Skip to contentSkip to conversation

Add a tool for configuration of DCB

PATCH Superseded Petr Machata 4 participants 27 messages Oct 30, 2020 – Nov 5, 2020

Subsystems (4)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#22 lore raw

PATCH v2 8/11 lib: parse_mapping: Update argc, argv on error

In reply to Petr Machata

Currently argc and argv are not updated unless parsing of all of the mapping was successful. However in that case, "ip link" will point at the wrong argument when complaining:

    # ip link add name eth0.100 link eth0 type vlan id 100 egress 1:1 2:foo
    Error: argument "1" is wrong: invalid egress-qos-map

Update argc and argv even in the case of parsing error, so that the right element is indicated.

---
 lib/utils.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
lib/utils.c +10 -5
index aba7cc0960cd..089bbde715da 100644
@@ -1778,6 +1778,7 @@ int parse_mapping(int *argcp, char ***argvp,
 {
 	int argc = *argcp;
 	char **argv = *argvp;
+	int ret = 0;
 
 	while (argc > 0) {
 		char *colon = strchr(*argv, ':');
@@ -1787,15 +1788,19 @@ int parse_mapping(int *argcp, char ***argvp,
 			break;
 		*colon = '\0';
 
-		if (get_u32(&key, *argv, 0))
-			return 1;
-		if (mapping_cb(key, colon + 1, mapping_cb_data))
-			return 1;
+		if (get_u32(&key, *argv, 0)) {
+			ret = 1;
+			break;
+		}
+		if (mapping_cb(key, colon + 1, mapping_cb_data)) {
+			ret = 1;
+			break;
+		}
 
 		argc--, argv++;
 	}
 
 	*argcp = argc;
 	*argvp = argv;
-	return 0;
+	return ret;
 }
Signature · 1 line
2.25.1
#21 lore raw

PATCH v2 7/11 lib: Extract from iplink_vlan a helper to parse key:value arrays

In reply to Petr Machata

VLAN netdevices have two similar attributes: ingress-qos-map and egress-qos-map. These attributes can be configured with a series of 802.1-priority-to-skb-priority (and vice versa) mappings. A reusable helper along those lines will be handy for configuration of various priority-to-tc, tc-to-algorithm, and other arrays in DCB.

Therefore extract the logic to a function parse_mapping(), move to utils.c, and dispatch to utils.c from iplink_vlan.c. That necessitates extraction of a VLAN-specific parse_qos_mapping(). Do that, and propagate addattr_l() return value up, unlike the original.

---

Notes:
v2:
- In parse_qos_mapping(), propagate return value from addattr_l() [Roman Mashak]

 include/utils.h  |  4 ++++
 ip/iplink_vlan.c | 36 +++++++++++++++---------------------
 lib/utils.c      | 28 ++++++++++++++++++++++++++++
 3 files changed, 47 insertions(+), 21 deletions(-)
include/utils.h +4 -0
index e3cdb098834a..1c72221ae92c 100644
@@ -330,4 +330,8 @@ int parse_one_of(const char *msg, const char *realval, const char * const *list,
 int parse_on_off(const char *msg, const char *realval, int *p_err);
 void print_on_off_bool(FILE *fp, const char *flag, bool val);
 
+int parse_mapping(int *argcp, char ***argvp,
+		  int (*mapping_cb)(__u32 key, char *value, void *data),
+		  void *mapping_cb_data);
+
 #endif /* __UTILS_H__ */
ip/iplink_vlan.c +15 -21
index 1e6817f5de3d..dadc349db16c 100644
@@ -49,36 +49,30 @@ static int on_off(const char *msg, const char *arg)
 	return -1;
 }
 
+static int parse_qos_mapping(__u32 key, char *value, void *data)
+{
+	struct nlmsghdr *n = data;
+	struct ifla_vlan_qos_mapping m = {
+		.from = key,
+	};
+
+	if (get_u32(&m.to, value, 0))
+		return 1;
+
+	return addattr_l(n, 1024, IFLA_VLAN_QOS_MAPPING, &m, sizeof(m));
+}
+
 static int vlan_parse_qos_map(int *argcp, char ***argvp, struct nlmsghdr *n,
 			      int attrtype)
 {
-	int argc = *argcp;
-	char **argv = *argvp;
-	struct ifla_vlan_qos_mapping m;
 	struct rtattr *tail;
 
 	tail = addattr_nest(n, 1024, attrtype);
 
-	while (argc > 0) {
-		char *colon = strchr(*argv, ':');
-
-		if (!colon)
-			break;
-		*colon = '\0';
-
-		if (get_u32(&m.from, *argv, 0))
-			return 1;
-		if (get_u32(&m.to, colon + 1, 0))
-			return 1;
-		argc--, argv++;
-
-		addattr_l(n, 1024, IFLA_VLAN_QOS_MAPPING, &m, sizeof(m));
-	}
+	if (parse_mapping(argcp, argvp, &parse_qos_mapping, n))
+		return 1;
 
 	addattr_nest_end(n, tail);
-
-	*argcp = argc;
-	*argvp = argv;
 	return 0;
 }
 
lib/utils.c +28 -0
index 8deec86ecbcd..aba7cc0960cd 100644
@@ -1771,3 +1771,31 @@ void print_on_off_bool(FILE *fp, const char *flag, bool val)
 	else
 		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
 }
+
+int parse_mapping(int *argcp, char ***argvp,
+		  int (*mapping_cb)(__u32 key, char *value, void *data),
+		  void *mapping_cb_data)
+{
+	int argc = *argcp;
+	char **argv = *argvp;
+
+	while (argc > 0) {
+		char *colon = strchr(*argv, ':');
+		__u32 key;
+
+		if (!colon)
+			break;
+		*colon = '\0';
+
+		if (get_u32(&key, *argv, 0))
+			return 1;
+		if (mapping_cb(key, colon + 1, mapping_cb_data))
+			return 1;
+
+		argc--, argv++;
+	}
+
+	*argcp = argc;
+	*argvp = argv;
+	return 0;
+}
Signature · 1 line
2.25.1
#20 lore raw

PATCH v2 6/11 lib: Extract from devlink/mnlg a helper, mnlu_socket_recv_run()

In reply to Petr Machata

Receiving a message in libmnl is a somewhat involved operation. Devlink's mnlg library has an implementation that is going to be handy for other tools as well. Extract it into a new helper.

---
 devlink/mnlg.c      | 56 ++---------------------------------------
 include/mnl_utils.h |  2 ++
 lib/mnl_utils.c     | 61 +++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 65 insertions(+), 54 deletions(-)
devlink/mnlg.c +2 -54
index 4995b7af06a3..21b10c5a5669 100644
@@ -28,7 +28,6 @@ struct mnlg_socket {
 	uint32_t id;
 	uint8_t version;
 	unsigned int seq;
-	unsigned int portid;
 };
 
 static struct nlmsghdr *__mnlg_msg_prepare(struct mnlg_socket *nlg, uint8_t cmd,
@@ -57,61 +56,10 @@ int mnlg_socket_send(struct mnlg_socket *nlg, const struct nlmsghdr *nlh)
 	return mnl_socket_sendto(nlg->nl, nlh, nlh->nlmsg_len);
 }
 
