[Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

[-- Attachment #1: sysctl_tcp_udp_mem_max_overflows --]
[-- Type: text/plain, Size: 2747 bytes --]

Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
maximized without overflowing.

Signed-off-by: Robin Holt <holt@sgi.com>
To: Willy Tarreau <w@1wt.eu>
To: linux-kernel@vger.kernel.org
To: netdev@vger.kernel.org
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: James Morris <jmorris@namei.org>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Patrick McHardy <kaber@trash.net>

---

 net/ipv4/tcp.c |    4 +++-
 net/ipv4/udp.c |    4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

Index: pv1010932/net/ipv4/tcp.c
===================================================================
--- pv1010932.orig/net/ipv4/tcp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/tcp.c	2010-10-02 06:12:35.445454593 -0500
@@ -3271,12 +3271,14 @@ void __init tcp_init(void)
 
 	/* 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.
+	 * memory, with a floor of 128 pages, and a ceiling that prevents an
+	 * integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_tcp_mem[0] = limit / 4 * 3;
 	sysctl_tcp_mem[1] = limit;
 	sysctl_tcp_mem[2] = sysctl_tcp_mem[0] * 2;
Index: pv1010932/net/ipv4/udp.c
===================================================================
--- pv1010932.orig/net/ipv4/udp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/udp.c	2010-10-02 06:12:35.453453784 -0500
@@ -2167,12 +2167,14 @@ void __init udp_init(void)
 	udp_table_init(&udp_table, "UDP");
 	/* Set the pressure threshold up by the same strategy of TCP. It is 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.
+	 * toward zero with the amount of memory, with a floor of 128 pages,
+	 * and a ceiling that prevents an integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_udp_mem[0] = limit / 4 * 3;
 	sysctl_udp_mem[1] = limit;
 	sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Eric Dumazet]

Le samedi 02 octobre 2010 à 06:24 -0500, Robin Holt a écrit :
> pièce jointe document texte brut (sysctl_tcp_udp_mem_max_overflows)
> Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> 
> On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> maximized without overflowing.
> 
> Signed-off-by: Robin Holt <holt@sgi.com>
> To: Willy Tarreau <w@1wt.eu>
> To: linux-kernel@vger.kernel.org
> To: netdev@vger.kernel.org
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
> Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
> Cc: James Morris <jmorris@namei.org>
> Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
> Cc: Patrick McHardy <kaber@trash.net>
> 
> ---
> 
>  net/ipv4/tcp.c |    4 +++-
>  net/ipv4/udp.c |    4 +++-
>  2 files changed, 6 insertions(+), 2 deletions(-)

Hi Mr SixteenTB

Strange, you mention sctp in changelog but I cant see the patch.

We could apply your patch (with sctp changes) for stable.

IMHO it would be better in long term to switch all these limits from
"int" to "long", now we have atomic_long_t primitives that have same
runtime cost than atomic_t ones. I am pretty sure one of your customer
will need more than 5TB of memory for tcp/udp buffers before year 2020,
dont you think ? (some further scalability works probably needed.)

Something like this (boot tested on my dev machine, not a 16TB one, as
you guessed ;) ) :

Based on linux-2.6 for your convenience, should probably sit first in
David net-next-2.6 ...

Note : this needs the "sysctl: fix min/max handling in
__do_proc_doulongvec_minmax()" I sent to Andrew some minutes ago.
or this triggers a BUG :

"echo "378912 505216 758989782400000" >/proc/sys/net/ipv4/tcp_mem


Thanks !

[PATCH] net: avoid limits overflow

Robin Holt tried to boot a 16TB machine and found some limits were
reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]

We can switch infrastructure to use long "instead" of "int", now
atomic_long_t primitives are available for free.

Reported-by: Robin Holt <holt@sgi.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 include/net/dn.h               |    2 +-
 include/net/sock.h             |    4 ++--
 include/net/tcp.h              |    6 +++---
 include/net/udp.h              |    4 ++--
 net/core/sock.c                |   14 +++++++-------
 net/decnet/af_decnet.c         |    2 +-
 net/decnet/sysctl_net_decnet.c |    4 ++--
 net/ipv4/proc.c                |    8 ++++----
 net/ipv4/sysctl_net_ipv4.c     |    5 ++---
 net/ipv4/tcp.c                 |    4 ++--
 net/ipv4/tcp_input.c           |   11 +++++++----
 net/ipv4/udp.c                 |    4 ++--
 net/sctp/protocol.c            |    2 +-
 net/sctp/socket.c              |    4 ++--
 net/sctp/sysctl.c              |    4 ++--
 15 files changed, 40 insertions(+), 38 deletions(-)

diff --git a/include/net/dn.h b/include/net/dn.h
index e5469f7..a514a3c 100644
--- a/include/net/dn.h
+++ b/include/net/dn.h
@@ -225,7 +225,7 @@ extern int decnet_di_count;
 extern int decnet_dr_count;
 extern int decnet_no_fc_max_cwnd;
 
-extern int sysctl_decnet_mem[3];
+extern long sysctl_decnet_mem[3];
 extern int sysctl_decnet_wmem[3];
 extern int sysctl_decnet_rmem[3];
 
diff --git a/include/net/sock.h b/include/net/sock.h
index adab9dc..5d84d86 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -762,7 +762,7 @@ struct proto {
 
