openwrt/package/network/services/ppp/patches/141-Expand-byte-count-statistics-to-64-bits-298.patch

From 81ad945630120cc1c27c8bb00503be42b76ff202 Mon Sep 17 00:00:00 2001
From: Jaco Kroon <jaco@uls.co.za>
Date: Thu, 13 Jan 2022 08:38:04 +0200
Subject: [PATCH] Expand byte count statistics to 64 bits (#298)

* Add Gigawords to radius packets where applicable.

IMPORTANT NOTE:  The ioctl() only supports 32-bit counters.  In order t
obtain 64-bit counters, these are now pulled in from sysfs (it's assumed
to be mounted on /sys which I'm assuming is standard).

It is unknown whether sysfs will be available everywhere, as such, keep
the ioctl() method in place, but attempt to detect wrap-overs.

If the sysfs mechanism fails, fail back to the ioctl().

Given maximum data rates, the intervals between calling this needs to be
such that no more than 4GB (2^32) bytes are sent or received in any
given interval.  Mostly important for radius plugin where data
accounting may be in effect.

Towards this, a timer interval on 25 seconds is set to force a ioctl()
poll irrespective of the rate of stats update calls.  This may be
important for especially radius that needs to provide interim-update
intervals, if the interim updates is too long and the counters could
wrap-over twice in a single interval.  At 25 seconds we should detect
all wraps up to an effective data rate of 1.37Gbps, which for my
purposes is adequate.

Possible downsides, 4 files are opened, read and closed every time
statistics is requested.  This results in 12 system calls every single
time statistics is required, compared to 1 for the ioctl.  Efficiency is
unknown, but as a rule of thumb fewer system calls are better, this is
however not a critical path in my opinion, so should not be a problem.
If required I can run a few benchmarks using gettimeofday() to measure
actual impact.

Signed-off-by: Jaco Kroon <jaco@uls.co.za>

* Use netlink if possible to obtain 64-bit stats.

This uses two system calls per round.

This should be preferred where available.  It seems the RTM_GETSTATS was
only added from 2016 some point (4.7.0 as per pali), which is in my
opinion old, but given experience with certain embedded systems does
need to be supported.

Signed-off-by: Jaco Kroon <jaco@uls.co.za>

Co-authored-by: Jaco Kroon <jaco@iewc.co.za>
---
 pppd/main.c                        |   5 +-
 pppd/plugins/radius/etc/dictionary |   2 +
 pppd/plugins/radius/radius.c       |  28 ++-
 pppd/plugins/radius/radiusclient.h |   2 +
 pppd/pppd.h                        |   9 +-
 pppd/sys-linux.c                   | 281 ++++++++++++++++++++++++++++-
 6 files changed, 313 insertions(+), 14 deletions(-)

--- a/pppd/main.c
+++ b/pppd/main.c
@@ -87,6 +87,7 @@
 #include <sys/socket.h>
 #include <netinet/in.h>
 #include <arpa/inet.h>
+#include <inttypes.h>
 
 #include "pppd.h"
 #include "magic.h"
@@ -1230,9 +1231,9 @@ update_link_stats(int u)
 
     slprintf(numbuf, sizeof(numbuf), "%u", link_connect_time);
     script_setenv("CONNECT_TIME", numbuf, 0);
-    slprintf(numbuf, sizeof(numbuf), "%u", link_stats.bytes_out);
+    snprintf(numbuf, sizeof(numbuf), "%" PRIu64, link_stats.bytes_out);
     script_setenv("BYTES_SENT", numbuf, 0);
-    slprintf(numbuf, sizeof(numbuf), "%u", link_stats.bytes_in);
+    snprintf(numbuf, sizeof(numbuf), "%" PRIu64, link_stats.bytes_in);
     script_setenv("BYTES_RCVD", numbuf, 0);
 }
 
--- a/pppd/plugins/radius/etc/dictionary
+++ b/pppd/plugins/radius/etc/dictionary
@@ -82,6 +82,8 @@ ATTRIBUTE	Acct-Session-Time	46	integer
 ATTRIBUTE	Acct-Input-Packets	47	integer
 ATTRIBUTE	Acct-Output-Packets	48	integer
 ATTRIBUTE	Acct-Terminate-Cause	49	integer
+ATTRIBUTE	Acct-Input-Gigawords	52	integer
+ATTRIBUTE	Acct-Output-Gigawords	53	integer
 ATTRIBUTE       Chap-Challenge          60      string
 ATTRIBUTE	NAS-Port-Type		61	integer
 ATTRIBUTE	Port-Limit		62	integer
--- a/pppd/plugins/radius/radius.c
+++ b/pppd/plugins/radius/radius.c
@@ -1020,12 +1020,22 @@ radius_acct_stop(void)
 	av_type = link_connect_time;
 	rc_avpair_add(&send, PW_ACCT_SESSION_TIME, &av_type, 0, VENDOR_NONE);
 
