* 4 converted to << 2 for networking code

* 4 converted to << 2 for networking code

[antirez]

[-- Attachment #1: Type: text/plain, Size: 560 bytes --]

Hi all,

The attached patch converts many occurences of '* 4' in the networking code
(often used to convert in bytes the TCP data offset and the IP header len)
to the faster '<< 2'. Since this was a quite repetitive work it's better
if someone double-check it before to apply the patch.
The patch is for linux-2.4.

Please, CC: me for replies since I'm not subscribed to the list.

regards,
antirez

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com

[-- Attachment #2: shift.diff --]
[-- Type: text/plain, Size: 23741 bytes --]

diff -ru linux-2.4-orig/net/ipv4/ip_fragment.c linux-2.4/net/ipv4/ip_fragment.c
--- linux-2.4-orig/net/ipv4/ip_fragment.c	Fri Dec 29 23:07:24 2000
+++ linux-2.4/net/ipv4/ip_fragment.c	Wed Jan 10 16:50:05 2001
@@ -383,7 +383,7 @@
 	flags = offset & ~IP_OFFSET;
 	offset &= IP_OFFSET;
 	offset <<= 3;		/* offset is in 8-byte chunks */
-	ihl = iph->ihl * 4;
+	ihl = iph->ihl << 2;
 
 	/* Determine the position of this fragment. */
 	end = offset + (ntohs(iph->tot_len) - ihl);
@@ -521,7 +521,7 @@
 	BUG_TRAP(FRAG_CB(head)->offset == 0);
 
 	/* Allocate a new buffer for the datagram. */
-	ihlen = head->nh.iph->ihl*4;
+	ihlen = head->nh.iph->ihl << 2;
 	len = ihlen + qp->len;
 
 	if(len > 65535)
diff -ru linux-2.4-orig/net/ipv4/ip_input.c linux-2.4/net/ipv4/ip_input.c
--- linux-2.4-orig/net/ipv4/ip_input.c	Mon Dec 11 21:37:04 2000
+++ linux-2.4/net/ipv4/ip_input.c	Wed Jan 10 16:50:31 2001
@@ -208,7 +208,7 @@
 			if(skb2 != NULL) {
 				ret = 1;
 				ipprot->handler(skb2,
-						ntohs(iph->tot_len) - (iph->ihl * 4));
+						ntohs(iph->tot_len) - (iph->ihl << 2));
 			}
 		}
 		ipprot = (struct inet_protocol *) ipprot->next;
@@ -226,7 +226,7 @@
 #endif /*CONFIG_NETFILTER_DEBUG*/
 
         /* Point into the IP datagram, just past the header. */
-        skb->h.raw = skb->nh.raw + iph->ihl*4;
+        skb->h.raw = skb->nh.raw + (iph->ihl << 2);
 
 	{
 		/* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
@@ -251,7 +251,7 @@
 				
 				/* Fast path... */
 				ret = ipprot->handler(skb, (ntohs(iph->tot_len) -
-							    (iph->ihl * 4)));
+							    (iph->ihl << 2)));
 
 				return ret;
 			} else {
diff -ru linux-2.4-orig/net/ipv4/ip_options.c linux-2.4/net/ipv4/ip_options.c
--- linux-2.4-orig/net/ipv4/ip_options.c	Wed Aug  9 22:51:09 2000
+++ linux-2.4/net/ipv4/ip_options.c	Wed Jan 10 16:41:12 2001
@@ -255,7 +255,7 @@
 		opt = &(IPCB(skb)->opt);
 		memset(opt, 0, sizeof(struct ip_options));
 		iph = skb->nh.raw;
-		opt->optlen = ((struct iphdr *)iph)->ihl*4 - sizeof(struct iphdr);
+		opt->optlen = (((struct iphdr *)iph)->ihl << 2) - sizeof(struct iphdr);
 		optptr = iph + sizeof(struct iphdr);
 		opt->is_data = 0;
 	} else {
diff -ru linux-2.4-orig/net/ipv4/ip_output.c linux-2.4/net/ipv4/ip_output.c
--- linux-2.4-orig/net/ipv4/ip_output.c	Fri Oct 27 20:03:14 2000
+++ linux-2.4/net/ipv4/ip_output.c	Wed Jan 10 16:51:18 2001
@@ -537,7 +537,7 @@
 						 ipc->addr, rt, offset);
 			}
 			iph->tos = sk->protinfo.af_inet.tos;
-			iph->tot_len = htons(fraglen - fragheaderlen + iph->ihl*4);
+			iph->tot_len = htons(fraglen - fragheaderlen + (iph->ihl << 2));
 			iph->frag_off = htons(offset>>3)|mf|df;
 			iph->id = id;
 			if (!mf) {
@@ -564,7 +564,7 @@
 			iph->saddr = rt->rt_src;
 			iph->daddr = rt->rt_dst;
 			iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
-			data += iph->ihl*4;
+			data += iph->ihl << 2;
 		}
 
 		/*
@@ -685,7 +685,7 @@
 		iph->daddr=rt->rt_dst;
 		iph->check=0;
 		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
-		err = getfrag(frag, ((char *)iph)+iph->ihl*4,0, length-iph->ihl*4);
+		err = getfrag(frag, ((char *)iph)+(iph->ihl << 2),0, length-(iph->ihl << 2));
 	}
 	else
 		err = getfrag(frag, (void *)iph, 0, length);
@@ -745,7 +745,7 @@
 	 *	Setup starting values.
 	 */
 
