I got tired of getting 15% packet loss [1] when doing
lots of DNS lookups on my ADSL link...
And that was even when limiting outgoing DNS traffic
to 200 Kbit/s (ADSL modem upstream speed is 512 Kbit/s).
I had used tcatm patch with 2.6.16 kernel and I was
happy with it.
Now I patched Linux kernel 2.6.22 and iproute-2.6.22
for tcatm. Seems to work (TM). Only HTB tested.
Now I get 0% packet loss when doing lots of DNS queries
(270pps) and DNS traffic is limited with
HTB/ESFQ to 504Kbit/s. I used "tc class add ... atm overhead 20"
but I am not sure is it really 20 (Sonera in Finland).
Without tcatm I had to have there 420Kbit and it still sucked.
I keep patches at
http://safari.iki.fi/tcatm/
I read one thread from this year where there were objections
about tcatm's some parts and then discussion petered out...
I and probably Russell Stuart would probably like to get
these patches fixed so that everyone is pleased
and these can be incorporated into kernel some year,
because I believe ADSL is very popular nowadays ( =) )
and people would probably like if traffic control was
actually usable for them ( =) )...
[1] ping -A 80.223.96.1
--
Do what you love because life is too short for anything else.
#
# include/linux/pkt_sched.h | 5 +++--
# include/net/sch_generic.h | 15 +++++++++++++++
# net/sched/act_police.c | 4 ++--
# net/sched/sch_cbq.c | 2 +-
# net/sched/sch_htb.c | 9 ++++-----
# net/sched/sch_tbf.c | 4 ++--
# 6 files changed, 27 insertions(+), 12 deletions(-)
#
--- linux-2.6.22/include/linux/pkt_sched.h.bak 2007-07-09 21:58:23.559346000
+0300
+++ linux-2.6.22/include/linux/pkt_sched.h 2007-07-18 21:46:00.084770053
+0300
@@ -77,8 +77,9 @@ struct tc_ratespec
{
unsigned char cell_log;
unsigned char __reserved;
- unsigned short feature;
- short addend;
+ unsigned short feature; /* Always 0 in pre-atm patch kernels */
+ char cell_align; /* Always 0 in pre-atm patch kernels */
+ unsigned char __reserved2;
unsigned short mpu;
__u32 rate;
};
--- linux-2.6.22/include/net/sch_generic.h.bak 2007-07-09 02:32:17.000000000
+0300
+++ linux-2.6.22/include/net/sch_generic.h 2007-07-18 21:44:40.024580754
+0300
@@ -302,4 +302,19 @@ drop:
return NET_XMIT_DROP;
}
+/* Lookup a qdisc_rate_table to determine how long it will take to send a
+ * packet given its size.
+ */
+static inline u32 qdisc_l2t(struct qdisc_rate_table* rtab, int pktlen)
+{
+ int slot = pktlen + rtab->rate.cell_align;
+
+ if (slot < 0)
+ slot = 0;
+ slot >>= rtab->rate.cell_log;
+ if (slot > 255)
+ return rtab->data[255] + 1;
+ return rtab->data[slot];
+}
+
#endif
--- linux-2.6.22/net/sched/act_police.c.bak 2007-07-09 02:32:17.000000000
+0300
+++ linux-2.6.22/net/sched/act_police.c 2007-07-18 21:42:49.275936447 +0300
@@ -32,8 +32,8 @@
#include <net/act_api.h>
#include <net/netlink.h>
-#define L2T(p,L) ((p)->tcfp_R_tab->data[(L)>>(p)->tcfp_R_tab->rate.cell_log])
-#define L2T_P(p,L) ((p)->tcfp_P_tab->data[(L)>>(p)->tcfp_P_tab->rate.cell_log])
+#define L2T(p,L) qdisc_l2t((p)->tcfp_R_tab,L)
+#define L2T_P(p,L) qdisc_l2t((p)->tcfp_P_tab,L)
#define POL_TAB_MASK 15
static struct tcf_common *tcf_police_ht[POL_TAB_MASK + 1];
--- linux-2.6.22/net/sched/sch_cbq.c.bak 2007-07-09 02:32:17.000000000
+0300
+++ linux-2.6.22/net/sched/sch_cbq.c 2007-07-18 21:51:12.794420373 +0300
@@ -192,7 +192,7 @@ struct cbq_sched_data
};
-#define L2T(cl,len) ((cl)->R_tab->data[(len)>>(cl)->R_tab->rate.cell_log])
+#define L2T(cl,len) qdisc_l2t((cl)->R_tab,len)
static __inline__ unsigned cbq_hash(u32 h)
--- linux-2.6.22/net/sched/sch_htb.c.bak 2007-07-09 21:17:53.417438000