-	av_type = link_stats.bytes_out;
+	av_type = link_stats.bytes_out & 0xFFFFFFFF;
 	rc_avpair_add(&send, PW_ACCT_OUTPUT_OCTETS, &av_type, 0, VENDOR_NONE);
 
-	av_type = link_stats.bytes_in;
+	if (link_stats.bytes_out > 0xFFFFFFFF) {
+	    av_type = link_stats.bytes_out >> 32;
+	    rc_avpair_add(&send, PW_ACCT_OUTPUT_GIGAWORDS, &av_type, 0, VENDOR_NONE);
+	}
+
+	av_type = link_stats.bytes_in & 0xFFFFFFFF;
 	rc_avpair_add(&send, PW_ACCT_INPUT_OCTETS, &av_type, 0, VENDOR_NONE);
 
+	if (link_stats.bytes_in > 0xFFFFFFFF) {
+	    av_type = link_stats.bytes_in >> 32;
+	    rc_avpair_add(&send, PW_ACCT_INPUT_GIGAWORDS, &av_type, 0, VENDOR_NONE);
+	}
+
 	av_type = link_stats.pkts_out;
 	rc_avpair_add(&send, PW_ACCT_OUTPUT_PACKETS, &av_type, 0, VENDOR_NONE);
 
@@ -1172,12 +1182,22 @@ radius_acct_interim(void *ignored)
 	av_type = link_connect_time;
 	rc_avpair_add(&send, PW_ACCT_SESSION_TIME, &av_type, 0, VENDOR_NONE);
 
-	av_type = link_stats.bytes_out;
+	av_type = link_stats.bytes_out & 0xFFFFFFFF;
 	rc_avpair_add(&send, PW_ACCT_OUTPUT_OCTETS, &av_type, 0, VENDOR_NONE);
 
-	av_type = link_stats.bytes_in;
+	if (link_stats.bytes_out > 0xFFFFFFFF) {
+	    av_type = link_stats.bytes_out >> 32;
+	    rc_avpair_add(&send, PW_ACCT_OUTPUT_GIGAWORDS, &av_type, 0, VENDOR_NONE);
+	}
+
+	av_type = link_stats.bytes_in & 0xFFFFFFFF;
 	rc_avpair_add(&send, PW_ACCT_INPUT_OCTETS, &av_type, 0, VENDOR_NONE);
 
+	if (link_stats.bytes_in > 0xFFFFFFFF) {
+	    av_type = link_stats.bytes_in >> 32;
+	    rc_avpair_add(&send, PW_ACCT_INPUT_GIGAWORDS, &av_type, 0, VENDOR_NONE);
+	}
+
 	av_type = link_stats.pkts_out;
 	rc_avpair_add(&send, PW_ACCT_OUTPUT_PACKETS, &av_type, 0, VENDOR_NONE);
 
--- a/pppd/plugins/radius/radiusclient.h
+++ b/pppd/plugins/radius/radiusclient.h
@@ -184,6 +184,8 @@ typedef struct pw_auth_hdr
 #define PW_ACCT_LINK_COUNT		51	/* integer */
 
 /* From RFC 2869 */
+#define PW_ACCT_INPUT_GIGAWORDS         52	/* integer */
+#define PW_ACCT_OUTPUT_GIGAWORDS        53	/* integer */
 #define PW_ACCT_INTERIM_INTERVAL        85	/* integer */
 