-static int mnlg_cb_noop(const struct nlmsghdr *nlh, void *data)
-{
-	return MNL_CB_OK;
-}
-
-static int mnlg_cb_error(const struct nlmsghdr *nlh, void *data)
-{
-	const struct nlmsgerr *err = mnl_nlmsg_get_payload(nlh);
-
-	/* Netlink subsystems returns the errno value with different signess */
-	if (err->error < 0)
-		errno = -err->error;
-	else
-		errno = err->error;
-
-	if (nl_dump_ext_ack(nlh, NULL))
-		return MNL_CB_ERROR;
-
-	return err->error == 0 ? MNL_CB_STOP : MNL_CB_ERROR;
-}
-
-static int mnlg_cb_stop(const struct nlmsghdr *nlh, void *data)
-{
-	int len = *(int *)NLMSG_DATA(nlh);
-
-	if (len < 0) {
-		errno = -len;
-		nl_dump_ext_ack_done(nlh, len);
-		return MNL_CB_ERROR;
-	}
-	return MNL_CB_STOP;
-}
-
-static mnl_cb_t mnlg_cb_array[NLMSG_MIN_TYPE] = {
-	[NLMSG_NOOP]	= mnlg_cb_noop,
-	[NLMSG_ERROR]	= mnlg_cb_error,
-	[NLMSG_DONE]	= mnlg_cb_stop,
-	[NLMSG_OVERRUN]	= mnlg_cb_noop,
-};
-
 int mnlg_socket_recv_run(struct mnlg_socket *nlg, mnl_cb_t data_cb, void *data)
 {
-	int err;
-
-	do {
-		err = mnl_socket_recvfrom(nlg->nl, nlg->buf,
-					  MNL_SOCKET_BUFFER_SIZE);
-		if (err <= 0)
-			break;
-		err = mnl_cb_run2(nlg->buf, err, nlg->seq, nlg->portid,
-				  data_cb, data, mnlg_cb_array,
-				  ARRAY_SIZE(mnlg_cb_array));
-	} while (err > 0);
-
-	return err;
+	return mnlu_socket_recv_run(nlg->nl, nlg->seq, nlg->buf, MNL_SOCKET_BUFFER_SIZE,
+				    data_cb, data);
 }
 
 struct group_info {
include/mnl_utils.h +2 -0
index 86ce30f49a94..fa826ef1f8fe 100644
@@ -5,5 +5,7 @@
 struct mnl_socket *mnlu_socket_open(int bus);
 struct nlmsghdr *mnlu_msg_prepare(void *buf, uint32_t nlmsg_type, uint16_t flags,
 				  void *extra_header, size_t extra_header_size);
+int mnlu_socket_recv_run(struct mnl_socket *nl, unsigned int seq, void *buf, size_t buf_size,
+			 mnl_cb_t cb, void *data);
 
 #endif /* __MNL_UTILS_H__ */
lib/mnl_utils.c +61 -0

75 lines not on this page · raw message

Signature · 1 line
2.25.1
#23 lore raw

PATCH v2 9/11 lib: parse_mapping: Recognize a keyword "all"

In reply to Petr Machata

The DCB tool will have to provide an interface to a number of fixed-size arrays. Unlike the egress- and ingress-qos-map, it makes good sense to have an interface to set all members to the same value. For example to set strict priority on all TCs besides select few, or to reset allocated bandwidth to all zeroes, again besides several explicitly-given ones.

To support this usage, extend the parse_mapping() with a boolean that determines whether this special use is supported. If "all" is given and recognized, mapping_cb is called with the key of -1.

Have iplink_vlan pass false for allow_all.

---
 include/utils.h  | 2 +-
 ip/iplink_vlan.c | 2 +-
 lib/utils.c      | 6 ++++--
 3 files changed, 6 insertions(+), 4 deletions(-)
include/utils.h +1 -1
index 1c72221ae92c..8ec1b7ab0d8d 100644
@@ -330,7 +330,7 @@ int parse_one_of(const char *msg, const char *realval, const char * const *list,
 int parse_on_off(const char *msg, const char *realval, int *p_err);
 void print_on_off_bool(FILE *fp, const char *flag, bool val);
 
-int parse_mapping(int *argcp, char ***argvp,
+int parse_mapping(int *argcp, char ***argvp, bool allow_all,
 		  int (*mapping_cb)(__u32 key, char *value, void *data),
 		  void *mapping_cb_data);
 
ip/iplink_vlan.c +1 -1
index dadc349db16c..1426f2afca23 100644
@@ -69,7 +69,7 @@ static int vlan_parse_qos_map(int *argcp, char ***argvp, struct nlmsghdr *n,
 
 	tail = addattr_nest(n, 1024, attrtype);
 
-	if (parse_mapping(argcp, argvp, &parse_qos_mapping, n))
+	if (parse_mapping(argcp, argvp, false, &parse_qos_mapping, n))
 		return 1;
 
 	addattr_nest_end(n, tail);
lib/utils.c +4 -2
index 089bbde715da..3b3b42b15013 100644
@@ -1772,7 +1772,7 @@ void print_on_off_bool(FILE *fp, const char *flag, bool val)
 		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
 }
 
-int parse_mapping(int *argcp, char ***argvp,
+int parse_mapping(int *argcp, char ***argvp, bool allow_all,
 		  int (*mapping_cb)(__u32 key, char *value, void *data),
 		  void *mapping_cb_data)
 {
@@ -1788,7 +1788,9 @@ int parse_mapping(int *argcp, char ***argvp,
 			break;
 		*colon = '\0';
 
-		if (get_u32(&key, *argv, 0)) {
+		if (allow_all && matches(*argv, "all") == 0) {
+			key = (__u32) -1;
+		} else if (get_u32(&key, *argv, 0)) {
 			ret = 1;
 			break;
 		}
Signature · 1 line
2.25.1
#24 lore raw

PATCH v2 10/11 Add skeleton of a new tool, dcb

In reply to Petr Machata

The Linux DCB interface allows configuration of a broad range of hardware-specific attributes, such as TC scheduling, flow control, per-port buffer configuration, TC rate, etc. Add a new tool to show that configuration and tweak it.

DCB allows configuration of several objects, and possibly could expand to pre-standard CEE interfaces. Therefore the tool itself is a lean shell that dispatches to subtools each dedicated to one of the objects.

---
 Makefile       |   2 +-
 dcb/Makefile   |  24 +++
 dcb/dcb.c      | 401 +++++++++++++++++++++++++++++++++++++++++++++++++
 dcb/dcb.h      |  35 +++++
 man/man8/dcb.8 | 103 +++++++++++++
 5 files changed, 564 insertions(+), 1 deletion(-)
 create mode 100644 dcb/Makefile
 create mode 100644 dcb/dcb.c
 create mode 100644 dcb/dcb.h
 create mode 100644 man/man8/dcb.8
Makefile +1 -1
index 5b040415a12b..e64c65992585 100644
@@ -55,7 +55,7 @@ WFLAGS += -Wmissing-declarations -Wold-style-definition -Wformat=2
 CFLAGS := $(WFLAGS) $(CCOPTS) -I../include -I../include/uapi $(DEFINES) $(CFLAGS)
 YACCFLAGS = -d -t -v
 
-SUBDIRS=lib ip tc bridge misc netem genl tipc devlink rdma man
+SUBDIRS=lib ip tc bridge misc netem genl tipc devlink rdma dcb man
 
 LIBNETLINK=../lib/libutil.a ../lib/libnetlink.a
 LDLIBS += $(LIBNETLINK)
dcb/Makefile new file +24 -0
new file mode 100644
index 000000000000..9966c8f0bfa4
@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: GPL-2.0
+include ../config.mk
+
+TARGETS :=
+
+ifeq ($(HAVE_MNL),y)
+
+DCBOBJ = dcb.o
+TARGETS += dcb
+
+endif
+
+all: $(TARGETS) $(LIBS)
+
+dcb: $(DCBOBJ) $(LIBNETLINK)
+	$(QUIET_LINK)$(CC) $^ $(LDFLAGS) $(LDLIBS) -o $@
+
+install: all
+	for i in $(TARGETS); \
+	do install -m 0755 $$i $(DESTDIR)$(SBINDIR); \
+	done
+
+clean:
+	rm -f $(DCBOBJ) $(TARGETS)
dcb/dcb.c new file +401 -0

401 lines not on this page · raw message

dcb/dcb.h new file +35 -0
new file mode 100644
index 000000000000..1d31a0f94652
@@ -0,0 +1,35 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __DCB_H__
+#define __DCB_H__ 1
+
+#include <stdbool.h>
+#include <stddef.h>
+
+/* dcb.c */
+
+struct dcb {
+	char *buf;
+	struct mnl_socket *nl;
+	bool json_output;
+};
+
+int dcb_parse_mapping(__u32 key, __u8 value, __u8 max_value, __u8 *array,
+		      const char *what_key, const char *what_value);
+int dcb_cmd_parse_dev(struct dcb *dcb, int argc, char **argv,
+		      int (*and_then)(struct dcb *dcb, const char *dev,
+				      int argc, char **argv),
+		      void (*help)(void));
+
+int dcb_get_attribute(struct dcb *dcb, const char *dev, int attr,
+		      void *data, size_t data_len);
+int dcb_set_attribute(struct dcb *dcb, const char *dev, int attr,
+		      const void *data, size_t data_len);
+
+void dcb_print_named_array(FILE *fp, const char *fp_name, const char *json_name,
+			   const __u8 *array, size_t size,
+			   void (*print_array)(FILE *, const __u8 *, size_t));
+void dcb_print_array_num(FILE *fp, const __u8 *array, size_t size);
+void dcb_print_array_kw(FILE *fp, const __u8 *array, size_t array_size,
+			const char *const kw[], size_t kw_size);
+
+#endif /* __DCB_H__ */
man/man8/dcb.8 new file +103 -0

103 lines not on this page · raw message

Signature · 1 line
2.25.1
#25 lore raw

PATCH v2 11/11 dcb: Add a subtool for the DCB ETS object

In reply to Petr Machata

ETS, for "Enhanced Transmission Selection", is a set of configurations that permit configuration of mapping of priorities to traffic classes, traffic selection algorithm to use per traffic class, bandwidth allocation, etc.

