ip: Support filter links with no VF info

Commit Message

renmingshuai April 10, 2024, 2:51 a.m. UTC

Kernel has add IFLA_EXT_MASK attribute for indicating that certain
extended ifinfo values are requested by the user application. The ip
link show cmd always request VFs extended ifinfo.

In this case, RTM_GETLINK for greater than about 220 VFs truncates
IFLA_VFINFO_LIST due to the maximum reach of nlattr's nla_len being
exceeded. As a result, ip link show command only show the truncated
VFs info sucn as:

    #ip link show dev eth0
    1: eth0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 ...
        link/ether ...
        vf 0     link/ether 00:00:00:00:00:00 brd ff:ff:ff:ff:ff:ff ...
    Truncated VF list: eth0

This patch add novf to support filter links with no VF info:
ip link show novf

v1->v2:
  - use a one word option instead of an option with on/off.
  - fix the issue that break the changes made for the link filter
    already done for VF's.

Signed-off-by: Mingshuai Ren <[e-mail]>
Reviewed-by: Stephen Hemminger <[e-mail]>
---
 ip/ip_common.h        |  1 +
 ip/ipaddress.c        | 15 ++++++++++-----
 ip/iplink.c           |  2 +-
 man/man8/ip-link.8.in |  7 ++++++-
 4 files changed, 18 insertions(+), 7 deletions(-)

Comments

David Ahern April 10, 2024, 2:47 p.m. UTC | #1

On 4/9/24 8:51 PM, renmingshuai wrote:
> @@ -2139,6 +2141,7 @@ static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
>  	ipaddr_reset_filter(oneline, 0);
>  	filter.showqueue = 1;
>  	filter.family = preferred_family;
> +	filter.vfinfo = 0;
>  
>  	if (action == IPADD_FLUSH) {
>  		if (argc <= 0) {
> @@ -2221,6 +2224,8 @@ static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
>  				invarg("\"proto\" value is invalid\n", *argv);
>  			filter.have_proto = true;
>  			filter.proto = proto;
> +		} else if (strcmp(*argv, "novf") == 0) {
> +			filter.vfinfo = -1;
>  		} else {
>  			if (strcmp(*argv, "dev") == 0)
>  				NEXT_ARG();

The reverse logic is how other filters work. Meaning vfinfo is set, add
the RTEXT_FILTER_VF flag (default for backwards compatibility). From
there, "novf" set vfinfo to 0 and flag is not added.

renmingshuai April 11, 2024, 2:57 a.m. UTC | #2

> > @@ -2139,6 +2141,7 @@ static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
> >  	ipaddr_reset_filter(oneline, 0);
> >  	filter.showqueue = 1;
> >  	filter.family = preferred_family;
> > +	filter.vfinfo = 0;
> >  
> >  	if (action == IPADD_FLUSH) {
> >  		if (argc <= 0) {
> > @@ -2221,6 +2224,8 @@ static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
> >  				invarg("\"proto\" value is invalid\n", *argv);
> >  			filter.have_proto = true;
> >  			filter.proto = proto;
> > +		} else if (strcmp(*argv, "novf") == 0) {
> > +			filter.vfinfo = -1;
> >  		} else {
> >  			if (strcmp(*argv, "dev") == 0)
> >  				NEXT_ARG();
> 
> The reverse logic is how other filters work. Meaning vfinfo is set, add
> the RTEXT_FILTER_VF flag (default for backwards compatibility). From
> there, "novf" set vfinfo to 0 and flag is not added.

As you suggested, I've modified it and submitted a new patch.

