Merge branch 'pktgen-imix'

Nick Richardson says:

====================
pktgen: Add IMIX mode

Adds internet mix (IMIX) mode to pktgen. Internet mix is
included in many user-space network perf testing tools. It allows
for the user to specify a distribution of discrete packet sizes to be
generated. This type of test is common among vendors when perf testing
their devices.
[RFC link: https://datatracker.ietf.org/doc/html/rfc2544#section-9.1]

This allows users to get a
more complete picture of how their device will perform in the
real-world.

This feature adds a command that allows users to specify an imix
distribution in the following format:
  imix_weights size_1,weight_1 size_2,weight_2 ... size_n,weight_n

The distribution of packets with size_i will be
(weight_i / total_weights) where
total_weights = weight_1 + weight_2 + ... + weight_n

For example:
  imix_weights 40,7 576,4 1500,1

The pkt_size "40" will account for 7 / (7 + 4 + 1) = ~58% of the total
packets sent.

This patch was tested with the following:
1. imix_weights = 40,7 576,4 1500,1
2. imix_weights = 0,7 576,4 1500,1
  - Packet size of 0 is resized to the minimum, 42
3. imix_weights = 40,7 576,4 1500,1 count = 0
  - Zero count.
  - Runs until user stops pktgen.
Invalid Configurations
1. clone_skb = 200 imix_weights = 40,7 576,4 1500,1
    - Returns error code -524 (-ENOTSUPP) when setting imix_weights
2. len(imix_weights) > MAX_IMIX_ENTRIES
    - Returns -7 (-E2BIG)

This patch is split into three parts, each provide different aspects of
required functionality:
  1. Parse internet mix input.
  2. Add IMIX Distribution representation.
  3. Process and output IMIX results.

Changes in v2:
* Remove __ prefix outside of uAPI.
* Use seq_puts instead of seq_printf where necessary.
* Reorder variable declaration.
* Return -EINVAL instead of -ENOTSUPP when using IMIX with clone_skb > 0
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

net/core/pktgen.c

@@ -175,6 +175,9 @@
 #define IP_NAME_SZ 32
 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
 #define MPLS_STACK_BOTTOM htonl(0x00000100)
+/* Max number of internet mix entries that can be specified in imix_weights. */
+#define MAX_IMIX_ENTRIES 20
+#define IMIX_PRECISION 100 /* Precision of IMIX distribution */
 
 #define func_enter() pr_debug("entering %s\n", __func__);
 
@@ -242,6 +245,12 @@ static char *pkt_flag_names[] = {
 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
 
+struct imix_pkt {
+	u64 size;
+	u64 weight;
+	u64 count_so_far;
+};
+
 struct flow_state {
 	__be32 cur_daddr;
 	int count;
@@ -343,6 +352,12 @@ struct pktgen_dev {
 	__u8 traffic_class;  /* ditto for the (former) Traffic Class in IPv6
 				(see RFC 3260, sec. 4) */
 
+	/* IMIX */
+	unsigned int n_imix_entries;
+	struct imix_pkt imix_entries[MAX_IMIX_ENTRIES];
+	/* Maps 0-IMIX_PRECISION range to imix_entry based on probability*/
+	__u8 imix_distribution[IMIX_PRECISION];
+
 	/* MPLS */
 	unsigned int nr_labels;	/* Depth of stack, 0 = no MPLS */
 	__be32 labels[MAX_MPLS_LABELS];
@@ -471,6 +486,7 @@ static void pktgen_stop_all_threads(struct pktgen_net *pn);
 
 static void pktgen_stop(struct pktgen_thread *t);
 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
+static void fill_imix_distribution(struct pktgen_dev *pkt_dev);
 
 /* Module parameters, defaults. */
 static int pg_count_d __read_mostly = 1000;
@@ -552,6 +568,16 @@ static int pktgen_if_show(struct seq_file *seq, void *v)
 		   (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
 		   pkt_dev->max_pkt_size);
 
+	if (pkt_dev->n_imix_entries > 0) {
+		seq_puts(seq, "     imix_weights: ");
+		for (i = 0; i < pkt_dev->n_imix_entries; i++) {
+			seq_printf(seq, "%llu,%llu ",
+				   pkt_dev->imix_entries[i].size,
+				   pkt_dev->imix_entries[i].weight);
+		}
+		seq_puts(seq, "\n");
+	}
+
 	seq_printf(seq,
 		   "     frags: %d  delay: %llu  clone_skb: %d  ifname: %s\n",
 		   pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
@@ -669,6 +695,18 @@ static int pktgen_if_show(struct seq_file *seq, void *v)
 		   (unsigned long long)pkt_dev->sofar,
 		   (unsigned long long)pkt_dev->errors);
 
+	if (pkt_dev->n_imix_entries > 0) {
+		int i;
+
+		seq_puts(seq, "     imix_size_counts: ");
+		for (i = 0; i < pkt_dev->n_imix_entries; i++) {
+			seq_printf(seq, "%llu,%llu ",
+				   pkt_dev->imix_entries[i].size,
+				   pkt_dev->imix_entries[i].count_so_far);
+		}
+		seq_puts(seq, "\n");
+	}
+
 	seq_printf(seq,
 		   "     started: %lluus  stopped: %lluus idle: %lluus\n",
 		   (unsigned long long) ktime_to_us(pkt_dev->started_at),
@@ -792,6 +830,62 @@ static int strn_len(const char __user * user_buffer, unsigned int maxlen)
 	return i;
 }
 