Add a dcb subtool to allow showing and tweaking of individual ETS configuration options. For example:

    # dcb ets show dev eni1np1
    willing on ets_cap 8 cbs off
    tc-bw 0:0 1:0 2:0 3:0 4:100 5:0 6:0 7:0
    pg-bw 0:0 1:0 2:0 3:0 4:0 5:0 6:0 7:0
    tc-tsa 0:strict 1:strict 2:strict 3:strict 4:ets 5:strict 6:strict 7:strict
    prio-tc 0:1 1:3 2:5 3:0 4:0 5:0 6:0 7:0
    reco-tc-bw 0:0 1:0 2:0 3:0 4:0 5:0 6:0 7:0
    reco-tc-tsa 0:strict 1:strict 2:strict 3:strict 4:strict 5:strict 6:strict 7:strict
    reco-prio-tc 0:0 1:0 2:0 3:0 4:0 5:0 6:0 7:0
---
 dcb/Makefile       |   2 +-
 dcb/dcb.c          |   4 +-
 dcb/dcb.h          |   4 +
 dcb/dcb_ets.c      | 430 +++++++++++++++++++++++++++++++++++++++++++++
 man/man8/dcb-ets.8 | 185 +++++++++++++++++++
 man/man8/dcb.8     |  11 ++
 6 files changed, 634 insertions(+), 2 deletions(-)
 create mode 100644 dcb/dcb_ets.c
 create mode 100644 man/man8/dcb-ets.8
dcb/Makefile +1 -1
index 9966c8f0bfa4..895817163562 100644
@@ -5,7 +5,7 @@ TARGETS :=
 
 ifeq ($(HAVE_MNL),y)
 
-DCBOBJ = dcb.o
+DCBOBJ = dcb.o dcb_ets.o
 TARGETS += dcb
 
 endif
dcb/dcb.c +3 -1
index 26c1f91e815a..a776aa542f93 100644
@@ -286,7 +286,7 @@ static void dcb_help(void)
 	fprintf(stderr,
 		"Usage: dcb [ OPTIONS ] OBJECT { COMMAND | help }\n"
 		"       dcb [ -f[orce] ] -b[atch] filename -N[etns] netnsname\n"
-		"where  OBJECT :=\n"
+		"where  OBJECT := ets\n"
 		"       OPTIONS := { -V[ersion] | -j[son] | -p[retty] | -v[erbose] }\n");
 }
 
@@ -295,6 +295,8 @@ static int dcb_cmd(struct dcb *dcb, int argc, char **argv)
 	if (!argc || matches(*argv, "help") == 0) {
 		dcb_help();
 		return 0;
+	} else if (matches(*argv, "ets") == 0) {
+		return dcb_cmd_ets(dcb, argc - 1, argv + 1);
 	}
 
 	fprintf(stderr, "Object \"%s\" is unknown\n", *argv);
dcb/dcb.h +4 -0
index 1d31a0f94652..9b46e09178be 100644
@@ -32,4 +32,8 @@ void dcb_print_array_num(FILE *fp, const __u8 *array, size_t size);
 void dcb_print_array_kw(FILE *fp, const __u8 *array, size_t array_size,
 			const char *const kw[], size_t kw_size);
 
+/* dcb_ets.c */
+
+int dcb_cmd_ets(struct dcb *dcb, int argc, char **argv);
+
 #endif /* __DCB_H__ */
dcb/dcb_ets.c new file +430 -0

430 lines not on this page · raw message

man/man8/dcb-ets.8 new file +185 -0

185 lines not on this page · raw message

man/man8/dcb.8 +11 -0
index 25ddf204d60e..ec116f6ce265 100644
@@ -6,6 +6,13 @@ dcb \- show / manipulate DCB (Data Center Bridging) settings
 .ad l
 .in +8
 
+.ti -8
+.B dcb
+.RI "[ " OPTIONS " ] "
+.B ets
+.RI "{ " COMMAND " | " help " }"
+.sp
+
 .ti -8
 .B dcb
 .RB "[ " -force " ] "
@@ -46,6 +53,10 @@ When combined with -j generate a pretty JSON output.
 
 .SH OBJECTS
 
+.TP
+.B ets
+- Configuration of ETS (Enhanced Transmission Selection)
+
 .SH COMMANDS
 
 A \fICOMMAND\fR specifies the action to perform on the object. The set of
Signature · 1 line
2.25.1
#0 lore raw

v2 cover Add a tool for configuration of DCB

The Linux DCB interface allows configuration of a broad range of hardware-specific attributes, such as TC scheduling, flow control, per-port buffer configuration, TC rate, etc.

Currently a common libre tool for configuration of DCB is OpenLLDP. This suite contains a daemon that uses Linux DCB interface to configure HW according to the DCB TLVs exchanged over an interface. The daemon can also be controlled by a client, through which the user can adjust and view the configuration. The downside of using OpenLLDP is that it is somewhat heavyweight and difficult to use in scripts, and does not support extensions such as buffer and rate commands.

For access to many HW features, one would be perfectly fine with a fire-and-forget tool along the lines of "ip" or "tc". For scripting in particular, this would be ideal. This author is aware of one such tool, mlnx_qos from Mellanox OFED scripts collection[1].

The downside here is that the tool is very verbose, the command line language is awkward to use, it is not packaged in Linux distros, and generally has the appearance of a very vendor-specific tool, despite not being one.

This patchset addresses the above issues by providing a seed of a clean, well-documented, easily usable, extensible fire-and-forget tool for DCB configuration:

    # dcb ets set dev eni1np1 \
                  tc-tsa all:strict 0:ets 1:ets 2:ets \
		  tc-bw all:0 0:33 1:33 2:34
    # dcb ets show dev eni1np1 tc-tsa tc-bw
    tc-tsa 0:ets 1:ets 2:ets 3:strict 4:strict 5:strict 6:strict 7:strict
    tc-bw 0:33 1:33 2:34 3:0 4:0 5:0 6:0 7:0
    # dcb ets set dev eni1np1 tc-bw 1:30 2:37
    # dcb -j ets show dev eni1np1 | jq '.tc_bw[2]'
    37

The patchset proceeds as follows:

- Many tools in iproute2 have an option to work in batch mode, where the commands to run are given in a file. The code to handle batching is largely the same independent of the tool in question. In patch #1, add a helper to handle the batching, and migrate individual tools to use it.

- A number of configuration options come in a form of an on-off switch. This in turn can be considered a special case of parsing one of a given set of strings. In patch #2, extract helpers to parse one of a number of strings, on top of which build an on-off parser.

Currently each tool open-codes the logic to parse the on-off toggle. A future patch set will migrate instances of this code over to the new helpers.

- The on/off toggles from previous list item sometimes need to be dumped. While in the FP output, one typically wishes to maintain consistency with the command line and show actual strings, "on" and "off", in JSON output one would rather use booleans. This logic is somewhat annoying to have to open-code time and again. Therefore in patch #3, add a helper to do just that.

