txring_sz = NXDOM_MAX(nxdom, tx_slots);
}
NXDOM_DEF(nxdom, tx_slots) = sk_net_txring_sz;
}
if (sk_net_rxring_sz != 0) {
if (sk_net_rxring_sz < NXDOM_MIN(nxdom, rx_slots)) {
sk_net_rxring_sz = NXDOM_MIN(nxdom, rx_slots);
} else if (sk_net_rxring_sz > NXDOM_MAX(nxdom, rx_slots)) {
sk_net_rxring_sz = NXDOM_MAX(nxdom, rx_slots);
}
NXDOM_DEF(nxdom, rx_slots) = sk_net_rxring_sz;
}
}
#endif /* DEVELOPMENT || DEBUG */
/* verify that parameters are sane */
ASSERT(NXDOM_MAX(nxdom, ports) > 0);
ASSERT(NXDOM_MAX(nxdom, ports) <= NEXUS_PORT_MAX);
ASSERT_NXDOM_PARAMS(nxdom, ports);
ASSERT_NXDOM_PARAMS(nxdom, tx_rings);
ASSERT_NXDOM_PARAMS(nxdom, rx_rings);
ASSERT(NXDOM_MAX(nxdom, tx_slots) > 0);
ASSERT_NXDOM_PARAMS(nxdom, tx_slots);
ASSERT(NXDOM_MAX(nxdom, rx_slots) > 0);
ASSERT_NXDOM_PARAMS(nxdom, rx_slots);
ASSERT_NXDOM_PARAMS(nxdom, buf_size);
ASSERT_NXDOM_PARAMS(nxdom, meta_size);
ASSERT_NXDOM_PARAMS(nxdom, pipes);
ASSERT_NXDOM_PARAMS(nxdom, extensions);
/* these must exist */
ASSERT(nxdom->nxdom_bind_port != NULL);
ASSERT(nxdom->nxdom_unbind_port != NULL);
ASSERT(nxdom->nxdom_connect != NULL);
ASSERT(nxdom->nxdom_disconnect != NULL);
ASSERT(nxdom->nxdom_defunct != NULL);
ASSERT(nxdom->nxdom_defunct_finalize != NULL);
STAILQ_INSERT_TAIL(&nexus_domains, nxdom, nxdom_link);
nxdom->nxdom_flags |= NEXUSDOMF_ATTACHED;
}
#undef VERIFY_NXDOM_PARAMS
static void
nxdom_detach(struct nxdom *nxdom)
{
SK_LOCK_ASSERT_HELD();
ASSERT(nxdom->nxdom_flags & NEXUSDOMF_ATTACHED);
STAILQ_REMOVE(&nexus_domains, nxdom, nxdom, nxdom_link);
nxdom->nxdom_flags &= ~NEXUSDOMF_ATTACHED;
}
static void
nxdom_init(struct nxdom *nxdom)
{
ASSERT(nxdom->nxdom_flags & NEXUSDOMF_ATTACHED);
SK_LOCK_ASSERT_HELD();
if (!(nxdom->nxdom_flags & NEXUSDOMF_INITIALIZED)) {
if (nxdom->nxdom_init != NULL) {
nxdom->nxdom_init(nxdom);
}
nxdom->nxdom_flags |= NEXUSDOMF_INITIALIZED;
}
}
static void
nxdom_terminate(struct nxdom *nxdom)
{
ASSERT(nxdom->nxdom_flags & NEXUSDOMF_ATTACHED);
SK_LOCK_ASSERT_HELD();
if ((nxdom->nxdom_flags & NEXUSDOMF_INITIALIZED) &&
!(nxdom->nxdom_flags & NEXUSDOMF_TERMINATED)) {
if (nxdom->nxdom_terminate != NULL) {
nxdom->nxdom_terminate(nxdom);
}
nxdom->nxdom_flags |= NEXUSDOMF_TERMINATED;
}
}
static void
nxdom_fini(struct nxdom *nxdom)
{
ASSERT(nxdom->nxdom_flags & NEXUSDOMF_ATTACHED);
if (nxdom->nxdom_flags & NEXUSDOMF_INITIALIZED) {
if (nxdom->nxdom_fini != NULL) {
nxdom->nxdom_fini(nxdom);
}
nxdom->nxdom_flags &= ~NEXUSDOMF_INITIALIZED;
}
}
int
nxdom_prov_add(struct nxdom *nxdom,
struct kern_nexus_domain_provider *nxdom_prov)
{
struct kern_nexus_domain_provider *nxprov1;
nexus_type_t type = nxdom->nxdom_type;
boolean_t builtin;
int err = 0;
SK_LOCK_ASSERT_HELD();
ASSERT(type < NEXUS_TYPE_MAX);
builtin = !(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT);
STAILQ_FOREACH(nxprov1, &nxdom->nxdom_prov_head, nxdom_prov_link) {
/*
* We can be a little more strict in the kernel and
* avoid namespace collision (even though each domain
* provider has UUID; this also guarantees that external
* providers won't conflict with the builtin ones.
*/
if (strcmp(nxprov1->nxdom_prov_name,
nxdom_prov->nxdom_prov_name) == 0) {
return EEXIST;
}
}
VERIFY(!(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_ATTACHED));
VERIFY(!(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_INITIALIZED));
uuid_generate_random(nxdom_prov->nxdom_prov_uuid);
nxdom_prov->nxdom_prov_dom = nxdom;
if (nxdom_prov->nxdom_prov_init != NULL) {
err = nxdom_prov->nxdom_prov_init(nxdom_prov);
}
if (err == 0) {
nxdom_prov->nxdom_prov_flags |=
(NXDOMPROVF_ATTACHED | NXDOMPROVF_INITIALIZED);
STAILQ_INSERT_TAIL(&nxdom->nxdom_prov_head, nxdom_prov,
nxdom_prov_link);
/* for being in the list */
nxdom_prov_retain_locked(nxdom_prov);
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_DEFAULT) {
VERIFY(builtin && nxdom_prov_default[type] == NULL);
nxdom_prov_default[type] = nxdom_prov;
/* for being in the array */
nxdom_prov_retain_locked(nxdom_prov);
}
SK_D("nxdom_prov 0x%llx (%s) dom %s",
SK_KVA(nxdom_prov), nxdom_prov->nxdom_prov_name,
nxdom->nxdom_name);
} else {
uuid_clear(nxdom_prov->nxdom_prov_uuid);
nxdom_prov->nxdom_prov_dom = NULL;
}
return err;
}
void
nxdom_prov_del(struct kern_nexus_domain_provider *nxdom_prov)
{
struct nxdom *nxdom = nxdom_prov->nxdom_prov_dom;
nexus_type_t type = nxdom->nxdom_type;
SK_LOCK_ASSERT_HELD();
ASSERT(type < NEXUS_TYPE_MAX);
ASSERT(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_ATTACHED);
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_DETACHING) {
return;
}
SK_D("nxdom_prov 0x%llx (%s:%s)", SK_KVA(nxdom_prov), nxdom->nxdom_name,
nxdom_prov->nxdom_prov_name);
/* keep the reference around for the detaching list (see below) */
STAILQ_REMOVE(&nxdom->nxdom_prov_head, nxdom_prov,
kern_nexus_domain_provider, nxdom_prov_link);
nxdom_prov->nxdom_prov_flags &= ~NXDOMPROVF_ATTACHED;
nxdom_prov->nxdom_prov_flags |= NXDOMPROVF_DETACHING;
/* there can only be one default and it must match this one */
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_DEFAULT) {
ASSERT(!(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT));
VERIFY(nxdom_prov_default[type] == nxdom_prov);
nxdom_prov_default[type] = NULL;
/*
* Release reference held for the array; this must
* not be the last reference, as there is still at
* least one which we kept for the detaching list.
*/
VERIFY(!nxdom_prov_release_locked(nxdom_prov));
}
/* add to detaching list and wake up detacher */
nxprov_detaching_enqueue(nxdom_prov);
}
static void
nxdom_del_provider_final(struct kern_nexus_domain_provider *nxdom_prov)
{
#if (DEBUG || DEVELOPMENT)
struct nxdom *nxdom = nxdom_prov->nxdom_prov_dom;
#endif /* DEBUG || DEVELOPMENT */
SK_LOCK_ASSERT_HELD();
ASSERT((nxdom_prov->nxdom_prov_flags & (NXDOMPROVF_ATTACHED |
NXDOMPROVF_DETACHING)) == NXDOMPROVF_DETACHING);
ASSERT(nxdom != NULL);
SK_D("nxdom_prov 0x%llx (%s:%s)", SK_KVA(nxdom_prov), nxdom->nxdom_name,
nxdom_prov->nxdom_prov_name);
nxdom_prov->nxdom_prov_flags &= ~NXDOMPROVF_DETACHING;
/*
* Release reference held for detaching list; if this is the last
* reference, the domain provider's nxdom_prov_fini() callback will
* be called (if applicable) within the detacher thread's context.
