Skip to contentSkip to conversation

lib: Replace 16384 netlink buf size by macro

PATCH In review Vadim Kochan 3 participants 4 messages Feb 27, 2015 – Feb 28, 2015

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 lib: Replace 16384 netlink buf size by macro

From: Vadim Kochan <[email protected]>

Replaced hard coded 16384 netlink buffer size by NLBUF_SIZE, which can be overridden by -DNLBUF_SIZE.

---
v4:
Added "iproute2" in the subject

v3:
Fixed "overrided" -> overridden" in the
commit message : suggested by Sergei Shtylyov

v2:
Get rid of NL_BUF: suggested by Jiri Pirko

 include/libnetlink.h | 4 ++++
 ip/iplink.c          | 2 +-
 lib/libnetlink.c     | 8 ++++----
 misc/ss.c            | 2 +-
 4 files changed, 10 insertions(+), 6 deletions(-)
include/libnetlink.h +4 -0
index 898275b..1ef7f1f 100644
@@ -170,5 +170,9 @@ extern int rtnl_from_file(FILE *, rtnl_filter_t handler,
  * messages from dump file */
 #define NLMSG_TSTAMP	15
 
+#ifndef NLBUF_SIZE
+#define NLBUF_SIZE 16384
+#endif
+
 #endif /* __LIBNETLINK_H__ */
 
ip/iplink.c +1 -1
index 5893ee4..db60afa 100644
@@ -770,7 +770,7 @@ int iplink_get(unsigned int flags, char *name, __u32 filt_mask)
 {
 	int len;
 	struct iplink_req req;
-	char answer[16384];
+	char answer[NLBUF_SIZE];
 
 	memset(&req, 0, sizeof(req));
 
lib/libnetlink.c +4 -4
index 77e07ef..666e277 100644
@@ -194,7 +194,7 @@ int rtnl_dump_filter_l(struct rtnl_handle *rth,
 		.msg_iov = &iov,
 		.msg_iovlen = 1,
 	};
-	char buf[16384];
+	char buf[NLBUF_SIZE];
 	int dump_intr = 0;
 
 	iov.iov_base = buf;
@@ -317,7 +317,7 @@ int rtnl_talk(struct rtnl_handle *rtnl, struct nlmsghdr *n, pid_t peer,
 		.msg_iov = &iov,
 		.msg_iovlen = 1,
 	};
-	char   buf[16384];
+	char buf[NLBUF_SIZE];
 
 	memset(&nladdr, 0, sizeof(nladdr));
 	nladdr.nl_family = AF_NETLINK;
@@ -432,7 +432,7 @@ int rtnl_listen(struct rtnl_handle *rtnl,
 		.msg_iov = &iov,
 		.msg_iovlen = 1,
 	};
-	char   buf[16384];
+	char buf[NLBUF_SIZE];
 
 	memset(&nladdr, 0, sizeof(nladdr));
 	nladdr.nl_family = AF_NETLINK;
@@ -498,7 +498,7 @@ int rtnl_from_file(FILE *rtnl, rtnl_filter_t handler,
 {
 	int status;
 	struct sockaddr_nl nladdr;
-	char   buf[16384];
+	char buf[NLBUF_SIZE];
 	struct nlmsghdr *h = (void*)buf;
 
 	memset(&nladdr, 0, sizeof(nladdr));
misc/ss.c +1 -1
index 21a4366..036ede8 100644
@@ -2222,7 +2222,7 @@ Exit:
 static int tcp_show_netlink_file(struct filter *f)
 {
 	FILE	*fp;
-	char	buf[16384];
+	char buf[NLBUF_SIZE];
 
 	if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
 		perror("fopen($TCPDIAG_FILE)");
Signature · 1 line
2.2.2
maintainer
#1 lore raw

From: Vadim Kochan <[email protected]>
Date: Fri, 27 Feb 2015 08:31:40 +0200

Quoted 6 lines

From: Vadim Kochan <[email protected]>

Replaced hard coded 16384 netlink buffer size by NLBUF_SIZE, which can be overridden by -DNLBUF_SIZE.

I would like to know what problem is solved by changing this to any other value?

And then I would like to know why other potential solutions to this problem were not considered, such as making the buffer size dynamic and run-time selectable, perhaps with a size choosen in a manner to avoid whatever problem you hit with the current size.

#2 lore raw
Quoted 10 lines · David Miller

From: Vadim Kochan <[email protected]>
Date: Fri, 27 Feb 2015 08:31:40 +0200

Quoted 6 lines

From: Vadim Kochan <[email protected]>

Replaced hard coded 16384 netlink buffer size by NLBUF_SIZE, which can be overridden by -DNLBUF_SIZE.

Quoted 8 lines

I would like to know what problem is solved by changing this to any other value?

And then I would like to know why other potential solutions to this problem were not considered, such as making the buffer size dynamic and run-time selectable, perhaps with a size choosen in a manner to avoid whatever problem you hit with the current size.

Its just for the case when it might be changed in the future then it can be easy found and changed. Just reject it if it looks useless. About dynamic changing I think it makes sense to do it because memset is used on the each netlink 16384 sized msg.