SCTP: Rewrite of sctp buffer management code

index 01c6364..03579d5 100644
@@ -106,25 +106,41 @@ static int sctp_autobind(struct sock *sk);
 static void sctp_sock_migrate(struct sock *, struct sock *,
 			      struct sctp_association *, sctp_socket_type_t);
 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
+extern int sysctl_sctp_mem[3];
+extern int sysctl_sctp_rmem[3];
+extern int sysctl_sctp_wmem[3];
 
+int sctp_memory_pressure;
+atomic_t sctp_memory_allocated;
+atomic_t sctp_sockets_allocated;
+
+static void sctp_enter_memory_pressure(void)
+{
+	sctp_memory_pressure = 1;
+}
+ 
 /* Get the sndbuf space available at the time on the association.  */
 static inline int sctp_wspace(struct sctp_association *asoc)
 {
-	struct sock *sk = asoc->base.sk;
-	int amt = 0;
+ 	int amt;
 
-	if (asoc->ep->sndbuf_policy) {
-		/* make sure that no association uses more than sk_sndbuf */
-		amt = sk->sk_sndbuf - asoc->sndbuf_used;
+	if (asoc->ep->sndbuf_policy)
+		amt = asoc->sndbuf_used;
+	else
+		amt = atomic_read(&asoc->base.sk->sk_wmem_alloc);
+
+	if (amt >= asoc->base.sk->sk_sndbuf) {
+		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
+			amt = 0;
+		else {
+			amt = sk_stream_wspace(asoc->base.sk);
+			if (amt < 0)
+				amt = 0;
+		}
 	} else {
-		/* do socket level accounting */
-		amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
+ 		amt = asoc->base.sk->sk_sndbuf - amt;
 	}
-
-	if (amt < 0)
-		amt = 0;
-
-	return amt;
+  	return amt;
 }
 
 /* Increment the used sndbuf space count of the corresponding association by

@@ -155,6 +171,7 @@ static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
 				sizeof(struct sctp_chunk);
 
 	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
+	sk_charge_skb(sk, chunk->skb);
 }
 
 /* Verify that this is a valid address. */

@@ -3314,6 +3331,7 @@ SCTP_STATIC int sctp_init_sock(struct sock *sk)
 	sp->hmac = NULL;
 
 	SCTP_DBG_OBJCNT_INC(sock);
+	atomic_inc(&sctp_sockets_allocated);
 	return 0;
 }
 

@@ -3327,7 +3345,7 @@ SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
 	/* Release our hold on the endpoint. */
 	ep = sctp_sk(sk)->ep;
 	sctp_endpoint_free(ep);
-
+	atomic_dec(&sctp_sockets_allocated);
 	return 0;
 }
 

@@ -5747,6 +5765,12 @@ static void sctp_wfree(struct sk_buff *skb)
 
 	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
 
+	/*
+	 * This undoes what is done via sk_charge_skb
+	 */
+	sk->sk_wmem_queued   -= skb->truesize;
+	sk->sk_forward_alloc += skb->truesize;
+
 	sock_wfree(skb);
 	__sctp_write_space(asoc);
 

@@ -5764,6 +5788,11 @@ void sctp_sock_rfree(struct sk_buff *skb)
 	struct sctp_ulpevent *event = sctp_skb2event(skb);
 
 	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
+
+	/*
+	 * Mimic the behavior of sk_stream_rfree
+	 */
+	sk->sk_forward_alloc += event->rmem_len;
 }
 
 

@@ -6153,6 +6182,7 @@ static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
 	sctp_release_sock(newsk);
 }
 
+
 /* This proto struct describes the ULP interface for SCTP.  */
 struct proto sctp_prot = {
 	.name        =	"SCTP",

@@ -6175,6 +6205,12 @@ struct proto sctp_prot = {
 	.unhash      =	sctp_unhash,
 	.get_port    =	sctp_get_port,
 	.obj_size    =  sizeof(struct sctp_sock),
+	.sysctl_mem  =  sysctl_sctp_mem,
+	.sysctl_rmem =  sysctl_sctp_rmem,
+	.sysctl_wmem =  sysctl_sctp_wmem,
+	.memory_pressure = &sctp_memory_pressure,
+	.enter_memory_pressure = sctp_enter_memory_pressure,
+	.memory_allocated = &sctp_memory_allocated,
 };
 
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)

@@ -6199,5 +6235,11 @@ struct proto sctpv6_prot = {
 	.unhash		= sctp_unhash,
 	.get_port	= sctp_get_port,
 	.obj_size	= sizeof(struct sctp6_sock),
+	.sysctl_mem	= sysctl_sctp_mem,
+	.sysctl_rmem	= sysctl_sctp_rmem,
+	.sysctl_wmem	= sysctl_sctp_wmem,
+	.memory_pressure = &sctp_memory_pressure,
+	.enter_memory_pressure = sctp_enter_memory_pressure,
+	.memory_allocated = &sctp_memory_allocated,
 };
 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */

index d529045..46d7d09 100644
@@ -468,6 +468,11 @@ static inline void sctp_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
 	skb->sk = sk;
 	skb->destructor = sctp_sock_rfree;
 	atomic_add(event->rmem_len, &sk->sk_rmem_alloc);
+	/*
+	 * This mimics the behavior of
+	 * sk_stream_set_owner_r
+	 */
+	sk->sk_forward_alloc -= event->rmem_len;
 }
 