* Otherwise, this will occur when the last nexus provider for that
* domain provider has been released.
*/
(void) nxdom_prov_release_locked(nxdom_prov);
}
struct nxdom *
nxdom_find(nexus_type_t type)
{
struct nxdom *nxdom;
SK_LOCK_ASSERT_HELD();
ASSERT(type < NEXUS_TYPE_MAX);
STAILQ_FOREACH(nxdom, &nexus_domains, nxdom_link) {
if (nxdom->nxdom_type == type) {
break;
}
}
return nxdom;
}
struct kern_nexus_domain_provider *
nxdom_prov_find(const struct nxdom *nxdom, const char *name)
{
struct kern_nexus_domain_provider *nxdom_prov = NULL;
SK_LOCK_ASSERT_HELD();
if (name != NULL) {
STAILQ_FOREACH(nxdom_prov, &nxdom->nxdom_prov_head,
nxdom_prov_link) {
if (strcmp(nxdom_prov->nxdom_prov_name, name) == 0) {
break;
}
}
}
if (nxdom_prov != NULL) {
nxdom_prov_retain_locked(nxdom_prov); /* for caller */
}
return nxdom_prov;
}
struct kern_nexus_domain_provider *
nxdom_prov_find_uuid(const uuid_t dom_prov_uuid)
{
struct kern_nexus_domain_provider *nxdom_prov = NULL;
struct nxdom *nxdom;
SK_LOCK_ASSERT_HELD();
ASSERT(dom_prov_uuid != NULL && !uuid_is_null(dom_prov_uuid));
STAILQ_FOREACH(nxdom, &nexus_domains, nxdom_link) {
STAILQ_FOREACH(nxdom_prov, &nxdom->nxdom_prov_head,
nxdom_prov_link) {
ASSERT(!uuid_is_null(nxdom_prov->nxdom_prov_uuid));
if (uuid_compare(nxdom_prov->nxdom_prov_uuid,
dom_prov_uuid) == 0) {
break;
}
}
if (nxdom_prov != NULL) {
nxdom_prov_retain_locked(nxdom_prov); /* for caller */
break;
}
}
return nxdom_prov;
}
errno_t
kern_nexus_register_domain_provider(const nexus_type_t type,
const nexus_domain_provider_name_t name,
const struct kern_nexus_domain_provider_init *init,
const uint32_t init_len, uuid_t *dom_prov_uuid)
{
struct kern_nexus_domain_provider *nxdom_prov = NULL;
struct nxdom *nxdom;
errno_t err = 0;
_CASSERT(sizeof(*init) == sizeof(nxdom_prov->nxdom_prov_ext));
if (type >= NEXUS_TYPE_MAX || dom_prov_uuid == NULL) {
return EINVAL;
}
uuid_clear(*dom_prov_uuid);
if (name == NULL || init == NULL || init_len < sizeof(*init) ||
init->nxdpi_version != KERN_NEXUS_DOMAIN_PROVIDER_CURRENT_VERSION) {
return EINVAL;
}
/*
* init, fini are required.
*/
if (init->nxdpi_init == NULL || init->nxdpi_fini == NULL) {
return EINVAL;
}
SK_LOCK();
if (nxdom_prov_default[type] == NULL) {
err = ENXIO;
goto done;
}
nxdom = nxdom_find(type);
if (nxdom == NULL) {
err = ENXIO;
goto done;
}
/*
* Allow only kernel pipe and netif external domain providers for
* now, until we understand the implications and requirements for
* supporting other domain types. For all other types, using
* the built-in domain providers and registering nexus should
* suffice.
*/
if (nxdom->nxdom_type != NEXUS_TYPE_KERNEL_PIPE &&
nxdom->nxdom_type != NEXUS_TYPE_NET_IF) {
err = EINVAL;
goto done;
}
nxdom_prov = nxdom_prov_alloc(Z_WAITOK);
/*
* Point all callback routines to the default provider for this
* domain; for nxdom_prov{init,fini}, refer to externally-provided
* callback routines, if applicable.
*/
bcopy(init, &nxdom_prov->nxdom_prov_ext, sizeof(*init));
bcopy(&nxdom_prov_default[type]->nxdom_prov_cb,
&nxdom_prov->nxdom_prov_cb, sizeof(struct nxdom_prov_cb));
nxdom_prov->nxdom_prov_flags |= NXDOMPROVF_EXT;
nxdom_prov->nxdom_prov_init = nxdom_prov_ext_init;
nxdom_prov->nxdom_prov_fini = nxdom_prov_ext_fini;
(void) snprintf(nxdom_prov->nxdom_prov_name,
sizeof(nxdom_prov->nxdom_prov_name), "%s", name);
ASSERT(!(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_DEFAULT));
err = nxdom_prov_add(nxdom, nxdom_prov);
if (err != 0) {
nxdom_prov_free(nxdom_prov);
nxdom_prov = NULL;
}
done:
if (nxdom_prov != NULL) {
ASSERT(err == 0 && !uuid_is_null(nxdom_prov->nxdom_prov_uuid));
uuid_copy(*dom_prov_uuid, nxdom_prov->nxdom_prov_uuid);
}
SK_UNLOCK();
return err;
}
errno_t
kern_nexus_deregister_domain_provider(const uuid_t dom_prov_uuid)
{
struct kern_nexus_domain_provider *nxdom_prov = NULL;
errno_t err = 0;
if (dom_prov_uuid == NULL || uuid_is_null(dom_prov_uuid)) {
return EINVAL;
}
SK_LOCK();
nxdom_prov = nxdom_prov_find_uuid(dom_prov_uuid);
if (nxdom_prov == NULL) {
err = ENXIO;
goto done;
}
/* don't allow external request for built-in domain providers */
if (!(nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT)) {
err = EINVAL;
goto done;
}
/* schedule this to be deleted */
nxdom_prov_del(nxdom_prov);
done:
/* release reference from nxdom_prov_find_uuid */
if (nxdom_prov != NULL) {
(void) nxdom_prov_release_locked(nxdom_prov);
}
SK_UNLOCK();
return err;
}
errno_t
kern_nexus_get_default_domain_provider(const nexus_type_t type,
uuid_t *dom_prov_uuid)
{
struct kern_nexus_domain_provider *nxdom_prov;
if (type >= NEXUS_TYPE_MAX || dom_prov_uuid == NULL) {
return EINVAL;
}
uuid_clear(*dom_prov_uuid);
/* no lock is needed; array is immutable */
if ((nxdom_prov = nxdom_prov_default[type]) == NULL) {
return ENXIO;
}
uuid_copy(*dom_prov_uuid, nxdom_prov->nxdom_prov_uuid);
return 0;
}
static int
nxdom_prov_ext_init(struct kern_nexus_domain_provider *nxdom_prov)
{
int err = 0;
SK_D("initializing %s", nxdom_prov->nxdom_prov_name);
ASSERT(nxdom_prov->nxdom_prov_ext.nxdpi_init != NULL);
if ((err = nxdom_prov->nxdom_prov_ext.nxdpi_init(nxdom_prov)) == 0) {
nxdom_prov->nxdom_prov_flags |= NXDOMPROVF_EXT_INITED;
}
return err;
}
static void
nxdom_prov_ext_fini(struct kern_nexus_domain_provider *nxdom_prov)
{
SK_D("destroying %s", nxdom_prov->nxdom_prov_name);
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT_INITED) {
ASSERT(nxdom_prov->nxdom_prov_ext.nxdpi_fini != NULL);
nxdom_prov->nxdom_prov_ext.nxdpi_fini(nxdom_prov);
nxdom_prov->nxdom_prov_flags &= ~NXDOMPROVF_EXT_INITED;
}
}
static struct nexus_attr *
nxa_alloc(zalloc_flags_t how)
{
return zalloc_flags(nxa_zone, how | Z_ZERO);
}
static void
nxa_free(struct nexus_attr *nxa)
{
SK_DF(SK_VERB_MEM, "nxa 0x%llx FREE", SK_KVA(nxa));
zfree(nxa_zone, nxa);
}
errno_t
kern_nexus_attr_create(nexus_attr_t *nxa)
{
errno_t err = 0;
if (nxa == NULL) {
err = EINVAL;
} else {
*nxa = nxa_alloc(Z_WAITOK);
}
return err;
}
errno_t
kern_nexus_attr_clone(const nexus_attr_t nxa, nexus_attr_t *nnxa)
{
errno_t err = 0;
if (nnxa == NULL) {
err = EINVAL;
} else {
err = kern_nexus_attr_create(nnxa);
if (err == 0 && nxa != NULL) {
ASSERT(*nnxa != NULL);
bcopy(nxa, *nnxa, sizeof(**nnxa));
}
}
return err;
}
errno_t
kern_nexus_attr_set(const nexus_attr_t nxa,
const nexus_attr_type_t type, const uint64_t value)
{
return __nexus_attr_set(nxa, type, value);
}
errno_t
kern_nexus_attr_get(nexus_attr_t nxa, const nexus_attr_type_t type,
uint64_t *value)
{
return __nexus_attr_get(nxa, type, value);
}
void
kern_nexus_attr_destroy(nexus_attr_t nxa)
{
nxa_free(nxa);
}
static struct nexus_controller *
ncd_alloc(zalloc_flags_t how)
{
return zalloc_flags(ncd_zone, how | Z_ZERO);
}
static void
ncd_free(struct nexus_controller *ncd)
{
SK_DF(SK_VERB_MEM, "ncd 0x%llx FREE", SK_KVA(ncd));
zfree(ncd_zone, ncd);
}
nexus_controller_t
kern_nexus_shared_controller(void)