 /*	Merit Experimental Extensions */
--- a/pppd/pppd.h
+++ b/pppd/pppd.h
@@ -53,6 +53,7 @@
 #include <stdlib.h>		/* for encrypt */
 #include <unistd.h>		/* for setkey */
 #include <stdarg.h>
+#include <stdint.h>
 #include <limits.h>		/* for NGROUPS_MAX */
 #include <sys/param.h>		/* for MAXPATHLEN and BSD4_4, if defined */
 #include <sys/types.h>		/* for u_int32_t, if defined */
@@ -173,8 +174,8 @@ struct permitted_ip {
  * pppd needs.
  */
 struct pppd_stats {
-    unsigned int	bytes_in;
-    unsigned int	bytes_out;
+    uint64_t		bytes_in;
+    uint64_t		bytes_out;
     unsigned int	pkts_in;
     unsigned int	pkts_out;
 };
@@ -347,7 +348,7 @@ extern char *max_tls_version;
 extern unsigned int maxoctets;	     /* Maximum octetes per session (in bytes) */
 extern int       maxoctets_dir;      /* Direction :
 				      0 - in+out (default)
-				      1 - in 
+				      1 - in
 				      2 - out
 				      3 - max(in,out) */
 extern int       maxoctets_timeout;  /* Timeout for check of octets limit */
@@ -356,7 +357,7 @@ extern int       maxoctets_timeout;  /*
 #define PPP_OCTETS_DIRECTION_OUT        2
 #define PPP_OCTETS_DIRECTION_MAXOVERAL  3
 /* same as previos, but little different on RADIUS side */
-#define PPP_OCTETS_DIRECTION_MAXSESSION 4	
+#define PPP_OCTETS_DIRECTION_MAXSESSION 4
 #endif
 
 #ifdef PPP_FILTER
--- a/pppd/sys-linux.c
+++ b/pppd/sys-linux.c
@@ -79,6 +79,7 @@
 #include <sys/sysmacros.h>
 
 #include <errno.h>
+#include <stddef.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <syslog.h>
@@ -92,6 +93,7 @@
 #include <ctype.h>
 #include <termios.h>
 #include <unistd.h>
+#include <limits.h>
 
 /* This is in netdevice.h. However, this compile will fail miserably if
    you attempt to include netdevice.h because it has so many references
@@ -121,9 +123,19 @@
 #include <linux/ppp_defs.h>
 #include <linux/if_ppp.h>
 
-#ifdef INET6
 #include <linux/netlink.h>
 #include <linux/rtnetlink.h>
+#include <linux/if_link.h>
+/* Attempt at retaining compile-support with older than 4.7 kernels, or kernels
+ * where RTM_NEWSTATS isn't defined for whatever reason.
+ */
+#ifndef RTM_NEWSTATS
+#define RTM_NEWSTATS 92
+#define RTM_GETSTATS 94
+#define IFLA_STATS_LINK_64 1
+#endif
+
+#ifdef INET6
 #include <linux/if_addr.h>
 /* glibc versions prior to 2.24 do not define SOL_NETLINK */
 #ifndef SOL_NETLINK
@@ -1407,11 +1419,17 @@ get_idle_time(int u, struct ppp_idle *ip
 
 /********************************************************************
  *
- * get_ppp_stats - return statistics for the link.
+ * get_ppp_stats_iocl - return statistics for the link, using the ioctl() method,
+ * this only supports 32-bit counters, so need to count the wraps.
  */
-int
-get_ppp_stats(int u, struct pppd_stats *stats)
+static int
+get_ppp_stats_ioctl(int u, struct pppd_stats *stats)
 {
+    static u_int32_t previbytes = 0;
+    static u_int32_t prevobytes = 0;
+    static u_int32_t iwraps = 0;
+    static u_int32_t owraps = 0;
+
     struct ifpppstatsreq req;
 