 /* Tests if the list has one and only one entry. */

index 1404a9e..aba9258 100644
@@ -103,6 +103,7 @@ static struct sctp_endpoint *sctp_endpoint_init(struct sctp_endpoint *ep,
 
 	/* Use SCTP specific send buffer space queues.  */
 	ep->sndbuf_policy = sctp_sndbuf_policy;
+
 	sk->sk_write_space = sctp_write_space;
 	sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
 

index e98579b..a746ebf 100644
@@ -51,6 +51,7 @@
 #include <linux/netdevice.h>
 #include <linux/inetdevice.h>
 #include <linux/seq_file.h>
+#include <linux/bootmem.h>
 #include <net/protocol.h>
 #include <net/ip.h>
 #include <net/ipv6.h>

@@ -82,6 +83,10 @@ static struct sctp_af *sctp_af_v6_specific;
 struct kmem_cache *sctp_chunk_cachep __read_mostly;
 struct kmem_cache *sctp_bucket_cachep __read_mostly;
 
+extern int sysctl_sctp_mem[3];
+extern int sysctl_sctp_rmem[3];
+extern int sysctl_sctp_wmem[3];
+
 /* Return the address of the control sock. */
 struct sock *sctp_get_ctl_sock(void)
 {

@@ -969,6 +974,8 @@ SCTP_STATIC __init int sctp_init(void)
 	int i;
 	int status = -EINVAL;
 	unsigned long goal;
+	unsigned long limit;
+	int max_share;
 	int order;
 
 	/* SCTP_DEBUG sanity check. */

@@ -1059,6 +1066,31 @@ SCTP_STATIC __init int sctp_init(void)
 	/* Initialize handle used for association ids. */
 	idr_init(&sctp_assocs_id);
 
+	/* Set the pressure threshold to be a fraction of global memory that
+	 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
+	 * memory, with a floor of 128 pages.
+	 * Note this initalizes the data in sctpv6_prot too
+	 * Unabashedly stolen from tcp_init
+	 */
+	limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
+	limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
+	limit = max(limit, 128UL);
+	sysctl_sctp_mem[0] = limit / 4 * 3;
+	sysctl_sctp_mem[1] = limit;
+	sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
+
+	/* Set per-socket limits to no more than 1/128 the pressure threshold*/
+	limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
+	max_share = min(4UL*1024*1024, limit);
+
+	sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
+	sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
+	sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
+
+	sysctl_sctp_wmem[0] = SK_STREAM_MEM_QUANTUM;
+	sysctl_sctp_wmem[1] = 16*1024;
+	sysctl_sctp_wmem[2] = max(64*1024, max_share);
+
 	/* Size and allocate the association hash table.
 	 * The methodology is similar to that of the tcp hash tables.
 	 */

100 lines not on this page · raw message

128 lines not on this page · raw message

index e2c679b..ba75ef4 100644
@@ -52,6 +52,15 @@ static int int_max = INT_MAX;
 static long sack_timer_min = 1;
 static long sack_timer_max = 500;
 
+int sysctl_sctp_mem[3];
+int sysctl_sctp_rmem[3];
+int sysctl_sctp_wmem[3];
+
+/*
+ * per assoc memory limitationf for sends
+ */
+int sysctl_sctp_wmem[3];
+
 static ctl_table sctp_table[] = {
 	{
 		.ctl_name	= NET_SCTP_RTO_INITIAL,

@@ -226,6 +235,30 @@ static ctl_table sctp_table[] = {
 		.extra1         = &sack_timer_min,
 		.extra2         = &sack_timer_max,
 	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "sctp_mem",
+		.data		= &sysctl_sctp_mem,
+		.maxlen		= sizeof(sysctl_sctp_mem),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "sctp_rmem",
+		.data		= &sysctl_sctp_rmem,
+		.maxlen		= sizeof(sysctl_sctp_rmem),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "sctp_wmem",
+		.data		= &sysctl_sctp_wmem,
+		.maxlen		= sizeof(sysctl_sctp_wmem),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
 	{ .ctl_name = 0 }
 };
 

index bfecb35..5dc094b 100644
@@ -685,6 +685,24 @@ struct sctp_ulpevent *sctp_ulpevent_make_rcvmsg(struct sctp_association *asoc,
 	struct sctp_ulpevent *event = NULL;
 	struct sk_buff *skb;
 	size_t padding, len;
+	int rx_count;
+
+	/*
+	 * check to see if we need to make space for this
+	 * new skb, expand the rcvbuffer if needed, or drop
+	 * the frame
+	 */
+	if (asoc->ep->rcvbuf_policy)
+		rx_count = atomic_read(&asoc->rmem_alloc);
+	else
+		rx_count = atomic_read(&asoc->base.sk->sk_rmem_alloc);
+
+	if (rx_count >= asoc->base.sk->sk_rcvbuf) {
+
+		if ((asoc->base.sk->sk_userlocks & SOCK_RCVBUF_LOCK) ||
+		   (!sk_stream_rmem_schedule(asoc->base.sk, chunk->skb)))
+			goto fail;
+	}
 
 	/* Clone the original skb, sharing the data.  */
 	skb = skb_clone(chunk->skb, gfp);

index 34eb977..33e49b5 100644
@@ -978,6 +978,7 @@ void sctp_ulpq_renege(struct sctp_ulpq *ulpq, struct sctp_chunk *chunk,
 		sctp_ulpq_partial_delivery(ulpq, chunk, gfp);
 	}
 
+	sk_stream_mem_reclaim(asoc->base.sk);
 	return;
 }