{
return &kernnxctl;
}
errno_t
kern_nexus_controller_create(nexus_controller_t *ncd)
{
struct nxctl *nxctl = NULL;
uuid_t nxctl_uuid;
errno_t err = 0;
uuid_generate_random(nxctl_uuid);
if (ncd == NULL) {
err = EINVAL;
goto done;
} else {
*ncd = NULL;
}
nxctl = nxctl_create(kernproc, NULL, nxctl_uuid, &err);
if (nxctl == NULL) {
ASSERT(err != 0);
goto done;
}
*ncd = ncd_alloc(Z_WAITOK);
(*ncd)->ncd_nxctl = nxctl; /* ref from nxctl_create */
done:
if (err != 0) {
if (nxctl != NULL) {
nxctl_dtor(nxctl);
nxctl = NULL;
}
if (ncd != NULL && *ncd != NULL) {
ncd_free(*ncd);
*ncd = NULL;
}
}
return err;
}
#define NXPI_INVALID_CB_PAIRS(cb1, cb2) \
(!(init->nxpi_##cb1 == NULL && init->nxpi_##cb2 == NULL) && \
((init->nxpi_##cb1 == NULL) ^ (init->nxpi_##cb2 == NULL)))
static errno_t
nexus_controller_register_provider_validate_init_params(
const struct kern_nexus_provider_init *init, uint32_t init_len,
nexus_type_t nxdom_type)
{
errno_t err = 0;
struct kern_nexus_netif_provider_init *netif_init;
_CASSERT(__builtin_offsetof(struct kern_nexus_provider_init,
nxpi_version) == 0);
_CASSERT(sizeof(init->nxpi_version) == sizeof(uint32_t));
if (init == NULL) {
return 0;
}
if (init_len < sizeof(uint32_t)) {
return EINVAL;
}
switch (init->nxpi_version) {
case KERN_NEXUS_PROVIDER_VERSION_1:
if (init_len != sizeof(struct kern_nexus_provider_init)) {
err = EINVAL;
break;
}
/*
* sync_{tx,rx} callbacks are required; the rest of the
* callback pairs are optional, but must be symmetrical.
*/
if (init->nxpi_sync_tx == NULL || init->nxpi_sync_rx == NULL ||
init->nxpi_pre_connect == NULL ||
init->nxpi_connected == NULL ||
init->nxpi_pre_disconnect == NULL ||
init->nxpi_disconnected == NULL ||
NXPI_INVALID_CB_PAIRS(ring_init, ring_fini) ||
NXPI_INVALID_CB_PAIRS(slot_init, slot_fini)) {
err = EINVAL;
break;
}
/*
* Tx doorbell interface is only supported for netif and
* Tx doorbell is mandatory for netif
*/
if (((init->nxpi_tx_doorbell != NULL) &&
(nxdom_type != NEXUS_TYPE_NET_IF)) ||
((nxdom_type == NEXUS_TYPE_NET_IF) &&
(init->nxpi_tx_doorbell == NULL))) {
err = EINVAL;
break;
}
/*
* Capabilities configuration interface is only supported for
* netif.
*/
if ((init->nxpi_config_capab != NULL) &&
(nxdom_type != NEXUS_TYPE_NET_IF)) {
err = EINVAL;
break;
}
break;
case KERN_NEXUS_PROVIDER_VERSION_NETIF:
if (init_len != sizeof(struct kern_nexus_netif_provider_init)) {
err = EINVAL;
break;
}
if (nxdom_type != NEXUS_TYPE_NET_IF) {
err = EINVAL;
break;
}
netif_init =
__DECONST(struct kern_nexus_netif_provider_init *, init);
if (netif_init->nxnpi_pre_connect == NULL ||
netif_init->nxnpi_connected == NULL ||
netif_init->nxnpi_pre_disconnect == NULL ||
netif_init->nxnpi_disconnected == NULL ||
netif_init->nxnpi_qset_init == NULL ||
netif_init->nxnpi_qset_fini == NULL ||
netif_init->nxnpi_queue_init == NULL ||
netif_init->nxnpi_queue_fini == NULL ||
netif_init->nxnpi_tx_qset_notify == NULL ||
netif_init->nxnpi_config_capab == NULL) {
err = EINVAL;
break;
}
break;
default:
err = EINVAL;
break;
}
return err;
}
errno_t
kern_nexus_controller_register_provider(const nexus_controller_t ncd,
const uuid_t dom_prov_uuid, const nexus_name_t name,
const struct kern_nexus_provider_init *init, uint32_t init_len,
const nexus_attr_t nxa, uuid_t *prov_uuid)
{
struct kern_nexus_domain_provider *nxdom_prov = NULL;
struct kern_nexus_provider *nxprov = NULL;
nexus_type_t nxdom_type;
struct nxprov_reg reg;
struct nxctl *nxctl;
errno_t err = 0;
if (prov_uuid == NULL) {
return EINVAL;
}
uuid_clear(*prov_uuid);
if (ncd == NULL ||
dom_prov_uuid == NULL || uuid_is_null(dom_prov_uuid)) {
return EINVAL;
}
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
SK_LOCK();
nxdom_prov = nxdom_prov_find_uuid(dom_prov_uuid);
if (nxdom_prov == NULL) {
SK_UNLOCK();
err = ENXIO;
goto done;
}
nxdom_type = nxdom_prov->nxdom_prov_dom->nxdom_type;
ASSERT(nxdom_type < NEXUS_TYPE_MAX);
err = nexus_controller_register_provider_validate_init_params(init,
init_len, nxdom_type);
if (err != 0) {
SK_UNLOCK();
err = EINVAL;
goto done;
}
if ((err = __nexus_provider_reg_prepare(&reg, name,
nxdom_type, nxa)) != 0) {
SK_UNLOCK();
goto done;
}
if (init && init->nxpi_version == KERN_NEXUS_PROVIDER_VERSION_NETIF) {
reg.nxpreg_params.nxp_flags |= NXPF_NETIF_LLINK;
}
/* callee will hold reference on nxdom_prov upon success */
if ((nxprov = nxprov_create_kern(nxctl, nxdom_prov, &reg,
init, &err)) == NULL) {
SK_UNLOCK();
ASSERT(err != 0);
goto done;
}
SK_UNLOCK();
uuid_copy(*prov_uuid, nxprov->nxprov_uuid);
done:
SK_LOCK_ASSERT_NOTHELD();
NXCTL_UNLOCK(nxctl);
if (err != 0 && nxprov != NULL) {
err = nxprov_close(nxprov, FALSE);
}
/* release extra ref from nxprov_create_kern */
if (nxprov != NULL) {
nxprov_release(nxprov);
}
/* release extra ref from nxdom_prov_find_uuid */
if (nxdom_prov != NULL) {
(void) nxdom_prov_release(nxdom_prov);
}
return err;
}
#undef NXPI_INVALID_CB_PAIRS
errno_t
kern_nexus_controller_deregister_provider(const nexus_controller_t ncd,
const uuid_t prov_uuid)
{
errno_t err;
if (ncd == NULL || prov_uuid == NULL || uuid_is_null(prov_uuid)) {
err = EINVAL;
} else {
struct nxctl *nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxprov_destroy(nxctl, prov_uuid);
NXCTL_UNLOCK(nxctl);
}
return err;
}
errno_t
kern_nexus_controller_alloc_provider_instance(const nexus_controller_t ncd,
const uuid_t prov_uuid, const void *nx_ctx,
nexus_ctx_release_fn_t nx_ctx_release, uuid_t *nx_uuid,
const struct kern_nexus_init *init)
{
struct kern_nexus *nx = NULL;
struct nxctl *nxctl;
errno_t err = 0;
if (ncd == NULL || prov_uuid == NULL || uuid_is_null(prov_uuid) ||
nx_uuid == NULL || init == NULL ||
init->nxi_version != KERN_NEXUS_CURRENT_VERSION ||
(init->nxi_rx_pbufpool != NULL &&
init->nxi_rx_pbufpool != init->nxi_tx_pbufpool)) {
err = EINVAL;
goto done;
}
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
nx = nx_create(nxctl, prov_uuid, NEXUS_TYPE_UNDEFINED, nx_ctx,
nx_ctx_release, init->nxi_tx_pbufpool, init->nxi_rx_pbufpool, &err);
NXCTL_UNLOCK(nxctl);
if (nx == NULL) {
ASSERT(err != 0);
goto done;
}
ASSERT(err == 0);
uuid_copy(*nx_uuid, nx->nx_uuid);
done:
/* release extra ref from nx_create */
if (nx != NULL) {
(void) nx_release(nx);
}
return err;
}
errno_t
kern_nexus_controller_alloc_net_provider_instance(
const nexus_controller_t ncd, const uuid_t prov_uuid, const void *nx_ctx,
nexus_ctx_release_fn_t nx_ctx_release, uuid_t *nx_uuid,
const struct kern_nexus_net_init *init, struct ifnet **pifp)
{
struct kern_nexus *nx = NULL;
struct ifnet *ifp = NULL;
struct nxctl *nxctl;
boolean_t nxctl_locked = FALSE;
errno_t err = 0;
if (ncd == NULL || prov_uuid == NULL || uuid_is_null(prov_uuid) ||
nx_uuid == NULL || init == NULL ||
init->nxneti_version != KERN_NEXUS_NET_CURRENT_VERSION ||
init->nxneti_eparams == NULL || pifp == NULL) {
err = EINVAL;
goto done;
}
/*
* Skywalk native interface doesn't support legacy model.