 	/* Memory pressure */
 	void			(*enter_memory_pressure)(struct sock *sk);
-	atomic_t		*memory_allocated;	/* Current allocated memory. */
+	atomic_long_t		*memory_allocated;	/* Current allocated memory. */
 	struct percpu_counter	*sockets_allocated;	/* Current number of sockets. */
 	/*
 	 * Pressure flag: try to collapse.
@@ -771,7 +771,7 @@ struct proto {
 	 * is strict, actions are advisory and have some latency.
 	 */
 	int			*memory_pressure;
-	int			*sysctl_mem;
+	long			*sysctl_mem;
 	int			*sysctl_wmem;
 	int			*sysctl_rmem;
 	int			max_header;
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 3e4b33e..3f05403 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -224,7 +224,7 @@ extern int sysctl_tcp_fack;
 extern int sysctl_tcp_reordering;
 extern int sysctl_tcp_ecn;
 extern int sysctl_tcp_dsack;
-extern int sysctl_tcp_mem[3];
+extern long sysctl_tcp_mem[3];
 extern int sysctl_tcp_wmem[3];
 extern int sysctl_tcp_rmem[3];
 extern int sysctl_tcp_app_win;
@@ -247,7 +247,7 @@ extern int sysctl_tcp_cookie_size;
 extern int sysctl_tcp_thin_linear_timeouts;
 extern int sysctl_tcp_thin_dupack;
 
-extern atomic_t tcp_memory_allocated;
+extern atomic_long_t tcp_memory_allocated;
 extern struct percpu_counter tcp_sockets_allocated;
 extern int tcp_memory_pressure;
 
@@ -280,7 +280,7 @@ static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
 	}
 
 	if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
-	    atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2])
+	    atomic_long_read(&tcp_memory_allocated) > sysctl_tcp_mem[2])
 		return true;
 	return false;
 }
diff --git a/include/net/udp.h b/include/net/udp.h
index a184d34..e686e01 100644
--- a/include/net/udp.h
+++ b/include/net/udp.h
@@ -105,10 +105,10 @@ static inline struct udp_hslot *udp_hashslot2(struct udp_table *table,
 
 extern struct proto udp_prot;
 
-extern atomic_t udp_memory_allocated;
+extern atomic_long_t udp_memory_allocated;
 
 /* sysctl variables for udp */
-extern int sysctl_udp_mem[3];
+extern long sysctl_udp_mem[3];
 extern int sysctl_udp_rmem_min;
 extern int sysctl_udp_wmem_min;
 
diff --git a/net/core/sock.c b/net/core/sock.c
index ef30e9d..a2af957 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1646,10 +1646,10 @@ int __sk_mem_schedule(struct sock *sk, int size, int kind)
 {
 	struct proto *prot = sk->sk_prot;
 	int amt = sk_mem_pages(size);
-	int allocated;
+	long allocated;
 
 	sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
-	allocated = atomic_add_return(amt, prot->memory_allocated);
+	allocated = atomic_long_add_return(amt, prot->memory_allocated);
 
 	/* Under limit. */
 	if (allocated <= prot->sysctl_mem[0]) {
@@ -1707,7 +1707,7 @@ suppress_allocation:
 
 	/* Alas. Undo changes. */
 	sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
-	atomic_sub(amt, prot->memory_allocated);
+	atomic_long_sub(amt, prot->memory_allocated);
 	return 0;
 }
 EXPORT_SYMBOL(__sk_mem_schedule);
@@ -1720,12 +1720,12 @@ void __sk_mem_reclaim(struct sock *sk)
 {
 	struct proto *prot = sk->sk_prot;
 
-	atomic_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
+	atomic_long_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
 		   prot->memory_allocated);
 	sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
 
 	if (prot->memory_pressure && *prot->memory_pressure &&
-	    (atomic_read(prot->memory_allocated) < prot->sysctl_mem[0]))
+	    (atomic_long_read(prot->memory_allocated) < prot->sysctl_mem[0]))
 		*prot->memory_pressure = 0;
 }
 EXPORT_SYMBOL(__sk_mem_reclaim);
@@ -2445,12 +2445,12 @@ static char proto_method_implemented(const void *method)
 
 static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
 {
-	seq_printf(seq, "%-9s %4u %6d  %6d   %-3s %6u   %-3s  %-10s "
+	seq_printf(seq, "%-9s %4u %6d  %6ld   %-3s %6u   %-3s  %-10s "
 			"%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
 		   proto->name,
 		   proto->obj_size,
 		   sock_prot_inuse_get(seq_file_net(seq), proto),
-		   proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
+		   proto->memory_allocated != NULL ? atomic_long_read(proto->memory_allocated) : -1L,
 		   proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
 		   proto->max_header,
 		   proto->slab == NULL ? "no" : "yes",
diff --git a/net/decnet/af_decnet.c b/net/decnet/af_decnet.c
index d6b93d1..a76b78d 100644
--- a/net/decnet/af_decnet.c
+++ b/net/decnet/af_decnet.c
@@ -155,7 +155,7 @@ static const struct proto_ops dn_proto_ops;
 static DEFINE_RWLOCK(dn_hash_lock);
 static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE];
 static struct hlist_head dn_wild_sk;
-static atomic_t decnet_memory_allocated;
+static atomic_long_t decnet_memory_allocated;
 
 static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen, int flags);
 static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags);
diff --git a/net/decnet/sysctl_net_decnet.c b/net/decnet/sysctl_net_decnet.c
index be3eb8e..28f8b5e 100644
--- a/net/decnet/sysctl_net_decnet.c
+++ b/net/decnet/sysctl_net_decnet.c
@@ -38,7 +38,7 @@ int decnet_log_martians = 1;
 int decnet_no_fc_max_cwnd = NSP_MIN_WINDOW;
 
 /* Reasonable defaults, I hope, based on tcp's defaults */
-int sysctl_decnet_mem[3] = { 768 << 3, 1024 << 3, 1536 << 3 };
+long sysctl_decnet_mem[3] = { 768 << 3, 1024 << 3, 1536 << 3 };
 int sysctl_decnet_wmem[3] = { 4 * 1024, 16 * 1024, 128 * 1024 };
 int sysctl_decnet_rmem[3] = { 4 * 1024, 87380, 87380 * 2 };
 