+/* Parses imix entries from user buffer.
+ * The user buffer should consist of imix entries separated by spaces
+ * where each entry consists of size and weight delimited by commas.
+ * "size1,weight_1 size2,weight_2 ... size_n,weight_n" for example.
+ */
+static ssize_t get_imix_entries(const char __user *buffer,
+				struct pktgen_dev *pkt_dev)
+{
+	const int max_digits = 10;
+	int i = 0;
+	long len;
+	char c;
+
+	pkt_dev->n_imix_entries = 0;
+
+	do {
+		unsigned long weight;
+		unsigned long size;
+
+		len = num_arg(&buffer[i], max_digits, &size);
+		if (len < 0)
+			return len;
+		i += len;
+		if (get_user(c, &buffer[i]))
+			return -EFAULT;
+		/* Check for comma between size_i and weight_i */
+		if (c != ',')
+			return -EINVAL;
+		i++;
+
+		if (size < 14 + 20 + 8)
+			size = 14 + 20 + 8;
+
+		len = num_arg(&buffer[i], max_digits, &weight);
+		if (len < 0)
+			return len;
+		if (weight <= 0)
+			return -EINVAL;
+
+		pkt_dev->imix_entries[pkt_dev->n_imix_entries].size = size;
+		pkt_dev->imix_entries[pkt_dev->n_imix_entries].weight = weight;
+
+		i += len;
+		if (get_user(c, &buffer[i]))
+			return -EFAULT;
+
+		i++;
+		pkt_dev->n_imix_entries++;
+
+		if (pkt_dev->n_imix_entries > MAX_IMIX_ENTRIES)
+			return -E2BIG;
+	} while (c == ' ');
+
+	return i;
+}
+
 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
 {
 	unsigned int n = 0;
@@ -960,6 +1054,20 @@ static ssize_t pktgen_if_write(struct file *file,
 		return count;
 	}
 
+	if (!strcmp(name, "imix_weights")) {
+		if (pkt_dev->clone_skb > 0)
+			return -EINVAL;
+
+		len = get_imix_entries(&user_buffer[i], pkt_dev);
+		if (len < 0)
+			return len;
+
+		fill_imix_distribution(pkt_dev);
+
+		i += len;
+		return count;
+	}
+
 	if (!strcmp(name, "debug")) {
 		len = num_arg(&user_buffer[i], 10, &value);
 		if (len < 0)
@@ -1082,10 +1190,16 @@ static ssize_t pktgen_if_write(struct file *file,
 		len = num_arg(&user_buffer[i], 10, &value);
 		if (len < 0)
 			return len;
+		/* clone_skb is not supported for netif_receive xmit_mode and
+		 * IMIX mode.
+		 */
 		if ((value > 0) &&
 		    ((pkt_dev->xmit_mode == M_NETIF_RECEIVE) ||
 		     !(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
 			return -ENOTSUPP;
+		if (value > 0 && pkt_dev->n_imix_entries > 0)
+			return -EINVAL;
+
 		i += len;
 		pkt_dev->clone_skb = value;
 
@@ -2472,6 +2586,14 @@ static void mod_cur_headers(struct pktgen_dev *pkt_dev)
 				t = pkt_dev->min_pkt_size;
 		}
 		pkt_dev->cur_pkt_size = t;
+	} else if (pkt_dev->n_imix_entries > 0) {
+		struct imix_pkt *entry;
+		__u32 t = prandom_u32() % IMIX_PRECISION;
+		__u8 entry_index = pkt_dev->imix_distribution[t];
+
+		entry = &pkt_dev->imix_entries[entry_index];
+		entry->count_so_far++;
+		pkt_dev->cur_pkt_size = entry->size;
 	}
 
 	set_cur_queue_map(pkt_dev);
@@ -2540,6 +2662,33 @@ static void free_SAs(struct pktgen_dev *pkt_dev)
 	}
 }
 
+static void fill_imix_distribution(struct pktgen_dev *pkt_dev)
+{
+	int cumulative_probabilites[MAX_IMIX_ENTRIES];
+	int j = 0;
+	__u64 cumulative_prob = 0;
+	__u64 total_weight = 0;
+	int i = 0;
+
+	for (i = 0; i < pkt_dev->n_imix_entries; i++)
+		total_weight += pkt_dev->imix_entries[i].weight;
+
+	/* Fill cumulative_probabilites with sum of normalized probabilities */
+	for (i = 0; i < pkt_dev->n_imix_entries - 1; i++) {
+		cumulative_prob += div64_u64(pkt_dev->imix_entries[i].weight *
+						     IMIX_PRECISION,
+					     total_weight);
+		cumulative_probabilites[i] = cumulative_prob;
+	}
+	cumulative_probabilites[pkt_dev->n_imix_entries - 1] = 100;
+
+	for (i = 0; i < IMIX_PRECISION; i++) {
+		if (i == cumulative_probabilites[j])
+			j++;
+		pkt_dev->imix_distribution[i] = j;
+	}
+}
+
 static int process_ipsec(struct pktgen_dev *pkt_dev,
 			      struct sk_buff *skb, __be16 protocol)
 {
@@ -3140,7 +3289,19 @@ static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
 	pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
 			ktime_to_ns(elapsed));
 
-	bps = pps * 8 * pkt_dev->cur_pkt_size;
+	if (pkt_dev->n_imix_entries > 0) {
+		int i;
+		struct imix_pkt *entry;
+
+		bps = 0;
+		for (i = 0; i < pkt_dev->n_imix_entries; i++) {
+			entry = &pkt_dev->imix_entries[i];
+			bps += entry->size * entry->count_so_far;
+		}
+		bps = div64_u64(bps * 8 * NSEC_PER_SEC, ktime_to_ns(elapsed));
+	} else {
+		bps = pps * 8 * pkt_dev->cur_pkt_size;
+	}
 
 	mbps = bps;
 	do_div(mbps, 1000000);