X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/de355530ae67247cbd0da700edb3a2a1dae884c2..7ee9d059c4eecf68ae4f8b0fb99ae2471eda79af:/bsd/netinet/ip_dummynet.c diff --git a/bsd/netinet/ip_dummynet.c b/bsd/netinet/ip_dummynet.c index 0979e45d3..048cff004 100644 --- a/bsd/netinet/ip_dummynet.c +++ b/bsd/netinet/ip_dummynet.c @@ -1,25 +1,32 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2010 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * 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. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * 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 OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * 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_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ - * Copyright (c) 1998-2001 Luigi Rizzo, Universita` di Pisa +/* + * Copyright (c) 1998-2002 Luigi Rizzo, Universita` di Pisa * Portions Copyright (c) 2000 Akamba Corp. * All rights reserved * @@ -44,11 +51,10 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * $FreeBSD: src/sys/netinet/ip_dummynet.c,v 1.24.2.11 2001/02/09 23:18:08 luigi Exp $ + * $FreeBSD: src/sys/netinet/ip_dummynet.c,v 1.84 2004/08/25 09:31:30 pjd Exp $ */ -#define DEB(x) -#define DDB(x) x +#define DUMMYNET_DEBUG /* * This module implements IP dummynet, a bandwidth limiter/delay emulator @@ -60,10 +66,7 @@ * + scheduler and dummynet functions; * + configuration and initialization. * - * NOTA BENE: critical sections are protected by splimp()/splx() - * pairs. One would think that splnet() is enough as for most of - * the netinet code, but it is not so because when used with - * bridging, dummynet is invoked at splimp(). + * NOTA BENE: critical sections are protected by the "dummynet lock". * * Most important Changes: * @@ -86,8 +89,10 @@ #include #include #include +//#include #include #include +#include #include #include #include @@ -96,17 +101,16 @@ #include #include -#if BRIDGE -#include /* for struct arpcom */ -#include -#endif - /* * We keep a private variable for the simulation time, but we could * probably use an existing one ("softticks" in sys/kern/kern_timer.c) */ static dn_key curr_time = 0 ; /* current simulation time */ +/* this is for the timer that fires to call dummynet() - we only enable the timer when + there are packets to process, otherwise it's disabled */ +static int timer_enabled = 0; + static int dn_hash_size = 64 ; /* default hash size */ /* statistics on number of queue searches and search steps */ @@ -118,6 +122,8 @@ static int red_lookup_depth = 256; /* RED - default lookup table depth */ static int red_avg_pkt_size = 512; /* RED - default medium packet size */ static int red_max_pkt_size = 1500; /* RED - default max packet size */ +static int serialize = 0; + /* * Three heaps contain queues and pipes that the scheduler handles: * @@ -134,64 +140,102 @@ static int heap_init(struct dn_heap *h, int size) ; static int heap_insert (struct dn_heap *h, dn_key key1, void *p); static void heap_extract(struct dn_heap *h, void *obj); -static void transmit_event(struct dn_pipe *pipe); -static void ready_event(struct dn_flow_queue *q); -static struct dn_pipe *all_pipes = NULL ; /* list of all pipes */ -static struct dn_flow_set *all_flow_sets = NULL ;/* list of all flow_sets */ +static void transmit_event(struct dn_pipe *pipe, struct mbuf **head, + struct mbuf **tail); +static void ready_event(struct dn_flow_queue *q, struct mbuf **head, + struct mbuf **tail); +static void ready_event_wfq(struct dn_pipe *p, struct mbuf **head, + struct mbuf **tail); + +/* + * Packets are retrieved from queues in Dummynet in chains instead of + * packet-by-packet. The entire list of packets is first dequeued and + * sent out by the following function. + */ +static void dummynet_send(struct mbuf *m); + +#define HASHSIZE 16 +#define HASH(num) ((((num) >> 8) ^ ((num) >> 4) ^ (num)) & 0x0f) +static struct dn_pipe_head pipehash[HASHSIZE]; /* all pipes */ +static struct dn_flow_set_head flowsethash[HASHSIZE]; /* all flowsets */ -#if SYSCTL_NODE + +#ifdef SYSCTL_NODE SYSCTL_NODE(_net_inet_ip, OID_AUTO, dummynet, - CTLFLAG_RW, 0, "Dummynet"); + CTLFLAG_RW | CTLFLAG_LOCKED, 0, "Dummynet"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, hash_size, - CTLFLAG_RW, &dn_hash_size, 0, "Default hash table size"); -SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, curr_time, - CTLFLAG_RD, &curr_time, 0, "Current tick"); + CTLFLAG_RW | CTLFLAG_LOCKED, &dn_hash_size, 0, "Default hash table size"); +SYSCTL_QUAD(_net_inet_ip_dummynet, OID_AUTO, curr_time, + CTLFLAG_RD | CTLFLAG_LOCKED, &curr_time, "Current tick"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, ready_heap, - CTLFLAG_RD, &ready_heap.size, 0, "Size of ready heap"); + CTLFLAG_RD | CTLFLAG_LOCKED, &ready_heap.size, 0, "Size of ready heap"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, extract_heap, - CTLFLAG_RD, &extract_heap.size, 0, "Size of extract heap"); + CTLFLAG_RD | CTLFLAG_LOCKED, &extract_heap.size, 0, "Size of extract heap"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, searches, - CTLFLAG_RD, &searches, 0, "Number of queue searches"); + CTLFLAG_RD | CTLFLAG_LOCKED, &searches, 0, "Number of queue searches"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, search_steps, - CTLFLAG_RD, &search_steps, 0, "Number of queue search steps"); + CTLFLAG_RD | CTLFLAG_LOCKED, &search_steps, 0, "Number of queue search steps"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, expire, - CTLFLAG_RW, &pipe_expire, 0, "Expire queue if empty"); + CTLFLAG_RW | CTLFLAG_LOCKED, &pipe_expire, 0, "Expire queue if empty"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, max_chain_len, - CTLFLAG_RW, &dn_max_ratio, 0, + CTLFLAG_RW | CTLFLAG_LOCKED, &dn_max_ratio, 0, "Max ratio between dynamic queues and buckets"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, red_lookup_depth, - CTLFLAG_RD, &red_lookup_depth, 0, "Depth of RED lookup table"); + CTLFLAG_RD | CTLFLAG_LOCKED, &red_lookup_depth, 0, "Depth of RED lookup table"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, red_avg_pkt_size, - CTLFLAG_RD, &red_avg_pkt_size, 0, "RED Medium packet size"); + CTLFLAG_RD | CTLFLAG_LOCKED, &red_avg_pkt_size, 0, "RED Medium packet size"); SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, red_max_pkt_size, - CTLFLAG_RD, &red_max_pkt_size, 0, "RED Max packet size"); + CTLFLAG_RD | CTLFLAG_LOCKED, &red_max_pkt_size, 0, "RED Max packet size"); +#endif + +#ifdef DUMMYNET_DEBUG +int dummynet_debug = 0; +#ifdef SYSCTL_NODE +SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, debug, CTLFLAG_RW | CTLFLAG_LOCKED, &dummynet_debug, + 0, "control debugging printfs"); +#endif +#define DPRINTF(X) if (dummynet_debug) printf X +#else +#define DPRINTF(X) #endif +/* contrary to the comment above random(), it does not actually + * return a value [0, 2^31 - 1], which breaks plr amongst other + * things. Masking it should work even if the behavior of + * the function is fixed. + */ +#define MY_RANDOM (random() & 0x7FFFFFFF) + +/* dummynet lock */ +static lck_grp_t *dn_mutex_grp; +static lck_grp_attr_t *dn_mutex_grp_attr; +static lck_attr_t *dn_mutex_attr; +static lck_mtx_t *dn_mutex; + static int config_pipe(struct dn_pipe *p); static int ip_dn_ctl(struct sockopt *sopt); -static void rt_unref(struct rtentry *); static void dummynet(void *); static void dummynet_flush(void); void dummynet_drain(void); +static ip_dn_io_t dummynet_io; +static void dn_rule_delete(void *); + int if_tx_rdy(struct ifnet *ifp); -/* - * ip_fw_chain is used when deleting a pipe, because ipfw rules can - * hold references to the pipe. - */ -extern LIST_HEAD (ip_fw_head, ip_fw_chain) ip_fw_chain_head; +static void cp_flow_set_to_64_user(struct dn_flow_set *set, struct dn_flow_set_64 *fs_bp); +static void cp_queue_to_64_user( struct dn_flow_queue *q, struct dn_flow_queue_64 *qp); +static char *cp_pipe_to_64_user(struct dn_pipe *p, struct dn_pipe_64 *pipe_bp); +static char* dn_copy_set_64(struct dn_flow_set *set, char *bp); +static int cp_pipe_from_user_64( struct sockopt *sopt, struct dn_pipe *p ); + +static void cp_flow_set_to_32_user(struct dn_flow_set *set, struct dn_flow_set_32 *fs_bp); +static void cp_queue_to_32_user( struct dn_flow_queue *q, struct dn_flow_queue_32 *qp); +static char *cp_pipe_to_32_user(struct dn_pipe *p, struct dn_pipe_32 *pipe_bp); +static char* dn_copy_set_32(struct dn_flow_set *set, char *bp); +static int cp_pipe_from_user_32( struct sockopt *sopt, struct dn_pipe *p ); -static void -rt_unref(struct rtentry *rt) -{ - if (rt == NULL) - return ; - if (rt->rt_refcnt <= 0) - printf("-- warning, refcnt now %ld, decreasing\n", rt->rt_refcnt); - rtfree(rt); -} /* * Heap management functions. @@ -212,25 +256,301 @@ rt_unref(struct rtentry *rt) #define HEAP_SWAP(a, b, buffer) { buffer = a ; a = b ; b = buffer ; } #define HEAP_INCREMENT 15 + +int cp_pipe_from_user_32( struct sockopt *sopt, struct dn_pipe *p ) +{ + struct dn_pipe_32 user_pipe_32; + int error=0; + + error = sooptcopyin(sopt, &user_pipe_32, sizeof(struct dn_pipe_32), sizeof(struct dn_pipe_32)); + if ( !error ){ + p->pipe_nr = user_pipe_32.pipe_nr; + p->bandwidth = user_pipe_32.bandwidth; + p->delay = user_pipe_32.delay; + p->V = user_pipe_32.V; + p->sum = user_pipe_32.sum; + p->numbytes = user_pipe_32.numbytes; + p->sched_time = user_pipe_32.sched_time; + bcopy( user_pipe_32.if_name, p->if_name, IFNAMSIZ); + p->ready = user_pipe_32.ready; + + p->fs.fs_nr = user_pipe_32.fs.fs_nr; + p->fs.flags_fs = user_pipe_32.fs.flags_fs; + p->fs.parent_nr = user_pipe_32.fs.parent_nr; + p->fs.weight = user_pipe_32.fs.weight; + p->fs.qsize = user_pipe_32.fs.qsize; + p->fs.plr = user_pipe_32.fs.plr; + p->fs.flow_mask = user_pipe_32.fs.flow_mask; + p->fs.rq_size = user_pipe_32.fs.rq_size; + p->fs.rq_elements = user_pipe_32.fs.rq_elements; + p->fs.last_expired = user_pipe_32.fs.last_expired; + p->fs.backlogged = user_pipe_32.fs.backlogged; + p->fs.w_q = user_pipe_32.fs.w_q; + p->fs.max_th = user_pipe_32.fs.max_th; + p->fs.min_th = user_pipe_32.fs.min_th; + p->fs.max_p = user_pipe_32.fs.max_p; + p->fs.c_1 = user_pipe_32.fs.c_1; + p->fs.c_2 = user_pipe_32.fs.c_2; + p->fs.c_3 = user_pipe_32.fs.c_3; + p->fs.c_4 = user_pipe_32.fs.c_4; + p->fs.lookup_depth = user_pipe_32.fs.lookup_depth; + p->fs.lookup_step = user_pipe_32.fs.lookup_step; + p->fs.lookup_weight = user_pipe_32.fs.lookup_weight; + p->fs.avg_pkt_size = user_pipe_32.fs.avg_pkt_size; + p->fs.max_pkt_size = user_pipe_32.fs.max_pkt_size; + } + return error; +} + + +int cp_pipe_from_user_64( struct sockopt *sopt, struct dn_pipe *p ) +{ + struct dn_pipe_64 user_pipe_64; + int error=0; + + error = sooptcopyin(sopt, &user_pipe_64, sizeof(struct dn_pipe_64), sizeof(struct dn_pipe_64)); + if ( !error ){ + p->pipe_nr = user_pipe_64.pipe_nr; + p->bandwidth = user_pipe_64.bandwidth; + p->delay = user_pipe_64.delay; + p->V = user_pipe_64.V; + p->sum = user_pipe_64.sum; + p->numbytes = user_pipe_64.numbytes; + p->sched_time = user_pipe_64.sched_time; + bcopy( user_pipe_64.if_name, p->if_name, IFNAMSIZ); + p->ready = user_pipe_64.ready; + + p->fs.fs_nr = user_pipe_64.fs.fs_nr; + p->fs.flags_fs = user_pipe_64.fs.flags_fs; + p->fs.parent_nr = user_pipe_64.fs.parent_nr; + p->fs.weight = user_pipe_64.fs.weight; + p->fs.qsize = user_pipe_64.fs.qsize; + p->fs.plr = user_pipe_64.fs.plr; + p->fs.flow_mask = user_pipe_64.fs.flow_mask; + p->fs.rq_size = user_pipe_64.fs.rq_size; + p->fs.rq_elements = user_pipe_64.fs.rq_elements; + p->fs.last_expired = user_pipe_64.fs.last_expired; + p->fs.backlogged = user_pipe_64.fs.backlogged; + p->fs.w_q = user_pipe_64.fs.w_q; + p->fs.max_th = user_pipe_64.fs.max_th; + p->fs.min_th = user_pipe_64.fs.min_th; + p->fs.max_p = user_pipe_64.fs.max_p; + p->fs.c_1 = user_pipe_64.fs.c_1; + p->fs.c_2 = user_pipe_64.fs.c_2; + p->fs.c_3 = user_pipe_64.fs.c_3; + p->fs.c_4 = user_pipe_64.fs.c_4; + p->fs.lookup_depth = user_pipe_64.fs.lookup_depth; + p->fs.lookup_step = user_pipe_64.fs.lookup_step; + p->fs.lookup_weight = user_pipe_64.fs.lookup_weight; + p->fs.avg_pkt_size = user_pipe_64.fs.avg_pkt_size; + p->fs.max_pkt_size = user_pipe_64.fs.max_pkt_size; + } + return error; +} + +static void +cp_flow_set_to_32_user(struct dn_flow_set *set, struct dn_flow_set_32 *fs_bp) +{ + fs_bp->fs_nr = set->fs_nr; + fs_bp->flags_fs = set->flags_fs ; + fs_bp->parent_nr = set->parent_nr ; + fs_bp->weight = set->weight ; + fs_bp->qsize = set->qsize ; + fs_bp->plr = set->plr ; + fs_bp->flow_mask = set->flow_mask ; + fs_bp->rq_size = set->rq_size ; + fs_bp->rq_elements = set->rq_elements ; + fs_bp->last_expired = set->last_expired ; + fs_bp->backlogged = set->backlogged ; + fs_bp->w_q = set->w_q ; + fs_bp->max_th = set->max_th ; + fs_bp->min_th = set->min_th ; + fs_bp->max_p = set->max_p ; + fs_bp->c_1 = set->c_1 ; + fs_bp->c_2 = set->c_2 ; + fs_bp->c_3 = set->c_3 ; + fs_bp->c_4 = set->c_4 ; + fs_bp->w_q_lookup = CAST_DOWN_EXPLICIT(user32_addr_t, set->w_q_lookup) ; + fs_bp->lookup_depth = set->lookup_depth ; + fs_bp->lookup_step = set->lookup_step ; + fs_bp->lookup_weight = set->lookup_weight ; + fs_bp->avg_pkt_size = set->avg_pkt_size ; + fs_bp->max_pkt_size = set->max_pkt_size ; +} + +static void +cp_flow_set_to_64_user(struct dn_flow_set *set, struct dn_flow_set_64 *fs_bp) +{ + fs_bp->fs_nr = set->fs_nr; + fs_bp->flags_fs = set->flags_fs ; + fs_bp->parent_nr = set->parent_nr ; + fs_bp->weight = set->weight ; + fs_bp->qsize = set->qsize ; + fs_bp->plr = set->plr ; + fs_bp->flow_mask = set->flow_mask ; + fs_bp->rq_size = set->rq_size ; + fs_bp->rq_elements = set->rq_elements ; + fs_bp->last_expired = set->last_expired ; + fs_bp->backlogged = set->backlogged ; + fs_bp->w_q = set->w_q ; + fs_bp->max_th = set->max_th ; + fs_bp->min_th = set->min_th ; + fs_bp->max_p = set->max_p ; + fs_bp->c_1 = set->c_1 ; + fs_bp->c_2 = set->c_2 ; + fs_bp->c_3 = set->c_3 ; + fs_bp->c_4 = set->c_4 ; + fs_bp->w_q_lookup = CAST_DOWN(user64_addr_t, set->w_q_lookup) ; + fs_bp->lookup_depth = set->lookup_depth ; + fs_bp->lookup_step = set->lookup_step ; + fs_bp->lookup_weight = set->lookup_weight ; + fs_bp->avg_pkt_size = set->avg_pkt_size ; + fs_bp->max_pkt_size = set->max_pkt_size ; +} + +static +void cp_queue_to_32_user( struct dn_flow_queue *q, struct dn_flow_queue_32 *qp) +{ + qp->id = q->id; + qp->len = q->len; + qp->len_bytes = q->len_bytes; + qp->numbytes = q->numbytes; + qp->tot_pkts = q->tot_pkts; + qp->tot_bytes = q->tot_bytes; + qp->drops = q->drops; + qp->hash_slot = q->hash_slot; + qp->avg = q->avg; + qp->count = q->count; + qp->random = q->random; + qp->q_time = q->q_time; + qp->heap_pos = q->heap_pos; + qp->sched_time = q->sched_time; + qp->S = q->S; + qp->F = q->F; +} + +static +void cp_queue_to_64_user( struct dn_flow_queue *q, struct dn_flow_queue_64 *qp) +{ + qp->id = q->id; + qp->len = q->len; + qp->len_bytes = q->len_bytes; + qp->numbytes = q->numbytes; + qp->tot_pkts = q->tot_pkts; + qp->tot_bytes = q->tot_bytes; + qp->drops = q->drops; + qp->hash_slot = q->hash_slot; + qp->avg = q->avg; + qp->count = q->count; + qp->random = q->random; + qp->q_time = q->q_time; + qp->heap_pos = q->heap_pos; + qp->sched_time = q->sched_time; + qp->S = q->S; + qp->F = q->F; +} + +static +char *cp_pipe_to_32_user(struct dn_pipe *p, struct dn_pipe_32 *pipe_bp) +{ + char *bp; + + pipe_bp->pipe_nr = p->pipe_nr; + pipe_bp->bandwidth = p->bandwidth; + pipe_bp->delay = p->delay; + bcopy( &(p->scheduler_heap), &(pipe_bp->scheduler_heap), sizeof(struct dn_heap_32)); + pipe_bp->scheduler_heap.p = CAST_DOWN_EXPLICIT(user32_addr_t, pipe_bp->scheduler_heap.p); + bcopy( &(p->not_eligible_heap), &(pipe_bp->not_eligible_heap), sizeof(struct dn_heap_32)); + pipe_bp->not_eligible_heap.p = CAST_DOWN_EXPLICIT(user32_addr_t, pipe_bp->not_eligible_heap.p); + bcopy( &(p->idle_heap), &(pipe_bp->idle_heap), sizeof(struct dn_heap_32)); + pipe_bp->idle_heap.p = CAST_DOWN_EXPLICIT(user32_addr_t, pipe_bp->idle_heap.p); + pipe_bp->V = p->V; + pipe_bp->sum = p->sum; + pipe_bp->numbytes = p->numbytes; + pipe_bp->sched_time = p->sched_time; + bcopy( p->if_name, pipe_bp->if_name, IFNAMSIZ); + pipe_bp->ifp = CAST_DOWN_EXPLICIT(user32_addr_t, p->ifp); + pipe_bp->ready = p->ready; + + cp_flow_set_to_32_user( &(p->fs), &(pipe_bp->fs)); + + pipe_bp->delay = (pipe_bp->delay * 1000) / (hz*10) ; + /* + * XXX the following is a hack based on ->next being the + * first field in dn_pipe and dn_flow_set. The correct + * solution would be to move the dn_flow_set to the beginning + * of struct dn_pipe. + */ + pipe_bp->next = CAST_DOWN_EXPLICIT( user32_addr_t, DN_IS_PIPE ); + /* clean pointers */ + pipe_bp->head = pipe_bp->tail = (user32_addr_t) 0 ; + pipe_bp->fs.next = (user32_addr_t)0 ; + pipe_bp->fs.pipe = (user32_addr_t)0 ; + pipe_bp->fs.rq = (user32_addr_t)0 ; + bp = ((char *)pipe_bp) + sizeof(struct dn_pipe_32); + return( dn_copy_set_32( &(p->fs), bp) ); +} + +static +char *cp_pipe_to_64_user(struct dn_pipe *p, struct dn_pipe_64 *pipe_bp) +{ + char *bp; + + pipe_bp->pipe_nr = p->pipe_nr; + pipe_bp->bandwidth = p->bandwidth; + pipe_bp->delay = p->delay; + bcopy( &(p->scheduler_heap), &(pipe_bp->scheduler_heap), sizeof(struct dn_heap_64)); + pipe_bp->scheduler_heap.p = CAST_DOWN(user64_addr_t, pipe_bp->scheduler_heap.p); + bcopy( &(p->not_eligible_heap), &(pipe_bp->not_eligible_heap), sizeof(struct dn_heap_64)); + pipe_bp->not_eligible_heap.p = CAST_DOWN(user64_addr_t, pipe_bp->not_eligible_heap.p); + bcopy( &(p->idle_heap), &(pipe_bp->idle_heap), sizeof(struct dn_heap_64)); + pipe_bp->idle_heap.p = CAST_DOWN(user64_addr_t, pipe_bp->idle_heap.p); + pipe_bp->V = p->V; + pipe_bp->sum = p->sum; + pipe_bp->numbytes = p->numbytes; + pipe_bp->sched_time = p->sched_time; + bcopy( p->if_name, pipe_bp->if_name, IFNAMSIZ); + pipe_bp->ifp = CAST_DOWN(user64_addr_t, p->ifp); + pipe_bp->ready = p->ready; + + cp_flow_set_to_64_user( &(p->fs), &(pipe_bp->fs)); + + pipe_bp->delay = (pipe_bp->delay * 1000) / (hz*10) ; + /* + * XXX the following is a hack based on ->next being the + * first field in dn_pipe and dn_flow_set. The correct + * solution would be to move the dn_flow_set to the beginning + * of struct dn_pipe. + */ + pipe_bp->next = CAST_DOWN( user64_addr_t, DN_IS_PIPE ); + /* clean pointers */ + pipe_bp->head = pipe_bp->tail = USER_ADDR_NULL ; + pipe_bp->fs.next = USER_ADDR_NULL ; + pipe_bp->fs.pipe = USER_ADDR_NULL ; + pipe_bp->fs.rq = USER_ADDR_NULL ; + bp = ((char *)pipe_bp) + sizeof(struct dn_pipe_64); + return( dn_copy_set_64( &(p->fs), bp) ); +} + static int heap_init(struct dn_heap *h, int new_size) -{ +{ struct dn_heap_entry *p; if (h->size >= new_size ) { - printf("heap_init, Bogus call, have %d want %d\n", + printf("dummynet: heap_init, Bogus call, have %d want %d\n", h->size, new_size); return 0 ; - } + } new_size = (new_size + HEAP_INCREMENT ) & ~HEAP_INCREMENT ; - p = _MALLOC(new_size * sizeof(*p), M_IPFW, M_DONTWAIT ); + p = _MALLOC(new_size * sizeof(*p), M_DUMMYNET, M_DONTWAIT ); if (p == NULL) { - printf(" heap_init, resize %d failed\n", new_size ); + printf("dummynet: heap_init, resize %d failed\n", new_size ); return 1 ; /* error */ } if (h->size > 0) { bcopy(h->p, p, h->size * sizeof(*p) ); - FREE(h->p, M_IPFW); + FREE(h->p, M_DUMMYNET); } h->p = p ; h->size = new_size ; @@ -258,7 +578,7 @@ heap_init(struct dn_heap *h, int new_size) *((int *)((char *)(heap->p[node].object) + heap->offset)) = -1 ; static int heap_insert(struct dn_heap *h, dn_key key1, void *p) -{ +{ int son = h->elements ; if (p == NULL) /* data already there, set starting point */ @@ -277,7 +597,7 @@ heap_insert(struct dn_heap *h, dn_key key1, void *p) struct dn_heap_entry tmp ; if (DN_KEY_LT( h->p[father].key, h->p[son].key ) ) - break ; /* found right position */ + break ; /* found right position */ /* son smaller than father, swap and repeat */ HEAP_SWAP(h->p[son], h->p[father], tmp) ; SET_OFFSET(h, son); @@ -292,84 +612,43 @@ heap_insert(struct dn_heap *h, dn_key key1, void *p) */ static void heap_extract(struct dn_heap *h, void *obj) -{ - int child, father, max = h->elements - 1 ; +{ + int child, father, maxelt = h->elements - 1 ; - if (max < 0) { - printf("warning, extract from empty heap 0x%p\n", h); + if (maxelt < 0) { + printf("dummynet: warning, extract from empty heap 0x%p\n", h); return ; } father = 0 ; /* default: move up smallest child */ if (obj != NULL) { /* extract specific element, index is at offset */ if (h->offset <= 0) - panic("*** heap_extract from middle not supported on this heap!!!