*/
if ((init->nxneti_eparams->start != NULL) ||
(init->nxneti_eparams->flags & IFNET_INIT_LEGACY) ||
(init->nxneti_eparams->flags & IFNET_INIT_INPUT_POLL)) {
err = EINVAL;
goto done;
}
/* create an embryonic ifnet */
err = ifnet_allocate_extended(init->nxneti_eparams, &ifp);
if (err != 0) {
goto done;
}
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
nxctl_locked = TRUE;
nx = nx_create(nxctl, prov_uuid, NEXUS_TYPE_NET_IF, nx_ctx,
nx_ctx_release, init->nxneti_tx_pbufpool, init->nxneti_rx_pbufpool,
&err);
if (nx == NULL) {
ASSERT(err != 0);
goto done;
}
if (NX_LLINK_PROV(nx)) {
if (init->nxneti_llink == NULL) {
SK_ERR("logical link configuration required");
err = EINVAL;
goto done;
}
err = nx_netif_default_llink_config(NX_NETIF_PRIVATE(nx),
init->nxneti_llink);
if (err != 0) {
goto done;
}
}
/* prepare this ifnet instance if needed */
if (init->nxneti_prepare != NULL) {
err = init->nxneti_prepare(nx, ifp);
if (err != 0) {
goto done;
}
}
/* attach embryonic ifnet to nexus */
err = _kern_nexus_ifattach(nxctl, nx->nx_uuid, ifp, NULL, FALSE, NULL);
if (err != 0) {
goto done;
}
/* and finalize the ifnet attach */
ASSERT(nxctl_locked);
NXCTL_UNLOCK(nxctl);
nxctl_locked = FALSE;
err = ifnet_attach(ifp, init->nxneti_lladdr);
if (err != 0) {
goto done;
}
ASSERT(err == 0);
/*
* Return ifnet reference held by ifnet_allocate_extended();
* caller is expected to retain this reference until its ifnet
* detach callback is called.
*/
*pifp = ifp;
uuid_copy(*nx_uuid, nx->nx_uuid);
done:
if (nxctl_locked) {
NXCTL_UNLOCK(nxctl);
}
/* release extra ref from nx_create */
if (nx != NULL) {
SK_LOCK();
if (err != 0) {
(void) nx_close(nx, TRUE);
}
(void) nx_release_locked(nx);
SK_UNLOCK();
}
if (err != 0 && ifp != NULL) {
ifnet_release(ifp);
}
return err;
}
errno_t
kern_nexus_controller_free_provider_instance(const nexus_controller_t ncd,
const uuid_t nx_uuid)
{
errno_t err;
if (ncd == NULL || nx_uuid == NULL || uuid_is_null(nx_uuid)) {
err = EINVAL;
} else {
struct nxctl *nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nx_destroy(nxctl, nx_uuid);
NXCTL_UNLOCK(nxctl);
}
return err;
}
errno_t
kern_nexus_controller_bind_provider_instance(const nexus_controller_t ncd,
const uuid_t nx_uuid, nexus_port_t *port, const pid_t pid,
const uuid_t exec_uuid, const void *key, const uint32_t key_len,
const uint32_t bind_flags)
{
struct nx_bind_req nbr;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (ncd == NULL || nx_uuid == NULL || uuid_is_null(nx_uuid) ||
port == NULL) {
return EINVAL;
}
__nexus_bind_req_prepare(&nbr, nx_uuid, *port, pid, exec_uuid,
key, key_len, bind_flags);
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_BIND;
sopt.sopt_val = (user_addr_t)&nbr;
sopt.sopt_valsize = sizeof(nbr);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_set_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
if (err == 0) {
*port = nbr.nb_port;
}
return err;
}
errno_t
kern_nexus_controller_unbind_provider_instance(const nexus_controller_t ncd,
const uuid_t nx_uuid, const nexus_port_t port)
{
struct nx_unbind_req nbu;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (ncd == NULL || nx_uuid == NULL || uuid_is_null(nx_uuid)) {
return EINVAL;
}
__nexus_unbind_req_prepare(&nbu, nx_uuid, port);
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_UNBIND;
sopt.sopt_val = (user_addr_t)&nbu;
sopt.sopt_valsize = sizeof(nbu);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_set_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
return err;
}
errno_t
kern_nexus_controller_read_provider_attr(const nexus_controller_t ncd,
const uuid_t prov_uuid, nexus_attr_t nxa)
{
struct nxprov_reg_ent nre;
struct nxprov_params *p = &nre.npre_prov_params;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (ncd == NULL || prov_uuid == NULL || uuid_is_null(prov_uuid) ||
nxa == NULL) {
return EINVAL;
}
bzero(&nre, sizeof(nre));
bcopy(prov_uuid, nre.npre_prov_uuid, sizeof(uuid_t));
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_GET;
sopt.sopt_name = NXOPT_NEXUS_PROV_ENTRY;
sopt.sopt_val = (user_addr_t)&nre;
sopt.sopt_valsize = sizeof(nre);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_get_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
if (err == 0) {
__nexus_attr_from_params(nxa, p);
}
return err;
}
void
kern_nexus_controller_destroy(nexus_controller_t ncd)
{
struct nxctl *nxctl;
if (ncd == NULL) {
return;
}
nxctl = ncd->ncd_nxctl;
ASSERT(nxctl != NULL);
ncd->ncd_nxctl = NULL;
nxctl_dtor(nxctl);
ncd_free(ncd);
}
void *
kern_nexus_get_context(const kern_nexus_t nx)
{
return nx->nx_ctx;
}
void
kern_nexus_stop(const kern_nexus_t nx)
{
SK_LOCK();
nx_stop(nx);
SK_UNLOCK();
}
errno_t
kern_nexus_get_pbufpool(const kern_nexus_t nx, kern_pbufpool_t *ptx_pp,
kern_pbufpool_t *prx_pp)
{
kern_pbufpool_t tpp = NULL, rpp = NULL;
int err = 0;
if (ptx_pp == NULL && prx_pp == NULL) {
return EINVAL;
}
if (NX_DOM_PROV(nx)->nxdom_prov_nx_mem_info == NULL) {
err = ENOTSUP;
} else {
err = NX_DOM_PROV(nx)->nxdom_prov_nx_mem_info(nx, &tpp, &rpp);
}
if (ptx_pp != NULL) {
*ptx_pp = tpp;
}
if (prx_pp != NULL) {
*prx_pp = rpp;
}
return err;
}
static int
_kern_nexus_ifattach(struct nxctl *nxctl, const uuid_t nx_uuid,
struct ifnet *ifp, const uuid_t nx_uuid_attachee, boolean_t host,
uuid_t *nx_if_uuid)
{
struct nx_cfg_req ncr;
struct nx_spec_req nsr;
struct sockopt sopt;
int err = 0;
NXCTL_LOCK_ASSERT_HELD(nxctl);
if (nx_uuid == NULL || uuid_is_null(nx_uuid)) {
return EINVAL;
}
bzero(&nsr, sizeof(nsr));
if (ifp != NULL) {
if (nx_uuid_attachee != NULL) {
return EINVAL;
}
nsr.nsr_flags = NXSPECREQ_IFP;
nsr.nsr_ifp = ifp;
} else {
if (nx_uuid_attachee == NULL) {
return EINVAL;
}
nsr.nsr_flags = NXSPECREQ_UUID;
if (host) {
nsr.nsr_flags |= NXSPECREQ_HOST;
}
uuid_copy(nsr.nsr_uuid, nx_uuid_attachee);
}
__nexus_config_req_prepare(&ncr, nx_uuid, NXCFG_CMD_ATTACH,
&nsr, sizeof(nsr));
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_CONFIG;
sopt.sopt_val = (user_addr_t)&ncr;
sopt.sopt_valsize = sizeof(ncr);
sopt.sopt_p = kernproc;
err = nxctl_set_opt(nxctl, &sopt);
if (err == 0 && nx_if_uuid != NULL) {
uuid_copy(*nx_if_uuid, nsr.nsr_if_uuid);