+0300
+++ linux-2.6.22/net/sched/sch_htb.c 2007-07-18 21:50:08.602465126 +0300
@@ -157,12 +157,11 @@ struct htb_class {
static inline long L2T(struct htb_class *cl, struct qdisc_rate_table *rate,
int size)
{
- int slot = size >> rate->rate.cell_log;
- if (slot > 255) {
+ long result = qdisc_l2t(rate, size);
+
+ if (result > rate->data[255])
cl->xstats.giants++;
- slot = 255;
- }
- return rate->data[slot];
+ return result;
}
struct htb_sched {
--- linux-2.6.22/net/sched/sch_tbf.c.bak 2007-07-09 02:32:17.000000000
+0300
+++ linux-2.6.22/net/sched/sch_tbf.c 2007-07-18 21:52:10.665281840 +0300
@@ -132,8 +132,8 @@ struct tbf_sched_data
struct qdisc_watchdog watchdog; /* Watchdog timer */
};
-#define L2T(q,L) ((q)->R_tab->data[(L)>>(q)->R_tab->rate.cell_log])
-#define L2T_P(q,L) ((q)->P_tab->data[(L)>>(q)->P_tab->rate.cell_log])
+#define L2T(q,L) qdisc_l2t((q)->R_tab,L)
+#define L2T_P(q,L) qdisc_l2t((q)->P_tab,L)
static int tbf_enqueue(struct sk_buff *skb, struct Qdisc* sch)
{
#
# include/linux/pkt_sched.h | 5 +-
# tc/m_police.c | 37 +++++++++------
# tc/q_cbq.c | 52 ++++++++++++++-------
# tc/q_htb.c | 39 +++++++---------
# tc/q_tbf.c | 53 ++++++++++++++--------
# tc/tc_core.c | 109
++++++++++++++++++++++++++++++++++++++++++----
# tc/tc_core.h | 7 ++
# 7 files changed, 217 insertions(+), 85 deletions(-)
#
--- iproute-2.6.22/include/linux/pkt_sched.h.bak 2007-07-13
20:33:32.760700000 +0300
+++ iproute-2.6.22/include/linux/pkt_sched.h 2007-07-18 22:11:55.408999963
+0300
@@ -77,8 +77,9 @@ struct tc_ratespec
{
unsigned char cell_log;
unsigned char __reserved;
- unsigned short feature;
- short addend;
+ unsigned short feature; /* Always 0 in pre-atm patch kernels */
+ char cell_align; /* Always 0 in pre-atm patch kernels */
+ unsigned char __reserved2;
unsigned short mpu;
__u32 rate;
};
--- iproute-2.6.22/tc/m_police.c.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/m_police.c 2007-07-18 22:48:35.435620446 +0300
@@ -35,7 +35,7 @@ struct action_util police_action_util =
static void usage(void)
{
fprintf(stderr, "Usage: ... police rate BPS burst BYTES[/BYTES] [ mtu
BYTES[/BYTES] ]\n");
- fprintf(stderr, " [ peakrate BPS ] [ avrate BPS ]\n");
+ fprintf(stderr, " [ peakrate BPS ] [ avrate BPS ] [
overhead OVERHEAD ] [ atm ]\n");
fprintf(stderr, " [ ACTIONTERM ]\n");
fprintf(stderr, "Old Syntax ACTIONTERM := action
<EXCEEDACT>[/NOTEXCEEDACT] \n");
fprintf(stderr, "New Syntax ACTIONTERM := conform-exceed
<EXCEEDACT>[/NOTEXCEEDACT] \n");
@@ -132,7 +132,10 @@ int act_parse_police(struct action_util
__u32 ptab[256];
__u32 avrate = 0;
int presult = 0;
- unsigned buffer=0, mtu=0, mpu=0;
+ unsigned buffer=0, mtu=0;
+ __u8 mpu=0;
+ __s8 overhead=0;
+ int atm=0;
int Rcell_log=-1, Pcell_log = -1;
struct rtattr *tail;
@@ -182,7 +185,7 @@ int act_parse_police(struct action_util
fprintf(stderr, "Double \"mpu\" spec\n");
return -1;
}
- if (get_size(&mpu, *argv)) {
+ if (get_u8(&mpu, *argv, 10)) {
explain1("mpu");
return -1;
}
@@ -196,6 +199,18 @@ int act_parse_police(struct action_util
explain1("rate");
return -1;
}
+ } else if (strcmp(*argv, "overhead") == 0) {
+ NEXT_ARG();
+ if (p.rate.rate) {
+ fprintf(stderr, "Double \"overhead\" spec\n");
+ return -1;
+ }
+ if (get_s8(&overhead, *argv, 10)) {
+ explain1("overhead");
+ return -1;
+ }
+ } else if (strcmp(*argv, "atm") == 0) {
+ atm = 1;
} else if (strcmp(*argv, "avrate") == 0) {
NEXT_ARG();
if (avrate) {
@@ -261,22 +276,14 @@ int act_parse_police(struct action_util
}
if (p.rate.rate) {