-	hlen = iph->ihl * 4;
+	hlen = iph->ihl << 2;
 	left = ntohs(iph->tot_len) - hlen;	/* Space per frame */
 	mtu = rt->u.dst.pmtu - hlen;	/* Size of data space */
 	ptr = raw + hlen;			/* Where to start from */
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_core.c linux-2.4/net/ipv4/netfilter/ip_conntrack_core.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_core.c	Thu Aug 10 21:35:15 2000
+++ linux-2.4/net/ipv4/netfilter/ip_conntrack_core.c	Wed Jan 10 16:46:23 2001
@@ -123,7 +123,7 @@
 		return 0;
 	}
 	/* Guarantee 8 protocol bytes: if more wanted, use len param */
-	else if (iph->ihl * 4 + 8 > len)
+	else if ((iph->ihl << 2) + 8 > len)
 		return 0;
 
 	tuple->src.ip = iph->saddr;
@@ -308,9 +308,9 @@
 	iph = skb->nh.iph;
 	hdr = (struct icmphdr *)((u_int32_t *)iph + iph->ihl);
 	inner = (struct iphdr *)(hdr + 1);
-	datalen = skb->len - iph->ihl*4 - sizeof(*hdr);
+	datalen = skb->len - (iph->ihl << 2) - sizeof(*hdr);
 
-	if (skb->len < iph->ihl * 4 + sizeof(struct icmphdr)) {
+	if (skb->len < (iph->ihl << 2) + sizeof(struct icmphdr)) {
 		DEBUGP("icmp_error_track: too short\n");
 		return NULL;
 	}
@@ -337,7 +337,7 @@
 
 	innerproto = find_proto(inner->protocol);
 	/* Are they talking about one of our connections? */
-	if (inner->ihl * 4 + 8 > datalen
+	if ((inner->ihl << 2) + 8 > datalen
 	    || !get_tuple(inner, datalen, &origtuple, innerproto)) {
 		DEBUGP("icmp_error: ! get_tuple p=%u (%u*4+%u dlen=%u)\n",
 		       inner->protocol, inner->ihl, 8,
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_ftp.c linux-2.4/net/ipv4/netfilter/ip_conntrack_ftp.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_ftp.c	Thu Aug 10 21:35:15 2000
+++ linux-2.4/net/ipv4/netfilter/ip_conntrack_ftp.c	Wed Jan 10 16:47:37 2001
@@ -104,10 +104,10 @@
 		enum ip_conntrack_info ctinfo)
 {
 	/* tcplen not negative guaranteed by ip_conntrack_tcp.c */
-	struct tcphdr *tcph = (void *)iph + iph->ihl * 4;
-	const char *data = (const char *)tcph + tcph->doff * 4;
-	unsigned int tcplen = len - iph->ihl * 4;
-	unsigned int datalen = tcplen - tcph->doff * 4;
+	struct tcphdr *tcph = (void *)iph + (iph->ihl << 2);
+	const char *data = (const char *)tcph + (tcph->doff << 2);
+	unsigned int tcplen = len - (iph->ihl << 2);
+	unsigned int datalen = tcplen - (tcph->doff << 2);
 	u_int32_t old_seq_aft_nl;
 	int old_seq_aft_nl_set;
 	u_int32_t array[6] = { 0 };
@@ -124,7 +124,7 @@
 	}
 
 	/* Not whole TCP header? */
-	if (tcplen < sizeof(struct tcphdr) || tcplen < tcph->doff*4) {
+	if (tcplen < sizeof(struct tcphdr) || tcplen < (tcph->doff << 2)) {
 		DEBUGP("ftp: tcplen = %u\n", (unsigned)tcplen);
 		return NF_ACCEPT;
 	}
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_proto_tcp.c linux-2.4/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_proto_tcp.c	Fri Aug  4 22:07:24 2000
+++ linux-2.4/net/ipv4/netfilter/ip_conntrack_proto_tcp.c	Wed Jan 10 16:49:20 2001
@@ -153,7 +153,7 @@
 
 	/* We're guaranteed to have the base header, but maybe not the
            options. */
-	if (len < (iph->ihl + tcph->doff) * 4) {
+	if (len < ((iph->ihl + tcph->doff) << 2)) {
 		DEBUGP("ip_conntrack_tcp: Truncated packet.\n");
 		return -1;
 	}
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_standalone.c linux-2.4/net/ipv4/netfilter/ip_conntrack_standalone.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_conntrack_standalone.c	Thu Aug 10 21:35:15 2000
+++ linux-2.4/net/ipv4/netfilter/ip_conntrack_standalone.c	Wed Jan 10 16:51:59 2001
@@ -217,7 +217,7 @@
 {
 	/* root is playing with raw sockets. */
 	if ((*pskb)->len < sizeof(struct iphdr)
-	    || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
+	    || ((*pskb)->nh.iph->ihl << 2) < sizeof(struct iphdr)) {
 		if (net_ratelimit())
 			printk("ipt_hook: happy cracking.\n");
 		return NF_ACCEPT;
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_nat_core.c linux-2.4/net/ipv4/netfilter/ip_nat_core.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_nat_core.c	Thu Aug 10 21:35:15 2000
+++ linux-2.4/net/ipv4/netfilter/ip_nat_core.c	Wed Jan 10 16:52:27 2001
@@ -749,7 +749,7 @@
 	unsigned int i;
 	struct ip_nat_info *info = &conntrack->nat.info;
 