}
return err;
}
int
kern_nexus_ifattach(nexus_controller_t ncd, const uuid_t nx_uuid,
struct ifnet *ifp, const uuid_t nx_uuid_attachee, boolean_t host,
uuid_t *nx_if_uuid)
{
struct nxctl *nxctl;
int err = 0;
if (ncd == NULL) {
return EINVAL;
}
nxctl = ncd->ncd_nxctl;
ASSERT(nxctl != NULL);
NXCTL_LOCK(nxctl);
err = _kern_nexus_ifattach(nxctl, nx_uuid, ifp, nx_uuid_attachee,
host, nx_if_uuid);
NXCTL_UNLOCK(nxctl);
return err;
}
int
kern_nexus_ifdetach(const nexus_controller_t ncd,
const uuid_t nx_uuid, const uuid_t nx_if_uuid)
{
struct nx_cfg_req ncr;
struct nx_spec_req nsr;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (ncd == NULL || nx_uuid == NULL || uuid_is_null(nx_uuid) ||
nx_if_uuid == NULL || uuid_is_null(nx_if_uuid)) {
return EINVAL;
}
bzero(&nsr, sizeof(nsr));
uuid_copy(nsr.nsr_if_uuid, nx_if_uuid);
__nexus_config_req_prepare(&ncr, nx_uuid, NXCFG_CMD_DETACH,
&nsr, sizeof(nsr));
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_CONFIG;
sopt.sopt_val = (user_addr_t)&ncr;
sopt.sopt_valsize = sizeof(ncr);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_set_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
return err;
}
int
kern_nexus_get_netif_instance(struct ifnet *ifp, uuid_t nx_uuid)
{
struct nexus_netif_adapter *if_na;
int err = 0;
SK_LOCK();
if_na = ifp->if_na;
if (if_na != NULL) {
uuid_copy(nx_uuid, if_na->nifna_up.na_nx->nx_uuid);
} else {
err = ENXIO;
}
SK_UNLOCK();
if (err != 0) {
uuid_clear(nx_uuid);
}
return err;
}
int
kern_nexus_get_flowswitch_instance(struct ifnet *ifp, uuid_t nx_uuid)
{
struct nexus_netif_adapter *if_na;
struct nx_flowswitch *fsw = NULL;
int err = 0;
SK_LOCK();
if_na = ifp->if_na;
if (if_na != NULL) {
fsw = ifp->if_na->nifna_netif->nif_fsw;
}
if (fsw != NULL) {
uuid_copy(nx_uuid, fsw->fsw_nx->nx_uuid);
} else {
err = ENXIO;
}
SK_UNLOCK();
if (err != 0) {
uuid_clear(nx_uuid);
}
return err;
}
static void
kern_nexus_netagent_add(struct kern_nexus *nx, void *arg0)
{
#pragma unused(arg0)
nx_fsw_netagent_add(nx);
}
static void
kern_nexus_netagent_remove(struct kern_nexus *nx, void *arg0)
{
#pragma unused(arg0)
nx_fsw_netagent_remove(nx);
}
static void
kern_nexus_netagent_update(struct kern_nexus *nx, void *arg0)
{
#pragma unused(arg0)
nx_fsw_netagent_update(nx);
}
void
kern_nexus_register_netagents(void)
{
kern_nexus_walktree(kern_nexus_netagent_add, NULL, FALSE);
}
void
kern_nexus_deregister_netagents(void)
{
kern_nexus_walktree(kern_nexus_netagent_remove, NULL, FALSE);
}
void
kern_nexus_update_netagents(void)
{
kern_nexus_walktree(kern_nexus_netagent_update, NULL, FALSE);
}
static int
_interface_add_remove_netagent(struct ifnet *ifp, bool add)
{
struct nexus_netif_adapter *if_na;
int err = ENXIO;
SK_LOCK();
if_na = ifp->if_na;
if (if_na != NULL) {
struct nx_flowswitch *fsw;
fsw = if_na->nifna_netif->nif_fsw;
if (fsw != NULL) {
if (add) {
err = nx_fsw_netagent_add(fsw->fsw_nx);
} else {
err = nx_fsw_netagent_remove(fsw->fsw_nx);
}
}
}
SK_UNLOCK();
return err;
}
int
kern_nexus_interface_add_netagent(struct ifnet *ifp)
{
return _interface_add_remove_netagent(ifp, true);
}
int
kern_nexus_interface_remove_netagent(struct ifnet *ifp)
{
return _interface_add_remove_netagent(ifp, false);
}
int
kern_nexus_set_netif_input_tbr_rate(struct ifnet *ifp, uint64_t rate)
{
/* input tbr is only functional with active netif attachment */
if (ifp->if_na == NULL) {
if (rate != 0) {
return EINVAL;
} else {
return 0;
}
}
ifp->if_na->nifna_netif->nif_input_rate = rate;
return 0;
}
int
kern_nexus_set_if_netem_params(const nexus_controller_t ncd,
const uuid_t nx_uuid, void *data, size_t data_len)
{
struct nx_cfg_req ncr;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (nx_uuid == NULL || uuid_is_null(nx_uuid) ||
data_len < sizeof(struct if_netem_params)) {
return EINVAL;
}
__nexus_config_req_prepare(&ncr, nx_uuid, NXCFG_CMD_NETEM,
data, data_len);
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_CONFIG;
sopt.sopt_val = (user_addr_t)&ncr;
sopt.sopt_valsize = sizeof(ncr);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_set_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
return err;
}
static int
_kern_nexus_flow_config(const nexus_controller_t ncd, const uuid_t nx_uuid,
const nxcfg_cmd_t cmd, void *data, size_t data_len)
{
struct nx_cfg_req ncr;
struct sockopt sopt;
struct nxctl *nxctl;
int err = 0;
if (nx_uuid == NULL || uuid_is_null(nx_uuid) ||
data_len < sizeof(struct nx_flow_req)) {
return EINVAL;
}
__nexus_config_req_prepare(&ncr, nx_uuid, cmd, data, data_len);
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_name = NXOPT_NEXUS_CONFIG;
sopt.sopt_val = (user_addr_t)&ncr;
sopt.sopt_valsize = sizeof(ncr);
sopt.sopt_p = kernproc;
nxctl = ncd->ncd_nxctl;
NXCTL_LOCK(nxctl);
err = nxctl_set_opt(nxctl, &sopt);
NXCTL_UNLOCK(nxctl);
return err;
}
int
kern_nexus_flow_add(const nexus_controller_t ncd, const uuid_t nx_uuid,
void *data, size_t data_len)
{
return _kern_nexus_flow_config(ncd, nx_uuid, NXCFG_CMD_FLOW_ADD, data,
data_len);
}
int
kern_nexus_flow_del(const nexus_controller_t ncd, const uuid_t nx_uuid,
void *data, size_t data_len)
{
return _kern_nexus_flow_config(ncd, nx_uuid, NXCFG_CMD_FLOW_DEL, data,
data_len);
}
static struct kern_nexus_domain_provider *
nxdom_prov_alloc(zalloc_flags_t how)
{
SK_LOCK_ASSERT_HELD();
return zalloc_flags(nxdom_prov_zone, how | Z_ZERO);
}
static void
nxdom_prov_free(struct kern_nexus_domain_provider *nxdom_prov)
{
SK_LOCK_ASSERT_HELD();
ASSERT(nxdom_prov->nxdom_prov_refcnt == 0);
ASSERT(!(nxdom_prov->nxdom_prov_flags &
(NXDOMPROVF_ATTACHED | NXDOMPROVF_DETACHING)));
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_INITIALIZED) {
/*
* Tell the domain provider that we're done with this
* instance, and it is now free to go away.
*/
if (nxdom_prov->nxdom_prov_fini != NULL) {
nxdom_prov->nxdom_prov_fini(nxdom_prov);
}
nxdom_prov->nxdom_prov_flags &= ~NXDOMPROVF_INITIALIZED;
}
uuid_clear(nxdom_prov->nxdom_prov_uuid);
nxdom_prov->nxdom_prov_dom = NULL;
SK_DF(SK_VERB_MEM, "nxdom_prov 0x%llx %s", SK_KVA(nxdom_prov),
((nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT) ?