     memset (&req, 0, sizeof (req));
@@ -1426,7 +1444,262 @@ get_ppp_stats(int u, struct pppd_stats *
     stats->bytes_out = req.stats.p.ppp_obytes;
     stats->pkts_in = req.stats.p.ppp_ipackets;
     stats->pkts_out = req.stats.p.ppp_opackets;
+
+    if (stats->bytes_in < previbytes)
+	++iwraps;
+    if (stats->bytes_out < prevobytes)
+	++owraps;
+
+    previbytes = stats->bytes_in;
+    prevobytes = stats->bytes_out;
+
+    stats->bytes_in += (uint64_t)iwraps << 32;
+    stats->bytes_out += (uint64_t)owraps << 32;
+
+    return 1;
+}
+
+/********************************************************************
+ * get_ppp_stats_rtnetlink - return statistics for the link, using rtnetlink
+ * This provides native 64-bit counters.
+ */
+static int
+get_ppp_stats_rtnetlink(int u, struct pppd_stats *stats)
+{
+    static int rtnl_fd = -1;
+
+    struct sockaddr_nl nladdr;
+    struct {
+        struct nlmsghdr nlh;
+        struct if_stats_msg ifsm;
+    } nlreq;
+    struct nlresp {
+        struct nlmsghdr nlh;
+	union {
+	    struct {
+		struct nlmsgerr nlerr;
+		char __end_err[0];
+	    };
+	    struct {
+		struct rtmsg rth;
+		struct  {
+		    /* We only case about these first fields from rtnl_link_stats64 */
+		    uint64_t rx_packets;
+		    uint64_t tx_packets;
+		    uint64_t rx_bytes;
+		    uint64_t tx_bytes;
+		} stats;
+		char __end_stats[0];
+	    };
+	};
+    } nlresp;
+    ssize_t nlresplen;
+    struct iovec iov;
+    struct msghdr msg;
+
+    memset(&nladdr, 0, sizeof(nladdr));
+    nladdr.nl_family = AF_NETLINK;
+
+    if (rtnl_fd < 0) {
+	rtnl_fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
+	if (rtnl_fd < 0) {
+	    error("get_ppp_stats_rtnetlink: error creating NETLINK socket: %m (line %d)", __LINE__);
+	    return 0;
+	}
+
+	if (bind(rtnl_fd, (struct sockaddr *)&nladdr, sizeof(nladdr)) < 0) {
+	    error("get_ppp_stats_rtnetlink: bind(AF_NETLINK): %m (line %d)", __LINE__);
+	    goto err;
+	}
+    }
+
+    memset(&nlreq, 0, sizeof(nlreq));
+    nlreq.nlh.nlmsg_len = sizeof(nlreq);
+    nlreq.nlh.nlmsg_type = RTM_GETSTATS;
+    nlreq.nlh.nlmsg_flags = NLM_F_REQUEST;
+
+    nlreq.ifsm.ifindex = if_nametoindex(ifname);
+    nlreq.ifsm.filter_mask = IFLA_STATS_LINK_64;
+
+    memset(&iov, 0, sizeof(iov));
+    iov.iov_base = &nlreq;
+    iov.iov_len = sizeof(nlreq);
+
+    memset(&msg, 0, sizeof(msg));
+    msg.msg_name = &nladdr;
+    msg.msg_namelen = sizeof(nladdr);
+    msg.msg_iov = &iov;
+    msg.msg_iovlen = 1;
+
+    if (sendmsg(rtnl_fd, &msg, 0) < 0) {
+        error("get_ppp_stats_rtnetlink: sendmsg(RTM_GETSTATS): %m (line %d)", __LINE__);
+	goto err;
+    }
+
+    /* We just need to repoint to IOV ... everything else stays the same */
+    iov.iov_base = &nlresp;
+    iov.iov_len = sizeof(nlresp);
+
+    nlresplen = recvmsg(rtnl_fd, &msg, 0);
+
+    if (nlresplen < 0) {
+        error("get_ppp_stats_rtnetlink: recvmsg(RTM_GETSTATS): %m (line %d)", __LINE__);
+	goto err;
+    }
+
+    if (nlresplen < sizeof(nlresp.nlh)) {
+	error("get_ppp_stats_rtnetlink: Netlink response message was incomplete (line %d)", __LINE__);
+	goto err;
+    }
+
+    if (nlresp.nlh.nlmsg_type == NLMSG_ERROR) {
+	if (nlresplen < offsetof(struct nlresp, __end_err)) {
+	    if (kernel_version >= KVERSION(4,7,0))
+		error("get_ppp_stats_rtnetlink: Netlink responded with error: %s (line %d)", strerror(-nlresp.nlerr.error), __LINE__);
+	} else {
+	    error("get_ppp_stats_rtnetlink: Netlink responded with an error message, but the nlmsgerr structure is incomplete (line %d).",
+		    __LINE__);
+	}
+	goto err;
+    }
+
+    if (nlresp.nlh.nlmsg_type != RTM_NEWSTATS) {
+	error("get_ppp_stats_rtnetlink: Expected RTM_NEWSTATS response, found something else (mlmsg_type %d, line %d)",
+		nlresp.nlh.nlmsg_type, __LINE__);
+	goto err;
+    }
+
+    if (nlresplen < offsetof(struct nlresp, __end_stats)) {