-	IP_NF_ASSERT(skb->len >= iph->ihl*4 + sizeof(struct icmphdr));
+	IP_NF_ASSERT(skb->len >= (iph->ihl << 2) + sizeof(struct icmphdr));
 	/* Must be RELATED */
 	IP_NF_ASSERT(skb->nfct - (struct ip_conntrack *)skb->nfct->master
 		     == IP_CT_RELATED
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_nat_ftp.c linux-2.4/net/ipv4/netfilter/ip_nat_ftp.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_nat_ftp.c	Sun Sep 17 19:15:00 2000
+++ linux-2.4/net/ipv4/netfilter/ip_nat_ftp.c	Wed Jan 10 16:55:21 2001
@@ -112,9 +112,9 @@
 	sprintf(buffer, "%u,%u,%u,%u,%u,%u",
 		NIPQUAD(newip), port>>8, port&0xFF);
 
-	tcplen = (*pskb)->len - iph->ihl * 4;
+	tcplen = (*pskb)->len - (iph->ihl << 2);
 	newtcplen = tcplen - matchlen + strlen(buffer);
-	newlen = iph->ihl*4 + newtcplen;
+	newlen = (iph->ihl << 2) + newtcplen;
 
 	/* So there I am, in the middle of my `netfilter-is-wonderful'
 	   talk in Sydney, and someone asks `What happens if you try
@@ -144,8 +144,8 @@
 		}
 	}
 
-	tcph = (void *)iph + iph->ihl*4;
-	data = (void *)tcph + tcph->doff*4;
+	tcph = (void *)iph + (iph->ihl << 2);
+	data = (void *)tcph + (tcph->doff << 2);
 
 	DEBUGP("Mapping `%.*s' [%u %u %u] to new `%s' [%u]\n",
 		 (int)matchlen, data+matchoff,
@@ -183,11 +183,11 @@
 
 	/* Fix checksums */
 	iph->tot_len = htons(newlen);
-	(*pskb)->csum = csum_partial((char *)tcph + tcph->doff*4,
-				     newtcplen - tcph->doff*4, 0);
+	(*pskb)->csum = csum_partial((char *)tcph + (tcph->doff << 2),
+				     newtcplen - (tcph->doff << 2), 0);
 	tcph->check = 0;
 	tcph->check = tcp_v4_check(tcph, newtcplen, iph->saddr, iph->daddr,
-				   csum_partial((char *)tcph, tcph->doff*4,
+				   csum_partial((char *)tcph, tcph->doff << 2,
 						(*pskb)->csum));
 	ip_send_check(iph);
 	return 1;
@@ -202,8 +202,8 @@
 	u_int8_t *opt = (u_int8_t *)tcph;
 
 	DEBUGP("Seeking SACKPERM in SYN packet (doff = %u).\n",
-	       tcph->doff * 4);
-	for (i = sizeof(struct tcphdr); i < tcph->doff * 4;) {
+	       tcph->doff << 2);
+	for (i = sizeof(struct tcphdr); i < (tcph->doff << 2);) {
 		DEBUGP("%u ", opt[i]);
 		switch (opt[i]) {
 		case TCPOPT_NOP:
@@ -227,7 +227,7 @@
 	DEBUGP("Found SACKPERM at offset %u.\n", i);
 
 	/* Must be within TCP header, and valid SACK perm. */
-	if (i + opt[i+1] <= tcph->doff*4 && opt[i+1] == 2) {
+	if (i + opt[i+1] <= (tcph->doff << 2) && opt[i+1] == 2) {
 		/* Replace with NOPs. */
 		tcph->check
 			= ip_nat_cheat_check(*((u_int16_t *)(opt + i))^0xFFFF,
@@ -245,7 +245,7 @@
 {
 	u_int32_t newip;
 	struct iphdr *iph = (*pskb)->nh.iph;
-	struct tcphdr *tcph = (void *)iph + iph->ihl*4;
+	struct tcphdr *tcph = (void *)iph + (iph->ihl << 2);
 	u_int16_t port;
 	struct ip_conntrack_tuple tuple;
 	/* Don't care about source port */
@@ -304,7 +304,7 @@
 			 struct sk_buff **pskb)
 {
 	struct iphdr *iph = (*pskb)->nh.iph;
-	struct tcphdr *tcph = (void *)iph + iph->ihl*4;
+	struct tcphdr *tcph = (void *)iph + (iph->ihl << 2);
 	u_int32_t newseq, newack;
 	unsigned int datalen;
 	int dir;
@@ -331,7 +331,7 @@
 		return NF_ACCEPT;
 	}
 
-	datalen = (*pskb)->len - iph->ihl * 4 - tcph->doff * 4;
+	datalen = (*pskb)->len - (iph->ihl << 2) - (tcph->doff << 2);
 	score = 0;
 	LOCK_BH(&ip_ftp_lock);
 	if (ct_ftp_info->len) {
@@ -361,7 +361,7 @@
 
 			/* skb may have been reallocated */
 			iph = (*pskb)->nh.iph;
-			tcph = (void *)iph + iph->ihl*4;
+			tcph = (void *)iph + (iph->ihl << 2);
 		}
 	}
 
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_nat_standalone.c linux-2.4/net/ipv4/netfilter/ip_nat_standalone.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_nat_standalone.c	Mon Oct 30 23:27:49 2000
+++ linux-2.4/net/ipv4/netfilter/ip_nat_standalone.c	Wed Jan 10 16:55:51 2001
@@ -139,7 +139,7 @@
 {
 	/* root is playing with raw sockets. */
 	if ((*pskb)->len < sizeof(struct iphdr)
-	    || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
+	    || ((*pskb)->nh.iph->ihl << 2) < sizeof(struct iphdr))
 		return NF_ACCEPT;
 