\n"); + panic("dummynet: heap_extract from middle not supported on this heap!!!\n"); father = *((int *)((char *)obj + h->offset)) ; if (father < 0 || father >= h->elements) { printf("dummynet: heap_extract, father %d out of bound 0..%d\n", father, h->elements); - panic("heap_extract"); + panic("dummynet: heap_extract"); } } RESET_OFFSET(h, father); child = HEAP_LEFT(father) ; /* left child */ - while (child <= max) { /* valid entry */ - if (child != max && DN_KEY_LT(h->p[child+1].key, h->p[child].key) ) + while (child <= maxelt) { /* valid entry */ + if (child != maxelt && DN_KEY_LT(h->p[child+1].key, h->p[child].key) ) child = child+1 ; /* take right child, otherwise left */ h->p[father] = h->p[child] ; SET_OFFSET(h, father); father = child ; child = HEAP_LEFT(child) ; /* left child for next loop */ - } + } h->elements-- ; - if (father != max) { + if (father != maxelt) { /* * Fill hole with last entry and bubble up, reusing the insert code */ - h->p[father] = h->p[max] ; + h->p[father] = h->p[maxelt] ; heap_insert(h, father, NULL); /* this one cannot fail */ } -} - -#if 0 -/* - * change object position and update references - * XXX this one is never used! - */ -static void -heap_move(struct dn_heap *h, dn_key new_key, void *object) -{ - int temp; - int i ; - int max = h->elements-1 ; - struct dn_heap_entry buf ; - - if (h->offset <= 0) - panic("cannot move items on this heap"); - - i = *((int *)((char *)object + h->offset)); - if (DN_KEY_LT(new_key, h->p[i].key) ) { /* must move up */ - h->p[i].key = new_key ; - for (; i>0 && DN_KEY_LT(new_key, h->p[(temp = HEAP_FATHER(i))].key) ; - i = temp ) { /* bubble up */ - HEAP_SWAP(h->p[i], h->p[temp], buf) ; - SET_OFFSET(h, i); - } - } else { /* must move down */ - h->p[i].key = new_key ; - while ( (temp = HEAP_LEFT(i)) <= max ) { /* found left child */ - if ((temp != max) && DN_KEY_GT(h->p[temp].key, h->p[temp+1].key)) - temp++ ; /* select child with min key */ - if (DN_KEY_GT(new_key, h->p[temp].key)) { /* go down */ - HEAP_SWAP(h->p[i], h->p[temp], buf) ; - SET_OFFSET(h, i); - } else - break ; - i = temp ; - } - } - SET_OFFSET(h, i); } -#endif /* heap_move, unused */ /* * heapify() will reorganize data inside an array to maintain the @@ -391,14 +670,31 @@ static void heap_free(struct dn_heap *h) { if (h->size >0 ) - FREE(h->p, M_IPFW); - bzero(h, sizeof(*h) ); + FREE(h->p, M_DUMMYNET); + bzero(h, sizeof(*h)); } /* * --- end of heap management functions --- */ +/* + * Return the mbuf tag holding the dummynet state. As an optimization + * this is assumed to be the first tag on the list. If this turns out + * wrong we'll need to search the list. + */ +static struct dn_pkt_tag * +dn_tag_get(struct mbuf *m) +{ + struct m_tag *mtag = m_tag_first(m); +/* KASSERT(mtag != NULL && + mtag->m_tag_id == KERNEL_MODULE_TAG_ID && + mtag->m_tag_type == KERNEL_TAG_TYPE_DUMMYNET, + ("packet on dummynet queue w/o dummynet tag!")); +*/ + return (struct dn_pkt_tag *)(mtag+1); +} + /* * Scheduler functions: * @@ -418,78 +714,43 @@ heap_free(struct dn_heap *h) * invocations of the procedures. */ static void -transmit_event(struct dn_pipe *pipe) +transmit_event(struct dn_pipe *pipe, struct mbuf **head, struct mbuf **tail) { - struct dn_pkt *pkt ; - - while ( (pkt = pipe->head) && DN_KEY_LEQ(pkt->output_time, curr_time) ) { - /* - * first unlink, then call procedures, since ip_input() can invoke - * ip_output() and viceversa, thus causing nested calls - */ - pipe->head = DN_NEXT(pkt) ; - - /* - * The actual mbuf is preceded by a struct dn_pkt, resembling an mbuf - * (NOT A REAL one, just a small block of malloc'ed memory) with - * m_type = MT_DUMMYNET - * m_next = actual mbuf to be processed by ip_input/output - * m_data = the matching rule - * and some other fields. - * The block IS FREED HERE because it contains parameters passed - * to the called routine. - */ - switch (pkt->dn_dir) { - case DN_TO_IP_OUT: - (void)ip_output((struct mbuf *)pkt, NULL, NULL, 0, NULL); - rt_unref (pkt->ro.ro_rt) ; - break ; - - case DN_TO_IP_IN : - ip_input((struct mbuf *)pkt) ; - break ; - -#if BRIDGE - case DN_TO_BDG_FWD : { - struct mbuf *m = (struct mbuf *)pkt ; - struct ether_header *eh; - - if (pkt->dn_m->m_len < ETHER_HDR_LEN - && (pkt->dn_m = m_pullup(pkt->dn_m, ETHER_HDR_LEN)) == NULL) { - printf("dummynet/bridge: pullup fail, dropping pkt\n"); - break; - } - /* - * same as ether_input, make eh be a pointer into the mbuf - */ - eh = mtod(pkt->dn_m, struct ether_header *); - m_adj(pkt->dn_m, ETHER_HDR_LEN); - /* - * bdg_forward() wants a pointer to the pseudo-mbuf-header, but - * on return it will supply the pointer to the actual packet - * (originally pkt->dn_m, but could be something else now) if - * it has not consumed it. - */ - m = bdg_forward(m, eh, pkt->ifp); - if (m) - m_freem(m); - } - break ; -#endif - - default: - printf("dummynet: bad switch %d!\n", pkt->dn_dir); - m_freem(pkt->dn_m); - break ; - } - FREE(pkt, M_IPFW); + struct mbuf *m ; + struct dn_pkt_tag *pkt ; + u_int64_t schedule_time; + + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + ASSERT(serialize >= 0); + if (serialize == 0) { + while ((m = pipe->head) != NULL) { + pkt = dn_tag_get(m); + if (!DN_KEY_LEQ(pkt->output_time, curr_time)) + break; + + pipe->head = m->m_nextpkt; + if (*tail != NULL) + (*tail)->m_nextpkt = m; + else + *head = m; + *tail = m; + } + + if (*tail != NULL) + (*tail)->m_nextpkt = NULL; + } + + schedule_time = DN_KEY_LEQ(pkt->output_time, curr_time) ? + curr_time+1 : pkt->output_time; + + /* if there are leftover packets, put the pipe into the heap for next ready event */ + if ((m = pipe->head) != NULL) { + pkt = dn_tag_get(m); + /* XXX should check errors on heap_insert, by draining the + * whole pipe p and hoping in the future we are more successful + */ + heap_insert(&extract_heap, schedule_time, pipe); } - /* if there are leftover packets, put into the heap for next event */ - if ( (pkt = pipe->head) ) - heap_insert(&extract_heap, pkt->output_time, pipe ) ; - /* XXX should check errors on heap_insert, by draining the - * whole pipe p and hoping in the future we are more successful - */ } /* @@ -497,8 +758,15 @@ transmit_event(struct dn_pipe *pipe) * before being able to transmit a packet. The credit is taken from * either a pipe (WF2Q) or a flow_queue (per-flow queueing) */ -#define SET_TICKS(pkt, q, p) \ - (pkt->dn_m->m_pkthdr.len*8*hz - (q)->numbytes + p->bandwidth - 1 ) / \ + +/* hz is 100, which gives a granularity of 10ms in the old timer. + * The timer has been changed to fire every 1ms, so the use of + * hz has been modified here. All instances of hz have been left + * in place but adjusted by a factor of 10 so that hz is functionally + * equal to 1000. + */ +#define SET_TICKS(_m, q, p) \ + ((_m)->m_pkthdr.len*8*(hz*10) - (q)->numbytes + p->bandwidth - 1 ) / \ p->bandwidth ; /* @@ -506,21 +774,23 @@ transmit_event(struct dn_pipe *pipe) * and put into delay line (p_queue) */ static void -move_pkt(struct dn_pkt *pkt, struct dn_flow_queue *q, +move_pkt(struct mbuf *pkt, struct dn_flow_queue *q, struct dn_pipe *p, int len) { - q->head = DN_NEXT(pkt) ; + struct dn_pkt_tag *dt = dn_tag_get(pkt); + + q->head = pkt->m_nextpkt ; q->len-- ; q->len_bytes -= len ; - pkt->output_time = curr_time + p->delay ; + dt->output_time = curr_time + p->delay ; if (p->head == NULL) p->head = pkt; else - DN_NEXT(p->tail) = pkt; + p->tail->m_nextpkt = pkt; p->tail = pkt; - DN_NEXT(p->tail) = NULL; + p->tail->m_nextpkt = NULL; } /* @@ -531,15 +801,17 @@ move_pkt(struct dn_pkt *pkt, struct dn_flow_queue *q, * if there are leftover packets reinsert the pkt in the scheduler. */ static void -ready_event(struct dn_flow_queue *q) +ready_event(struct dn_flow_queue *q, struct mbuf **head, struct mbuf **tail) { - struct dn_pkt *pkt; + struct mbuf *pkt; struct dn_pipe *p = q->fs->pipe ; int p_was_empty ; + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + if (p == NULL) { - printf("ready_event- pipe is gone\n"); - return ; + printf("dummynet: ready_event pipe is gone\n"); + return ; } p_was_empty = (p->head == NULL) ; @@ -553,8 +825,8 @@ ready_event(struct dn_flow_queue *q) */ q->numbytes += ( curr_time - q->sched_time ) * p->bandwidth; while ( (pkt = q->head) != NULL ) { - int len = pkt->dn_m->m_pkthdr.len; - int len_scaled = p->bandwidth ? len*8*hz : 0 ; + int len = pkt->m_pkthdr.len; + int len_scaled = p->bandwidth ? len*8*(hz*10) : 0 ; if (len_scaled > q->numbytes ) break ; q->numbytes -= len_scaled ; @@ -574,14 +846,16 @@ ready_event(struct dn_flow_queue *q) /* XXX should check errors on heap_insert, and drain the whole * queue on error hoping next time we are luckier. */ - } else /* RED needs to know when the queue becomes