[PATCH] iproute2: tc add mqprio qdisc support

[PATCH] iproute2: tc add mqprio qdisc support

[John Fastabend]

Add mqprio qdisc support. Output matches the following,

# ./tc/tc qdisc
qdisc mq 0: dev eth1 root
qdisc mq 0: dev eth2 root
qdisc mqprio 8001: dev eth3 root  tc 8 map 0 1 2 3 4 5 6 7 1 1 1 1 1 1 1 1
             queues:(0:7) (8:15) (16:23) (24:31) (32:39) (40:47) (48:55) (56:63)

And usage is,

# ./tc/tc qdisc add dev eth3 root mqprio help
Usage: ... mclass [num_tc NUMBER] [map P0 P1...]
                  [offset txq0 txq1 ...] [count cnt0 cnt1 ...] [hw 1|0]

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
---

 include/linux/pkt_sched.h |   12 ++++
 tc/Makefile               |    1 
 tc/q_mqprio.c             |  125 +++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 138 insertions(+), 0 deletions(-)
 create mode 100644 tc/q_mqprio.c

diff --git a/include/linux/pkt_sched.h b/include/linux/pkt_sched.h
index 2cfa4bc..776cd93 100644
--- a/include/linux/pkt_sched.h
+++ b/include/linux/pkt_sched.h
@@ -481,4 +481,16 @@ struct tc_drr_stats {
 	__u32	deficit;
 };
 
+/* MQPRIO */
+#define TC_QOPT_BITMASK 15
+#define TC_QOPT_MAX_QUEUE 16
+
+struct tc_mqprio_qopt {
+	__u8	num_tc;
+	__u8	prio_tc_map[TC_QOPT_BITMASK + 1];
+	__u8	hw;
+	__u16	count[TC_QOPT_MAX_QUEUE];
+	__u16	offset[TC_QOPT_MAX_QUEUE];
+};
+
 #endif
diff --git a/tc/Makefile b/tc/Makefile
index 101cc83..df372c6 100644
--- a/tc/Makefile
+++ b/tc/Makefile
@@ -43,6 +43,7 @@ TCMODULES += em_nbyte.o
 TCMODULES += em_cmp.o
 TCMODULES += em_u32.o
 TCMODULES += em_meta.o
+TCMODULES += q_mqprio.o
 
 TCSO :=
 ifeq ($(TC_CONFIG_ATM),y)
diff --git a/tc/q_mqprio.c b/tc/q_mqprio.c
new file mode 100644
index 0000000..8b2c006
--- /dev/null
+++ b/tc/q_mqprio.c
@@ -0,0 +1,125 @@
+/*
+ * q_mqprio.c	MQ prio qdisc
+ *
+ *		This program is free software; you can redistribute it and/or
+ *		modify it under the terms of the GNU General Public License
+ *		as published by the Free Software Foundation; either version
+ *		2 of the License, or (at your option) any later version.
+ *
+ * Author:	John Fastabend, <john.r.fastabend@intel.com>
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <syslog.h>
+#include <fcntl.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <string.h>
+
+#include "utils.h"
+#include "tc_util.h"
+
+static void explain(void)
+{
+	fprintf(stderr, "Usage: ... mqprio [num_tc NUMBER] [map P0 P1 ...]\n");
+	fprintf(stderr, "                  [offset txq0 txq1 ...] ");
+	fprintf(stderr, "[count cnt0,cnt1 ...] [hw 1|0]\n");
+}
+
+static int mqprio_parse_opt(struct qdisc_util *qu, int argc,
+			    char **argv, struct nlmsghdr *n)
+{
+	int idx;
+	struct tc_mqprio_qopt opt = {
+				     8,
+				     {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 1, 1, 3, 3, 3, 3},
+				     1,
+				    };
+
+	while (argc > 0) {
+		idx = 0;
+		if (strcmp(*argv, "num_tc") == 0) {
+			NEXT_ARG();
+			if (get_u8(&opt.num_tc, *argv, 10)) {
+				fprintf(stderr, "Illegal \"num_tc\"\n");
+				return -1;
+			}
+		} else if (strcmp(*argv, "map") == 0) {
+			while (idx < TC_QOPT_MAX_QUEUE && NEXT_ARG_OK()) {
+				NEXT_ARG();
+				if (get_u8(&opt.prio_tc_map[idx], *argv, 10)) {
+					PREV_ARG();
+					break;
+				}
+				idx++;
+			}
+			for ( ; idx < TC_QOPT_MAX_QUEUE; idx++)
+				opt.prio_tc_map[idx] = 0;
+		} else if (strcmp(*argv, "offset") == 0) {
+			while (idx < TC_QOPT_MAX_QUEUE && NEXT_ARG_OK()) {
+				NEXT_ARG();
+				if (get_u16(&opt.offset[idx], *argv, 10)) {
+					PREV_ARG();
+					break;
+				}
+				idx++;
+			}
+		} else if (strcmp(*argv, "count") == 0) {
+			while (idx < TC_QOPT_MAX_QUEUE && NEXT_ARG_OK()) {
+				NEXT_ARG();
+				if (get_u16(&opt.count[idx], *argv, 10)) {
+					PREV_ARG();
+					break;
+				}
+				idx++;
+			}
+		} else if (strcmp(*argv, "hw") == 0) {
+			NEXT_ARG();
+			if (get_u8(&opt.hw, *argv, 10)) {
+				fprintf(stderr, "Illegal \"hw\"\n");
+				return -1;
+			}
+			idx++;
+		} else if (strcmp(*argv, "help") == 0) {
+			explain();
+			return -1;
+		} else {
+			fprintf(stderr, "Unknown argument\n");
+			return -1;
+		}
+		argc--; argv++;
+	}
+
+	addattr_l(n, 1024, TCA_OPTIONS, &opt, sizeof(opt));
+	return 0;
+}
+
+int mqprio_print_opt(struct qdisc_util *qu, FILE *f, struct rtattr *opt)
+{
+	int i;
+	struct tc_mqprio_qopt *qopt;
+
+	if (opt == NULL)
+		return 0;
+
+	qopt = RTA_DATA(opt);
+
+	fprintf(f, " tc %u map ", qopt->num_tc);
+	for (i = 0; i <= TC_PRIO_MAX; i++)
+		fprintf(f, "%d ", qopt->prio_tc_map[i]);
+	fprintf(f, "\n             queues:");
+	for (i = 0; i < qopt->num_tc; i++)
+		fprintf(f, "(%i:%i) ", qopt->offset[i],
+			qopt->offset[i] + qopt->count[i] - 1);
+	return 0;
+}
+
+struct qdisc_util mqprio_qdisc_util = {
+	.id		= "mqprio",
+	.parse_qopt	= mqprio_parse_opt,
+	.print_qopt	= mqprio_print_opt,
+};
+