 	/* We can hit fragment here; forwarded packets get
@@ -202,7 +202,7 @@
 
 	/* root is playing with raw sockets. */
 	if ((*pskb)->len < sizeof(struct iphdr)
-	    || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
+	    || ((*pskb)->nh.iph->ihl << 2) < sizeof(struct iphdr))
 		return NF_ACCEPT;
 
 	saddr = (*pskb)->nh.iph->saddr;
diff -ru linux-2.4-orig/net/ipv4/netfilter/ip_tables.c linux-2.4/net/ipv4/netfilter/ip_tables.c
--- linux-2.4-orig/net/ipv4/netfilter/ip_tables.c	Tue Dec  5 03:43:02 2000
+++ linux-2.4/net/ipv4/netfilter/ip_tables.c	Wed Jan 10 16:56:27 2001
@@ -267,7 +267,7 @@
 	/* Initialization */
 	ip = (*pskb)->nh.iph;
 	protohdr = (u_int32_t *)ip + ip->ihl;
-	datalen = (*pskb)->len - ip->ihl * 4;
+	datalen = (*pskb)->len - (ip->ihl << 2);
 	indev = in ? in->name : nulldevname;
 	outdev = out ? out->name : nulldevname;
 	/* We handle fragments by dealing with the first fragment as
@@ -373,7 +373,7 @@
 				/* Target might have changed stuff. */
 				ip = (*pskb)->nh.iph;
 				protohdr = (u_int32_t *)ip + ip->ihl;
-				datalen = (*pskb)->len - ip->ihl * 4;
+				datalen = (*pskb)->len - (ip->ihl << 2);
 
 				if (verdict == IPT_CONTINUE)
 					e = (void *)e + e->next_offset;
@@ -1450,12 +1450,12 @@
 
 	duprintf("tcp_match: finding option\n");
 	/* If we don't have the whole header, drop packet. */
-	if (tcp->doff * 4 > datalen) {
+	if ((tcp->doff << 2) > datalen) {
 		*hotdrop = 1;
 		return 0;
 	}
 
-	while (i < tcp->doff * 4) {
+	while (i < (tcp->doff << 2)) {
 		if (opt[i] == option) return !invert;
 		if (opt[i] < 2) i++;
 		else i += opt[i+1]?:1;
diff -ru linux-2.4-orig/net/ipv4/netfilter/ipchains_core.c linux-2.4/net/ipv4/netfilter/ipchains_core.c
--- linux-2.4-orig/net/ipv4/netfilter/ipchains_core.c	Fri Apr 14 02:19:57 2000
+++ linux-2.4/net/ipv4/netfilter/ipchains_core.c	Wed Jan 10 16:56:50 2001
@@ -1679,7 +1679,7 @@
 		      struct sk_buff **pskb)
 {
 	/* Locally generated bogus packets by root. <SIGH>. */
-	if (((struct iphdr *)phdr)->ihl * 4 < sizeof(struct iphdr)
+	if ((((struct iphdr *)phdr)->ihl << 2) < sizeof(struct iphdr)
 	    || (*pskb)->len < sizeof(struct iphdr))
 		return FW_ACCEPT;
 	return ip_fw_check(phdr, dev->name,
diff -ru linux-2.4-orig/net/ipv4/netfilter/ipt_LOG.c linux-2.4/net/ipv4/netfilter/ipt_LOG.c
--- linux-2.4-orig/net/ipv4/netfilter/ipt_LOG.c	Mon Jan  1 18:54:07 2001
+++ linux-2.4/net/ipv4/netfilter/ipt_LOG.c	Wed Jan 10 16:57:48 2001
@@ -32,7 +32,7 @@
 			struct iphdr *iph, unsigned int len, int recurse)
 {
 	void *protoh = (u_int32_t *)iph + iph->ihl;
-	unsigned int datalen = len - iph->ihl * 4;
+	unsigned int datalen = len - (iph->ihl << 2);
 
 	/* Important fields:
 	 * TOS, len, DF/MF, fragment offset, TTL, src, dst, options. */
@@ -58,12 +58,12 @@
 		printk("FRAG:%u ", ntohs(iph->frag_off) & IP_OFFSET);
 
 	if ((info->logflags & IPT_LOG_IPOPT)
-	    && iph->ihl * 4 != sizeof(struct iphdr)) {
+	    && (iph->ihl << 2) != sizeof(struct iphdr)) {
 		unsigned int i;
 
 		/* Max length: 127 "OPT (" 15*4*2chars ") " */
 		printk("OPT (");
-		for (i = sizeof(struct iphdr); i < iph->ihl * 4; i++)
+		for (i = sizeof(struct iphdr); i < (iph->ihl << 2); i++)
 			printk("%02X", ((u_int8_t *)iph)[i]);
 		printk(") ");
 	}
@@ -112,12 +112,12 @@
 		printk("URGP=%u ", ntohs(tcph->urg_ptr));
 
 		if ((info->logflags & IPT_LOG_TCPOPT)
-		    && tcph->doff * 4 != sizeof(struct tcphdr)) {
+		    && (tcph->doff << 2) != sizeof(struct tcphdr)) {
 			unsigned int i;
 