- The DCB tool is built on top of libmnl. Several routines will be basically the same in DCB as they are currently in devlink. In patches #4-#6, extract them to a new module, mnl_utils, for easy reuse.

- Much of DCB is built around arrays. A syntax similar to the iplink_vlan's ingress-qos-map / egress-qos-map is very handy for describing changes done to such arrays. Therefore in patch #7, extract a helper, parse_mapping(), which manages parsing of key-value arrays. In patch #8, fix a buglet in the helper, and in patch #9, extend it to allow setting of all array elements in one go.

- In patch #10, add a skeleton of "dcb", which contains common helpers and dispatches to subtools for handling of individual objects. The skeleton is empty as of this patch.

In patch #11, add "dcb_ets", a module for handling of specifically DCB ETS objects.

The intention is to gradually add handlers for at least PFC, APP, peer configuration, buffers and rates.

[1] https://github.com/Mellanox/mlnx-tools/tree/master/ofed_scripts

v2:
- A new function, print_on_off_bool(), has been introduced for showing on-off toggles in both FP and JSON modes.
[Jakub Kicinski, Stephen Hemminger]

- This prompted refactoring in several existing files, and pushed the number of patches in the set too high. The cleanup patches have therefore been moved out to another patchset, which will follow after this one.

- When dumping JSON, format keys so that they are valid jq identifiers. E.g. "tc_tsa" instead of "tc-tsa". Additionally, do not dump arrays as objects with string indices, but as true arrays. This allows for more natural access to individual items, e.g.:

    # dcb ets -j show dev eth0 | jq '.tc_tsa[3]'
  Instead of:
    # dcb ets -j show dev eth0 | jq '.["tc-tsa"]["3"]'

- Patch #4:
- Add SPDX-License-Identifier

- Patch #7:
- In parse_qos_mapping(), propagate return value from addattr_l() [Roman Mashak]

Petr Machata (11):
Unify batch processing across tools
lib: Add parse_one_of(), parse_on_off()
lib: utils: Add print_on_off_bool()
lib: Extract from devlink/mnlg a helper, mnlu_socket_open()
lib: Extract from devlink/mnlg a helper, mnlu_msg_prepare()
lib: Extract from devlink/mnlg a helper, mnlu_socket_recv_run()
lib: Extract from iplink_vlan a helper to parse key:value arrays
lib: parse_mapping: Update argc, argv on error
lib: parse_mapping: Recognize a keyword "all"
Add skeleton of a new tool, dcb
dcb: Add a subtool for the DCB ETS object

 Makefile            |   2 +-
 bridge/bridge.c     |  38 +---
 dcb/Makefile        |  24 +++
 dcb/dcb.c           | 403 +++++++++++++++++++++++++++++++++++++++++
 dcb/dcb.h           |  39 ++++
 dcb/dcb_ets.c       | 430 ++++++++++++++++++++++++++++++++++++++++++++
 devlink/Makefile    |   2 +-
 devlink/devlink.c   |  41 +----
 devlink/mnlg.c      |  93 ++--------
 include/mnl_utils.h |  11 ++
 include/utils.h     |  12 ++
 ip/ip.c             |  46 ++---
 ip/iplink_vlan.c    |  36 ++--
 ip/ipmacsec.c       |  52 ++----
 lib/Makefile        |   2 +-
 lib/mnl_utils.c     | 110 ++++++++++++
 lib/utils.c         | 111 ++++++++++++
 man/man8/dcb-ets.8  | 185 +++++++++++++++++++
 man/man8/dcb.8      | 114 ++++++++++++
 rdma/rdma.c         |  38 +---
 tc/tc.c             |  38 +---
 21 files changed, 1518 insertions(+), 309 deletions(-)
 create mode 100644 dcb/Makefile
 create mode 100644 dcb/dcb.c
 create mode 100644 dcb/dcb.h
 create mode 100644 dcb/dcb_ets.c
 create mode 100644 include/mnl_utils.h
 create mode 100644 lib/mnl_utils.c
 create mode 100644 man/man8/dcb-ets.8
 create mode 100644 man/man8/dcb.8
Signature · 1 line
2.25.1
#19 lore raw

PATCH v2 5/11 lib: Extract from devlink/mnlg a helper, mnlu_msg_prepare()

Allocation of a new netlink message with the two usual headers is reusable with other netlink netlink message types. Extract it into a helper, mnlu_msg_prepare(). Take the second header as an argument, instead of passing in parameters to initialize it, and copy it in.

---
 devlink/mnlg.c      | 18 ++++++------------
 include/mnl_utils.h |  2 ++
 lib/mnl_utils.c     | 19 +++++++++++++++++++
 3 files changed, 27 insertions(+), 12 deletions(-)
devlink/mnlg.c +6 -12
index 9817bbad5e7d..4995b7af06a3 100644
@@ -14,7 +14,6 @@
 #include <string.h>
 #include <errno.h>
 #include <unistd.h>
-#include <time.h>
 #include <libmnl/libmnl.h>
 #include <linux/genetlink.h>
 
@@ -36,19 +35,14 @@ static struct nlmsghdr *__mnlg_msg_prepare(struct mnlg_socket *nlg, uint8_t cmd,
 					   uint16_t flags, uint32_t id,
 					   uint8_t version)
 {
+	struct genlmsghdr genl = {
+		.cmd = cmd,
+		.version = version,
+	};
 	struct nlmsghdr *nlh;
-	struct genlmsghdr *genl;
-
-	nlh = mnl_nlmsg_put_header(nlg->buf);
-	nlh->nlmsg_type	= id;
-	nlh->nlmsg_flags = flags;
-	nlg->seq = time(NULL);
-	nlh->nlmsg_seq = nlg->seq;
-
-	genl = mnl_nlmsg_put_extra_header(nlh, sizeof(struct genlmsghdr));
-	genl->cmd = cmd;
-	genl->version = version;
 
+	nlh = mnlu_msg_prepare(nlg->buf, id, flags, &genl, sizeof(genl));
+	nlg->seq = nlh->nlmsg_seq;
 	return nlh;
 }
 
include/mnl_utils.h +2 -0
index 10a064afdfe8..86ce30f49a94 100644
@@ -3,5 +3,7 @@
 #define __MNL_UTILS_H__ 1
 
 struct mnl_socket *mnlu_socket_open(int bus);
+struct nlmsghdr *mnlu_msg_prepare(void *buf, uint32_t nlmsg_type, uint16_t flags,
+				  void *extra_header, size_t extra_header_size);
 
 #endif /* __MNL_UTILS_H__ */
lib/mnl_utils.c +19 -0
index 2426912aa511..61e8060ecbca 100644
@@ -3,6 +3,8 @@
  * mnl_utils.c	Helpers for working with libmnl.
  */
 
+#include <string.h>
+#include <time.h>
 #include <libmnl/libmnl.h>
 
 #include "mnl_utils.h"
@@ -28,3 +30,20 @@ err_bind:
 	mnl_socket_close(nl);
 	return NULL;
 }
+
+struct nlmsghdr *mnlu_msg_prepare(void *buf, uint32_t nlmsg_type, uint16_t flags,
+				  void *extra_header, size_t extra_header_size)
+{
+	struct nlmsghdr *nlh;
+	void *eh;
+
+	nlh = mnl_nlmsg_put_header(buf);
+	nlh->nlmsg_type = nlmsg_type;
+	nlh->nlmsg_flags = flags;
+	nlh->nlmsg_seq = time(NULL);
+
+	eh = mnl_nlmsg_put_extra_header(nlh, extra_header_size);
+	memcpy(eh, extra_header, extra_header_size);
+
+	return nlh;
+}
Signature · 1 line
2.25.1
#1 lore raw

PATCH v2 1/11 Unify batch processing across tools

In reply to Petr Machata

The code for handling batches is largely the same across iproute2 tools. Extract a helper to handle the batch, and adjust the tools to dispatch to this helper. Sandwitch the invocation between prologue / epilogue code specific for each tool.