"FREE" : "DESTROY"));
if (nxdom_prov->nxdom_prov_flags & NXDOMPROVF_EXT) {
zfree(nxdom_prov_zone, nxdom_prov);
}
}
void
nxdom_prov_retain_locked(struct kern_nexus_domain_provider *nxdom_prov)
{
SK_LOCK_ASSERT_HELD();
nxdom_prov->nxdom_prov_refcnt++;
ASSERT(nxdom_prov->nxdom_prov_refcnt != 0);
}
void
nxdom_prov_retain(struct kern_nexus_domain_provider *nxdom_prov)
{
SK_LOCK();
nxdom_prov_retain_locked(nxdom_prov);
SK_UNLOCK();
}
static int
nxdom_prov_params_default(struct kern_nexus_domain_provider *nxdom_prov,
const uint32_t req, const struct nxprov_params *nxp0,
struct nxprov_params *nxp, struct skmem_region_params srp[SKMEM_REGIONS],
uint32_t pp_region_config_flags)
{
struct nxdom *nxdom = nxdom_prov->nxdom_prov_dom;
return nxprov_params_adjust(nxdom_prov, req, nxp0, nxp, srp,
nxdom, nxdom, nxdom, pp_region_config_flags, NULL);
}
int
nxdom_prov_validate_params(struct kern_nexus_domain_provider *nxdom_prov,
const struct nxprov_reg *reg, struct nxprov_params *nxp,
struct skmem_region_params srp[SKMEM_REGIONS], const uint32_t oflags,
uint32_t pp_region_config_flags)
{
const struct nxprov_params *nxp0 = &reg->nxpreg_params;
const uint32_t req = reg->nxpreg_requested;
int i, err = 0;
ASSERT(reg->nxpreg_version == NXPROV_REG_CURRENT_VERSION &&
nxp0->nxp_namelen != 0 &&
nxp0->nxp_namelen <= sizeof(nexus_name_t));
/* fill in with default values and let the nexus override them */
bzero(nxp, sizeof(*nxp));
bcopy(&nxp0->nxp_name, &nxp->nxp_name, sizeof(nxp->nxp_name));
nxp->nxp_name[sizeof(nxp->nxp_name) - 1] = '\0';
nxp->nxp_namelen = nxp0->nxp_namelen;
nxp->nxp_type = nxp0->nxp_type;
nxp->nxp_md_type = nxdom_prov->nxdom_prov_dom->nxdom_md_type;
nxp->nxp_md_subtype = nxdom_prov->nxdom_prov_dom->nxdom_md_subtype;
nxp->nxp_flags = (nxp0->nxp_flags & NXPF_MASK);
nxp->nxp_flags |= oflags; /* override */
nxp->nxp_format = nxp0->nxp_format;
nxp->nxp_ifindex = nxp0->nxp_ifindex;
nxp->nxp_reject_on_close = nxp0->nxp_reject_on_close;
/* inherit default region parameters */
for (i = 0; i < SKMEM_REGIONS; i++) {
srp[i] = *skmem_get_default(i);
}
if (nxdom_prov->nxdom_prov_params != NULL) {
err = nxdom_prov->nxdom_prov_params(nxdom_prov, req, nxp0,
nxp, srp, pp_region_config_flags);
} else {
err = nxdom_prov_params_default(nxdom_prov, req, nxp0,
nxp, srp, pp_region_config_flags);
}
return err;
}
boolean_t
nxdom_prov_release_locked(struct kern_nexus_domain_provider *nxdom_prov)
{
int oldref = nxdom_prov->nxdom_prov_refcnt;
SK_LOCK_ASSERT_HELD();
ASSERT(nxdom_prov->nxdom_prov_refcnt != 0);
if (--nxdom_prov->nxdom_prov_refcnt == 0) {
nxdom_prov_free(nxdom_prov);
}
return oldref == 1;
}
boolean_t
nxdom_prov_release(struct kern_nexus_domain_provider *nxdom_prov)
{
boolean_t lastref;
SK_LOCK();
lastref = nxdom_prov_release_locked(nxdom_prov);
SK_UNLOCK();
return lastref;
}
static uint32_t
nxprov_bound_var(uint32_t *v, uint32_t dflt, uint32_t lo, uint32_t hi,
const char *msg)
{
#pragma unused(msg)
uint32_t oldv = *v;
const char *op = NULL;
if (dflt < lo) {
dflt = lo;
}
if (dflt > hi) {
dflt = hi;
}
if (oldv < lo) {
*v = dflt;
op = "bump";
} else if (oldv > hi) {
*v = hi;
op = "clamp";
}
#if SK_LOG
if (op != NULL && msg != NULL) {
SK_ERR("%s %s to %u (was %u)", op, msg, *v, oldv);
}
#endif /* SK_LOG */
return *v;
}
#define NXPROV_PARAMS_ADJUST(flag, param) do { \
uint32_t _v0, _v; \
if (req & (flag)) \
_v = nxp0->nxp_##param; \
else \
_v = NXDOM_DEF(nxdom_def, param); \
_v0 = _v; \
if (nxprov_bound_var(&_v, NXDOM_DEF(nxdom_def, param), \
NXDOM_MIN(nxdom_min, param), NXDOM_MAX(nxdom_max, param), \
"nxp_" #param) < _v0) { \
err = ENOMEM; \
goto error; \
} \
nxp->nxp_##param = _v; \
} while (0)
#define MUL(x, y, z) do { \
if (__builtin_mul_overflow((x), (y), (z))) { \
overflowline = __LINE__; \
goto error; \
} \
} while (0)
#define ADD(x, y, z) do { \
if (__builtin_add_overflow((x), (y), (z))) { \
overflowline = __LINE__; \
goto error; \
} \
} while (0)
int
nxprov_params_adjust(struct kern_nexus_domain_provider *nxdom_prov,
const uint32_t req, const struct nxprov_params *nxp0,
struct nxprov_params *nxp, struct skmem_region_params srp[SKMEM_REGIONS],
const struct nxdom *nxdom_def, const struct nxdom *nxdom_min,
const struct nxdom *nxdom_max, uint32_t pp_region_config_flags,
int (*adjust_fn)(const struct kern_nexus_domain_provider *,
const struct nxprov_params *, struct nxprov_adjusted_params *))
{
uint32_t buf_cnt;
uint32_t stats_size;
uint32_t flowadv_max;
uint32_t nexusadv_size;
uint32_t capabs;
uint32_t tx_rings, rx_rings;
uint32_t alloc_rings = 0, free_rings = 0, ev_rings = 0;
uint32_t tx_slots, rx_slots;
uint32_t alloc_slots = 0, free_slots = 0, ev_slots = 0;
uint32_t buf_size, buf_region_segment_size, max_buffers = 0;
uint32_t tmp1, tmp2, tmp3, tmp4xpipes, tmpsumrings;
uint32_t tmpsumall, tmp4xpipesplusrings;
uint32_t large_buf_size;
int overflowline = 0;
int err = 0;
NXPROV_PARAMS_ADJUST(NXPREQ_TX_RINGS, tx_rings);
NXPROV_PARAMS_ADJUST(NXPREQ_RX_RINGS, rx_rings);
NXPROV_PARAMS_ADJUST(NXPREQ_TX_SLOTS, tx_slots);
NXPROV_PARAMS_ADJUST(NXPREQ_RX_SLOTS, rx_slots);
NXPROV_PARAMS_ADJUST(NXPREQ_BUF_SIZE, buf_size);
NXPROV_PARAMS_ADJUST(NXPREQ_LARGE_BUF_SIZE, large_buf_size);
NXPROV_PARAMS_ADJUST(NXPREQ_STATS_SIZE, stats_size);
NXPROV_PARAMS_ADJUST(NXPREQ_FLOWADV_MAX, flowadv_max);
NXPROV_PARAMS_ADJUST(NXPREQ_NEXUSADV_SIZE, nexusadv_size);
NXPROV_PARAMS_ADJUST(NXPREQ_PIPES, pipes);
NXPROV_PARAMS_ADJUST(NXPREQ_EXTENSIONS, extensions);
NXPROV_PARAMS_ADJUST(NXPREQ_MHINTS, mhints);
NXPROV_PARAMS_ADJUST(NXPREQ_CAPABILITIES, capabilities);
NXPROV_PARAMS_ADJUST(NXPREQ_QMAP, qmap);
NXPROV_PARAMS_ADJUST(NXPREQ_MAX_FRAGS, max_frags);
capabs = NXDOM_DEF(nxdom_def, capabilities);
if (req & NXPREQ_USER_CHANNEL) {
if (nxp->nxp_flags & NXPF_USER_CHANNEL) {
capabs |= NXPCAP_USER_CHANNEL;
} else {
capabs &= ~NXPCAP_USER_CHANNEL;
}
} else {
if (capabs & NXPCAP_USER_CHANNEL) {
nxp->nxp_flags |= NXPF_USER_CHANNEL;
} else {
nxp->nxp_flags &= ~NXPF_USER_CHANNEL;
}
}