Re: [PATCH] iproute2: tc add mqprio qdisc support

[Stephen Hemminger]

On Tue, 12 Apr 2011 08:57:27 -0700
John Fastabend <john.r.fastabend@intel.com> wrote:

> Add mqprio qdisc support. Output matches the following,
> 
> # ./tc/tc qdisc
> qdisc mq 0: dev eth1 root
> qdisc mq 0: dev eth2 root
> qdisc mqprio 8001: dev eth3 root  tc 8 map 0 1 2 3 4 5 6 7 1 1 1 1 1 1 1 1
>              queues:(0:7) (8:15) (16:23) (24:31) (32:39) (40:47) (48:55) (56:63)
> 
> And usage is,
> 
> # ./tc/tc qdisc add dev eth3 root mqprio help
> Usage: ... mclass [num_tc NUMBER] [map P0 P1...]
>                   [offset txq0 txq1 ...] [count cnt0 cnt1 ...] [hw 1|0]
> 
> Signed-off-by: John Fastabend <john.r.fastabend@intel.com>

Applied to net-next branch.



-- 

Re: [PATCH] iproute2: tc add mqprio qdisc support

[Ben Hutchings]

I know that this has already been applied, but:

On Tue, 2011-04-12 at 08:57 -0700, John Fastabend wrote:
> Add mqprio qdisc support. Output matches the following,
> 
> # ./tc/tc qdisc
> qdisc mq 0: dev eth1 root
> qdisc mq 0: dev eth2 root
> qdisc mqprio 8001: dev eth3 root  tc 8 map 0 1 2 3 4 5 6 7 1 1 1 1 1 1 1 1
>              queues:(0:7) (8:15) (16:23) (24:31) (32:39) (40:47) (48:55) (56:63)
> 
> And usage is,
> 
> # ./tc/tc qdisc add dev eth3 root mqprio help
> Usage: ... mclass [num_tc NUMBER] [map P0 P1...]

mclass?

>                   [offset txq0 txq1 ...] [count cnt0 cnt1 ...] [hw 1|0]

Of course I wrote something similar to this, but I never finished it
off, so thanks.

I don't think it makes sense to require count and offset to be specified
as separate lists.  The arguments could be interleaved but that adds
more opportunity for error.  Since offsets have to be in order and you
generally don't want to have gaps then the offsets could normally be
inferred.  So maybe something like:

	queues cnt0[@txq0] cnt1[@txq1] ...