+	error("get_ppp_stats_rtnetlink: Obtained an insufficiently sized rtnl_link_stats64 struct from the kernel (line %d).", __LINE__);
+	goto err;
+    }
+
+    stats->bytes_in  = nlresp.stats.rx_bytes;
+    stats->bytes_out = nlresp.stats.tx_bytes;
+    stats->pkts_in   = nlresp.stats.rx_packets;
+    stats->pkts_out  = nlresp.stats.tx_packets;
+
     return 1;
+err:
+    close(rtnl_fd);
+    rtnl_fd = -1;
+    return 0;
+}
+
+/********************************************************************
+ * get_ppp_stats_sysfs - return statistics for the link, using the files in sysfs,
+ * this provides native 64-bit counters.
+ */
+static int
+get_ppp_stats_sysfs(int u, struct pppd_stats *stats)
+{
+    char fname[PATH_MAX+1];
+    char buf[21], *err; /* 2^64 < 10^20 */
+    int blen, fd, rlen;
+    unsigned long long val;
+
+    struct {
+	const char* fname;
+	void* ptr;
+	unsigned size;
+    } slist[] = {
+#define statfield(fn, field)	{ .fname = #fn, .ptr = &stats->field, .size = sizeof(stats->field) }
+	statfield(rx_bytes, bytes_in),
+	statfield(tx_bytes, bytes_out),
+	statfield(rx_packets, pkts_in),
+	statfield(tx_packets, pkts_out),
+#undef statfield
+    };
+
+    blen = snprintf(fname, sizeof(fname), "/sys/class/net/%s/statistics/", ifname);
+    if (blen >= sizeof(fname))
+	return 0; /* ifname max 15, so this should be impossible */
+
+    for (int i = 0; i < sizeof(slist) / sizeof(*slist); ++i) {
+	if (snprintf(fname + blen, sizeof(fname) - blen, "%s", slist[i].fname) >= sizeof(fname) - blen) {
+	    fname[blen] = 0;
+	    error("sysfs stats: filename %s/%s overflowed PATH_MAX", fname, slist[i].fname);
+	    return 0;
+	}
+
+	fd = open(fname, O_RDONLY);
+	if (fd < 0) {
+	    error("%s: %m", fname);
+	    return 0;
+	}
+
+	rlen = read(fd, buf, sizeof(buf) - 1);
+	close(fd);
+	if (rlen < 0) {
+	    error("%s: %m", fname);
+	    return 0;
+	}
+	/* trim trailing \n if present */
+	while (rlen > 0 && buf[rlen-1] == '\n')
+	    rlen--;
+	buf[rlen] = 0;
+
+	errno = 0;
+	val = strtoull(buf, &err, 10);
+	if (*buf < '0' || *buf > '9' || errno != 0 || *err) {
+	    error("string to number conversion error converting %s (from %s) for remaining string %s%s%s",
+		    buf, fname, err, errno ? ": " : "", errno ? strerror(errno) : "");
+	    return 0;
+	}
+	switch (slist[i].size) {
+#define stattype(type)	case sizeof(type): *(type*)slist[i].ptr = (type)val; break
+	    stattype(uint64_t);
+	    stattype(uint32_t);
+	    stattype(uint16_t);
+	    stattype(uint8_t);
+#undef stattype
+	default:
+	    error("Don't know how to store stats for %s of size %u", slist[i].fname, slist[i].size);
+	    return 0;
+	}
+    }
+
+    return 1;
+}
+
+/********************************************************************
+ * Periodic timer function to be used to keep stats up to date in case of ioctl
+ * polling.
+ *
+ * Given the 25s interval this should be fine up to data rates of 1.37Gbps.
+ * If you do change the timer, remember to also bring the get_ppp_stats (which
+ * sets up the initial trigger) as well.
+ */
+static void
+ppp_stats_poller(void* u)
+{
+    struct pppd_stats dummy;
+    get_ppp_stats_ioctl((long)u, &dummy);
+    TIMEOUT(ppp_stats_poller, u, 25);
+}
+
+/********************************************************************
+ * get_ppp_stats - return statistics for the link.
+ */
+int get_ppp_stats(int u, struct pppd_stats *stats)
+{
+    static int (*func)(int, struct pppd_stats*) = NULL;
+
+    if (!func) {
+	if (get_ppp_stats_rtnetlink(u, stats)) {
+	    func = get_ppp_stats_rtnetlink;
+	    return 1;
+	}
+	if (get_ppp_stats_sysfs(u, stats)) {
+	    func = get_ppp_stats_sysfs;
+	    return 1;
+	}
+	warn("statistics falling back to ioctl which only supports 32-bit counters");
+	func = get_ppp_stats_ioctl;
+	TIMEOUT(ppp_stats_poller, (void*)(long)u, 25);
+    }
+
+    return func(u, stats);
 }
 
 /********************************************************************