empty */ + } else { /* RED needs to know when the queue becomes empty */ q->q_time = curr_time; + q->numbytes = 0; + } /* * If the delay line was empty call transmit_event(p) now. * Otherwise, the scheduler will take care of it. */ if (p_was_empty) - transmit_event(p); + transmit_event(p, head, tail); } /* @@ -593,20 +867,23 @@ ready_event(struct dn_flow_queue *q) * there is an additional delay. */ static void -ready_event_wfq(struct dn_pipe *p) +ready_event_wfq(struct dn_pipe *p, struct mbuf **head, struct mbuf **tail) { int p_was_empty = (p->head == NULL) ; struct dn_heap *sch = &(p->scheduler_heap); struct dn_heap *neh = &(p->not_eligible_heap) ; + int64_t p_numbytes = p->numbytes; + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + if (p->if_name[0] == 0) /* tx clock is simulated */ - p->numbytes += ( curr_time - p->sched_time ) * p->bandwidth; + p_numbytes += ( curr_time - p->sched_time ) * p->bandwidth; else { /* tx clock is for real, the ifq must be empty or this is a NOP */ if (p->ifp && p->ifp->if_snd.ifq_head != NULL) return ; else { - DEB(printf("pipe %d ready from %s --\n", - p->pipe_nr, p->if_name);) + DPRINTF(("dummynet: pipe %d ready from %s --\n", + p->pipe_nr, p->if_name)); } } @@ -614,16 +891,16 @@ ready_event_wfq(struct dn_pipe *p) * While we have backlogged traffic AND credit, we need to do * something on the queue. */ - while ( p->numbytes >=0 && (sch->elements>0 || neh->elements >0) ) { + while ( p_numbytes >=0 && (sch->elements>0 || neh->elements >0) ) { if (sch->elements > 0) { /* have some eligible pkts to send out */ struct dn_flow_queue *q = sch->p[0].object ; - struct dn_pkt *pkt = q->head; - struct dn_flow_set *fs = q->fs; - u_int64_t len = pkt->dn_m->m_pkthdr.len; - int len_scaled = p->bandwidth ? len*8*hz : 0 ; + struct mbuf *pkt = q->head; + struct dn_flow_set *fs = q->fs; + u_int64_t len = pkt->m_pkthdr.len; + int len_scaled = p->bandwidth ? len*8*(hz*10) : 0 ; heap_extract(sch, NULL); /* remove queue from heap */ - p->numbytes -= len_scaled ; + p_numbytes -= len_scaled ; move_pkt(pkt, q, p, len); p->V += (len<sum ; /* update V */ @@ -636,7 +913,7 @@ ready_event_wfq(struct dn_pipe *p) * update F and position in backlogged queue, then * put flow in not_eligible_heap (we will fix this later). */ - len = (q->head)->dn_m->m_pkthdr.len; + len = (q->head)->m_pkthdr.len; q->F += (len<weight ; if (DN_KEY_LEQ(q->S, p->V)) heap_insert(neh, q->S, q); @@ -660,11 +937,11 @@ ready_event_wfq(struct dn_pipe *p) } if (p->if_name[0] != '\0') {/* tx clock is from a real thing */ - p->numbytes = -1 ; /* mark not ready for I/O */ + p_numbytes = -1 ; /* mark not ready for I/O */ break ; } } - if (sch->elements == 0 && neh->elements == 0 && p->numbytes >= 0 + if (sch->elements == 0 && neh->elements == 0 && p_numbytes >= 0 && p->idle_heap.elements > 0) { /* * no traffic and no events scheduled. We can get rid of idle-heap. @@ -686,68 +963,90 @@ ready_event_wfq(struct dn_pipe *p) * If we are under credit, schedule the next ready event. * Also fix the delivery time of the last packet. */ - if (p->if_name[0]==0 && p->numbytes < 0) { /* this implies bandwidth >0 */ + if (p->if_name[0]==0 && p_numbytes < 0) { /* this implies bandwidth >0 */ dn_key t=0 ; /* number of ticks i have to wait */ if (p->bandwidth > 0) - t = ( p->bandwidth -1 - p->numbytes) / p->bandwidth ; - p->tail->output_time += t ; + t = ( p->bandwidth -1 - p_numbytes) / p->bandwidth ; + dn_tag_get(p->tail)->output_time += t ; p->sched_time = curr_time ; heap_insert(&wfq_ready_heap, curr_time + t, (void *)p); /* XXX should check errors on heap_insert, and drain the whole * queue on error hoping next time we are luckier. */ } + + /* Fit (adjust if necessary) 64bit result into 32bit variable. */ + if (p_numbytes > INT_MAX) + p->numbytes = INT_MAX; + else if (p_numbytes < INT_MIN) + p->numbytes = INT_MIN; + else + p->numbytes = p_numbytes; + /* * If the delay line was empty call transmit_event(p) now. * Otherwise, the scheduler will take care of it. */ if (p_was_empty) - transmit_event(p); + transmit_event(p, head, tail); + } /* - * This is called once per tick, or HZ times per second. It is used to + * This is called every 1ms. It is used to * increment the current tick counter and schedule expired events. */ static void -dummynet(void * __unused unused) +dummynet(__unused void * unused) { void *p ; /* generic parameter to handler */ struct dn_heap *h ; - int s ; struct dn_heap *heaps[3]; + struct mbuf *head = NULL, *tail = NULL; int i; struct dn_pipe *pe ; + struct timespec ts; + struct timeval tv; heaps[0] = &ready_heap ; /* fixed-rate queues */ heaps[1] = &wfq_ready_heap ; /* wfq queues */ heaps[2] = &extract_heap ; /* delay line */ - s = splimp(); /* see note on top, splnet() is not enough */ - curr_time++ ; + + lck_mtx_lock(dn_mutex); + + /* make all time measurements in milliseconds (ms) - + * here we convert secs and usecs to msecs (just divide the + * usecs and take the closest whole number). + */ + microuptime(&tv); + curr_time = (tv.tv_sec * 1000) + (tv.tv_usec / 1000); + for (i=0; i < 3 ; i++) { h = heaps[i]; while (h->elements > 0 && DN_KEY_LEQ(h->p[0].key, curr_time) ) { - DDB(if (h->p[0].key > curr_time) - printf("-- dummynet: warning, heap %d is %d ticks late\n", - i, (int)(curr_time - h->p[0].key));) - p = h->p[0].object ; /* store a copy before heap_extract */ - heap_extract(h, NULL); /* need to extract before processing */ - if (i == 0) - ready_event(p) ; - else if (i == 1) { - struct dn_pipe *pipe = p; - if (pipe->if_name[0] != '\0') - printf("*** bad ready_event_wfq for pipe %s\n", - pipe->if_name); - else - ready_event_wfq(p) ; - } else - transmit_event(p); + if (h->p[0].key > curr_time) + printf("dummynet: warning, heap %d is %d ticks late\n", + i, (int)(curr_time - h->p[0].key)); + p = h->p[0].object ; /* store a copy before heap_extract */ + heap_extract(h, NULL); /* need to extract before processing */ + if (i == 0) + ready_event(p, &head, &tail) ; + else if (i == 1) { + struct dn_pipe *pipe = p; + if (pipe->if_name[0] != '\0') + printf("dummynet: bad ready_event_wfq for pipe %s\n", + pipe->if_name); + else + ready_event_wfq(p, &head, &tail) ; + } else { + transmit_event(p, &head, &tail); + } } } /* sweep pipes trying to expire idle flow_queues */ - for (pe = all_pipes; pe ; pe = pe->next ) + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(pe, &pipehash[i], next) if (pe->idle_heap.elements > 0 && DN_KEY_LT(pe->idle_heap.p[0].key, pe->V) ) { struct dn_flow_queue *q = pe->idle_heap.p[0].object ; @@ -756,9 +1055,71 @@ dummynet(void * __unused unused) q->S = q->F + 1 ; /* mark timestamp as invalid */ pe->sum -= q->fs->weight ; } - splx(s); - timeout(dummynet, NULL, 1); + + /* check the heaps to see if there's still stuff in there, and + * only set the timer if there are packets to process + */ + timer_enabled = 0; + for (i=0; i < 3 ; i++) { + h = heaps[i]; + if (h->elements > 0) { // set the timer + ts.tv_sec = 0; + ts.tv_nsec = 1 * 1000000; // 1ms + timer_enabled = 1; + bsd_timeout(dummynet, NULL, &ts); + break; + } + } + + if (head != NULL) + serialize++; + + lck_mtx_unlock(dn_mutex); + + /* Send out the de-queued list of ready-to-send packets */ + if (head != NULL) { + dummynet_send(head); + lck_mtx_lock(dn_mutex); + serialize--; + lck_mtx_unlock(dn_mutex); + } +} + + +static void +dummynet_send(struct mbuf *m) +{ + struct dn_pkt_tag *pkt; + struct mbuf *n; + + for (; m != NULL; m = n) { + n = m->m_nextpkt; + m->m_nextpkt = NULL; + pkt = dn_tag_get(m); + + switch (pkt->dn_dir) { + case DN_TO_IP_OUT: { + struct route tmp_rt = pkt->ro; + (void)ip_output(m, NULL, &tmp_rt, pkt->flags, NULL, NULL); + if (tmp_rt.ro_rt) { + rtfree(tmp_rt.ro_rt); + tmp_rt.ro_rt = NULL; + } + break ; + } + case DN_TO_IP_IN : + proto_inject(PF_INET, m); + break ; + + default: + printf("dummynet: bad switch %d!\n", pkt->dn_dir); + m_freem(m); + break ; + } + } } + + /* * called by an interface when tx_rdy occurs. @@ -767,26 +1128,45 @@ int if_tx_rdy(struct ifnet *ifp) { struct dn_pipe *p; - - for (p = all_pipes; p ; p = p->next ) - if (p->ifp == ifp) - break ; + struct mbuf *head = NULL, *tail = NULL; + int i; + + lck_mtx_lock(dn_mutex); + + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(p, &pipehash[i], next) + if (p->ifp == ifp) + break ; if (p == NULL) { char buf[32]; - sprintf(buf, "%s%d",ifp->if_name, ifp->if_unit); - for (p = all_pipes; p ; p = p->next ) + snprintf(buf, sizeof(buf), "%s%d",ifp->if_name, ifp->if_unit); + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(p, &pipehash[i], next) if (!strcmp(p->if_name, buf) ) { p->ifp = ifp ; - DEB(printf("++ tx rdy from %s (now found)\n", buf);) + DPRINTF(("dummynet: ++ tx rdy from %s (now found)\n", buf)); break ; } } if (p != NULL) { - DEB(printf("++ tx rdy from %s%d - qlen %d\n", ifp->if_name, - ifp->if_unit, ifp->if_snd.ifq_len);) + DPRINTF(("dummynet: ++ tx rdy from %s%d - qlen %d\n", ifp->if_name, + ifp->if_unit, ifp->if_snd.ifq_len)); p->numbytes = 0 ; /* mark ready for I/O */ - ready_event_wfq(p); + ready_event_wfq(p, &head, &tail); } + + if (head != NULL) { + serialize++; + } + + lck_mtx_unlock(dn_mutex); + + + /* Send out the de-queued list of ready-to-send packets */ + if (head != NULL) { + dummynet_send(head); + serialize--; + } return 0; } @@ -799,10 +1179,14 @@ expire_queues(struct dn_flow_set *fs) { struct dn_flow_queue *q, *prev ; int i, initial_elements = fs->rq_elements ; + struct timeval timenow; + + /* reviewed for getmicrotime usage */ + getmicrotime(&timenow); - if (fs->last_expired == time_second) + if (fs->last_expired == timenow.tv_sec) return 0 ; - fs->last_expired = time_second ; + fs->last_expired = timenow.tv_sec ; for (i = 0 ; i <= fs->rq_size ; i++) /* last one is overflow */ for (prev=NULL, q = fs->rq[i] ; q != NULL ; ) if (q->head != NULL || q->S != q->F+1) { @@ -816,7 +1200,7 @@ expire_queues(struct dn_flow_set *fs) else fs->rq[i] = q = q->next ; fs->rq_elements-- ; - FREE(old_q, M_IPFW); + FREE(old_q, M_DUMMYNET); } return initial_elements - fs->rq_elements ; } @@ -839,12 +1223,11 @@ create_queue(struct dn_flow_set *fs, int i) if ( fs->rq[i] != NULL ) return fs->rq[i] ; } - q = _MALLOC(sizeof(*q), M_IPFW, M_DONTWAIT) ; + q = _MALLOC(sizeof(*q), M_DUMMYNET, M_DONTWAIT | M_ZERO); if (q == NULL) { - printf("sorry, cannot allocate queue for new flow\n"); + printf("dummynet: sorry, cannot allocate queue for new flow\n"); return NULL ; } - bzero(q, sizeof(*q) ); /* needed */ q->fs = fs ; q->hash_slot = i ; q->next = fs->rq[i] ; @@ -860,7 +1243,7 @@ create_queue(struct dn_flow_set *fs, int i) * so that further searches take less time. */ static struct dn_flow_queue * -find_queue(struct dn_flow_set *fs) +find_queue(struct dn_flow_set *fs, struct ipfw_flow_id *id) { int i = 0 ; /* we need i and q for new allocations */ struct dn_flow_queue *q, *prev; @@ -869,25 +1252,30 @@ find_queue(struct dn_flow_set *fs) q = fs->rq[0] ; else { /* first, do the masking */ - last_pkt.dst_ip &= fs->flow_mask.dst_ip ; - last_pkt.src_ip &= fs->flow_mask.src_ip ; - last_pkt.dst_port &= fs->flow_mask.dst_port ; - last_pkt.src_port &= fs->flow_mask.src_port ; - last_pkt.proto &= fs->flow_mask.proto ; - last_pkt.flags = 0 ; /* we don't care about this one */ + id->dst_ip &= fs->flow_mask.dst_ip ; + id->src_ip &= fs->flow_mask.src_ip ; + id->dst_port &= fs->flow_mask.dst_port ; + id->src_port &= fs->flow_mask.src_port ; + id->proto &= fs->flow_mask.proto ; + id->flags = 0 ; /* we don't care about this one */ /* then, hash function */ - i = ( (last_pkt.dst_ip) & 0xffff ) ^ - ( (last_pkt.dst_ip >> 15) & 0xffff ) ^ - ( (last_pkt.src_ip << 1) & 0xffff ) ^ - ( (last_pkt.src_ip >> 16 ) & 0xffff ) ^ - (last_pkt.dst_port << 1) ^ (last_pkt.src_port) ^ - (last_pkt.proto ); + i = ( (id->dst_ip) & 0xffff ) ^ + ( (id->dst_ip >> 15) & 0xffff ) ^ + ( (id->src_ip << 1) & 0xffff ) ^ + ( (id->src_ip >> 16 ) & 0xffff ) ^ + (id->dst_port << 1) ^ (id->src_port) ^ + (id->proto ); i = i % fs->rq_size ; /* finally, scan the current list for a match */ searches++ ; for (prev=NULL, q = fs->rq[i] ; q ; ) { search_steps++; - if (bcmp(&last_pkt, &(q->id), sizeof(q->id) ) == 0) + if (id->dst_ip == q->id.dst_ip && + id->src_ip == q->id.src_ip && + id->dst_port == q->id.dst_port && + id->src_port == q->id.src_port && + id->proto == q->id.proto && + id->flags == q->id.flags) break ; /* found */ else if (pipe_expire && q->head == NULL && q->S == q->F+1 ) { /* entry is idle and not in any heap, expire it */ @@ -898,7 +1286,7 @@ find_queue(struct dn_flow_set *fs) else fs->rq[i] = q = q->next ; fs->rq_elements-- ; - FREE(old_q, M_IPFW); + FREE(old_q, M_DUMMYNET); continue ; } prev = q ; @@ -913,7 +1301,7 @@ find_queue(struct dn_flow_set *fs) if (q == NULL) { /* no match, need to allocate a new entry */ q = create_queue(fs, i); if (q != NULL) - q->id = last_pkt ; + q->id = *id ; } return q ; } @@ -923,27 +1311,27 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) { /* * RED algorithm - * + * * RED calculates the average queue size (avg) using a low-pass filter * with an exponential weighted (w_q) moving average: * avg <- (1-w_q) * avg + w_q * q_size * where q_size is the queue length (measured in bytes or * packets). - * + * * If q_size == 0, we compute the idle time for the link, and set * avg = (1 - w_q)^(idle/s) * where s is the time needed for transmitting a medium-sized packet. - * + * * Now, if avg < min_th the packet is enqueued. * If avg > max_th the packet is dropped. Otherwise, the packet is * dropped with probability P function of avg. - * + * */ int64_t p_b = 0; /* queue in bytes or packets ? */ u_int q_size = (fs->flags_fs & DN_QSIZE_IS_BYTES) ? q->len_bytes : q->len; - DEB(printf("\n%d q: %2u ", (int) curr_time, q_size);) + DPRINTF(("\ndummynet: %d q: %2u ", (int) curr_time, q_size)); /* average queue size estimation */ if (q_size != 0) { @@ -969,7 +1357,7 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) SCALE_MUL(q->avg, fs->w_q_lookup[t]) : 0; } } - DEB(printf("avg: %u ", SCALE_VAL(q->avg));) + DPRINTF(("dummynet: avg: %u ", SCALE_VAL(q->avg))); /* should i drop ? */ @@ -988,7 +1376,7 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) p_b = SCALE_MUL((int64_t) fs->c_3, (int64_t) q->avg) - fs->c_4; } else { q->count = -1; - printf("- drop"); + DPRINTF(("dummynet: - drop")); return 1 ; } } else if (q->avg > fs->min_th) { @@ -1002,7 +1390,7 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) if (fs->flags_fs & DN_QSIZE_IS_BYTES) p_b = (p_b * len) / fs->max_pkt_size; if (++q->count == 0) - q->random = random() & 0xffff; + q->random = MY_RANDOM & 0xffff; else { /* * q->count counts packets arrived since last drop, so a greater @@ -1010,9 +1398,9 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) */ if (SCALE_MUL(p_b, SCALE((int64_t) q->count)) > q->random) { q->count = 0; - DEB(printf("- red drop");) + DPRINTF(("dummynet: - red drop")); /* after a drop we calculate a new random value */ - q->random = random() & 0xffff; + q->random = MY_RANDOM & 0xffff; return 1; /* drop */ } } @@ -1022,65 +1410,108 @@ red_drops(struct dn_flow_set *fs, struct dn_flow_queue *q, int len) static __inline struct dn_flow_set * -locate_flowset(int pipe_nr, struct ip_fw_chain *rule) +locate_flowset(int fs_nr) { - struct dn_flow_set *fs = NULL ; + struct dn_flow_set *fs; + SLIST_FOREACH(fs, &flowsethash[HASH(fs_nr)], next) + if (fs->fs_nr == fs_nr) + return fs ; + + return (NULL); +} - if ( (rule->rule->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_QUEUE ) - for (fs=all_flow_sets; fs && fs->fs_nr != pipe_nr; fs=fs->next) - ; - else { - struct dn_pipe *p1; - for (p1 = all_pipes; p1 && p1->pipe_nr != pipe_nr; p1 = p1->next) - ; - if (p1 != NULL) - fs = &(p1->fs) ; - } - if (fs != NULL) - rule->rule->pipe_ptr = fs ; /* record for the future */ - return fs ; +static __inline struct dn_pipe * +locate_pipe(int pipe_nr) +{ + struct dn_pipe *pipe; + + SLIST_FOREACH(pipe, &pipehash[HASH(pipe_nr)], next) + if (pipe->pipe_nr == pipe_nr) + return (pipe); + + return (NULL); } + + /* * dummynet hook for packets. Below 'pipe' is a pipe or a queue * depending on whether WF2Q or fixed bw is used. + * + * pipe_nr pipe or queue the packet is destined for. + * dir where shall we send the packet after dummynet. + * m the mbuf with the packet + * ifp the 'ifp' parameter from the caller. + * NULL in ip_input, destination interface in ip_output, + * real_dst in bdg_forward + * ro route parameter (only used in ip_output, NULL otherwise) + * dst destination address, only used by ip_output + * rule matching rule, in case of multiple passes + * flags flags from the caller, only used in ip_output + * */ -int -dummynet_io(int pipe_nr, int dir, /* pipe_nr can also be a fs_nr */ - struct mbuf *m, struct ifnet *ifp, struct route *ro, - struct sockaddr_in *dst, - struct ip_fw_chain *rule, int flags) +static int +dummynet_io(struct mbuf *m, int pipe_nr, int dir, struct ip_fw_args *fwa) { - struct dn_pkt *pkt; - struct dn_flow_set *fs; + struct mbuf *head = NULL, *tail = NULL; + struct dn_pkt_tag *pkt; + struct m_tag *mtag; + struct dn_flow_set *fs = NULL; struct dn_pipe *pipe ; u_int64_t len = m->m_pkthdr.len ; struct dn_flow_queue *q = NULL ; - int s ; - - s = splimp(); + int is_pipe; + struct timespec ts; + struct timeval tv; + +#if IPFW2 + ipfw_insn *cmd = fwa->rule->cmd + fwa->rule->act_ofs; + + if (cmd->opcode == O_LOG) + cmd += F_LEN(cmd); + is_pipe = (cmd->opcode == O_PIPE); +#else + is_pipe = (fwa->rule->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_PIPE; +#endif pipe_nr &= 0xffff ; - if ( (fs = rule->rule->pipe_ptr) == NULL ) { - fs = locate_flowset(pipe_nr, rule); - if (fs == NULL) - goto dropit ; /* this queue/pipe does not exist! */ + lck_mtx_lock(dn_mutex); + + /* make all time measurements in milliseconds (ms) - + * here we convert secs and usecs to msecs (just divide the + * usecs and take the closest whole number). + */ + microuptime(&tv); + curr_time = (tv.tv_sec * 1000) + (tv.tv_usec / 1000); + + /* + * This is a dummynet rule, so we expect an O_PIPE or O_QUEUE rule. + */ + if (is_pipe) { + pipe = locate_pipe(pipe_nr); + if (pipe != NULL) + fs = &(pipe->fs); + } else + fs = locate_flowset(pipe_nr); + + + if (fs == NULL){ + goto dropit ; /* this queue/pipe does not exist! */ } pipe = fs->pipe ; if (pipe == NULL) { /* must be a queue, try find a matching pipe */ - for (pipe = all_pipes; pipe && pipe->pipe_nr != fs->parent_nr; - pipe = pipe->next) - ; + pipe = locate_pipe(fs->parent_nr); + if (pipe != NULL) fs->pipe = pipe ; else { - printf("No pipe %d for queue %d, drop pkt\n", + printf("dummynet: no pipe %d for queue %d, drop pkt\n", fs->parent_nr, fs->fs_nr); goto dropit ; } } - q = find_queue(fs); + q = find_queue(fs, &(fwa->f_id)); if ( q == NULL ) goto dropit ; /* cannot allocate queue */ /* @@ -1088,7 +1519,7 @@ dummynet_io(int pipe_nr, int dir, /* pipe_nr can also be a fs_nr */ */ q->tot_bytes += len ; q->tot_pkts++ ; - if ( fs->plr && random() < fs->plr ) + if ( fs->plr && (MY_RANDOM < fs->plr) ) goto dropit ; /* random pkt drop */ if ( fs->flags_fs & DN_QSIZE_IS_BYTES) { if (q->len_bytes > fs->qsize) @@ -1100,59 +1531,64 @@ dummynet_io(int pipe_nr, int dir, /* pipe_nr can also be a fs_nr */ if ( fs->flags_fs & DN_IS_RED && red_drops(fs, q, len) ) goto dropit ; - pkt = (struct dn_pkt *)_MALLOC(sizeof (*pkt), M_IPFW, M_NOWAIT) ; - if ( pkt == NULL ) - goto dropit ; /* cannot allocate packet header */ + /* XXX expensive to zero, see if we can remove it*/ + mtag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DUMMYNET, + sizeof(struct dn_pkt_tag), M_NOWAIT, m); + if ( mtag == NULL ) + goto dropit ; /* cannot allocate packet header */ + m_tag_prepend(m, mtag); /* attach to mbuf chain */ + + pkt = (struct dn_pkt_tag *)(mtag+1); + bzero(pkt, sizeof(struct dn_pkt_tag)); /* ok, i can handle the pkt now... */ - bzero(pkt, sizeof(*pkt) ); /* XXX expensive, see if we can remove it*/ /* build and enqueue packet + parameters */ - pkt->hdr.mh_type = MT_DUMMYNET ; - (struct ip_fw_chain *)pkt->hdr.mh_data = rule ; - DN_NEXT(pkt) = NULL; - pkt->dn_m = m; + pkt->rule = fwa->rule ; pkt->dn_dir = dir ; - pkt->ifp = ifp; + pkt->ifp = fwa->oif; if (dir == DN_TO_IP_OUT) { /* * We need to copy *ro because for ICMP pkts (and maybe others) * the caller passed a pointer into the stack; dst might also be * a pointer into *ro so it needs to be updated. */ - pkt->ro = *ro; - if (ro->ro_rt) - rtref(ro->ro_rt); - if (dst == (struct sockaddr_in *)&ro->ro_dst) /* dst points into ro */ - dst = (struct sockaddr_in *)&(pkt->ro.ro_dst) ; - - pkt->dn_dst = dst; - pkt->flags = flags ; - } + pkt->ro = *(fwa->ro); + if (fwa->ro->ro_rt) + RT_ADDREF(fwa->ro->ro_rt); + + if (fwa->dst == (struct sockaddr_in *)&fwa->ro->ro_dst) /* dst points into ro */ + fwa->dst = (struct sockaddr_in *)&(pkt->ro.ro_dst) ; + + bcopy (fwa->dst, &pkt->dn_dst, sizeof(pkt->dn_dst)); + pkt->flags = fwa->flags; + if (fwa->ipoa != NULL) + pkt->ipoa = *(fwa->ipoa); + } if (q->head == NULL) - q->head = pkt; + q->head = m; else - DN_NEXT(q->tail) = pkt; - q->tail = pkt; + q->tail->m_nextpkt = m; + q->tail = m; q->len++; q->len_bytes += len ; - if ( q->head != pkt ) /* flow was not idle, we are done */ + if ( q->head != m ) /* flow was not idle, we are done */ goto done; /* * If we reach this point the flow was previously idle, so we need * to schedule it. This involves different actions for fixed-rate or * WF2Q queues. */ - if ( (rule->rule->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_PIPE ) { + if (is_pipe) { /* * Fixed-rate queue: just insert into the ready_heap. */ dn_key t = 0 ; - if (pipe->bandwidth) - t = SET_TICKS(pkt, q, pipe); + if (pipe->bandwidth) + t = SET_TICKS(m, q, pipe); q->sched_time = curr_time ; - if (t == 0) /* must process it now */ - ready_event( q ); + if (t == 0) /* must process it now */ + ready_event( q , &head, &tail ); else heap_insert(&ready_heap, curr_time + t , q ); } else { @@ -1193,42 +1629,60 @@ dummynet_io(int pipe_nr, int dir, /* pipe_nr can also be a fs_nr */ */ if (DN_KEY_GT(q->S, pipe->V) ) { /* not eligible */ if (pipe->scheduler_heap.elements == 0) - printf("++ ouch! not eligible but empty scheduler!\n"); + printf("dummynet: ++ ouch! not eligible but empty scheduler!\n"); heap_insert(&(pipe->not_eligible_heap), q->S, q); } else { heap_insert(&(pipe->scheduler_heap), q->F, q); if (pipe->numbytes >= 0) { /* pipe is idle */ if (pipe->scheduler_heap.elements != 1) - printf("*** OUCH! pipe should have been idle!\n"); - DEB(printf("Waking up pipe %d at %d\n", - pipe->pipe_nr, (int)(q->F >> MY_M)); ) + printf("dummynet: OUCH! pipe should have been idle!\n"); + DPRINTF(("dummynet: waking up pipe %d at %d\n", + pipe->pipe_nr, (int)(q->F >> MY_M))); pipe->sched_time = curr_time ; - ready_event_wfq(pipe); + ready_event_wfq(pipe, &head, &tail); } } } done: - splx(s); + /* start the timer and set global if not already set */ + if (!timer_enabled) { + ts.tv_sec = 0; + ts.tv_nsec = 1 * 1000000; // 1ms + timer_enabled = 1; + bsd_timeout(dummynet, NULL, &ts); + } + + lck_mtx_unlock(dn_mutex); + + if (head != NULL) + dummynet_send(head); + return 0; dropit: - splx(s); if (q) q->drops++ ; + lck_mtx_unlock(dn_mutex); m_freem(m); - return ENOBUFS ; + return ( (fs && (fs->flags_fs & DN_NOERROR)) ? 0 : ENOBUFS); } /* - * Below, the rt_unref is only needed when (pkt->dn_dir == DN_TO_IP_OUT) + * Below, the rtfree is only needed when (pkt->dn_dir == DN_TO_IP_OUT) * Doing this would probably save us the initial bzero of dn_pkt */ -#define DN_FREE_PKT(pkt) { \ - struct dn_pkt *n = pkt ; \ - rt_unref ( n->ro.ro_rt ) ; \ - m_freem(n->dn_m); \ - pkt = DN_NEXT(n) ; \ - FREE(n, M_IPFW) ; } +#define DN_FREE_PKT(_m) do { \ + struct m_tag *tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DUMMYNET, NULL); \ + if (tag) { \ + struct dn_pkt_tag *n = (struct dn_pkt_tag *)(tag+1); \ + if (n->ro.ro_rt != NULL) { \ + rtfree(n->ro.ro_rt); \ + n->ro.ro_rt = NULL; \ + } \ + } \ + m_tag_delete(_m, tag); \ + m_freem(_m); \ +} while (0) /* * Dispose all packets and flow_queues on a flow_set. @@ -1239,16 +1693,22 @@ dropit: static void purge_flow_set(struct dn_flow_set *fs, int all) { - struct dn_pkt *pkt ; struct dn_flow_queue *q, *qn ; int i ; + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + for (i = 0 ; i <= fs->rq_size ; i++ ) { for (q = fs->rq[i] ; q ; q = qn ) { - for (pkt = q->head ; pkt ; ) - DN_FREE_PKT(pkt) ; + struct mbuf *m, *mnext; + + mnext = q->head; + while ((m = mnext) != NULL) { + mnext = m->m_nextpkt; + DN_FREE_PKT(m); + } qn = q->next ; - FREE(q, M_IPFW); + FREE(q, M_DUMMYNET); } fs->rq[i] = NULL ; } @@ -1256,12 +1716,12 @@ purge_flow_set(struct dn_flow_set *fs, int all) if (all) { /* RED - free lookup table */ if (fs->w_q_lookup) - FREE(fs->w_q_lookup, M_IPFW); + FREE(fs->w_q_lookup, M_DUMMYNET); if (fs->rq) - FREE(fs->rq, M_IPFW); + FREE(fs->rq, M_DUMMYNET); /* if this fs is not part of a pipe, free it */ if (fs->pipe && fs != &(fs->pipe->fs) ) - FREE(fs, M_IPFW); + FREE(fs, M_DUMMYNET); } } @@ -1273,12 +1733,15 @@ purge_flow_set(struct dn_flow_set *fs, int all) static void purge_pipe(struct dn_pipe *pipe) { - struct dn_pkt *pkt ; + struct mbuf *m, *mnext; purge_flow_set( &(pipe->fs), 1 ); - for (pkt = pipe->head ; pkt ; ) - DN_FREE_PKT(pkt) ; + mnext = pipe->head; + while ((m = mnext) != NULL) { + mnext = m->m_nextpkt; + DN_FREE_PKT(m); + } heap_free( &(pipe->scheduler_heap) ); heap_free( &(pipe->not_eligible_heap) ); @@ -1290,59 +1753,57 @@ purge_pipe(struct dn_pipe *pipe) * remove references from all ipfw rules to all pipes. */ static void -dummynet_flush() +dummynet_flush(void) { - struct dn_pipe *curr_p, *p ; - struct ip_fw_chain *chain ; - struct dn_flow_set *fs, *curr_fs; - int s ; + struct dn_pipe *pipe, *pipe1; + struct dn_flow_set *fs, *fs1; + int i; - s = splimp() ; + lck_mtx_lock(dn_mutex); /* remove all references to pipes ...*/ - LIST_FOREACH(chain, &ip_fw_chain_head, next) - chain->rule->pipe_ptr = NULL ; - /* prevent future matches... */ - p = all_pipes ; - all_pipes = NULL ; - fs = all_flow_sets ; - all_flow_sets = NULL ; - /* and free heaps so we don't have unwanted events */ - heap_free(&ready_heap); - heap_free(&wfq_ready_heap); - heap_free(&extract_heap); - splx(s) ; - /* - * Now purge all queued pkts and delete all pipes - */ - /* scan and purge all flow_sets. */ - for ( ; fs ; ) { - curr_fs = fs ; - fs = fs->next ; - purge_flow_set(curr_fs, 1); - } - for ( ; p ; ) { - purge_pipe(p); - curr_p = p ; - p = p->next ; - FREE(q, M_IPFW); - } + flush_pipe_ptrs(NULL); + + /* Free heaps so we don't have unwanted events. */ + heap_free(&ready_heap); + heap_free(&wfq_ready_heap); + heap_free(&extract_heap); + + /* + * Now purge all queued pkts and delete all pipes. + * + * XXXGL: can we merge the for(;;) cycles into one or not? + */ + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH_SAFE(fs, &flowsethash[i], next, fs1) { + SLIST_REMOVE(&flowsethash[i], fs, dn_flow_set, next); + purge_flow_set(fs, 1); + } + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH_SAFE(pipe, &pipehash[i], next, pipe1) { + SLIST_REMOVE(&pipehash[i], pipe, dn_pipe, next); + purge_pipe(pipe); + FREE(pipe, M_DUMMYNET); + } + lck_mtx_unlock(dn_mutex); } -extern struct ip_fw_chain *ip_fw_default_rule ; +extern struct ip_fw *ip_fw_default_rule ; static void dn_rule_delete_fs(struct dn_flow_set *fs, void *r) { int i ; struct dn_flow_queue *q ; - struct dn_pkt *pkt ; + struct mbuf *m ; for (i = 0 ; i <= fs->rq_size ; i++) /* last one is ovflow */ for (q = fs->rq[i] ; q ; q = q->next ) - for (pkt = q->head ; pkt ; pkt = DN_NEXT(pkt) ) - if (pkt->hdr.mh_data == r) - pkt->hdr.mh_data = (void *)ip_fw_default_rule ; + for (m = q->head ; m ; m = m->m_nextpkt ) { + struct dn_pkt_tag *pkt = dn_tag_get(m) ; + if (pkt->rule == r) + pkt->rule = ip_fw_default_rule ; + } } /* * when a firewall rule is deleted, scan all queues and remove the flow-id @@ -1352,30 +1813,40 @@ void dn_rule_delete(void *r) { struct dn_pipe *p ; - struct dn_pkt *pkt ; struct dn_flow_set *fs ; + struct dn_pkt_tag *pkt ; + struct mbuf *m ; + int i; + + lck_mtx_lock(dn_mutex); /* * If the rule references a queue (dn_flow_set), then scan * the flow set, otherwise scan pipes. Should do either, but doing * both does not harm. */ - for ( fs = all_flow_sets ; fs ; fs = fs->next ) - dn_rule_delete_fs(fs, r); - for ( p = all_pipes ; p ; p = p->next ) { - fs = &(p->fs) ; - dn_rule_delete_fs(fs, r); - for (pkt = p->head ; pkt ; pkt = DN_NEXT(pkt) ) - if (pkt->hdr.mh_data == r) - pkt->hdr.mh_data = (void *)ip_fw_default_rule ; - } + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(fs, &flowsethash[i], next) + dn_rule_delete_fs(fs, r); + + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(p, &pipehash[i], next) { + fs = &(p->fs); + dn_rule_delete_fs(fs, r); + for (m = p->head ; m ; m = m->m_nextpkt ) { + pkt = dn_tag_get(m); + if (pkt->rule == r) + pkt->rule = ip_fw_default_rule; + } + } + lck_mtx_unlock(dn_mutex); } /* * setup RED parameters */ static int -config_red(struct dn_flow_set *p, struct dn_flow_set * x) +config_red(struct dn_flow_set *p, struct dn_flow_set * x) { int i; @@ -1392,19 +1863,21 @@ config_red(struct dn_flow_set *p, struct dn_flow_set * x) } /* if the