---
 bridge/bridge.c   | 38 +++++++-------------------------------
 devlink/devlink.c | 41 +++++++----------------------------------
 include/utils.h   |  3 +++
 ip/ip.c           | 46 ++++++++++------------------------------------
 lib/utils.c       | 40 ++++++++++++++++++++++++++++++++++++++++
 rdma/rdma.c       | 38 +++++++-------------------------------
 tc/tc.c           | 38 +++++++-------------------------------
 7 files changed, 81 insertions(+), 163 deletions(-)
bridge/bridge.c +7 -31
index 453d689732bd..8f691cfdd466 100644
@@ -77,20 +77,14 @@ static int do_cmd(const char *argv0, int argc, char **argv)
 	return -1;
 }
 
-static int batch(const char *name)
+static int br_batch_cmd(int argc, char *argv[], void *data)
 {
-	char *line = NULL;
-	size_t len = 0;
-	int ret = EXIT_SUCCESS;
+	return do_cmd(argv[0], argc, argv);
+}
 
-	if (name && strcmp(name, "-") != 0) {
-		if (freopen(name, "r", stdin) == NULL) {
-			fprintf(stderr,
-				"Cannot open file \"%s\" for reading: %s\n",
-				name, strerror(errno));
-			return EXIT_FAILURE;
-		}
-	}
+static int batch(const char *name)
+{
+	int ret;
 
 	if (rtnl_open(&rth, 0) < 0) {
 		fprintf(stderr, "Cannot open rtnetlink\n");
@@ -99,25 +93,7 @@ static int batch(const char *name)
 
 	rtnl_set_strict_dump(&rth);
 
-	cmdlineno = 0;
-	while (getcmdline(&line, &len, stdin) != -1) {
-		char *largv[100];
-		int largc;
-
-		largc = makeargs(line, largv, 100);
-		if (largc == 0)
-			continue;       /* blank line */
-
-		if (do_cmd(largv[0], largc, largv)) {
-			fprintf(stderr, "Command failed %s:%d\n",
-				name, cmdlineno);
-			ret = EXIT_FAILURE;
-			if (!force)
-				break;
-		}
-	}
-	if (line)
-		free(line);
+	ret = do_batch(name, force, br_batch_cmd, NULL);
 
 	rtnl_close(&rth);
 	return ret;
devlink/devlink.c +7 -34

50 lines not on this page · raw message

include/utils.h +3 -0
index 7041c4612e46..085b17b1f6e3 100644
@@ -322,4 +322,7 @@ int get_time64(__s64 *time, const char *str);
 char *sprint_time(__u32 time, char *buf);
 char *sprint_time64(__s64 time, char *buf);
 
+int do_batch(const char *name, bool force,
+	     int (*cmd)(int argc, char *argv[], void *user), void *user);
+
 #endif /* __UTILS_H__ */
ip/ip.c +10 -36

66 lines not on this page · raw message

lib/utils.c +40 -0
index c98021d6ecad..9815e328c9e0 100644
@@ -1695,3 +1695,43 @@ char *sprint_time64(__s64 time, char *buf)
 	print_time64(buf, SPRINT_BSIZE-1, time);
 	return buf;
 }
+
+int do_batch(const char *name, bool force,
+	     int (*cmd)(int argc, char *argv[], void *data), void *data)
+{
+	char *line = NULL;
+	size_t len = 0;
+	int ret = EXIT_SUCCESS;
+
+	if (name && strcmp(name, "-") != 0) {
+		if (freopen(name, "r", stdin) == NULL) {
+			fprintf(stderr,
+				"Cannot open file \"%s\" for reading: %s\n",
+				name, strerror(errno));
+			return EXIT_FAILURE;
+		}
+	}
+
+	cmdlineno = 0;
+	while (getcmdline(&line, &len, stdin) != -1) {
+		char *largv[100];
+		int largc;
+
+		largc = makeargs(line, largv, 100);
+		if (!largc)
+			continue;	/* blank line */
+
+		if (cmd(largc, largv, data)) {
+			fprintf(stderr, "Command failed %s:%d\n",
+				name, cmdlineno);
+			ret = EXIT_FAILURE;
+			if (!force)
+				break;
+		}
+	}
+
+	if (line)
+		free(line);
+
+	return ret;
+}
rdma/rdma.c +7 -31
index 9ea2d17ffe9e..8dc2d3e344be 100644
@@ -41,40 +41,16 @@ static int rd_cmd(struct rd *rd, int argc, char **argv)
 	return rd_exec_cmd(rd, cmds, "object");
 }
 
-static int rd_batch(struct rd *rd, const char *name, bool force)
+static int rd_batch_cmd(int argc, char *argv[], void *data)
 {
-	char *line = NULL;
-	size_t len = 0;
-	int ret = 0;
-
-	if (name && strcmp(name, "-") != 0) {
-		if (!freopen(name, "r", stdin)) {
-			pr_err("Cannot open file \"%s\" for reading: %s\n",
-			       name, strerror(errno));
-			return errno;
-		}
-	}
+	struct rd *rd = data;
 
-	cmdlineno = 0;
-	while (getcmdline(&line, &len, stdin) != -1) {
-		char *largv[512];
-		int largc;
-
-		largc = makeargs(line, largv, ARRAY_SIZE(largv));
-		if (!largc)
-			continue;	/* blank line */
-
-		ret = rd_cmd(rd, largc, largv);
-		if (ret) {
-			pr_err("Command failed %s:%d\n", name, cmdlineno);
-			if (!force)
-				break;
-		}
-	}
-
-	free(line);
+	return rd_cmd(rd, argc, argv);
+}
 
-	return ret;
+static int rd_batch(struct rd *rd, const char *name, bool force)
+{
+	return do_batch(name, force, rd_batch_cmd, rd);
 }
 
 static int rd_init(struct rd *rd, char *filename)
tc/tc.c +7 -31
index 5d57054b45fb..01fe58d06202 100644
@@ -231,22 +231,16 @@ static int do_cmd(int argc, char **argv)
 	return -1;
 }
 
+static int tc_batch_cmd(int argc, char *argv[], void *data)
+{
+	return do_cmd(argc, argv);
+}
+
 static int batch(const char *name)
 {
-	char *line = NULL;
-	size_t len = 0;
-	int ret = 0;
+	int ret;
 
 	batch_mode = 1;
-	if (name && strcmp(name, "-") != 0) {
-		if (freopen(name, "r", stdin) == NULL) {
-			fprintf(stderr,
-				"Cannot open file \"%s\" for reading: %s\n",
-				name, strerror(errno));
-			return -1;
-		}
-	}
-
 	tc_core_init();
 
 	if (rtnl_open(&rth, 0) < 0) {
@@ -254,26 +248,8 @@ static int batch(const char *name)
 		return -1;
 	}
 
-	cmdlineno = 0;
-	while (getcmdline(&line, &len, stdin) != -1) {
-		char *largv[100];
-		int largc;
-
-		largc = makeargs(line, largv, 100);
-		if (largc == 0)
-			continue;	/* blank line */
-
-		if (do_cmd(largc, largv)) {
-			fprintf(stderr, "Command failed %s:%d\n",
-				name, cmdlineno);
-			ret = 1;
-			if (!force)
-				break;
-		}
-		fflush(stdout);
-	}
+	ret = do_batch(name, force, tc_batch_cmd, NULL);
 
-	free(line);
 	rtnl_close(&rth);
 	return ret;
 }
Signature · 1 line
2.25.1
#5 lore raw

PATCH v2 3/11 lib: utils: Add print_on_off_bool()

In reply to Petr Machata

The value of a number of booleans is shown as "on" and "off" in the plain output, and as an actual boolean in JSON mode. Add a function that does that.

---
 include/utils.h | 1 +
 lib/utils.c     | 8 ++++++++
 2 files changed, 9 insertions(+)
include/utils.h +1 -0
index bd62cdcd7122..e3cdb098834a 100644
@@ -328,5 +328,6 @@ int do_batch(const char *name, bool force,
 int parse_one_of(const char *msg, const char *realval, const char * const *list,
 		 size_t len, int *p_err);
 int parse_on_off(const char *msg, const char *realval, int *p_err);
+void print_on_off_bool(FILE *fp, const char *flag, bool val);
 
 #endif /* __UTILS_H__ */
lib/utils.c +8 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}
Signature · 1 line
2.25.1
#18 lore raw

PATCH v2 4/11 lib: Extract from devlink/mnlg a helper, mnlu_socket_open()

In reply to Petr Machata

This little dance of mnl_socket_open(), option setting, and bind, is the same regardless of tool. Extract into a new module that should hold helpers for working with libmnl, mnl_util.c.