 			/* Max length: 127 "OPT (" 15*4*2chars ") " */
 			printk("OPT (");
-			for (i =sizeof(struct tcphdr); i < tcph->doff * 4; i++)
+			for (i =sizeof(struct tcphdr); i < (tcph->doff << 2); i++)
 				printk("%02X", ((u_int8_t *)tcph)[i]);
 			printk(") ");
 		}
diff -ru linux-2.4-orig/net/ipv4/netfilter/ipt_REJECT.c linux-2.4/net/ipv4/netfilter/ipt_REJECT.c
--- linux-2.4-orig/net/ipv4/netfilter/ipt_REJECT.c	Tue Sep 19 17:31:53 2000
+++ linux-2.4/net/ipv4/netfilter/ipt_REJECT.c	Wed Jan 10 16:58:37 2001
@@ -36,7 +36,7 @@
 		return;
 
 	otcph = (struct tcphdr *)((u_int32_t*)oldskb->nh.iph + oldskb->nh.iph->ihl);
-	otcplen = oldskb->len - oldskb->nh.iph->ihl*4;
+	otcplen = oldskb->len - (oldskb->nh.iph->ihl << 2);
 
 	/* No RST for RST. */
 	if (otcph->rst)
@@ -73,7 +73,7 @@
 
 	/* Truncate to length (no data) */
 	tcph->doff = sizeof(struct tcphdr)/4;
-	skb_trim(nskb, nskb->nh.iph->ihl*4 + sizeof(struct tcphdr));
+	skb_trim(nskb, (nskb->nh.iph->ihl << 2) + sizeof(struct tcphdr));
 	nskb->nh.iph->tot_len = htons(nskb->len);
 
 	if (tcph->ack) {
@@ -170,7 +170,7 @@
     	case IPT_ICMP_ECHOREPLY: {
 		struct icmphdr *icmph  = (struct icmphdr *)
 			((u_int32_t *)(*pskb)->nh.iph + (*pskb)->nh.iph->ihl);
-		unsigned int datalen = (*pskb)->len - (*pskb)->nh.iph->ihl * 4;
+		unsigned int datalen = (*pskb)->len - ((*pskb)->nh.iph->ihl << 2);
 
 		/* Not non-head frags, or truncated */
 		if (((ntohs((*pskb)->nh.iph->frag_off) & IP_OFFSET) == 0)
diff -ru linux-2.4-orig/net/ipv4/netfilter/ipt_unclean.c linux-2.4/net/ipv4/netfilter/ipt_unclean.c
--- linux-2.4-orig/net/ipv4/netfilter/ipt_unclean.c	Fri Apr 28 00:43:15 2000
+++ linux-2.4/net/ipv4/netfilter/ipt_unclean.c	Wed Jan 10 17:00:51 2001
@@ -112,7 +112,7 @@
 				limpk("ICMP error internal way too short\n");
 				return 0;
 			}
-			if (datalen - 8 < inner->ihl*4 + 8) {
+			if (datalen - 8 < (inner->ihl << 2) + 8) {
 				limpk("ICMP error internal too short\n");
 				return 0;
 			}
@@ -155,7 +155,7 @@
 		u_int32_t arg = ntohl(icmph->un.gateway);
 
 		if (icmph->code == 0) {
-			if ((arg >> 24) >= iph->ihl*4) {
+			if ((arg >> 24) >= (iph->ihl << 2)) {
 				limpk("ICMP PARAMETERPROB ptr = %u\n",
 				      ntohl(icmph->un.gateway) >> 24);
 				return 0;
@@ -292,7 +292,7 @@
 	}
 
 	/* CHECK: Smaller than actual TCP hdr. */
-	if (datalen < tcph->doff * 4) {
+	if (datalen < (tcph->doff << 2)) {
 		if (!embedded) {
 			limpk("Packet length %u < actual TCP header.\n",
 			      datalen);
@@ -339,7 +339,7 @@
 		return 0;
 	}
 
-	for (i = sizeof(struct tcphdr); i < tcph->doff * 4; ) {
+	for (i = sizeof(struct tcphdr); i < (tcph->doff << 2); ) {
 		switch (opt[i]) {
 		case 0:
 			end_of_options = 1;
@@ -356,7 +356,7 @@
 				return 0;
 			}
 			/* CHECK: options at tail. */
-			else if (i+1 >= tcph->doff * 4) {
+			else if (i+1 >= (tcph->doff << 2)) {
 				limpk("TCP option %u at tail\n",
 				      opt[i]);
 				return 0;
@@ -368,7 +368,7 @@
 				return 0;
 			}
 			/* CHECK: oversize options. */
-			else if (opt[i+1] + i >= tcph->doff * 4) {
+			else if (opt[i+1] + i >= (tcph->doff << 2)) {
 				limpk("TCP option %u at %Zu too long\n",
 				      (unsigned int) opt[i], i);
 				return 0;
@@ -393,14 +393,14 @@
 
 	/* Should only happen for local outgoing raw-socket packets. */
 	/* CHECK: length >= ip header. */
-	if (length < sizeof(struct iphdr) || length < iph->ihl * 4) {
+	if (length < sizeof(struct iphdr) || length < (iph->ihl << 2)) {
 		limpk("Packet length %Zu < IP header.\n", length);
 		return 0;
 	}
 
 	offset = ntohs(iph->frag_off) & IP_OFFSET;
-	protoh = (void *)iph + iph->ihl * 4;
-	datalen = length - iph->ihl * 4;
+	protoh = (void *)iph + (iph->ihl << 2);
+	datalen = length - (iph->ihl << 2);
 
 	/* CHECK: Embedded fragment. */
 	if (embedded && offset) {
@@ -408,7 +408,7 @@
 		return 0;
 	}
 