[...]
> +static int mqprio_parse_opt(struct qdisc_util *qu, int argc,
> +			    char **argv, struct nlmsghdr *n)
> +{
> +	int idx;
> +	struct tc_mqprio_qopt opt = {
> +				     8,
> +				     {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 1, 1, 3, 3, 3, 3},
> +				     1,
> +				    };

It would be clearer to name the fields being initialised.

[...]
> +int mqprio_print_opt(struct qdisc_util *qu, FILE *f, struct rtattr *opt)
> +{
> +	int i;
> +	struct tc_mqprio_qopt *qopt;
> +
> +	if (opt == NULL)
> +		return 0;
> +
> +	qopt = RTA_DATA(opt);
> +
> +	fprintf(f, " tc %u map ", qopt->num_tc);
> +	for (i = 0; i <= TC_PRIO_MAX; i++)
> +		fprintf(f, "%d ", qopt->prio_tc_map[i]);
> +	fprintf(f, "\n             queues:");
> +	for (i = 0; i < qopt->num_tc; i++)
> +		fprintf(f, "(%i:%i) ", qopt->offset[i],
> +			qopt->offset[i] + qopt->count[i] - 1);
[...]

Shouldn't this output be consistent with the command-line syntax?

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH] iproute2: tc add mqprio qdisc support

[John Fastabend]

On 4/13/2011 4:10 PM, Ben Hutchings wrote:
> I know that this has already been applied, but:
> 
> On Tue, 2011-04-12 at 08:57 -0700, John Fastabend wrote:
>> Add mqprio qdisc support. Output matches the following,
>>
>> # ./tc/tc qdisc
>> qdisc mq 0: dev eth1 root
>> qdisc mq 0: dev eth2 root
>> qdisc mqprio 8001: dev eth3 root  tc 8 map 0 1 2 3 4 5 6 7 1 1 1 1 1 1 1 1
>>              queues:(0:7) (8:15) (16:23) (24:31) (32:39) (40:47) (48:55) (56:63)
>>
>> And usage is,
>>
>> # ./tc/tc qdisc add dev eth3 root mqprio help
>> Usage: ... mclass [num_tc NUMBER] [map P0 P1...]
> 
> mclass?

agh stupid typo in the description that was my working name
for the qdisc some time ago. The help in 'tc' is correct.

> 
>>                   [offset txq0 txq1 ...] [count cnt0 cnt1 ...] [hw 1|0]
> 
> Of course I wrote something similar to this, but I never finished it
> off, so thanks.
> 
> I don't think it makes sense to require count and offset to be specified
> as separate lists.  The arguments could be interleaved but that adds
> more opportunity for error.  Since offsets have to be in order and you
> generally don't want to have gaps then the offsets could normally be
> inferred.  So maybe something like:
> 
> 	queues cnt0[@txq0] cnt1[@txq1] ...

OK. I agree with you this is better.

> 
> [...]
>> +static int mqprio_parse_opt(struct qdisc_util *qu, int argc,
>> +			    char **argv, struct nlmsghdr *n)
>> +{
>> +	int idx;
>> +	struct tc_mqprio_qopt opt = {
>> +				     8,
>> +				     {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 1, 1, 3, 3, 3, 3},
>> +				     1,
>> +				    };
> 
> It would be clearer to name the fields being initialised.
> 

OK.

> [...]
>> +int mqprio_print_opt(struct qdisc_util *qu, FILE *f, struct rtattr *opt)
>> +{
>> +	int i;
>> +	struct tc_mqprio_qopt *qopt;
>> +
>> +	if (opt == NULL)
>> +		return 0;
>> +
>> +	qopt = RTA_DATA(opt);
>> +
>> +	fprintf(f, " tc %u map ", qopt->num_tc);
>> +	for (i = 0; i <= TC_PRIO_MAX; i++)
>> +		fprintf(f, "%d ", qopt->prio_tc_map[i]);
>> +	fprintf(f, "\n             queues:");
>> +	for (i = 0; i < qopt->num_tc; i++)
>> +		fprintf(f, "(%i:%i) ", qopt->offset[i],
>> +			qopt->offset[i] + qopt->count[i] - 1);
> [...]
> 
> Shouldn't this output be consistent with the command-line syntax?

I'm not sure, here's what it is now,

queues:(0:7) (8:15) (16:23) (24:31) (32:39) (40:47) (48:55) (56:63)

And here's what it would be with the change,

queues: 8@0 8@8 8@16 8@24 8@32 8@40 8@48 8@56

I like the first option with (#:#) it seems a bit more obvious to me
what the layout is. I'll get this fixed up tomorrow. Thanks for
taking a look Ben.

~John.