---

Notes:
v2:
- Add SPDX-License-Identifier

 devlink/Makefile    |  2 +-
 devlink/mnlg.c      | 19 ++++---------------
 include/mnl_utils.h |  7 +++++++
 lib/Makefile        |  2 +-
 lib/mnl_utils.c     | 30 ++++++++++++++++++++++++++++++
 5 files changed, 43 insertions(+), 17 deletions(-)
 create mode 100644 include/mnl_utils.h
 create mode 100644 lib/mnl_utils.c
devlink/Makefile +1 -1
index 7da7d1fa18d5..d540feb3c012 100644
@@ -12,7 +12,7 @@ endif
 
 all: $(TARGETS) $(LIBS)
 
-devlink: $(DEVLINKOBJ)
+devlink: $(DEVLINKOBJ) $(LIBNETLINK)
 	$(QUIET_LINK)$(CC) $^ $(LDFLAGS) $(LDLIBS) -o $@
 
 install: all
devlink/mnlg.c +4 -15
index c7d25e8713a1..9817bbad5e7d 100644
@@ -19,6 +19,7 @@
 #include <linux/genetlink.h>
 
 #include "libnetlink.h"
+#include "mnl_utils.h"
 #include "utils.h"
 #include "mnlg.h"
 
@@ -263,7 +264,6 @@ struct mnlg_socket *mnlg_socket_open(const char *family_name, uint8_t version)
 {
 	struct mnlg_socket *nlg;
 	struct nlmsghdr *nlh;
-	int one = 1;
 	int err;
 
 	nlg = malloc(sizeof(*nlg));
@@ -274,19 +274,9 @@ struct mnlg_socket *mnlg_socket_open(const char *family_name, uint8_t version)
 	if (!nlg->buf)
 		goto err_buf_alloc;
 
-	nlg->nl = mnl_socket_open(NETLINK_GENERIC);
+	nlg->nl = mnlu_socket_open(NETLINK_GENERIC);
 	if (!nlg->nl)
-		goto err_mnl_socket_open;
-
-	/* Older kernels may no support capped/extended ACK reporting */
-	mnl_socket_setsockopt(nlg->nl, NETLINK_CAP_ACK, &one, sizeof(one));
-	mnl_socket_setsockopt(nlg->nl, NETLINK_EXT_ACK, &one, sizeof(one));
-
-	err = mnl_socket_bind(nlg->nl, 0, MNL_SOCKET_AUTOPID);
-	if (err < 0)
-		goto err_mnl_socket_bind;
-
-	nlg->portid = mnl_socket_get_portid(nlg->nl);
+		goto err_socket_open;
 
 	nlh = __mnlg_msg_prepare(nlg, CTRL_CMD_GETFAMILY,
 				 NLM_F_REQUEST | NLM_F_ACK, GENL_ID_CTRL, 1);
@@ -305,9 +295,8 @@ struct mnlg_socket *mnlg_socket_open(const char *family_name, uint8_t version)
 
 err_mnlg_socket_recv_run:
 err_mnlg_socket_send:
-err_mnl_socket_bind:
 	mnl_socket_close(nlg->nl);
-err_mnl_socket_open:
+err_socket_open:
 	free(nlg->buf);
 err_buf_alloc:
 	free(nlg);
include/mnl_utils.h new file +7 -0
new file mode 100644
index 000000000000..10a064afdfe8
@@ -0,0 +1,7 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __MNL_UTILS_H__
+#define __MNL_UTILS_H__ 1
+
+struct mnl_socket *mnlu_socket_open(int bus);
+
+#endif /* __MNL_UTILS_H__ */
lib/Makefile +1 -1
index 7cba1857b7fa..13f4ee15373b 100644
@@ -7,7 +7,7 @@ UTILOBJ = utils.o rt_names.o ll_map.o ll_types.o ll_proto.o ll_addr.o \
 	inet_proto.o namespace.o json_writer.o json_print.o \
 	names.o color.o bpf.o exec.o fs.o cg_map.o
 
-NLOBJ=libgenl.o libnetlink.o
+NLOBJ=libgenl.o libnetlink.o mnl_utils.o
 
 all: libnetlink.a libutil.a
 
lib/mnl_utils.c new file +30 -0
new file mode 100644
index 000000000000..2426912aa511
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * mnl_utils.c	Helpers for working with libmnl.
+ */
+
+#include <libmnl/libmnl.h>
+
+#include "mnl_utils.h"
+
+struct mnl_socket *mnlu_socket_open(int bus)
+{
+	struct mnl_socket *nl;
+	int one = 1;
+
+	nl = mnl_socket_open(bus);
+	if (nl == NULL)
+		return NULL;
+
+	mnl_socket_setsockopt(nl, NETLINK_CAP_ACK, &one, sizeof(one));
+	mnl_socket_setsockopt(nl, NETLINK_EXT_ACK, &one, sizeof(one));
+
+	if (mnl_socket_bind(nl, 0, MNL_SOCKET_AUTOPID) < 0)
+		goto err_bind;
+
+	return nl;
+
+err_bind:
+	mnl_socket_close(nl);
+	return NULL;
+}
Signature · 1 line
2.25.1
#2 lore raw

PATCH v2 2/11 lib: Add parse_one_of(), parse_on_off()

In reply to Petr Machata

Take from the macsec code parse_one_of() and adapt so that it passes the primary result as the main return value, and error result through a pointer. That is the simplest way to make the code reusable across data types without introducing extra magic.

Also from macsec take the specialization of parse_one_of() for parsing specifically the strings "off" and "on".

Convert the macsec code to the new helpers.

---
 include/utils.h |  4 ++++
 ip/ipmacsec.c   | 52 +++++++++++--------------------------------------
 lib/utils.c     | 28 ++++++++++++++++++++++++++
 3 files changed, 43 insertions(+), 41 deletions(-)
include/utils.h +4 -0
index 085b17b1f6e3..bd62cdcd7122 100644
@@ -325,4 +325,8 @@ char *sprint_time64(__s64 time, char *buf);
 int do_batch(const char *name, bool force,
 	     int (*cmd)(int argc, char *argv[], void *user), void *user);
 
+int parse_one_of(const char *msg, const char *realval, const char * const *list,
+		 size_t len, int *p_err);
+int parse_on_off(const char *msg, const char *realval, int *p_err);
+
 #endif /* __UTILS_H__ */
ip/ipmacsec.c +11 -41

118 lines not on this page · raw message

lib/utils.c +28 -0
index 9815e328c9e0..930877ae0f0d 100644
@@ -1735,3 +1735,31 @@ int do_batch(const char *name, bool force,
 
 	return ret;
 }
+
+int parse_one_of(const char *msg, const char *realval, const char * const *list,
+		 size_t len, int *p_err)
+{
+	int i;
+
+	for (i = 0; i < len; i++) {
+		if (list[i] && matches(realval, list[i]) == 0) {
+			*p_err = 0;
+			return i;
+		}
+	}
+
+	fprintf(stderr, "Error: argument of \"%s\" must be one of ", msg);
+	for (i = 0; i < len; i++)
+		if (list[i])
+			fprintf(stderr, "\"%s\", ", list[i]);
+	fprintf(stderr, "not \"%s\"\n", realval);
+	*p_err = -EINVAL;
+	return 0;
+}
+
+int parse_on_off(const char *msg, const char *realval, int *p_err)
+{
+	static const char * const values_on_off[] = { "off", "on" };
+
+	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
+}
Signature · 1 line
2.25.1
#6 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to Petr Machata

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 7 lines · On Fri, 30 Oct 2020 13:29:50 +0100 Petr Machata <[email protected]> wrote:
lib/utils.c +7 -0
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

Please use print_string(PRINT_FP for non json case.
I am use fprintf(fp as indicator for code that has not been already converted to JSON.

And passing the FILE *fp is also indication of old code. Original iproute2 passed it around but it was always stdout.