-	for (i = sizeof(struct iphdr); i < iph->ihl * 4; ) {
+	for (i = sizeof(struct iphdr); i < (iph->ihl << 2); ) {
 		switch (opt[i]) {
 		case 0:
 			end_of_options = 1;
@@ -425,7 +425,7 @@
 				return 0;
 			}
 			/* CHECK: options at tail. */
-			else if (i+1 >= iph->ihl * 4) {
+			else if (i+1 >= (iph->ihl << 2)) {
 				limpk("IP option %u at tail\n",
 				      opt[i]);
 				return 0;
@@ -437,7 +437,7 @@
 				return 0;
 			}
 			/* CHECK: oversize options. */
-			else if (opt[i+1] + i >= iph->ihl * 4) {
+			else if (opt[i+1] + i >= (iph->ihl << 2)) {
 				limpk("IP option %u at %u too long\n",
 				      opt[i], i);
 				return 0;
diff -ru linux-2.4-orig/net/ipv4/netfilter/iptable_filter.c linux-2.4/net/ipv4/netfilter/iptable_filter.c
--- linux-2.4-orig/net/ipv4/netfilter/iptable_filter.c	Fri May 12 20:45:26 2000
+++ linux-2.4/net/ipv4/netfilter/iptable_filter.c	Wed Jan 10 17:01:15 2001
@@ -105,7 +105,7 @@
 {
 	/* root is playing with raw sockets. */
 	if ((*pskb)->len < sizeof(struct iphdr)
-	    || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
+	    || ((*pskb)->nh.iph->ihl << 2) < sizeof(struct iphdr)) {
 		if (net_ratelimit())
 			printk("ipt_hook: happy cracking.\n");
 		return NF_ACCEPT;
diff -ru linux-2.4-orig/net/ipv4/netfilter/iptable_mangle.c linux-2.4/net/ipv4/netfilter/iptable_mangle.c
--- linux-2.4-orig/net/ipv4/netfilter/iptable_mangle.c	Sat Sep 16 06:37:23 2000
+++ linux-2.4/net/ipv4/netfilter/iptable_mangle.c	Wed Jan 10 17:02:25 2001
@@ -134,7 +134,7 @@
 
 	/* root is playing with raw sockets. */
 	if ((*pskb)->len < sizeof(struct iphdr)
-	    || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
+	    || ((*pskb)->nh.iph->ihl << 2) < sizeof(struct iphdr)) {
 		if (net_ratelimit())
 			printk("ipt_hook: happy cracking.\n");
 		return NF_ACCEPT;
diff -ru linux-2.4-orig/net/ipv4/route.c linux-2.4/net/ipv4/route.c
--- linux-2.4-orig/net/ipv4/route.c	Tue Oct 10 19:33:52 2000
+++ linux-2.4/net/ipv4/route.c	Wed Jan 10 17:19:36 2001
@@ -2247,7 +2247,7 @@
 #ifdef CONFIG_SMP
 		if (smp_num_cpus > 1 || cpu_logical_map(0) != 0) {
 			int i;
-			int cnt = length/4;
+			int cnt = length >> 2;
 
 			for (i=0; i<smp_num_cpus; i++) {
 				int cpu = cpu_logical_map(i);
diff -ru linux-2.4-orig/net/ipv4/tcp.c linux-2.4/net/ipv4/tcp.c
--- linux-2.4-orig/net/ipv4/tcp.c	Wed Nov 29 06:53:45 2000
+++ linux-2.4/net/ipv4/tcp.c	Wed Jan 10 16:43:30 2001
@@ -1534,7 +1534,7 @@
 
 		err = 0;
 		if (!(flags&MSG_TRUNC)) {
-			err = memcpy_toiovec(msg->msg_iov, ((unsigned char *)skb->h.th) + skb->h.th->doff*4 + offset, used);
+			err = memcpy_toiovec(msg->msg_iov, ((unsigned char *)skb->h.th) + (skb->h.th->doff << 2) + offset, used);
 			if (err) {
 				/* Exception. Bailout! */
 				if (!copied)
diff -ru linux-2.4-orig/net/ipv4/tcp_input.c linux-2.4/net/ipv4/tcp_input.c
--- linux-2.4-orig/net/ipv4/tcp_input.c	Fri Dec 29 23:07:24 2000
+++ linux-2.4/net/ipv4/tcp_input.c	Wed Jan 10 16:44:40 2001
@@ -1988,7 +1988,7 @@
 {
 	unsigned char *ptr;
 	struct tcphdr *th = skb->h.th;
-	int length=(th->doff*4)-sizeof(struct tcphdr);
+	int length=(th->doff << 2)-sizeof(struct tcphdr);
 
 	ptr = (unsigned char *)(th + 1);
 	tp->saw_tstamp = 0;
@@ -2829,8 +2829,8 @@
 	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
 
 	th = skb->h.th;
-	skb_pull(skb, th->doff*4);
-	skb_trim(skb, len - (th->doff*4));
+	skb_pull(skb, th->doff << 2);
+	skb_trim(skb, len - (th->doff << 2));
 
         if (skb->len == 0 && !th->fin)
 		goto drop;
@@ -3052,7 +3052,7 @@
 
 	/* Do we wait for any urgent data? - normally not... */
 	if (tp->urg_data == TCP_URG_NOTYET) {
-		u32 ptr = tp->urg_seq - ntohl(th->seq) + (th->doff*4);
+		u32 ptr = tp->urg_seq - ntohl(th->seq) + (th->doff << 2);
 
