Modify the IP fragment reassembly code so that it uses a new UMA zone,

ipq_zone, to allocate fragment headers from, rather than using cast mbuf
storage.  This was one of the few remaining uses of mbuf storage for
local data structures that relied on dtom().  Implement the resource
limit on ipq's using UMA zone limits, but preserve current sysctl
semantics using a sysctl proc.

MFC after:	3 weeks

1 files changed, 81 insertions, 21 deletions

diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c
index dc9a92b..0b06255 100644
--- a/sys/netinet/ip_input.c
+++ b/sys/netinet/ip_input.c
@@ -107,17 +107,6 @@ SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
 	&ip_keepfaith,	0,
 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
 
-static int    nipq = 0;         /* total # of reass queues */
-static int    maxnipq;
-SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
-	&maxnipq, 0,
-	"Maximum number of IPv4 fragment reassembly queue entries");
-
-static int    maxfragsperpacket;
-SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
-	&maxfragsperpacket, 0,
-	"Maximum number of IPv4 fragments allowed per packet");
-
 static int	ip_sendsourcequench = 0;
 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
 	&ip_sendsourcequench, 0,
@@ -166,22 +155,38 @@ struct ipstat ipstat;
 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
 
-/* Packet reassembly stuff */
+/*
+ * IP datagram reassembly.
+ */
 #define IPREASS_NHASH_LOG2      6
 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
 #define IPREASS_HASH(x,y) \
 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
 
+static uma_zone_t ipq_zone;
 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
 static struct mtx ipqlock;
-struct callout ipport_tick_callout;
 
 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
 
+static void	maxnipq_update(void);
+
+static int	maxnipq;	/* Administrative limit on # reass queues. */
+static int	nipq = 0;	/* Total # of reass queues */
+SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0,
+	"Current number of IPv4 fragment reassembly queue entries");
+
+static int	maxfragsperpacket;
+SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
+	&maxfragsperpacket, 0,
+	"Maximum number of IPv4 fragments allowed per packet");
+
+struct callout	ipport_tick_callout;
+
 #ifdef IPCTL_DEFMTU
 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
     &ip_mtu, 0, "Default MTU");
@@ -249,6 +254,9 @@ ip_init()
 	    TAILQ_INIT(&ipq[i]);
 	maxnipq = nmbclusters / 32;
 	maxfragsperpacket = 16;
+	ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
+	    NULL, UMA_ALIGN_PTR, 0);
+	maxnipq_update();
 
 	/* Start ipport_tick. */
 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
@@ -747,6 +755,59 @@ bad:
 }
 
 /*
+ * After maxnipq has been updated, propagate the change to UMA.  The UMA zone
+ * max has slightly different semantics than the sysctl, for historical
+ * reasons.
+ */
+static void
+maxnipq_update(void)
+{
+
+	/*
+	 * -1 for unlimited allocation.
+	 */
+	if (maxnipq < 0)
+		uma_zone_set_max(ipq_zone, 0);
+	/*
+	 * Positive number for specific bound.
+	 */
+	if (maxnipq > 0)
+		uma_zone_set_max(ipq_zone, maxnipq);
+	/*
+	 * Zero specifies no further fragment queue allocation -- set the
+	 * bound very low, but rely on implementation elsewhere to actually
+	 * prevent allocation and reclaim current queues.
+	 */
+	if (maxnipq == 0)
+		uma_zone_set_max(ipq_zone, 1);
+}
+
+static int
+sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
+{
+	int error, i;
+
+	i = maxnipq;
+	error = sysctl_handle_int(oidp, &i, 0, req);
+	if (error || !req->newptr)
+		return (error);
+
+	/*
+	 * XXXRW: Might be a good idea to sanity check the argument and place
+	 * an extreme upper bound.
+	 */
+	if (i < -1)
+		return (EINVAL);
+	maxnipq = i;
+	maxnipq_update();
+	return (0);
+}
+
+SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
+    NULL, 0, sysctl_maxnipq, "I",
+    "Maximum number of IPv4 fragment reassembly queue entries");
+
+/*
  * Take incoming datagram fragment and try to reassemble it into
  * whole datagram.  If the argument is the first fragment or one
  * in between the function will return NULL and store the mbuf
@@ -800,9 +861,8 @@ ip_reass(struct mbuf *m)
 	fp = NULL;
 
 	/*
-	 * Enforce upper bound on number of fragmented packets
-	 * for which we attempt reassembly;
-	 * If maxnipq is -1, accept all fragments without limitation.
+	 * Attempt to trim the number of allocated fragment queues if it
+	 * exceeds the administrative limit.
 	 */
 	if ((nipq > maxnipq) && (maxnipq > 0)) {
 		/*
@@ -865,12 +925,12 @@ found:
 	 * If first fragment to arrive, create a reassembly queue.
 	 */
 	if (fp == NULL) {
-		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
+		fp = uma_zalloc(ipq_zone, M_NOWAIT);
+		if (fp == NULL)
 			goto dropfrag;
-		fp = mtod(t, struct ipq *);
 #ifdef MAC
 		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
-			m_free(t);
+			uma_zfree(ipq_zone, fp);
 			goto dropfrag;
 		}
 		mac_create_ipq(m, fp);
@@ -1038,7 +1098,7 @@ found:
 	ip->ip_dst = fp->ipq_dst;
 	TAILQ_REMOVE(head, fp, ipq_list);
 	nipq--;
-	(void) m_free(dtom(fp));
+	uma_zfree(ipq_zone, fp);
 	m->m_len += (ip->ip_hl << 2);
 	m->m_data -= (ip->ip_hl << 2);
 	/* some debugging cruft by sklower, below, will go away soon */
@@ -1079,7 +1139,7 @@ ip_freef(fhp, fp)
 		m_freem(q);
 	}
 	TAILQ_REMOVE(fhp, fp, ipq_list);
-	(void) m_free(dtom(fp));
+	uma_zfree(ipq_zone, fp);
 	nipq--;
 }