@@ -324,7 +324,7 @@ static ctl_table dn_table[] = {
 		.data = &sysctl_decnet_mem,
 		.maxlen = sizeof(sysctl_decnet_mem),
 		.mode = 0644,
-		.proc_handler = proc_dointvec,
+		.proc_handler = proc_doulongvec_minmax
 	},
 	{
 		.procname = "decnet_rmem",
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index 4ae1f20..1b48eb1 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -59,13 +59,13 @@ static int sockstat_seq_show(struct seq_file *seq, void *v)
 	local_bh_enable();
 
 	socket_seq_show(seq);
-	seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n",
+	seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n",
 		   sock_prot_inuse_get(net, &tcp_prot), orphans,
 		   tcp_death_row.tw_count, sockets,
-		   atomic_read(&tcp_memory_allocated));
-	seq_printf(seq, "UDP: inuse %d mem %d\n",
+		   atomic_long_read(&tcp_memory_allocated));
+	seq_printf(seq, "UDP: inuse %d mem %ld\n",
 		   sock_prot_inuse_get(net, &udp_prot),
-		   atomic_read(&udp_memory_allocated));
+		   atomic_long_read(&udp_memory_allocated));
 	seq_printf(seq, "UDPLITE: inuse %d\n",
 		   sock_prot_inuse_get(net, &udplite_prot));
 	seq_printf(seq, "RAW: inuse %d\n",
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index d96c1da..e91911d 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -398,7 +398,7 @@ static struct ctl_table ipv4_table[] = {
 		.data		= &sysctl_tcp_mem,
 		.maxlen		= sizeof(sysctl_tcp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec
+		.proc_handler	= proc_doulongvec_minmax
 	},
 	{
 		.procname	= "tcp_wmem",
@@ -602,8 +602,7 @@ static struct ctl_table ipv4_table[] = {
 		.data		= &sysctl_udp_mem,
 		.maxlen		= sizeof(sysctl_udp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= &zero
+		.proc_handler	= proc_doulongvec_minmax,
 	},
 	{
 		.procname	= "udp_rmem_min",
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index f115ea6..e88d7a0 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -282,7 +282,7 @@ int sysctl_tcp_fin_timeout __read_mostly = TCP_FIN_TIMEOUT;
 struct percpu_counter tcp_orphan_count;
 EXPORT_SYMBOL_GPL(tcp_orphan_count);
 
-int sysctl_tcp_mem[3] __read_mostly;
+long sysctl_tcp_mem[3] __read_mostly;
 int sysctl_tcp_wmem[3] __read_mostly;
 int sysctl_tcp_rmem[3] __read_mostly;
 
@@ -290,7 +290,7 @@ EXPORT_SYMBOL(sysctl_tcp_mem);
 EXPORT_SYMBOL(sysctl_tcp_rmem);
 EXPORT_SYMBOL(sysctl_tcp_wmem);
 
-atomic_t tcp_memory_allocated;	/* Current allocated memory. */
+atomic_long_t tcp_memory_allocated;	/* Current allocated memory. */
 EXPORT_SYMBOL(tcp_memory_allocated);
 
 /*
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index b55f60f..0f56fb4 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -259,8 +259,11 @@ static void tcp_fixup_sndbuf(struct sock *sk)
 	int sndmem = tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER + 16 +
 		     sizeof(struct sk_buff);
 
-	if (sk->sk_sndbuf < 3 * sndmem)
-		sk->sk_sndbuf = min(3 * sndmem, sysctl_tcp_wmem[2]);
+	if (sk->sk_sndbuf < 3 * sndmem) {
+		sk->sk_sndbuf = 3 * sndmem;
+		if (sk->sk_sndbuf > sysctl_tcp_wmem[2])
+			sk->sk_sndbuf = sysctl_tcp_wmem[2];
+	}
 }
 
 /* 2. Tuning advertised window (window_clamp, rcv_ssthresh)
@@ -396,7 +399,7 @@ static void tcp_clamp_window(struct sock *sk)
 	if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] &&
 	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
 	    !tcp_memory_pressure &&
-	    atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) {
+	    atomic_long_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) {
 		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
 				    sysctl_tcp_rmem[2]);
 	}
@@ -4870,7 +4873,7 @@ static int tcp_should_expand_sndbuf(struct sock *sk)
 		return 0;
 
 	/* If we are under soft global TCP memory pressure, do not expand.  */
-	if (atomic_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0])
+	if (atomic_long_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0])
 		return 0;
 
 	/* If we filled the congestion window, do not expand.  */
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index fb23c2e..ecfdf2e 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -110,7 +110,7 @@
 struct udp_table udp_table __read_mostly;
 EXPORT_SYMBOL(udp_table);
 
-int sysctl_udp_mem[3] __read_mostly;
+long sysctl_udp_mem[3] __read_mostly;
 EXPORT_SYMBOL(sysctl_udp_mem);
 
 int sysctl_udp_rmem_min __read_mostly;
@@ -119,7 +119,7 @@ EXPORT_SYMBOL(sysctl_udp_rmem_min);
 int sysctl_udp_wmem_min __read_mostly;
 EXPORT_SYMBOL(sysctl_udp_wmem_min);
 
-atomic_t udp_memory_allocated;
+atomic_long_t udp_memory_allocated;
 EXPORT_SYMBOL(udp_memory_allocated);
 
 #define MAX_UDP_PORTS 65536
diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
index 5027b83..bf95400 100644
--- a/net/sctp/protocol.c
+++ b/net/sctp/protocol.c
@@ -90,7 +90,7 @@ static struct sctp_af *sctp_af_v6_specific;
 struct kmem_cache *sctp_chunk_cachep __read_mostly;
 struct kmem_cache *sctp_bucket_cachep __read_mostly;
 