if (NXDOM_MIN(nxdom_min, capabilities) != 0 &&
!(capabs & NXDOM_MIN(nxdom_min, capabilities))) {
SK_ERR("%s: caps 0x%b < min 0x%b",
nxdom_prov->nxdom_prov_name, capabs, NXPCAP_BITS,
NXDOM_MIN(nxdom_min, capabilities), NXPCAP_BITS);
err = EINVAL;
goto error;
} else if (NXDOM_MAX(nxdom_max, capabilities) != 0 &&
(capabs & ~NXDOM_MAX(nxdom_max, capabilities))) {
SK_ERR("%s: caps 0x%b > max 0x%b",
nxdom_prov->nxdom_prov_name, capabs, NXPCAP_BITS,
NXDOM_MAX(nxdom_max, capabilities), NXPCAP_BITS);
err = EINVAL;
goto error;
}
stats_size = nxp->nxp_stats_size;
flowadv_max = nxp->nxp_flowadv_max;
nexusadv_size = nxp->nxp_nexusadv_size;
tx_rings = nxp->nxp_tx_rings;
rx_rings = nxp->nxp_rx_rings;
tx_slots = nxp->nxp_tx_slots;
rx_slots = nxp->nxp_rx_slots;
buf_size = nxp->nxp_buf_size;
large_buf_size = nxp->nxp_large_buf_size;
buf_region_segment_size = skmem_usr_buf_seg_size;
ASSERT(pp_region_config_flags & PP_REGION_CONFIG_MD_MAGAZINE_ENABLE);
if (adjust_fn != NULL) {
struct nxprov_adjusted_params adj = {
.adj_md_subtype = &nxp->nxp_md_subtype,
.adj_stats_size = &stats_size,
.adj_flowadv_max = &flowadv_max,
.adj_nexusadv_size = &nexusadv_size,
.adj_caps = &capabs,
.adj_tx_rings = &tx_rings,
.adj_rx_rings = &rx_rings,
.adj_tx_slots = &tx_slots,
.adj_rx_slots = &rx_slots,
.adj_alloc_rings = &alloc_rings,
.adj_free_rings = &free_rings,
.adj_alloc_slots = &alloc_slots,
.adj_free_slots = &free_slots,
.adj_buf_size = &buf_size,
.adj_buf_region_segment_size = &buf_region_segment_size,
.adj_pp_region_config_flags = &pp_region_config_flags,
.adj_max_frags = &nxp->nxp_max_frags,
.adj_event_rings = &ev_rings,
.adj_event_slots = &ev_slots,
.adj_max_buffers = &max_buffers,
.adj_large_buf_size = &large_buf_size,
};
err = adjust_fn(nxdom_prov, nxp, &adj);
if (err != 0) {
goto error;
}
ASSERT(capabs >= NXDOM_MIN(nxdom_min, capabilities));
ASSERT(capabs <= NXDOM_MAX(nxdom_max, capabilities));
}
if (nxp->nxp_max_frags > UINT16_MAX) {
SK_ERR("invalid configuration for max frags %d",
nxp->nxp_max_frags);
err = EINVAL;
}
if (nxp->nxp_type == NEXUS_TYPE_USER_PIPE) {
if (tx_rings != rx_rings) {
SK_ERR("invalid configuration: {rx,tx} rings must be"
"in pairs for user pipe rx_rings(%d) tx_rings(%d)",
rx_rings, tx_rings);
err = EINVAL;
}
} else {
if (nxp->nxp_pipes != 0) {
SK_ERR("invalid configuration: pipe configuration is"
"only valid for user pipe nexus, type %d, pipes %d",
nxp->nxp_type, nxp->nxp_pipes);
err = EINVAL;
}
}
if (err != 0) {
goto error;
}
/* leading and trailing guard pages (if applicable) */
if (sk_guard) {
srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_size = SKMEM_PAGE_SIZE;
srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_cnt = sk_headguard_sz;
skmem_region_params_config(&srp[SKMEM_REGION_GUARD_HEAD]);
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_size = SKMEM_PAGE_SIZE;
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_cnt = sk_tailguard_sz;
skmem_region_params_config(&srp[SKMEM_REGION_GUARD_TAIL]);
} else {
srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_size = 0;
srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_cnt = 0;
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_size = 0;
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_cnt = 0;
}
/* update to the adjusted/configured values */
nxp->nxp_buf_size = buf_size;
nxp->nxp_tx_slots = tx_slots;
nxp->nxp_rx_slots = rx_slots;
nxp->nxp_large_buf_size = large_buf_size;
SK_D("nxdom \"%s\" (0x%llx) type %d",
nxdom_prov->nxdom_prov_dom->nxdom_name,
SK_KVA(nxdom_prov->nxdom_prov_dom),
nxdom_prov->nxdom_prov_dom->nxdom_type);
SK_D("nxp \"%s\" (0x%llx) flags 0x%b",
nxp->nxp_name, SK_KVA(nxp), nxp->nxp_flags, NXPF_BITS);
SK_D(" req 0x%b rings %u/%u/%u/%u/%u slots %u/%u/%u/%u/%u buf %u "
"type %u subtype %u stats %u flowadv_max %u nexusadv_size %u "
"capabs 0x%b pipes %u extensions %u max_frags %u headguard %u "
"tailguard %u large_buf %u", req, NXPREQ_BITS, tx_rings, rx_rings,
alloc_rings, free_rings, ev_rings, tx_slots, rx_slots, alloc_slots,
free_slots, ev_slots, nxp->nxp_buf_size, nxp->nxp_md_type,
nxp->nxp_md_subtype, stats_size, flowadv_max, nexusadv_size,
capabs, NXPCAP_BITS, nxp->nxp_pipes, nxp->nxp_extensions,
nxp->nxp_max_frags, srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_size *
srp[SKMEM_REGION_GUARD_HEAD].srp_r_obj_cnt,
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_size *
srp[SKMEM_REGION_GUARD_TAIL].srp_r_obj_cnt,
nxp->nxp_large_buf_size);
/*
* tmp4xpipes = 4 * nxp->nxp_pipes
*/
MUL(4, nxp->nxp_pipes, &tmp4xpipes);
/*
* tmp4xpipesplusrings = tx_rings + (4 * nxp->nxp_pipes)
*/
VERIFY((tmp4xpipes == 0) || (rx_rings == tx_rings));
ADD(tx_rings, tmp4xpipes, &tmp4xpipesplusrings);
/*
* tmpsumrings = tx_rings + rx_rings + alloc_rings + free_rings +
* ev_rings
*/
ADD(tx_rings, rx_rings, &tmpsumrings);
ADD(tmpsumrings, alloc_rings, &tmpsumrings);
ADD(tmpsumrings, free_rings, &tmpsumrings);
ADD(tmpsumrings, ev_rings, &tmpsumrings);
/*
* tmpsumall = (tx_rings + rx_rings +
* alloc_rings + free_rings + ev_rings + (4 * nxp->nxp_pipes))
*/
ADD(tmpsumrings, tmp4xpipes, &tmpsumall);
/* possibly increase them to fit user request */
VERIFY(CHANNEL_SCHEMA_SIZE(tmpsumrings) <= UINT32_MAX);
srp[SKMEM_REGION_SCHEMA].srp_r_obj_size =
(uint32_t)CHANNEL_SCHEMA_SIZE(tmpsumrings);
/* worst case is one channel bound to each ring pair */
srp[SKMEM_REGION_SCHEMA].srp_r_obj_cnt = tmp4xpipesplusrings;
skmem_region_params_config(&srp[SKMEM_REGION_SCHEMA]);
srp[SKMEM_REGION_RING].srp_r_obj_size =
sizeof(struct __user_channel_ring);
/* each pipe endpoint needs two tx rings and two rx rings */
srp[SKMEM_REGION_RING].srp_r_obj_cnt = tmpsumall;
skmem_region_params_config(&srp[SKMEM_REGION_RING]);
/*
* For each pipe we only need the buffers for the "real" rings.
* On the other end, the pipe ring dimension may be different from
* the parent port ring dimension. As a compromise, we allocate twice
* the space actually needed if the pipe rings were the same size as
* the parent rings.