 		/* Is the urgent pointer pointing into this packet? */	 
 		if (ptr < len) {
diff -ru linux-2.4-orig/net/ipv4/tcp_ipv4.c linux-2.4/net/ipv4/tcp_ipv4.c
--- linux-2.4-orig/net/ipv4/tcp_ipv4.c	Mon Jan  1 20:01:58 2001
+++ linux-2.4/net/ipv4/tcp_ipv4.c	Wed Jan 10 17:20:09 2001
@@ -1117,7 +1117,7 @@
 	/* Swap the send and the receive. */
 	rep.th.dest = th->source;
 	rep.th.source = th->dest; 
-	rep.th.doff = arg.iov[0].iov_len/4;
+	rep.th.doff = arg.iov[0].iov_len >> 2;
 	rep.th.seq = htonl(seq);
 	rep.th.ack_seq = htonl(ack);
 	rep.th.ack = 1;
@@ -1629,7 +1629,7 @@
 
 	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
 	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
-				    len - th->doff*4);
+				    len - (th->doff << 2));
 	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
 	TCP_SKB_CB(skb)->when = 0;
 	TCP_SKB_CB(skb)->flags = skb->nh.iph->tos;
diff -ru linux-2.4-orig/net/ipv6/tcp_ipv6.c linux-2.4/net/ipv6/tcp_ipv6.c
--- linux-2.4-orig/net/ipv6/tcp_ipv6.c	Mon Jan  1 20:01:58 2001
+++ linux-2.4/net/ipv6/tcp_ipv6.c	Wed Jan 10 17:35:19 2001
@@ -1558,7 +1558,7 @@
 
 	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
 	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
-				    len - th->doff*4);
+				    len - (th->doff << 2));
 	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
 	TCP_SKB_CB(skb)->when = 0;
 	TCP_SKB_CB(skb)->flags = ip6_get_dsfield(skb->nh.ipv6h);

Re: * 4 converted to << 2 for networking code

[David S. Miller]

   Date: 	Wed, 10 Jan 2001 17:48:59 +0100
   From: antirez <antirez@invece.org>

   The attached patch converts many occurences of '* 4' in the
   networking code (often used to convert in bytes the TCP data offset
   and the IP header len) to the faster '<< 2'.

The compiler does this for you, check the assembler it outputs.

Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Brian Gerst]

antirez wrote:
> 
> Hi all,
> 
> The attached patch converts many occurences of '* 4' in the networking code
> (often used to convert in bytes the TCP data offset and the IP header len)
> to the faster '<< 2'. Since this was a quite repetitive work it's better
> if someone double-check it before to apply the patch.
> The patch is for linux-2.4.

This patch isn't really necessary, because GCC will automatically
convert multiplications and divisions by powers of two to use shifts.

--

				Brian Gerst
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[antirez]

On Wed, Jan 10, 2001 at 09:54:04AM -0500, Brian Gerst wrote:
> This patch isn't really necessary, because GCC will automatically
> convert multiplications and divisions by powers of two to use shifts.

Sure, but since many << 2 already exists in the net kernel code
I feel it's better to use just a format, and it seems more appropriate
to write << 2, just to reflect what we want.
Also some piece of kernel code may be used with compilers that does not
optimize power of two.

antirez
(CC me)

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Jakob Østergaard]

On Wed, Jan 10, 2001 at 06:03:22PM +0100, antirez wrote:
> On Wed, Jan 10, 2001 at 09:54:04AM -0500, Brian Gerst wrote:
> > This patch isn't really necessary, because GCC will automatically
> > convert multiplications and divisions by powers of two to use shifts.
> 
> Sure, but since many << 2 already exists in the net kernel code
> I feel it's better to use just a format, and it seems more appropriate
> to write << 2, just to reflect what we want.
> Also some piece of kernel code may be used with compilers that does not
> optimize power of two.

On most processors <<2 is slower than *4.   It's outright stupid to 
write <<2 when we mean *4 in order to optimize for one out of a
gazillion supported architectures - even more so when the compiler
for the one CPU where <<2 is faster, will actually generate a shift
instead of a multiply as a part of the standard optimization.

One question for the GCC people:  Will gcc change <<2 to *4 on other 
architectures ?    If so, then my case is not quite as strong of course.

-- 
................................................................
:   jakob@unthought.net   : And I see the elder races,         :
:.........................: putrid forms of man                :
:   Jakob Østergaard      : See him rise and claim the earth,  :
:        OZ9ABN           : his downfall is at hand.           :
:.........................:............{Konkhra}...............:
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Mike Harrold]

> 
> On Wed, Jan 10, 2001 at 06:03:22PM +0100, antirez wrote:
> > On Wed, Jan 10, 2001 at 09:54:04AM -0500, Brian Gerst wrote:
> > > This patch isn't really necessary, because GCC will automatically
> > > convert multiplications and divisions by powers of two to use shifts.
> > 
> > Sure, but since many << 2 already exists in the net kernel code
> > I feel it's better to use just a format, and it seems more appropriate
> > to write << 2, just to reflect what we want.
> > Also some piece of kernel code may be used with compilers that does not
> > optimize power of two.
> 
> On most processors <<2 is slower than *4.   It's outright stupid to 
> write <<2 when we mean *4 in order to optimize for one out of a
> gazillion supported architectures - even more so when the compiler
> for the one CPU where <<2 is faster, will actually generate a shift
> instead of a multiply as a part of the standard optimization.
> 
> One question for the GCC people:  Will gcc change <<2 to *4 on other 
> architectures ?    If so, then my case is not quite as strong of course.

Be careful. *4 is not a simple <<2 substitution (by the compiler) if
the variable is signed. *4 translates to 3 instructions (on x86) if
it's an int.