-int sysctl_sctp_mem[3];
+long sysctl_sctp_mem[3];
 int sysctl_sctp_rmem[3];
 int sysctl_sctp_wmem[3];
 
diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index ca44917..cc03b44 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -109,12 +109,12 @@ static void sctp_sock_migrate(struct sock *, struct sock *,
 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
 
 extern struct kmem_cache *sctp_bucket_cachep;
-extern int sysctl_sctp_mem[3];
+extern long sysctl_sctp_mem[3];
 extern int sysctl_sctp_rmem[3];
 extern int sysctl_sctp_wmem[3];
 
 static int sctp_memory_pressure;
-static atomic_t sctp_memory_allocated;
+static atomic_long_t sctp_memory_allocated;
 struct percpu_counter sctp_sockets_allocated;
 
 static void sctp_enter_memory_pressure(struct sock *sk)
diff --git a/net/sctp/sysctl.c b/net/sctp/sysctl.c
index 832590b..50cb57f 100644
--- a/net/sctp/sysctl.c
+++ b/net/sctp/sysctl.c
@@ -54,7 +54,7 @@ static int sack_timer_max = 500;
 static int addr_scope_max = 3; /* check sctp_scope_policy_t in include/net/sctp/constants.h for max entries */
 static int rwnd_scale_max = 16;
 
-extern int sysctl_sctp_mem[3];
+extern long sysctl_sctp_mem[3];
 extern int sysctl_sctp_rmem[3];
 extern int sysctl_sctp_wmem[3];
 
@@ -203,7 +203,7 @@ static ctl_table sctp_table[] = {
 		.data		= &sysctl_sctp_mem,
 		.maxlen		= sizeof(sysctl_sctp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec,
+		.proc_handler	= proc_doulongvec_minmax
 	},
 	{
 		.procname	= "sctp_rmem",

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

On Sat, Oct 02, 2010 at 03:22:16PM +0200, Eric Dumazet wrote:
> Le samedi 02 octobre 2010 à 06:24 -0500, Robin Holt a écrit :
...
> Strange, you mention sctp in changelog but I cant see the patch.

After looking at the patch, I realized it really belonged in a separate
change and sent that to the sctp mailing list without noticing I forgot
to Cc: lkml.

> We can switch infrastructure to use long "instead" of "int", now
> atomic_long_t primitives are available for free.
> 
> Reported-by: Robin Holt <holt@sgi.com>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>

Reviewed-by: Robin Holt <holt@sgi.com>

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[David Miller]

From: Robin Holt <holt@sgi.com>
Date: Sat, 02 Oct 2010 06:24:06 -0500

> Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> 
> On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> maximized without overflowing.
> 
> Signed-off-by: Robin Holt <holt@sgi.com>

Robin please resubmit this with the SCTP bits included.

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Sat, 02 Oct 2010 15:22:16 +0200

> [PATCH] net: avoid limits overflow
> 
> Robin Holt tried to boot a 16TB machine and found some limits were
> reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]
> 
> We can switch infrastructure to use long "instead" of "int", now
> atomic_long_t primitives are available for free.
> 
> Reported-by: Robin Holt <holt@sgi.com>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>

Eric please resubmit this when the sysctl fix is resolved.

Thanks.

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

On Tue, Oct 05, 2010 at 02:50:32PM -0700, David Miller wrote:
> From: Robin Holt <holt@sgi.com>
> Date: Sat, 02 Oct 2010 06:24:06 -0500
> 
> > Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> > 
> > On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> > sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> > maximized without overflowing.
> > 
> > Signed-off-by: Robin Holt <holt@sgi.com>
> 
> Robin please resubmit this with the SCTP bits included.

David,

I did not mean to blow you off.  I was not going to resubmit because
Eric's patch is the preferred direction.  I had been given an indication
that my patch was preferred, but that indicator changed.

Robin

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Eric Dumazet]

Le lundi 18 octobre 2010 à 12:43 -0500, Robin Holt a écrit :
> On Tue, Oct 05, 2010 at 02:50:32PM -0700, David Miller wrote:
> > From: Robin Holt <holt@sgi.com>
> > Date: Sat, 02 Oct 2010 06:24:06 -0500
> > 
> > > Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> > > 
> > > On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> > > sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> > > maximized without overflowing.
> > > 
> > > Signed-off-by: Robin Holt <holt@sgi.com>
> > 
> > Robin please resubmit this with the SCTP bits included.
> 
> David,
> 
> I did not mean to blow you off.  I was not going to resubmit because
> Eric's patch is the preferred direction.  I had been given an indication
> that my patch was preferred, but that indicator changed.
> 

Maybe there is some misunderstanding.

My patch was for net-next kernels (not before 2.6.37), while yours was
applicable to previous kernels.

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

On Mon, Oct 18, 2010 at 07:52:28PM +0200, Eric Dumazet wrote:
> Le lundi 18 octobre 2010 à 12:43 -0500, Robin Holt a écrit :
> > On Tue, Oct 05, 2010 at 02:50:32PM -0700, David Miller wrote:
> > > From: Robin Holt <holt@sgi.com>
> > > Date: Sat, 02 Oct 2010 06:24:06 -0500
> > > 
> > > > Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> > > > 
> > > > On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> > > > sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> > > > maximized without overflowing.
> > > > 
> > > > Signed-off-by: Robin Holt <holt@sgi.com>
> > > 
> > > Robin please resubmit this with the SCTP bits included.
> > 
> > David,
> > 
> > I did not mean to blow you off.  I was not going to resubmit because
> > Eric's patch is the preferred direction.  I had been given an indication
> > that my patch was preferred, but that indicator changed.
> > 
> 
> Maybe there is some misunderstanding.
> 
> My patch was for net-next kernels (not before 2.6.37), while yours was
> applicable to previous kernels.

Ah.  I will resubmit then.

Sorry for the confusion,
Robin

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

On Tue, Oct 05, 2010 at 02:50:55PM -0700, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Sat, 02 Oct 2010 15:22:16 +0200
> 
> > [PATCH] net: avoid limits overflow
> > 
> > Robin Holt tried to boot a 16TB machine and found some limits were
> > reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]
> > 
> > We can switch infrastructure to use long "instead" of "int", now
> > atomic_long_t primitives are available for free.
> > 
> > Reported-by: Robin Holt <holt@sgi.com>
> > Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> 
> Eric please resubmit this when the sysctl fix is resolved.

It looks like the sysctl fix is upstream.  Has this been resubmitted
and I missed it?