Patch

diff --git a/ip/ip_common.h b/ip/ip_common.h
index b65c2b41..d3645a2c 100644
--- a/ip/ip_common.h
+++ b/ip/ip_common.h
@@ -30,6 +30,7 @@  struct link_filter {
 	int target_nsid;
 	bool have_proto;
 	int proto;
+	int vfinfo;
 };
 
 const char *get_ip_lib_dir(void);
diff --git a/ip/ipaddress.c b/ip/ipaddress.c
index e536912f..e7e5ce76 100644
--- a/ip/ipaddress.c
+++ b/ip/ipaddress.c
@@ -2029,10 +2029,11 @@  static int ipaddr_flush(void)
 
 static int iplink_filter_req(struct nlmsghdr *nlh, int reqlen)
 {
-	__u32 filt_mask;
+	__u32 filt_mask = 0;
 	int err;
 
-	filt_mask = RTEXT_FILTER_VF;
+	if (!filter.vfinfo)
+		filt_mask |= RTEXT_FILTER_VF;
 	if (!show_stats)
 		filt_mask |= RTEXT_FILTER_SKIP_STATS;
 	err = addattr32(nlh, reqlen, IFLA_EXT_MASK, filt_mask);
@@ -2070,12 +2071,13 @@  static int ipaddr_link_get(int index, struct nlmsg_chain *linfo)
 		.i.ifi_family = filter.family,
 		.i.ifi_index = index,
 	};
-	__u32 filt_mask = RTEXT_FILTER_VF;
+	__u32 filt_mask = 0;
 	struct nlmsghdr *answer;
 
+	if (!filter.vfinfo)
+		filt_mask |= RTEXT_FILTER_VF;
 	if (!show_stats)
 		filt_mask |= RTEXT_FILTER_SKIP_STATS;
-
 	addattr32(&req.n, sizeof(req), IFLA_EXT_MASK, filt_mask);
 
 	if (rtnl_talk(&rth, &req.n, &answer) < 0) {
@@ -2139,6 +2141,7 @@  static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
 	ipaddr_reset_filter(oneline, 0);
 	filter.showqueue = 1;
 	filter.family = preferred_family;
+	filter.vfinfo = 0;
 
 	if (action == IPADD_FLUSH) {
 		if (argc <= 0) {
@@ -2221,6 +2224,8 @@  static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
 				invarg("\"proto\" value is invalid\n", *argv);
 			filter.have_proto = true;
 			filter.proto = proto;
+		} else if (strcmp(*argv, "novf") == 0) {
+			filter.vfinfo = -1;
 		} else {
 			if (strcmp(*argv, "dev") == 0)
 				NEXT_ARG();
@@ -2274,7 +2279,7 @@  static int ipaddr_list_flush_or_save(int argc, char **argv, int action)
 	 * the link device
 	 */
 	if (filter_dev && filter.group == -1 && do_link == 1) {
-		if (iplink_get(filter_dev, RTEXT_FILTER_VF) < 0) {
+		if (iplink_get(filter_dev, filter.vfinfo < 0 ? 0 : RTEXT_FILTER_VF) < 0) {
 			perror("Cannot send link get request");
 			delete_json_obj();
 			exit(1);
diff --git a/ip/iplink.c b/ip/iplink.c
index 95314af5..1bb4a998 100644
--- a/ip/iplink.c
+++ b/ip/iplink.c
@@ -111,7 +111,7 @@  void iplink_usage(void)
 		"		[ gro_max_size BYTES ] [ gro_ipv4_max_size BYTES ]\n"
 		"\n"
 		"	ip link show [ DEVICE | group GROUP ] [up] [master DEV] [vrf NAME] [type TYPE]\n"
-		"		[nomaster]\n"
+		"		[nomaster] [ novf ]\n"
 		"\n"
 		"	ip link xstats type TYPE [ ARGS ]\n"
 		"\n"
diff --git a/man/man8/ip-link.8.in b/man/man8/ip-link.8.in
index 31e2d7f0..066ad874 100644
--- a/man/man8/ip-link.8.in
+++ b/man/man8/ip-link.8.in
@@ -202,7 +202,8 @@  ip-link \- network device configuration
 .IR ETYPE " ] ["
 .B vrf
 .IR NAME " ] ["
-.BR nomaster " ]"
+.BR nomaster " ] ["
+.BR novf " ]"
 
 .ti -8
 .B ip link xstats
@@ -2898,6 +2899,10 @@  output.
 .B nomaster
 only show devices with no master
 
+.TP
+.B novf
+only show devices with no VF info
+
 .SS  ip link xstats - display extended statistics
 
 .TP