My feeling is that it shouldn't matter if you use <<2 or *4 even if the
compiler optimises - one would hope that the compiler would optimise to
the fastest in both directions.

Regards,

/Mike
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Chris Jones]

Mike Harrold <mharrold@cas.org> writes:

[...]

  My feeling is that it shouldn't matter if you use <<2 or *4 even if the
  compiler optimises - one would hope that the compiler would optimise to
  the fastest in both directions.

I agree this should be left to the compiler.  The programmer should write *4
when multiplying by 4 and <<2 when shifting left by 2.  In the case that
sparked this (converting counts of 32 bit units to counts of octets),
multiplication is the proper conversion operation.

(If performance is truly critical AND profiling shows that writing <<2 instead
of *4 makes a significant difference, doing the shift might be called for.  I
hardly believe that's the case here, and the fact that the code has to run on
several architectures and be compiled by various compilers makes it less likely
that this change is a clearcut win.)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Jamie Lokier]

Mike Harrold wrote:
> Be careful. *4 is not a simple <<2 substitution (by the compiler) if
> the variable is signed. *4 translates to 3 instructions (on x86) if
> it's an int.

I think you mean /4 is not the same as >>2 if the variable is signed.

In general, non-widening multiplies give the same result for signed and
unsigned variables.

-- Jamie
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Mike Harrold]

> 
> Mike Harrold wrote:
> > Be careful. *4 is not a simple <<2 substitution (by the compiler) if
> > the variable is signed. *4 translates to 3 instructions (on x86) if
> > it's an int.
> 
> I think you mean /4 is not the same as >>2 if the variable is signed.
> 
> In general, non-widening multiplies give the same result for signed and
> unsigned variables.
> 

I believe you're right. Teach me to post before morning coffee :)

/Mike

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[antirez]

On Wed, Jan 10, 2001 at 04:11:46PM +0100, Jakob ?stergaard wrote:
> On most processors <<2 is slower than *4.   It's outright stupid to 
> write <<2 when we mean *4 in order to optimize for one out of a
> gazillion supported architectures - even more so when the compiler
> for the one CPU where <<2 is faster, will actually generate a shift
> instead of a multiply as a part of the standard optimization.

Hug, ok, so all the << 2 already in should be changed in *4.
My point is that it is better to use only << 2 or *4, selecting
the better form.

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Pauline Middelink]

On Wed, 10 Jan 2001 around 18:25:46 +0100, antirez wrote:
> On Wed, Jan 10, 2001 at 04:11:46PM +0100, Jakob ?stergaard wrote:
> > On most processors <<2 is slower than *4.   It's outright stupid to 
> > write <<2 when we mean *4 in order to optimize for one out of a
> > gazillion supported architectures - even more so when the compiler
> > for the one CPU where <<2 is faster, will actually generate a shift
> > instead of a multiply as a part of the standard optimization.
> 
> Hug, ok, so all the << 2 already in should be changed in *4.
> My point is that it is better to use only << 2 or *4, selecting
> the better form.

Well, better not change things so they look like:

#define	MSG_BLA1	(1<<1)
#define	MSG_BLA2	(1*4)
#define	MSG_BLA3	(1<<3)

as a result... :)

    Met vriendelijke groet,
        Pauline Middelink
-- 
GPG Key fingerprint = 2D5B 87A7 DDA6 0378 5DEA  BD3B 9A50 B416 E2D0 C3C2
For more details look at my website http://www.polyware.nl/~middelink
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[H. Peter Anvin]

Followup to:  <20010110161146.A3252@unthought.net>
By author:    =?iso-8859-1?Q?Jakob_=D8stergaard?= <jakob@unthought.net>
In newsgroup: linux.dev.kernel
> 
> On most processors <<2 is slower than *4.
> 

That's a funny statement.  Which processors do you include in "most"?
That has not been my experience.

> It's outright stupid to write <<2 when we mean *4 in order to optimize for one out of a
> gazillion supported architectures - even more so when the compiler
> for the one CPU where <<2 is faster, will actually generate a shift
> instead of a multiply as a part of the standard optimization.
> 
> One question for the GCC people:  Will gcc change <<2 to *4 on other 
> architectures ?    If so, then my case is not quite as strong of course.
> 

gcc should consider the statements equivalent, and generate whichever
pattern is preferred.  On an i386 that may mean take a pattern such as

	foo = (bar << 2) + quux;

... and generate ...

	lea ecx,[esi*4+ebx]

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: * 4 converted to << 2 for networking code

[Matthias Andree]

On Wed, 10 Jan 2001, antirez wrote:

> Hi all,
> 
> The attached patch converts many occurences of '* 4' in the networking code
> (often used to convert in bytes the TCP data offset and the IP header len)
> to the faster '<< 2'. Since this was a quite repetitive work it's better
> if someone double-check it before to apply the patch.
> The patch is for linux-2.4.
> 
> Please, CC: me for replies since I'm not subscribed to the list.

You'd better be. As Alan Cox pointed out some time ago, this is the
compiler's duty, to be more precise, it's the peephole (local)
optimizer's task, if it's actually implemented in the core or the
backend, I cannot currently tell, and I don't care.

But see a minimal program 
main(int argc) {
        j(argc * 4);
}

run gcc -O2 -S thissource.c and then -O0 and compare. On an
amigaos-m68k configured gcc 2.7.2.3, this directly throws asll #2,d0
(arithmetic shift left by 2 positions, fill with trailing zeroes), -O2
saves some moving around. (I cannot read 386 assembly language, thus
m68k ;)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/