Robin

[PATCH] net: avoid limits overflow

[Eric Dumazet]

Le mercredi 10 novembre 2010 à 00:15 -0600, Robin Holt a écrit :
> On Tue, Oct 05, 2010 at 02:50:55PM -0700, David Miller wrote:
> > From: Eric Dumazet <eric.dumazet@gmail.com>
> > Date: Sat, 02 Oct 2010 15:22:16 +0200
> > 
> > > [PATCH] net: avoid limits overflow
> > > 
> > > Robin Holt tried to boot a 16TB machine and found some limits were
> > > reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]
> > > 
> > > We can switch infrastructure to use long "instead" of "int", now
> > > atomic_long_t primitives are available for free.
> > > 
> > > Reported-by: Robin Holt <holt@sgi.com>
> > > Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> > 
> > Eric please resubmit this when the sysctl fix is resolved.
> 
> It looks like the sysctl fix is upstream.  Has this been resubmitted
> and I missed it?
> 
> Robin

I believe it was in Andrew mm tree for a while, and Andrew sent it do
David the 21th of october, I am not sure what happened.

Here it is again for latest linux-2.6 tree, thanks for the headup !

[PATCH] net: avoid limits overflow

Robin Holt tried to boot a 16TB machine and found some limits were
reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]

We can switch infrastructure to use long "instead" of "int", now
atomic_long_t primitives are available for free.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Reported-by: Robin Holt <holt@sgi.com>
Reviewed-by: Robin Holt <holt@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
 include/net/dn.h               |    2 +-
 include/net/sock.h             |    4 ++--
 include/net/tcp.h              |    6 +++---
 include/net/udp.h              |    4 ++--
 net/core/sock.c                |   14 +++++++-------
 net/decnet/af_decnet.c         |    2 +-
 net/decnet/sysctl_net_decnet.c |    4 ++--
 net/ipv4/proc.c                |    8 ++++----
 net/ipv4/sysctl_net_ipv4.c     |    5 ++---
 net/ipv4/tcp.c                 |    4 ++--
 net/ipv4/tcp_input.c           |   11 +++++++----
 net/ipv4/udp.c                 |    4 ++--
 net/sctp/protocol.c            |    2 +-
 net/sctp/socket.c              |    4 ++--
 net/sctp/sysctl.c              |    4 ++--
 15 files changed, 40 insertions(+), 38 deletions(-)

diff --git a/include/net/dn.h b/include/net/dn.h
index e5469f7..a514a3c 100644
--- a/include/net/dn.h
+++ b/include/net/dn.h
@@ -225,7 +225,7 @@ extern int decnet_di_count;
 extern int decnet_dr_count;
 extern int decnet_no_fc_max_cwnd;
 
-extern int sysctl_decnet_mem[3];
+extern long sysctl_decnet_mem[3];
 extern int sysctl_decnet_wmem[3];
 extern int sysctl_decnet_rmem[3];
 
diff --git a/include/net/sock.h b/include/net/sock.h
index c7a7362..a6338d0 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -762,7 +762,7 @@ struct proto {
 
 	/* Memory pressure */
 	void			(*enter_memory_pressure)(struct sock *sk);
-	atomic_t		*memory_allocated;	/* Current allocated memory. */
+	atomic_long_t		*memory_allocated;	/* Current allocated memory. */
 	struct percpu_counter	*sockets_allocated;	/* Current number of sockets. */
 	/*
 	 * Pressure flag: try to collapse.
@@ -771,7 +771,7 @@ struct proto {
 	 * is strict, actions are advisory and have some latency.
 	 */
 	int			*memory_pressure;
-	int			*sysctl_mem;
+	long			*sysctl_mem;
 	int			*sysctl_wmem;
 	int			*sysctl_rmem;
 	int			max_header;
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 4fee042..e36c874 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -224,7 +224,7 @@ extern int sysctl_tcp_fack;
 extern int sysctl_tcp_reordering;
 extern int sysctl_tcp_ecn;
 extern int sysctl_tcp_dsack;
-extern int sysctl_tcp_mem[3];
+extern long sysctl_tcp_mem[3];
 extern int sysctl_tcp_wmem[3];
 extern int sysctl_tcp_rmem[3];
 extern int sysctl_tcp_app_win;
@@ -247,7 +247,7 @@ extern int sysctl_tcp_cookie_size;
 extern int sysctl_tcp_thin_linear_timeouts;
 extern int sysctl_tcp_thin_dupack;
 
-extern atomic_t tcp_memory_allocated;
+extern atomic_long_t tcp_memory_allocated;
 extern struct percpu_counter tcp_sockets_allocated;
 extern int tcp_memory_pressure;
 
@@ -280,7 +280,7 @@ static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
 	}
 
 	if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
-	    atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2])
+	    atomic_long_read(&tcp_memory_allocated) > sysctl_tcp_mem[2])
 		return true;
 	return false;
 }
diff --git a/include/net/udp.h b/include/net/udp.h
index 200b828..bb967dd 100644
--- a/include/net/udp.h
+++ b/include/net/udp.h
@@ -105,10 +105,10 @@ static inline struct udp_hslot *udp_hashslot2(struct udp_table *table,
 
 extern struct proto udp_prot;
 
-extern atomic_t udp_memory_allocated;
+extern atomic_long_t udp_memory_allocated;
 
 /* sysctl variables for udp */
-extern int sysctl_udp_mem[3];
+extern long sysctl_udp_mem[3];
 extern int sysctl_udp_rmem_min;
 extern int sysctl_udp_wmem_min;
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 3eed542..fb60801 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1653,10 +1653,10 @@ int __sk_mem_schedule(struct sock *sk, int size, int kind)
 {
 	struct proto *prot = sk->sk_prot;
 	int amt = sk_mem_pages(size);
-	int allocated;
+	long allocated;
 
 	sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
-	allocated = atomic_add_return(amt, prot->memory_allocated);
+	allocated = atomic_long_add_return(amt, prot->memory_allocated);
 
 	/* Under limit. */
 	if (allocated <= prot->sysctl_mem[0]) {
@@ -1714,7 +1714,7 @@ suppress_allocation:
 