lookup table already exist, free and create it again */ - if (x->w_q_lookup) - FREE(x->w_q_lookup, M_IPFW); + if (x->w_q_lookup) { + FREE(x->w_q_lookup, M_DUMMYNET); + x->w_q_lookup = NULL ; + } if (red_lookup_depth == 0) { - printf("\nnet.inet.ip.dummynet.red_lookup_depth must be > 0"); - FREE(x, M_IPFW); + printf("\ndummynet: net.inet.ip.dummynet.red_lookup_depth must be > 0\n"); + FREE(x, M_DUMMYNET); return EINVAL; } x->lookup_depth = red_lookup_depth; x->w_q_lookup = (u_int *) _MALLOC(x->lookup_depth * sizeof(int), - M_IPFW, M_DONTWAIT); + M_DUMMYNET, M_DONTWAIT); if (x->w_q_lookup == NULL) { - printf("sorry, cannot allocate red lookup table\n"); - FREE(x, M_IPFW); + printf("dummynet: sorry, cannot allocate red lookup table\n"); + FREE(x, M_DUMMYNET); return ENOSPC; } @@ -1433,18 +1906,17 @@ alloc_hash(struct dn_flow_set *x, struct dn_flow_set *pfs) l = dn_hash_size; if (l < 4) l = 4; - else if (l > 1024) - l = 1024; + else if (l > DN_MAX_HASH_SIZE) + l = DN_MAX_HASH_SIZE; x->rq_size = l; } else /* one is enough for null mask */ x->rq_size = 1; x->rq = _MALLOC((1 + x->rq_size) * sizeof(struct dn_flow_queue *), - M_IPFW, M_DONTWAIT); + M_DUMMYNET, M_DONTWAIT | M_ZERO); if (x->rq == NULL) { - printf("sorry, cannot allocate queue\n"); + printf("dummynet: sorry, cannot allocate queue\n"); return ENOSPC; } - bzero(x->rq, (1+x->rq_size) * sizeof(struct dn_flow_queue *)); x->rq_elements = 0; return 0 ; } @@ -1461,9 +1933,9 @@ set_fs_parms(struct dn_flow_set *x, struct dn_flow_set *src) x->qsize = 1024*1024 ; } else { if (x->qsize == 0) - x->qsize = 50 ; + x->qsize = 50; if (x->qsize > 100) - x->qsize = 50 ; + x->qsize = 50; } /* configuring RED */ if ( x->flags_fs & DN_IS_RED ) @@ -1474,84 +1946,91 @@ set_fs_parms(struct dn_flow_set *x, struct dn_flow_set *src) * setup pipe or queue parameters. */ -static int +static int config_pipe(struct dn_pipe *p) { - int s ; + int i, r; struct dn_flow_set *pfs = &(p->fs); + struct dn_flow_queue *q; - /* - * The config program passes parameters as follows: + /* + * The config program passes parameters as follows: * bw = bits/second (0 means no limits), * delay = ms, must be translated into ticks. * qsize = slots/bytes - */ - p->delay = ( p->delay * hz ) / 1000 ; + */ + p->delay = ( p->delay * (hz*10) ) / 1000 ; /* We need either a pipe number or a flow_set number */ if (p->pipe_nr == 0 && pfs->fs_nr == 0) return EINVAL ; if (p->pipe_nr != 0 && pfs->fs_nr != 0) return EINVAL ; if (p->pipe_nr != 0) { /* this is a pipe */ - struct dn_pipe *x, *a, *b; - /* locate pipe */ - for (a = NULL , b = all_pipes ; b && b->pipe_nr < p->pipe_nr ; - a = b , b = b->next) ; + struct dn_pipe *x, *b; + lck_mtx_lock(dn_mutex); + + /* locate pipe */ + b = locate_pipe(p->pipe_nr); + if (b == NULL || b->pipe_nr != p->pipe_nr) { /* new pipe */ - x = _MALLOC(sizeof(struct dn_pipe), M_IPFW, M_DONTWAIT) ; + x = _MALLOC(sizeof(struct dn_pipe), M_DUMMYNET, M_DONTWAIT | M_ZERO) ; if (x == NULL) { - printf("ip_dummynet.c: no memory for new pipe\n"); + lck_mtx_unlock(dn_mutex); + printf("dummynet: no memory for new pipe\n"); return ENOSPC; } - bzero(x, sizeof(struct dn_pipe)); x->pipe_nr = p->pipe_nr; x->fs.pipe = x ; /* idle_heap is the only one from which we extract from the middle. */ x->idle_heap.size = x->idle_heap.elements = 0 ; - x->idle_heap.offset=OFFSET_OF(struct dn_flow_queue, heap_pos); - } else + x->idle_heap.offset=offsetof(struct dn_flow_queue, heap_pos); + } else { x = b; + /* Flush accumulated credit for all queues */ + for (i = 0; i <= x->fs.rq_size; i++) + for (q = x->fs.rq[i]; q; q = q->next) + q->numbytes = 0; + } - x->bandwidth = p->bandwidth ; + x->bandwidth = p->bandwidth ; x->numbytes = 0; /* just in case... */ bcopy(p->if_name, x->if_name, sizeof(p->if_name) ); x->ifp = NULL ; /* reset interface ptr */ - x->delay = p->delay ; + x->delay = p->delay ; set_fs_parms(&(x->fs), pfs); if ( x->fs.rq == NULL ) { /* a new pipe */ - s = alloc_hash(&(x->fs), pfs) ; - if (s) { - FREE(x, M_IPFW); - return s ; + r = alloc_hash(&(x->fs), pfs) ; + if (r) { + lck_mtx_unlock(dn_mutex); + FREE(x, M_DUMMYNET); + return r ; } - s = splimp() ; - x->next = b ; - if (a == NULL) - all_pipes = x ; - else - a->next = x ; - splx(s); + SLIST_INSERT_HEAD(&pipehash[HASH(x->pipe_nr)], + x, next); } + lck_mtx_unlock(dn_mutex); } else { /* config queue */ - struct dn_flow_set *x, *a, *b ; + struct dn_flow_set *x, *b ; + lck_mtx_lock(dn_mutex); /* locate flow_set */ - for (a=NULL, b=all_flow_sets ; b && b->fs_nr < pfs->fs_nr ; - a = b , b = b->next) ; + b = locate_flowset(pfs->fs_nr); if (b == NULL || b->fs_nr != pfs->fs_nr) { /* new */ - if (pfs->parent_nr == 0) /* need link to a pipe */ - return EINVAL ; - x = _MALLOC(sizeof(struct dn_flow_set), M_IPFW, M_DONTWAIT); + if (pfs->parent_nr == 0) { /* need link to a pipe */ + lck_mtx_unlock(dn_mutex); + return EINVAL ; + } + x = _MALLOC(sizeof(struct dn_flow_set), M_DUMMYNET, M_DONTWAIT | M_ZERO); if (x == NULL) { - printf("ip_dummynet.c: no memory for new flow_set\n"); - return ENOSPC; + lck_mtx_unlock(dn_mutex); + printf("dummynet: no memory for new flow_set\n"); + return ENOSPC; } - bzero(x, sizeof(struct dn_flow_set)); x->fs_nr = pfs->fs_nr; x->parent_nr = pfs->parent_nr; x->weight = pfs->weight ; @@ -1561,26 +2040,25 @@ config_pipe(struct dn_pipe *p) x->weight = 100 ; } else { /* Change parent pipe not allowed; must delete and recreate */ - if (pfs->parent_nr != 0 && b->parent_nr != pfs->parent_nr) - return EINVAL ; + if (pfs->parent_nr != 0 && b->parent_nr != pfs->parent_nr) { + lck_mtx_unlock(dn_mutex); + return EINVAL ; + } x = b; } set_fs_parms(x, pfs); if ( x->rq == NULL ) { /* a new flow_set */ - s = alloc_hash(x, pfs) ; - if (s) { - FREE(x, M_IPFW); - return s ; + r = alloc_hash(x, pfs) ; + if (r) { + lck_mtx_unlock(dn_mutex); + FREE(x, M_DUMMYNET); + return r ; } - s = splimp() ; - x->next = b; - if (a == NULL) - all_flow_sets = x; - else - a->next = x; - splx(s); + SLIST_INSERT_HEAD(&flowsethash[HASH(x->fs_nr)], + x, next); } + lck_mtx_unlock(dn_mutex); } return 0 ; } @@ -1627,23 +2105,32 @@ pipe_remove_from_heap(struct dn_heap *h, struct dn_pipe *p) * drain all queues. Called in case of severe mbuf shortage. */ void -dummynet_drain() +dummynet_drain(void) { struct dn_flow_set *fs; struct dn_pipe *p; - struct dn_pkt *pkt; + struct mbuf *m, *mnext; + int i; + + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); heap_free(&ready_heap); heap_free(&wfq_ready_heap); heap_free(&extract_heap); /* remove all references to this pipe from flow_sets */ - for (fs = all_flow_sets; fs; fs= fs->next ) - purge_flow_set(fs, 0); + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(fs, &flowsethash[i], next) + purge_flow_set(fs, 0); + + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(p, &pipehash[i], next) { + purge_flow_set(&(p->fs), 0); - for (p = all_pipes; p; p= p->next ) { - purge_flow_set(&(p->fs), 0); - for (pkt = p->head ; pkt ; ) - DN_FREE_PKT(pkt) ; + mnext = p->head; + while ((m = mnext) != NULL) { + mnext = m->m_nextpkt; + DN_FREE_PKT(m); + } p->head = p->tail = NULL ; } } @@ -1651,71 +2138,66 @@ dummynet_drain() /* * Fully delete a pipe or a queue, cleaning up associated info. */ -static int +static int delete_pipe(struct dn_pipe *p) { - int s ; - struct ip_fw_chain *chain ; - if (p->pipe_nr == 0 && p->fs.fs_nr == 0) return EINVAL ; if (p->pipe_nr != 0 && p->fs.fs_nr != 0) return EINVAL ; if (p->pipe_nr != 0) { /* this is an old-style pipe */ - struct dn_pipe *a, *b; + struct dn_pipe *b; struct dn_flow_set *fs; + int i; + lck_mtx_lock(dn_mutex); /* locate pipe */ - for (a = NULL , b = all_pipes ; b && b->pipe_nr < p->pipe_nr ; - a = b , b = b->next) ; - if (b == NULL || (b->pipe_nr != p->pipe_nr) ) + b = locate_pipe(p->pipe_nr); + if(b == NULL){ + lck_mtx_unlock(dn_mutex); return EINVAL ; /* not found */ + } - s = splimp() ; + /* Unlink from list of pipes. */ + SLIST_REMOVE(&pipehash[HASH(b->pipe_nr)], b, dn_pipe, next); - /* unlink from list of pipes */ - if (a == NULL) - all_pipes = b->next ; - else - a->next = b->next ; /* remove references to this pipe from the ip_fw rules. */ - LIST_FOREACH(chain, &ip_fw_chain_head, next) - if (chain->rule->pipe_ptr == &(b->fs)) - chain->rule->pipe_ptr = NULL ; - - /* remove all references to this pipe from flow_sets */ - for (fs = all_flow_sets; fs; fs= fs->next ) - if (fs->pipe == b) { - printf("++ ref to pipe %d from fs %d\n", - p->pipe_nr, fs->fs_nr); - fs->pipe = NULL ; - purge_flow_set(fs, 0); - } + flush_pipe_ptrs(&(b->fs)); + + /* Remove all references to this pipe from flow_sets. */ + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(fs, &flowsethash[i], next) + if (fs->pipe == b) { + printf("dummynet: ++ ref to pipe %d from fs %d\n", + p->pipe_nr, fs->fs_nr); + fs->pipe = NULL ; + purge_flow_set(fs, 0); + } fs_remove_from_heap(&ready_heap, &(b->fs)); + purge_pipe(b); /* remove all data associated to this pipe */ /* remove reference to here from extract_heap and wfq_ready_heap */ pipe_remove_from_heap(&extract_heap, b); pipe_remove_from_heap(&wfq_ready_heap, b); - splx(s); - FREE(b, M_IPFW); + lck_mtx_unlock(dn_mutex); + + FREE(b, M_DUMMYNET); } else { /* this is a WF2Q queue (dn_flow_set) */ - struct dn_flow_set *a, *b; + struct dn_flow_set *b; + lck_mtx_lock(dn_mutex); /* locate set */ - for (a = NULL, b = all_flow_sets ; b && b->fs_nr < p->fs.fs_nr ; - a = b , b = b->next) ; - if (b == NULL || (b->fs_nr != p->fs.fs_nr) ) + b = locate_flowset(p->fs.fs_nr); + if (b == NULL) { + lck_mtx_unlock(dn_mutex); return EINVAL ; /* not found */ + } - s = splimp() ; - if (a == NULL) - all_flow_sets = b->next ; - else - a->next = b->next ; /* remove references to this flow_set from the ip_fw rules. */ - LIST_FOREACH(chain, &ip_fw_chain_head, next) - if (chain->rule->pipe_ptr == b) - chain->rule->pipe_ptr = NULL ; + flush_pipe_ptrs(b); + + /* Unlink from list of flowsets. */ + SLIST_REMOVE( &flowsethash[HASH(b->fs_nr)], b, dn_flow_set, next); if (b->pipe != NULL) { /* Update total weight on parent pipe and cleanup parent heaps */ @@ -1727,7 +2209,7 @@ delete_pipe(struct dn_pipe *p) #endif } purge_flow_set(b, 1); - splx(s); + lck_mtx_unlock(dn_mutex); } return 0 ; } @@ -1735,94 +2217,167 @@ delete_pipe(struct dn_pipe *p) /* * helper function used to copy data from kernel in DUMMYNET_GET */ -static char * -dn_copy_set(struct dn_flow_set *set, char *bp) +static +char* dn_copy_set_32(struct dn_flow_set *set, char *bp) { int i, copied = 0 ; - struct dn_flow_queue *q, *qp = (struct dn_flow_queue *)bp; + struct dn_flow_queue *q; + struct dn_flow_queue_32 *qp = (struct dn_flow_queue_32 *)bp; + + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + + for (i = 0 ; i <= set->rq_size ; i++) + for (q = set->rq[i] ; q ; q = q->next, qp++ ) { + if (q->hash_slot != i) + printf("dummynet: ++ at %d: wrong slot (have %d, " + "should be %d)\n", copied, q->hash_slot, i); + if (q->fs != set) + printf("dummynet: ++ at %d: wrong fs ptr (have %p, should be %p)\n", + i, q->fs, set); + copied++ ; + cp_queue_to_32_user( q, qp ); + /* cleanup pointers */ + qp->next = (user32_addr_t)0 ; + qp->head = qp->tail = (user32_addr_t)0 ; + qp->fs = (user32_addr_t)0 ; + } + if (copied != set->rq_elements) + printf("dummynet: ++ wrong count, have %d should be %d\n", + copied, set->rq_elements); + return (char *)qp ; +} +static +char* dn_copy_set_64(struct dn_flow_set *set, char *bp) +{ + int i, copied = 0 ; + struct dn_flow_queue *q; + struct dn_flow_queue_64 *qp = (struct dn_flow_queue_64 *)bp; + + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + for (i = 0 ; i <= set->rq_size ; i++) - for (q = set->rq[i] ; q ; q = q->next, qp++ ) { - if (q->hash_slot != i) - printf("++ at %d: wrong slot (have %d, " - "should be %d)\n", copied, q->hash_slot, i); - if (q->fs != set) - printf("++ at %d: wrong fs ptr (have %p, should be %p)\n", - i, q->fs, set); - copied++ ; - bcopy(q, qp, sizeof( *q ) ); - /* cleanup pointers */ - qp->next = NULL ; - qp->head = qp->tail = NULL ; - qp->fs = NULL ; - } + for (q = set->rq[i] ; q ; q = q->next, qp++ ) { + if (q->hash_slot != i) + printf("dummynet: ++ at %d: wrong slot (have %d, " + "should be %d)\n", copied, q->hash_slot, i); + if (q->fs != set) + printf("dummynet: ++ at %d: wrong fs ptr (have %p, should be %p)\n", + i, q->fs, set); + copied++ ; + //bcopy(q, qp, sizeof(*q)); + cp_queue_to_64_user( q, qp ); + /* cleanup pointers */ + qp->next = USER_ADDR_NULL ; + qp->head = qp->tail = USER_ADDR_NULL ; + qp->fs = USER_ADDR_NULL ; + } if (copied != set->rq_elements) - printf("++ wrong count, have %d should be %d\n", - copied, set->rq_elements); + printf("dummynet: ++ wrong count, have %d should be %d\n", + copied, set->rq_elements); return (char *)qp ; } +static size_t +dn_calc_size(int is64user) +{ + struct dn_flow_set *set ; + struct dn_pipe *p ; + size_t size = 0 ; + size_t pipesize; + size_t queuesize; + size_t setsize; + int i; + + lck_mtx_assert(dn_mutex, LCK_MTX_ASSERT_OWNED); + if ( is64user ){ + pipesize = sizeof(struct dn_pipe_64); + queuesize = sizeof(struct dn_flow_queue_64); + setsize = sizeof(struct dn_flow_set_64); + } + else { + pipesize = sizeof(struct dn_pipe_32); + queuesize = sizeof( struct dn_flow_queue_32 ); + setsize = sizeof(struct dn_flow_set_32); + } + /* + * compute size of data structures: list of pipes and flow_sets. + */ + for (i = 0; i < HASHSIZE; i++) { + SLIST_FOREACH(p, &pipehash[i], next) + size += sizeof(*p) + + p->fs.rq_elements * sizeof(struct dn_flow_queue); + SLIST_FOREACH(set, &flowsethash[i], next) + size += sizeof (*set) + + set->rq_elements * sizeof(struct dn_flow_queue); + } + return size; +} + static int dummynet_get(struct sockopt *sopt) { - char *buf, *bp ; /* bp is the "copy-pointer" */ + char *buf, *bp=NULL; /* bp is the "copy-pointer" */ size_t size ; struct dn_flow_set *set ; struct dn_pipe *p ; - int s, error=0 ; + int error=0, i ; + int is64user = 0; - s = splimp(); + /* XXX lock held too long */ + lck_mtx_lock(dn_mutex); /* - * compute size of data structures: list of pipes and flow_sets. + * XXX: Ugly, but we need to allocate memory with M_WAITOK flag and we + * cannot use this flag while holding a mutex. */ - for (p = all_pipes, size = 0 ; p ; p = p->next ) - size += sizeof( *p ) + - p->fs.rq_elements * sizeof(struct dn_flow_queue); - for (set = all_flow_sets ; set ; set = set->next ) - size += sizeof ( *set ) + - set->rq_elements * sizeof(struct dn_flow_queue); - buf = _MALLOC(size, M_TEMP, M_DONTWAIT); - if (buf == 0) { - splx(s); - return ENOBUFS ; + if (proc_is64bit(sopt->sopt_p)) + is64user = 1; + for (i = 0; i < 10; i++) { + size = dn_calc_size(is64user); + lck_mtx_unlock(dn_mutex); + buf = _MALLOC(size, M_TEMP, M_WAITOK); + if (buf == NULL) + return ENOBUFS; + lck_mtx_lock(dn_mutex); + if (size == dn_calc_size(is64user)) + break; + FREE(buf, M_TEMP); + buf = NULL; + } + if (buf == NULL) { + lck_mtx_unlock(dn_mutex); + return ENOBUFS ; } - for (p = all_pipes, bp = buf ; p ; p = p->next ) { - struct dn_pipe *pipe_bp = (struct dn_pipe *)bp ; - /* - * copy pipe descriptor into *bp, convert delay back to ms, - * then copy the flow_set descriptor(s) one at a time. - * After each flow_set, copy the queue descriptor it owns. - */ - bcopy(p, bp, sizeof( *p ) ); - pipe_bp->delay = (pipe_bp->delay * 1000) / hz ; - /* - * XXX the following is a hack based on ->next being the - * first field in dn_pipe and dn_flow_set. The correct - * solution would be to move the dn_flow_set to the beginning - * of struct dn_pipe. - */ - pipe_bp->next = (struct dn_pipe *)DN_IS_PIPE ; - /* clean pointers */ - pipe_bp->head = pipe_bp->tail = NULL ; - pipe_bp->fs.next = NULL ; - pipe_bp->fs.pipe = NULL ; - pipe_bp->fs.rq = NULL ; - bp += sizeof( *p ) ; - bp = dn_copy_set( &(p->fs), bp ); + bp = buf; + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(p, &pipehash[i], next) { + /* + * copy pipe descriptor into *bp, convert delay back to ms, + * then copy the flow_set descriptor(s) one at a time. + * After each flow_set, copy the queue descriptor it owns. + */ + if ( is64user ){ + bp = cp_pipe_to_64_user(p, (struct dn_pipe_64 *)bp); + } + else{ + bp = cp_pipe_to_32_user(p, (struct dn_pipe_32 *)bp); + } } - for (set = all_flow_sets ; set ; set = set->next ) { - struct dn_flow_set *fs_bp = (struct dn_flow_set *)bp ; - bcopy(set, bp, sizeof( *set ) ); - /* XXX same hack as above */ - fs_bp->next = (struct dn_flow_set *)DN_IS_QUEUE ; - fs_bp->pipe = NULL ; - fs_bp->rq = NULL ; - bp += sizeof( *set ) ; - bp = dn_copy_set( set, bp ); + for (i = 0; i < HASHSIZE; i++) + SLIST_FOREACH(set, &flowsethash[i], next) { + struct dn_flow_set_64 *fs_bp = (struct dn_flow_set_64 *)bp ; + cp_flow_set_to_64_user(set, fs_bp); + /* XXX same hack as above */ + fs_bp->next = CAST_DOWN(user64_addr_t, DN_IS_QUEUE); + fs_bp->pipe = USER_ADDR_NULL; + fs_bp->rq = USER_ADDR_NULL ; + bp += sizeof(struct dn_flow_set_64); + bp = dn_copy_set_64( set, bp ); } - splx(s); + lck_mtx_unlock(dn_mutex); + error = sooptcopyout(sopt, buf, size); FREE(buf, M_TEMP); return error ; @@ -1843,7 +2398,7 @@ ip_dn_ctl(struct sockopt *sopt) switch (sopt->sopt_name) { default : - printf("ip_dn_ctl -- unknown option %d", sopt->sopt_name); + printf("dummynet: -- unknown option %d", sopt->sopt_name); return EINVAL ; case IP_DUMMYNET_GET : @@ -1853,9 +2408,14 @@ ip_dn_ctl(struct sockopt *sopt) case IP_DUMMYNET_FLUSH : dummynet_flush() ; break ; + case IP_DUMMYNET_CONFIGURE : p = &tmp_pipe ; - error = sooptcopyin(sopt, p, sizeof *p, sizeof *p); + if (proc_is64bit(sopt->sopt_p)) + error = cp_pipe_from_user_64( sopt, p ); + else + error = cp_pipe_from_user_32( sopt, p ); + if (error) break ; error = config_pipe(p); @@ -1863,7 +2423,10 @@ ip_dn_ctl(struct sockopt *sopt) case IP_DUMMYNET_DEL : /* remove a pipe or queue */ p = &tmp_pipe ; - error = sooptcopyin(sopt, p, sizeof *p, sizeof *p); + if (proc_is64bit(sopt->sopt_p)) + error = cp_pipe_from_user_64( sopt, p ); + else + error = cp_pipe_from_user_32( sopt, p ); if (error) break ; @@ -1873,13 +2436,20 @@ ip_dn_ctl(struct sockopt *sopt) return error ; } -static void +void ip_dn_init(void) { - printf("DUMMYNET initialized (010124)\n"); - all_pipes = NULL ; - all_flow_sets = NULL ; - ready_heap.size = ready_heap.elements = 0 ; + /* setup locks */ + dn_mutex_grp_attr = lck_grp_attr_alloc_init(); + dn_mutex_grp = lck_grp_alloc_init("dn", dn_mutex_grp_attr); + dn_mutex_attr = lck_attr_alloc_init(); + + if ((dn_mutex = lck_mtx_alloc_init(dn_mutex_grp, dn_mutex_attr)) == NULL) { + printf("ip_dn_init: can't alloc dn_mutex\n"); + return; + } + + ready_heap.size = ready_heap.elements = 0 ; ready_heap.offset = 0 ; wfq_ready_heap.size = wfq_ready_heap.elements = 0 ; @@ -1888,37 +2458,6 @@ ip_dn_init(void) extract_heap.size = extract_heap.elements = 0 ; extract_heap.offset = 0 ; ip_dn_ctl_ptr = ip_dn_ctl; - timeout(dummynet, NULL, 1); -} - -static ip_dn_ctl_t *old_dn_ctl_ptr ; - -static int -dummynet_modevent(module_t mod, int type, void *data) -{ - int s ; - switch (type) { - case MOD_LOAD: - s = splimp(); - old_dn_ctl_ptr = ip_dn_ctl_ptr; - ip_dn_init(); - splx(s); - break; - case MOD_UNLOAD: - s = splimp(); - ip_dn_ctl_ptr = old_dn_ctl_ptr; - splx(s); - dummynet_flush(); - break ; - default: - break ; - } - return 0 ; + ip_dn_io_ptr = dummynet_io; + ip_dn_ruledel_ptr = dn_rule_delete; } - -static moduledata_t dummynet_mod = { - "dummynet", - dummynet_modevent, - NULL -} ; -DECLARE_MODULE(dummynet, dummynet_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);