#3 lore raw

Re: [PATCH iproute2-next v2 02/11] lib: Add parse_one_of(), parse_on_off()

In reply to Petr Machata
Quoted 27 lines · Petr Machata
lib/utils.c +18 -0
index 9815e328c9e0..930877ae0f0d 100644
@@ -1735,3 +1735,31 @@ int do_batch(const char *name, bool force,
 
 	return ret;
 }
+
+int parse_one_of(const char *msg, const char *realval, const char * const *list,
+		 size_t len, int *p_err)
+{
+	int i;
+
+	for (i = 0; i < len; i++) {
+		if (list[i] && matches(realval, list[i]) == 0) {
+			*p_err = 0;
+			return i;
+		}
+	}
+
+	fprintf(stderr, "Error: argument of \"%s\" must be one of ", msg);
+	for (i = 0; i < len; i++)
+		if (list[i])
+			fprintf(stderr, "\"%s\", ", list[i]);
+	fprintf(stderr, "not \"%s\"\n", realval);

lib/utils.c int do_batch(const char *name, bool force, · line 1765

Quoted 10 lines
lib/utils.c +10 -0
+	*p_err = -EINVAL;
+	return 0;
+}
+
+int parse_on_off(const char *msg, const char *realval, int *p_err)
+{
+	static const char * const values_on_off[] = { "off", "on" };
+
+	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
+}

This has weird semantics to me. You have a buried array of strings and returning the index of the one that matches. Let's use a 'bool' return for parse_on_off that makes it clear that the string is 'off' = false or 'on' = true.

#8 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to Petr Machata
Quoted 18 lines · Petr Machata
include/utils.h +1 -0
index bd62cdcd7122..e3cdb098834a 100644
@@ -328,5 +328,6 @@ int do_batch(const char *name, bool force,
 int parse_one_of(const char *msg, const char *realval, const char * const *list,
 		 size_t len, int *p_err);
 int parse_on_off(const char *msg, const char *realval, int *p_err);
+void print_on_off_bool(FILE *fp, const char *flag, bool val);
 
 #endif /* __UTILS_H__ */
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 10 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

#26 lore raw
In reply to Petr Machata
Quoted 4 lines · Petr Machata

The Linux DCB interface allows configuration of a broad range of hardware-specific attributes, such as TC scheduling, flow control, per-port buffer configuration, TC rate, etc.

...

Quoted 38 lines

The patchset proceeds as follows:

- Many tools in iproute2 have an option to work in batch mode, where the commands to run are given in a file. The code to handle batching is largely the same independent of the tool in question. In patch #1, add a helper to handle the batching, and migrate individual tools to use it.

- A number of configuration options come in a form of an on-off switch. This in turn can be considered a special case of parsing one of a given set of strings. In patch #2, extract helpers to parse one of a number of strings, on top of which build an on-off parser.

Currently each tool open-codes the logic to parse the on-off toggle. A future patch set will migrate instances of this code over to the new helpers.

- The on/off toggles from previous list item sometimes need to be dumped. While in the FP output, one typically wishes to maintain consistency with the command line and show actual strings, "on" and "off", in JSON output one would rather use booleans. This logic is somewhat annoying to have to open-code time and again. Therefore in patch #3, add a helper to do just that.

- The DCB tool is built on top of libmnl. Several routines will be basically the same in DCB as they are currently in devlink. In patches #4-#6, extract them to a new module, mnl_utils, for easy reuse.

- Much of DCB is built around arrays. A syntax similar to the iplink_vlan's ingress-qos-map / egress-qos-map is very handy for describing changes done to such arrays. Therefore in patch #7, extract a helper, parse_mapping(), which manages parsing of key-value arrays. In patch #8, fix a buglet in the helper, and in patch #9, extend it to allow setting of all array elements in one go.

- In patch #10, add a skeleton of "dcb", which contains common helpers and dispatches to subtools for handling of individual objects. The skeleton is empty as of this patch.

Quoted 8 lines

In patch #11, add "dcb_ets", a module for handling of specifically DCB ETS objects.

The intention is to gradually add handlers for at least PFC, APP, peer configuration, buffers and rates.

[1] https://github.com/Mellanox/mlnx-tools/tree/master/ofed_scripts

overall this looks really good to me. Thanks for taking the time to do the refactoring.

#7 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to Stephen Hemminger
Quoted 15 lines · Stephen Hemminger

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 7 lines · On Fri, 30 Oct 2020 13:29:50 +0100 Petr Machata <[email protected]> wrote:
lib/utils.c +7 -0
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

Please use print_string(PRINT_FP for non json case.
I am use fprintf(fp as indicator for code that has not been already converted to JSON.

Quoted 2 lines

And passing the FILE *fp is also indication of old code. Original iproute2 passed it around but it was always stdout.

Ack, will fix.

#9 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to David Ahern
Quoted 7 lines · David Ahern
Quoted 6 lines · Petr Machata
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
Quoted 14 lines

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 11 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

print_on_off() is already used in the RDMA tool, and actually outputs "on" and "off", unlike this. So I chose this instead.

I could rename the RDMA one though -- it's used in two places, whereas this is used in about two dozen instances across the codebase.

#4 lore raw

Re: [PATCH iproute2-next v2 02/11] lib: Add parse_one_of(), parse_on_off()

In reply to David Ahern
Quoted 13 lines · David Ahern

lib/utils.c int do_batch(const char *name, bool force, · line 1765

Quoted 7 lines · Petr Machata
lib/utils.c +6 -0
+int parse_on_off(const char *msg, const char *realval, int *p_err)
+{
+	static const char * const values_on_off[] = { "off", "on" };
+
+	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
+}

This has weird semantics to me. You have a buried array of strings and returning the index of the one that matches. Let's use a 'bool' return for parse_on_off that makes it clear that the string is 'off' = false or 'on' = true.

Agreed, it should return bool.

#10 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to Petr Machata
Quoted 25 lines · Petr Machata
Quoted 7 lines · David Ahern
Quoted 6 lines · Petr Machata
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
Quoted 14 lines

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 11 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

Quoted 6 lines

print_on_off() is already used in the RDMA tool, and actually outputs "on" and "off", unlike this. So I chose this instead.

I could rename the RDMA one though -- it's used in two places, whereas this is used in about two dozen instances across the codebase.

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

#11 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 29 lines · David Ahern
Quoted 28 lines · Petr Machata
Quoted 7 lines · David Ahern
Quoted 6 lines · Petr Machata
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
Quoted 14 lines

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 11 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

print_on_off() is already used in the RDMA tool, and actually outputs "on" and "off", unlike this. So I chose this instead.

Quoted 7 lines
Quoted 3 lines

I could rename the RDMA one though -- it's used in two places, whereas this is used in about two dozen instances across the codebase.

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

However I don't understand why print_on_off_bool() is implemented in utils.c and not in lib/json_print.c that properly handles JSON context, provide colorized output and doesn't require to supply FILE *fp.

Thanks

#12 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 3 lines · Leon Romanovsky

However I don't understand why print_on_off_bool() is implemented in utils.c and not in lib/json_print.c that properly handles JSON context, provide colorized output and doesn't require to supply FILE *fp.

Good point, I missed that earlier.

#13 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

In reply to Leon Romanovsky
Quoted 34 lines · Leon Romanovsky
Quoted 33 lines · David Ahern
Quoted 31 lines · Petr Machata
Quoted 7 lines · David Ahern
Quoted 6 lines · Petr Machata
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
Quoted 14 lines

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 11 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

print_on_off() is already used in the RDMA tool, and actually outputs "on" and "off", unlike this. So I chose this instead.

I could rename the RDMA one though -- it's used in two places, whereas this is used in about two dozen instances across the codebase.

Quoted 7 lines
Quoted 3 lines

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

The RDMAtool uses literal "on" and "off" as values in JSON, not booleans. Moving over to print_on_off_bool() would be a breaking change, which is problematic especially in JSON output.

Quoted 2 lines

However I don't understand why print_on_off_bool() is implemented in utils.c and not in lib/json_print.c that properly handles JSON context,

There's a whole lot of print_X functions for printing non-fundamental data types in utils.c. Seemed obvious to put it there. I can move it to json_print.c, no problem.