 	/* Alas. Undo changes. */
 	sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
-	atomic_sub(amt, prot->memory_allocated);
+	atomic_long_sub(amt, prot->memory_allocated);
 	return 0;
 }
 EXPORT_SYMBOL(__sk_mem_schedule);
@@ -1727,12 +1727,12 @@ void __sk_mem_reclaim(struct sock *sk)
 {
 	struct proto *prot = sk->sk_prot;
 
-	atomic_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
+	atomic_long_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
 		   prot->memory_allocated);
 	sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
 
 	if (prot->memory_pressure && *prot->memory_pressure &&
-	    (atomic_read(prot->memory_allocated) < prot->sysctl_mem[0]))
+	    (atomic_long_read(prot->memory_allocated) < prot->sysctl_mem[0]))
 		*prot->memory_pressure = 0;
 }
 EXPORT_SYMBOL(__sk_mem_reclaim);
@@ -2452,12 +2452,12 @@ static char proto_method_implemented(const void *method)
 
 static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
 {
-	seq_printf(seq, "%-9s %4u %6d  %6d   %-3s %6u   %-3s  %-10s "
+	seq_printf(seq, "%-9s %4u %6d  %6ld   %-3s %6u   %-3s  %-10s "
 			"%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
 		   proto->name,
 		   proto->obj_size,
 		   sock_prot_inuse_get(seq_file_net(seq), proto),
-		   proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
+		   proto->memory_allocated != NULL ? atomic_long_read(proto->memory_allocated) : -1L,
 		   proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
 		   proto->max_header,
 		   proto->slab == NULL ? "no" : "yes",
diff --git a/net/decnet/af_decnet.c b/net/decnet/af_decnet.c
index d6b93d1..a76b78d 100644
--- a/net/decnet/af_decnet.c
+++ b/net/decnet/af_decnet.c
@@ -155,7 +155,7 @@ static const struct proto_ops dn_proto_ops;
 static DEFINE_RWLOCK(dn_hash_lock);
 static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE];
 static struct hlist_head dn_wild_sk;
-static atomic_t decnet_memory_allocated;
+static atomic_long_t decnet_memory_allocated;
 
 static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen, int flags);
 static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags);
diff --git a/net/decnet/sysctl_net_decnet.c b/net/decnet/sysctl_net_decnet.c
index be3eb8e..28f8b5e 100644
--- a/net/decnet/sysctl_net_decnet.c
+++ b/net/decnet/sysctl_net_decnet.c
@@ -38,7 +38,7 @@ int decnet_log_martians = 1;
 int decnet_no_fc_max_cwnd = NSP_MIN_WINDOW;
 
 /* Reasonable defaults, I hope, based on tcp's defaults */
-int sysctl_decnet_mem[3] = { 768 << 3, 1024 << 3, 1536 << 3 };
+long sysctl_decnet_mem[3] = { 768 << 3, 1024 << 3, 1536 << 3 };
 int sysctl_decnet_wmem[3] = { 4 * 1024, 16 * 1024, 128 * 1024 };
 int sysctl_decnet_rmem[3] = { 4 * 1024, 87380, 87380 * 2 };
 
@@ -324,7 +324,7 @@ static ctl_table dn_table[] = {
 		.data = &sysctl_decnet_mem,
 		.maxlen = sizeof(sysctl_decnet_mem),
 		.mode = 0644,
-		.proc_handler = proc_dointvec,
+		.proc_handler = proc_doulongvec_minmax
 	},
 	{
 		.procname = "decnet_rmem",
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index 4ae1f20..1b48eb1 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -59,13 +59,13 @@ static int sockstat_seq_show(struct seq_file *seq, void *v)
 	local_bh_enable();
 
 	socket_seq_show(seq);
-	seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n",
+	seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n",
 		   sock_prot_inuse_get(net, &tcp_prot), orphans,
 		   tcp_death_row.tw_count, sockets,
-		   atomic_read(&tcp_memory_allocated));
-	seq_printf(seq, "UDP: inuse %d mem %d\n",
+		   atomic_long_read(&tcp_memory_allocated));
+	seq_printf(seq, "UDP: inuse %d mem %ld\n",
 		   sock_prot_inuse_get(net, &udp_prot),
-		   atomic_read(&udp_memory_allocated));
+		   atomic_long_read(&udp_memory_allocated));
 	seq_printf(seq, "UDPLITE: inuse %d\n",
 		   sock_prot_inuse_get(net, &udplite_prot));
 	seq_printf(seq, "RAW: inuse %d\n",
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index d96c1da..e91911d 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -398,7 +398,7 @@ static struct ctl_table ipv4_table[] = {
 		.data		= &sysctl_tcp_mem,
 		.maxlen		= sizeof(sysctl_tcp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec
+		.proc_handler	= proc_doulongvec_minmax
 	},
 	{
 		.procname	= "tcp_wmem",
@@ -602,8 +602,7 @@ static struct ctl_table ipv4_table[] = {
 		.data		= &sysctl_udp_mem,
 		.maxlen		= sizeof(sysctl_udp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= &zero
+		.proc_handler	= proc_doulongvec_minmax,
 	},
 	{
 		.procname	= "udp_rmem_min",
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 1664a05..245603c 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -282,7 +282,7 @@ int sysctl_tcp_fin_timeout __read_mostly = TCP_FIN_TIMEOUT;
 struct percpu_counter tcp_orphan_count;
 EXPORT_SYMBOL_GPL(tcp_orphan_count);
 
-int sysctl_tcp_mem[3] __read_mostly;
+long sysctl_tcp_mem[3] __read_mostly;
 int sysctl_tcp_wmem[3] __read_mostly;
 int sysctl_tcp_rmem[3] __read_mostly;
 
@@ -290,7 +290,7 @@ EXPORT_SYMBOL(sysctl_tcp_mem);
 EXPORT_SYMBOL(sysctl_tcp_rmem);
 EXPORT_SYMBOL(sysctl_tcp_wmem);
 