*
* buf_cnt = ((4 * nxp->nxp_pipes) + rx_rings) * rx_slots +
* ((4 * nxp->nxp_pipes) + tx_rings) * tx_slots +
* (ev_rings * ev_slots);
*/
if (nxp->nxp_type == NEXUS_TYPE_USER_PIPE) {
MUL(tmp4xpipesplusrings, rx_slots, &tmp1);
MUL(tmp4xpipesplusrings, tx_slots, &tmp2);
ASSERT(ev_rings == 0);
tmp3 = 0;
} else {
MUL(rx_rings, rx_slots, &tmp1);
MUL(tx_rings, tx_slots, &tmp2);
MUL(ev_rings, ev_slots, &tmp3);
}
ADD(tmp1, tmp2, &buf_cnt);
ADD(tmp3, buf_cnt, &buf_cnt);
if (nxp->nxp_max_frags > 1) {
pp_region_config_flags |= PP_REGION_CONFIG_BUFLET;
buf_cnt = MIN((((uint32_t)P2ROUNDUP(NX_MAX_AGGR_PKT_SIZE,
nxp->nxp_buf_size) / nxp->nxp_buf_size) * buf_cnt),
(buf_cnt * nxp->nxp_max_frags));
}
if (max_buffers != 0) {
buf_cnt = MIN(max_buffers, buf_cnt);
}
if ((nxp->nxp_flags & NXPF_USER_CHANNEL) == 0) {
pp_region_config_flags |= PP_REGION_CONFIG_KERNEL_ONLY;
}
/* # of metadata objects is same as the # of buffer objects */
ASSERT(buf_region_segment_size != 0);
pp_regions_params_adjust(srp, nxp->nxp_md_type, nxp->nxp_md_subtype,
buf_cnt, (uint16_t)nxp->nxp_max_frags, nxp->nxp_buf_size,
nxp->nxp_large_buf_size, buf_cnt, buf_region_segment_size,
pp_region_config_flags);
/* statistics region size */
if (stats_size != 0) {
srp[SKMEM_REGION_USTATS].srp_r_obj_size = stats_size;
srp[SKMEM_REGION_USTATS].srp_r_obj_cnt = 1;
skmem_region_params_config(&srp[SKMEM_REGION_USTATS]);
} else {
srp[SKMEM_REGION_USTATS].srp_r_obj_size = 0;
srp[SKMEM_REGION_USTATS].srp_r_obj_cnt = 0;
srp[SKMEM_REGION_USTATS].srp_c_obj_size = 0;
srp[SKMEM_REGION_USTATS].srp_c_obj_cnt = 0;
}
/* flow advisory region size */
if (flowadv_max != 0) {
_CASSERT(NX_FLOWADV_DEFAULT * sizeof(struct __flowadv_entry) <=
SKMEM_MIN_SEG_SIZE);
MUL(sizeof(struct __flowadv_entry), flowadv_max, &tmp1);
srp[SKMEM_REGION_FLOWADV].srp_r_obj_size = tmp1;
srp[SKMEM_REGION_FLOWADV].srp_r_obj_cnt = 1;
skmem_region_params_config(&srp[SKMEM_REGION_FLOWADV]);
} else {
srp[SKMEM_REGION_FLOWADV].srp_r_obj_size = 0;
srp[SKMEM_REGION_FLOWADV].srp_r_obj_cnt = 0;
srp[SKMEM_REGION_FLOWADV].srp_c_obj_size = 0;
srp[SKMEM_REGION_FLOWADV].srp_c_obj_cnt = 0;
}
/* nexus advisory region size */
if (nexusadv_size != 0) {
srp[SKMEM_REGION_NEXUSADV].srp_r_obj_size = nexusadv_size +
sizeof(struct __kern_nexus_adv_metadata);
srp[SKMEM_REGION_NEXUSADV].srp_r_obj_cnt = 1;
skmem_region_params_config(&srp[SKMEM_REGION_NEXUSADV]);
} else {
srp[SKMEM_REGION_NEXUSADV].srp_r_obj_size = 0;
srp[SKMEM_REGION_NEXUSADV].srp_r_obj_cnt = 0;
srp[SKMEM_REGION_NEXUSADV].srp_c_obj_size = 0;
srp[SKMEM_REGION_NEXUSADV].srp_c_obj_cnt = 0;
}
/* sysctls region is not applicable to nexus */
srp[SKMEM_REGION_SYSCTLS].srp_r_obj_size = 0;
srp[SKMEM_REGION_SYSCTLS].srp_r_obj_cnt = 0;
srp[SKMEM_REGION_SYSCTLS].srp_c_obj_size = 0;
srp[SKMEM_REGION_SYSCTLS].srp_c_obj_cnt = 0;
/*
* Since the tx/alloc/event slots share the same region and cache,
* we will use the same object size for both types of slots.
*/
srp[SKMEM_REGION_TXAKSD].srp_r_obj_size =
(MAX(MAX(tx_slots, alloc_slots), ev_slots)) * SLOT_DESC_SZ;
srp[SKMEM_REGION_TXAKSD].srp_r_obj_cnt = tx_rings + alloc_rings +
ev_rings;
skmem_region_params_config(&srp[SKMEM_REGION_TXAKSD]);
/* USD and KSD objects share the same size and count */
srp[SKMEM_REGION_TXAUSD].srp_r_obj_size =
srp[SKMEM_REGION_TXAKSD].srp_r_obj_size;
srp[SKMEM_REGION_TXAUSD].srp_r_obj_cnt =
srp[SKMEM_REGION_TXAKSD].srp_r_obj_cnt;
skmem_region_params_config(&srp[SKMEM_REGION_TXAUSD]);
/*
* Since the rx/free slots share the same region and cache,
* we will use the same object size for both types of slots.
*/
srp[SKMEM_REGION_RXFKSD].srp_r_obj_size =
MAX(rx_slots, free_slots) * SLOT_DESC_SZ;
srp[SKMEM_REGION_RXFKSD].srp_r_obj_cnt = rx_rings + free_rings;
skmem_region_params_config(&srp[SKMEM_REGION_RXFKSD]);
/* USD and KSD objects share the same size and count */
srp[SKMEM_REGION_RXFUSD].srp_r_obj_size =
srp[SKMEM_REGION_RXFKSD].srp_r_obj_size;
srp[SKMEM_REGION_RXFUSD].srp_r_obj_cnt =
srp[SKMEM_REGION_RXFKSD].srp_r_obj_cnt;
skmem_region_params_config(&srp[SKMEM_REGION_RXFUSD]);
/* update these based on the adjusted/configured values */
nxp->nxp_meta_size = srp[SKMEM_REGION_KMD].srp_c_obj_size;
nxp->nxp_stats_size = stats_size;
nxp->nxp_flowadv_max = flowadv_max;
nxp->nxp_nexusadv_size = nexusadv_size;
nxp->nxp_capabilities = capabs;
error:
if (overflowline) {
err = EOVERFLOW;
SK_ERR("math overflow in %s on line %d",
__func__, overflowline);
}
return err;
}
#undef ADD
#undef MUL
#undef NXPROV_PARAMS_ADJUST
static void
nxprov_detaching_enqueue(struct kern_nexus_domain_provider *nxdom_prov)
{
SK_LOCK_ASSERT_HELD();
ASSERT((nxdom_prov->nxdom_prov_flags & (NXDOMPROVF_ATTACHED |
NXDOMPROVF_DETACHING)) == NXDOMPROVF_DETACHING);
++nxprov_detaching_cnt;
ASSERT(nxprov_detaching_cnt != 0);
/*
* Insert this to the detaching list; caller is expected to
* have held a reference, most likely the same one that was
* used for the per-domain provider list.
*/
STAILQ_INSERT_TAIL(&nxprov_detaching_head, nxdom_prov,
nxdom_prov_detaching_link);
wakeup((caddr_t)&nxprov_detach_wchan);
}
static struct kern_nexus_domain_provider *
nxprov_detaching_dequeue(void)
{
struct kern_nexus_domain_provider *nxdom_prov;
SK_LOCK_ASSERT_HELD();
nxdom_prov = STAILQ_FIRST(&nxprov_detaching_head);
ASSERT(nxprov_detaching_cnt != 0 || nxdom_prov == NULL);
if (nxdom_prov != NULL) {
ASSERT((nxdom_prov->nxdom_prov_flags & (NXDOMPROVF_ATTACHED |
NXDOMPROVF_DETACHING)) == NXDOMPROVF_DETACHING);
ASSERT(nxprov_detaching_cnt != 0);
--nxprov_detaching_cnt;
STAILQ_REMOVE(&nxprov_detaching_head, nxdom_prov,
kern_nexus_domain_provider, nxdom_prov_detaching_link);
}
return nxdom_prov;
}
__attribute__((noreturn))
static void
nxprov_detacher(void *v, wait_result_t w)
{
#pragma unused(v, w)
SK_LOCK();
(void) msleep0(&nxprov_detach_wchan, &sk_lock, (PZERO - 1),
__func__, 0, nxprov_detacher_cont);
/*
* msleep0() shouldn't have returned as PCATCH was not set;
* therefore assert in this case.
*/
SK_UNLOCK();
VERIFY(0);
/* NOTREACHED */
__builtin_unreachable();
}
static int
nxprov_detacher_cont(int err)
{
#pragma unused(err)
struct kern_nexus_domain_provider *nxdom_prov;
for (;;) {
SK_LOCK_ASSERT_HELD();
while (nxprov_detaching_cnt == 0) {
(void) msleep0(&nxprov_detach_wchan, &sk_lock,
(PZERO - 1), __func__, 0, nxprov_detacher_cont);
/* NOTREACHED */
}
ASSERT(STAILQ_FIRST(&nxprov_detaching_head) != NULL);
nxdom_prov = nxprov_detaching_dequeue();
if (nxdom_prov != NULL) {
nxdom_del_provider_final(nxdom_prov);
}
}
}
Reference in a new issue View git blame Copy permalink