I think the current function does handle JSON context, what else do you have in mind?

Quoted 1 line

provide colorized output and doesn't require to supply FILE *fp.

Stephen Hemminger already pointed out the FILE *fp bit, I'll be removing it.

#14 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 41 lines · Petr Machata
Quoted 38 lines · Leon Romanovsky
Quoted 36 lines · David Ahern
Quoted 25 lines · Petr Machata
Quoted 7 lines · David Ahern
Quoted 6 lines · Petr Machata
lib/utils.c +0 -0
index 930877ae0f0d..8deec86ecbcd 100644
@@ -1763,3 +1763,11 @@ int parse_on_off(const char *msg, const char *realval, int *p_err)
 
Quoted 14 lines

lib/utils.c int parse_on_off(const char *msg, const char *realval, int *p_err) · line 1773

Quoted 11 lines
lib/utils.c +8 -0
 	return parse_one_of(msg, realval, values_on_off, ARRAY_SIZE(values_on_off), p_err);
 }
+
+void print_on_off_bool(FILE *fp, const char *flag, bool val)
+{
+	if (is_json_context())
+		print_bool(PRINT_JSON, flag, NULL, val);
+	else
+		fprintf(fp, "%s %s ", flag, val ? "on" : "off");
+}

I think print_on_off should be fine and aligns with parse_on_off once it returns a bool.

Quoted 6 lines

print_on_off() is already used in the RDMA tool, and actually outputs "on" and "off", unlike this. So I chose this instead.

I could rename the RDMA one though -- it's used in two places, whereas this is used in about two dozen instances across the codebase.

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

Quoted 7 lines
Quoted 3 lines

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

The RDMAtool uses literal "on" and "off" as values in JSON, not booleans. Moving over to print_on_off_bool() would be a breaking change, which is problematic especially in JSON output.

Nothing prohibits us from adding extra parameter to this new function/json logic/json type that will control JSON behavior. Personally, I don't think that json and stdout outputs should be different, e.g. 1/0 for the json and on/off for the stdout.

Quoted 7 lines
Quoted 3 lines

However I don't understand why print_on_off_bool() is implemented in utils.c and not in lib/json_print.c that properly handles JSON context,

There's a whole lot of print_X functions for printing non-fundamental data types in utils.c. Seemed obvious to put it there. I can move it to json_print.c, no problem.

It looks like it is worth to cleanup util.c and make sure that all prints are gathered in one place. Out of the scope for this series.

Quoted 3 lines

I think the current function does handle JSON context, what else do you have in mind?a

It handles, but does it twice, first time for is_json_context() and second time inside print_bool.

Quoted 5 lines
Quoted 2 lines

provide colorized output and doesn't require to supply FILE *fp.

Stephen Hemminger already pointed out the FILE *fp bit, I'll be removing it.

#15 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 14 lines · Leon Romanovsky
Quoted 13 lines · Petr Machata
Quoted 9 lines · Leon Romanovsky
Quoted 3 lines · David Ahern

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

Quoted 8 lines
Quoted 3 lines

The RDMAtool uses literal "on" and "off" as values in JSON, not booleans. Moving over to print_on_off_bool() would be a breaking change, which is problematic especially in JSON output.

Nothing prohibits us from adding extra parameter to this new function/json logic/json type that will control JSON behavior. Personally, I don't think that json and stdout outputs should be different, e.g. 1/0 for the json and on/off for the stdout.

Emitting on/off in JSON as true booleans (true / false, not 1 / 0) does make sense. It's programmatically-consumed interface, the values should be of the right type.

On the other hand, having a FP output use literal "on" and "off" makes sense as well. It's an obvious reference to the command line, you can actually cut'n'paste it back to shell and it will do the right thing.

Many places in iproute2 do do this dual output, and ideally all new instances would behave this way as well. So no toggles, please.

Quoted 5 lines
Quoted 2 lines

I think the current function does handle JSON context, what else do you have in mind?

It handles, but does it twice, first time for is_json_context() and second time inside print_bool.

Gotcha.

#16 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 21 lines · Petr Machata
Quoted 18 lines · Leon Romanovsky
Quoted 16 lines · Petr Machata
Quoted 9 lines · Leon Romanovsky
Quoted 3 lines · David Ahern

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

The RDMAtool uses literal "on" and "off" as values in JSON, not booleans. Moving over to print_on_off_bool() would be a breaking change, which is problematic especially in JSON output.

Quoted 8 lines
Quoted 4 lines

Nothing prohibits us from adding extra parameter to this new function/json logic/json type that will control JSON behavior. Personally, I don't think that json and stdout outputs should be different, e.g. 1/0 for the json and on/off for the stdout.

Emitting on/off in JSON as true booleans (true / false, not 1 / 0) does make sense. It's programmatically-consumed interface, the values should be of the right type.

As long as you don't need to use those fields to "set .." after that.

Quoted 4 lines

On the other hand, having a FP output use literal "on" and "off" makes sense as well. It's an obvious reference to the command line, you can actually cut'n'paste it back to shell and it will do the right thing.

Maybe it is not so bad to change RDMAtool to general function, this on/of print is not widely use yet, just need to decide what is the right one.

Quoted 3 lines

Many places in iproute2 do do this dual output, and ideally all new instances would behave this way as well. So no toggles, please.

Good example why all utilities in iproute2 are better to use same input/output code and any attempt to make custom variants should be banned.

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

I think the current function does handle JSON context, what else do you have in mind?

It handles, but does it twice, first time for is_json_context() and second time inside print_bool.

Gotcha.

#17 lore raw

Re: [PATCH iproute2-next v2 03/11] lib: utils: Add print_on_off_bool()

Quoted 31 lines · Leon Romanovsky
Quoted 21 lines · Petr Machata
Quoted 18 lines · Leon Romanovsky
Quoted 16 lines · Petr Machata
Quoted 9 lines · Leon Romanovsky
Quoted 3 lines · David Ahern

yes, the rdma utils are using generic function names. The rdma version should be renamed; perhaps rd_print_on_off. That seems to be once common prefix. Added Leon.

I made fast experiment and the output for the code proposed here and existed in the RDMAtool - result the same. So the good thing will be to delete the function from the RDMA after print_on_off_bool() will be improved.

The RDMAtool uses literal "on" and "off" as values in JSON, not booleans. Moving over to print_on_off_bool() would be a breaking change, which is problematic especially in JSON output.

Quoted 8 lines
Quoted 4 lines

Nothing prohibits us from adding extra parameter to this new function/json logic/json type that will control JSON behavior. Personally, I don't think that json and stdout outputs should be different, e.g. 1/0 for the json and on/off for the stdout.

Emitting on/off in JSON as true booleans (true / false, not 1 / 0) does make sense. It's programmatically-consumed interface, the values should be of the right type.

Quoted 8 lines

As long as you don't need to use those fields to "set .." after that.

Quoted 4 lines

On the other hand, having a FP output use literal "on" and "off" makes sense as well. It's an obvious reference to the command line, you can actually cut'n'paste it back to shell and it will do the right thing.

Maybe it is not so bad to change RDMAtool to general function, this on/of print is not widely use yet

OK, if you think the API breakage is acceptable, I'll roll this into the patchset.

Quoted 1 line

just need to decide what is the right one.

Yeah, it's kinda fuzzy. Where JSON has an obvious type to use, use it:
arrays should probably be arrays, numbers should probably be numbers. I'm not so sure about enums, but I guess represent them as strings? As numbers they will not be more meaningful or easy to consume, and it does not make sense to do arithmetic on enums.

On/off toggles could be considered enums. But they are also booleans. Representing on as true and off as false is straightforward and from this perspective booleans are the obvious type to use.

Quoted 6 lines
Quoted 2 lines

Many places in iproute2 do do this dual output, and ideally all new instances would behave this way as well. So no toggles, please.

Good example why all utilities in iproute2 are better to use same input/output code and any attempt to make custom variants should be banned.

Yes, I have a clean-up patch that converts these custom on/off helpers to the new central one. I'll send this together with other refactorings of this sort after this patch set.