-atomic_t tcp_memory_allocated;	/* Current allocated memory. */
+atomic_long_t tcp_memory_allocated;	/* Current allocated memory. */
 EXPORT_SYMBOL(tcp_memory_allocated);
 
 /*
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 3357f69..6d8ab1c 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -259,8 +259,11 @@ static void tcp_fixup_sndbuf(struct sock *sk)
 	int sndmem = tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER + 16 +
 		     sizeof(struct sk_buff);
 
-	if (sk->sk_sndbuf < 3 * sndmem)
-		sk->sk_sndbuf = min(3 * sndmem, sysctl_tcp_wmem[2]);
+	if (sk->sk_sndbuf < 3 * sndmem) {
+		sk->sk_sndbuf = 3 * sndmem;
+		if (sk->sk_sndbuf > sysctl_tcp_wmem[2])
+			sk->sk_sndbuf = sysctl_tcp_wmem[2];
+	}
 }
 
 /* 2. Tuning advertised window (window_clamp, rcv_ssthresh)
@@ -396,7 +399,7 @@ static void tcp_clamp_window(struct sock *sk)
 	if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] &&
 	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
 	    !tcp_memory_pressure &&
-	    atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) {
+	    atomic_long_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) {
 		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
 				    sysctl_tcp_rmem[2]);
 	}
@@ -4861,7 +4864,7 @@ static int tcp_should_expand_sndbuf(struct sock *sk)
 		return 0;
 
 	/* If we are under soft global TCP memory pressure, do not expand.  */
-	if (atomic_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0])
+	if (atomic_long_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0])
 		return 0;
 
 	/* If we filled the congestion window, do not expand.  */
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 28cb2d7..5e0a3a5 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -110,7 +110,7 @@
 struct udp_table udp_table __read_mostly;
 EXPORT_SYMBOL(udp_table);
 
-int sysctl_udp_mem[3] __read_mostly;
+long sysctl_udp_mem[3] __read_mostly;
 EXPORT_SYMBOL(sysctl_udp_mem);
 
 int sysctl_udp_rmem_min __read_mostly;
@@ -119,7 +119,7 @@ EXPORT_SYMBOL(sysctl_udp_rmem_min);
 int sysctl_udp_wmem_min __read_mostly;
 EXPORT_SYMBOL(sysctl_udp_wmem_min);
 
-atomic_t udp_memory_allocated;
+atomic_long_t udp_memory_allocated;
 EXPORT_SYMBOL(udp_memory_allocated);
 
 #define MAX_UDP_PORTS 65536
diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
index 1ef29c7..e58f947 100644
--- a/net/sctp/protocol.c
+++ b/net/sctp/protocol.c
@@ -92,7 +92,7 @@ static struct sctp_af *sctp_af_v6_specific;
 struct kmem_cache *sctp_chunk_cachep __read_mostly;
 struct kmem_cache *sctp_bucket_cachep __read_mostly;
 
-int sysctl_sctp_mem[3];
+long sysctl_sctp_mem[3];
 int sysctl_sctp_rmem[3];
 int sysctl_sctp_wmem[3];
 
diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index e34ca9c..6bd5543 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -111,12 +111,12 @@ static void sctp_sock_migrate(struct sock *, struct sock *,
 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
 
 extern struct kmem_cache *sctp_bucket_cachep;
-extern int sysctl_sctp_mem[3];
+extern long sysctl_sctp_mem[3];
 extern int sysctl_sctp_rmem[3];
 extern int sysctl_sctp_wmem[3];
 
 static int sctp_memory_pressure;
-static atomic_t sctp_memory_allocated;
+static atomic_long_t sctp_memory_allocated;
 struct percpu_counter sctp_sockets_allocated;
 
 static void sctp_enter_memory_pressure(struct sock *sk)
diff --git a/net/sctp/sysctl.c b/net/sctp/sysctl.c
index 832590b..50cb57f 100644
--- a/net/sctp/sysctl.c
+++ b/net/sctp/sysctl.c
@@ -54,7 +54,7 @@ static int sack_timer_max = 500;
 static int addr_scope_max = 3; /* check sctp_scope_policy_t in include/net/sctp/constants.h for max entries */
 static int rwnd_scale_max = 16;
 
-extern int sysctl_sctp_mem[3];
+extern long sysctl_sctp_mem[3];
 extern int sysctl_sctp_rmem[3];
 extern int sysctl_sctp_wmem[3];
 
@@ -203,7 +203,7 @@ static ctl_table sctp_table[] = {
 		.data		= &sysctl_sctp_mem,
 		.maxlen		= sizeof(sysctl_sctp_mem),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec,
+		.proc_handler	= proc_doulongvec_minmax
 	},
 	{
 		.procname	= "sctp_rmem",

Re: [PATCH] net: avoid limits overflow

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 10 Nov 2010 10:24:26 +0100

> [PATCH] net: avoid limits overflow
> 
> Robin Holt tried to boot a 16TB machine and found some limits were
> reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]
> 
> We can switch infrastructure to use long "instead" of "int", now
> atomic_long_t primitives are available for free.
> 
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> Reported-by: Robin Holt <holt@sgi.com>
> Reviewed-by: Robin Holt <holt@sgi.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Applied.

Re: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[David Miller]

From: Robin Holt <holt@sgi.com>
Date: Wed, 20 Oct 2010 07:03:37 -0500

> Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.
> 
> On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
> sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
> maximized without overflowing.
> 
> Signed-off-by: Robin Holt <holt@sgi.com>

I'll queue this up for -stable, thanks Robin!

[Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

[-- Attachment #1: sysctl_tcp_udp_mem_max_overflows --]
[-- Type: text/plain, Size: 4098 bytes --]

Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
maximized without overflowing.

Signed-off-by: Robin Holt <holt@sgi.com>
To: "David S. Miller" <davem@davemloft.net>
Cc: Willy Tarreau <w@1wt.eu>
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Cc: linux-sctp@vger.kernel.org
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: James Morris <jmorris@namei.org>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Vlad Yasevich <vladislav.yasevich@hp.com>
Cc: Sridhar Samudrala <sri@us.ibm.com>

---

David, I believe this is also intended for stable.  If you agree, please
add the stable maintainers to the Cc: list.

 net/ipv4/tcp.c      |    4 +++-
 net/ipv4/udp.c      |    4 +++-
 net/sctp/protocol.c |    4 +++-
 3 files changed, 9 insertions(+), 3 deletions(-)

Index: pv1010932/net/ipv4/tcp.c
===================================================================
--- pv1010932.orig/net/ipv4/tcp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/tcp.c	2010-10-02 06:12:35.445454593 -0500
@@ -3271,12 +3271,14 @@ void __init tcp_init(void)
 
 	/* 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.
+	 * memory, with a floor of 128 pages, and a ceiling that prevents an
+	 * integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_tcp_mem[0] = limit / 4 * 3;
 	sysctl_tcp_mem[1] = limit;
 	sysctl_tcp_mem[2] = sysctl_tcp_mem[0] * 2;
Index: pv1010932/net/ipv4/udp.c
===================================================================
--- pv1010932.orig/net/ipv4/udp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/udp.c	2010-10-02 06:12:35.453453784 -0500
@@ -2167,12 +2167,14 @@ void __init udp_init(void)
 	udp_table_init(&udp_table, "UDP");
 	/* Set the pressure threshold up by the same strategy of TCP. It is 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.
+	 * toward zero with the amount of memory, with a floor of 128 pages,
+	 * and a ceiling that prevents an integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_udp_mem[0] = limit / 4 * 3;
 	sysctl_udp_mem[1] = limit;
 	sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
Index: pv1010932/net/sctp/protocol.c
===================================================================
--- pv1010932.orig/net/sctp/protocol.c	2010-10-02 06:11:59.000000000 -0500
+++ pv1010932/net/sctp/protocol.c	2010-10-02 06:13:17.727949810 -0500
@@ -1162,7 +1162,8 @@ SCTP_STATIC __init int sctp_init(void)
 
 	/* 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.
+	 * memory, with a floor of 128 pages, and a ceiling that prevents an
+	 * integer overflow.
 	 * Note this initializes the data in sctpv6_prot too
 	 * Unabashedly stolen from tcp_init
 	 */
@@ -1170,6 +1171,7 @@ SCTP_STATIC __init int sctp_init(void)
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_sctp_mem[0] = limit / 4 * 3;
 	sysctl_sctp_mem[1] = limit;
 	sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;

[Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

[Robin Holt]

Subject: [Patch] Limit sysctl_tcp_mem and sysctl_udp_mem initializers to prevent integer overflows.

On a 16TB x86_64 machine, sysctl_tcp_mem[2], sysctl_udp_mem[2], and
sysctl_sctp_mem[2] can integer overflow.  Set limit such that they are
maximized without overflowing.

Signed-off-by: Robin Holt <holt@sgi.com>
To: "David S. Miller" <davem@davemloft.net>
Cc: Willy Tarreau <w@1wt.eu>
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Cc: linux-sctp@vger.kernel.org
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: James Morris <jmorris@namei.org>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Vlad Yasevich <vladislav.yasevich@hp.com>
Cc: Sridhar Samudrala <sri@us.ibm.com>

---

David, I believe this is also intended for stable.  If you agree, please
add the stable maintainers to the Cc: list.

 net/ipv4/tcp.c      |    4 +++-
 net/ipv4/udp.c      |    4 +++-
 net/sctp/protocol.c |    4 +++-
 3 files changed, 9 insertions(+), 3 deletions(-)

Index: pv1010932/net/ipv4/tcp.c
=================================--- pv1010932.orig/net/ipv4/tcp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/tcp.c	2010-10-02 06:12:35.445454593 -0500
@@ -3271,12 +3271,14 @@ void __init tcp_init(void)
 
 	/* 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.
+	 * memory, with a floor of 128 pages, and a ceiling that prevents an
+	 * integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_tcp_mem[0] = limit / 4 * 3;
 	sysctl_tcp_mem[1] = limit;
 	sysctl_tcp_mem[2] = sysctl_tcp_mem[0] * 2;
Index: pv1010932/net/ipv4/udp.c
=================================--- pv1010932.orig/net/ipv4/udp.c	2010-10-02 06:11:59.737449853 -0500
+++ pv1010932/net/ipv4/udp.c	2010-10-02 06:12:35.453453784 -0500
@@ -2167,12 +2167,14 @@ void __init udp_init(void)
 	udp_table_init(&udp_table, "UDP");
 	/* Set the pressure threshold up by the same strategy of TCP. It is 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.
+	 * toward zero with the amount of memory, with a floor of 128 pages,
+	 * and a ceiling that prevents an integer overflow.
 	 */
 	nr_pages = totalram_pages - totalhigh_pages;
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_udp_mem[0] = limit / 4 * 3;
 	sysctl_udp_mem[1] = limit;
 	sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
Index: pv1010932/net/sctp/protocol.c
=================================--- pv1010932.orig/net/sctp/protocol.c	2010-10-02 06:11:59.000000000 -0500
+++ pv1010932/net/sctp/protocol.c	2010-10-02 06:13:17.727949810 -0500
@@ -1162,7 +1162,8 @@ SCTP_STATIC __init int sctp_init(void)
 
 	/* 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.
+	 * memory, with a floor of 128 pages, and a ceiling that prevents an
+	 * integer overflow.
 	 * Note this initializes the data in sctpv6_prot too
 	 * Unabashedly stolen from tcp_init
 	 */
@@ -1170,6 +1171,7 @@ SCTP_STATIC __init int sctp_init(void)
 	limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
 	limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
 	limit = max(limit, 128UL);
+	limit = min(limit, INT_MAX * 4UL / 3 / 2);
 	sysctl_sctp_mem[0] = limit / 4 * 3;
 	sysctl_sctp_mem[1] = limit;
 	sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;