- if ((Rcell_log = tc_calc_rtable(p.rate.rate, rtab, Rcell_log,
mtu, mpu)) < 0) {
- fprintf(stderr, "TBF: failed to calculate rate
table.\n");
- return -1;
- }
+ tc_calc_ratespec(&p.rate, rtab, p.rate.rate, Rcell_log,
+ mtu, mpu, atm, overhead);
p.burst = tc_calc_xmittime(p.rate.rate, buffer);
- p.rate.cell_log = Rcell_log;
- p.rate.mpu = mpu;
}
p.mtu = mtu;
if (p.peakrate.rate) {
- if ((Pcell_log = tc_calc_rtable(p.peakrate.rate, ptab,
Pcell_log, mtu, mpu)) < 0) {
- fprintf(stderr, "POLICE: failed to calculate peak rate
table.\n");
- return -1;
- }
- p.peakrate.cell_log = Pcell_log;
- p.peakrate.mpu = mpu;
+ tc_calc_ratespec(&p.peakrate, ptab, p.peakrate.rate, Pcell_log,
+ mtu, mpu, atm, overhead);
}
tail = NLMSG_TAIL(n);
--- iproute-2.6.22/tc/q_cbq.c.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/q_cbq.c 2007-07-18 22:50:27.207791135 +0300
@@ -32,6 +32,7 @@ static void explain_class(void)
fprintf(stderr, " [ prio NUMBER ] [ cell BYTES ] [ ewma
LOG ]\n");
fprintf(stderr, " [ estimator INTERVAL TIME_CONSTANT
]\n");
fprintf(stderr, " [ split CLASSID ] [ defmap MASK/CHANGE
]\n");
+ fprintf(stderr, " [ overhead BYTES ] [ atm ]\n");
}
static void explain(void)
@@ -52,7 +53,10 @@ static int cbq_parse_opt(struct qdisc_ut
struct tc_ratespec r;
struct tc_cbq_lssopt lss;
__u32 rtab[256];
- unsigned mpu=0, avpkt=0, allot=0;
+ unsigned avpkt=0, allot=0;
+ __u8 mpu=0;
+ __s8 overhead=0;
+ int atm=0;
int cell_log=-1;
int ewma_log=-1;
struct rtattr *tail;
@@ -102,7 +106,7 @@ static int cbq_parse_opt(struct qdisc_ut
}
} else if (strcmp(*argv, "mpu") == 0) {
NEXT_ARG();
- if (get_size(&mpu, *argv)) {
+ if (get_u8(&mpu, *argv, 10)) {
explain1("mpu");
return -1;
}
@@ -113,6 +117,14 @@ static int cbq_parse_opt(struct qdisc_ut
explain1("allot");
return -1;
}
+ } else if (strcmp(*argv, "overhead") == 0) {
+ NEXT_ARG();
+ if (get_s8(&overhead, *argv, 10)) {
+ explain1("overhead");
+ return -1;
+ }
+ } else if (strcmp(*argv, "atm") == 0) {
+ atm = 1;
} else if (strcmp(*argv, "help") == 0) {
explain();
return -1;
@@ -137,12 +149,7 @@ static int cbq_parse_opt(struct qdisc_ut
if (allot < (avpkt*3)/2)
allot = (avpkt*3)/2;
- if ((cell_log = tc_calc_rtable(r.rate, rtab, cell_log, allot, mpu)) <
0) {
- fprintf(stderr, "CBQ: failed to calculate rate table.\n");
- return -1;
- }
- r.cell_log = cell_log;
- r.mpu = mpu;
+ tc_calc_ratespec(&r, rtab, r.rate, cell_log, allot, mpu, atm, overhead);
if (ewma_log < 0)
ewma_log = TC_CBQ_DEF_EWMA;
@@ -175,7 +182,9 @@ static int cbq_parse_class_opt(struct qd
struct tc_cbq_fopt fopt;
struct tc_cbq_ovl ovl;
__u32 rtab[256];
- unsigned mpu=0;
+ __u8 mpu=0;
+ __s8 overhead = 0;
+ int atm = 0;
int cell_log=-1;
int ewma_log=-1;
unsigned bndw = 0;
@@ -289,10 +298,18 @@ static int cbq_parse_class_opt(struct qd
lss.change |= TCF_CBQ_LSS_AVPKT;
} else if (strcmp(*argv, "mpu") == 0) {
NEXT_ARG();
- if (get_size(&mpu, *argv)) {
+ if (get_u8(&mpu, *argv, 10)) {
explain1("mpu");
return -1;
}
+ } else if (strcmp(*argv, "overhead") == 0) {
+ NEXT_ARG();
+ if (get_s8(&overhead, *argv, 10)) {
+ explain1("overhead");
+ return -1;
+ }
+ } else if (strcmp(*argv, "atm") == 0) {
+ atm = 1;
} else if (strcmp(*argv, "weight") == 0) {
NEXT_ARG();
if (get_size(&wrr.weight, *argv)) {
@@ -336,12 +353,7 @@ static int cbq_parse_class_opt(struct qd
unsigned pktsize = wrr.allot;
if (wrr.allot < (lss.avpkt*3)/2)
wrr.allot = (lss.avpkt*3)/2;
- if ((cell_log = tc_calc_rtable(r.rate, rtab, cell_log, pktsize,
mpu)) < 0) {
- fprintf(stderr, "CBQ: failed to calculate rate
table.\n");
- return -1;
- }
- r.cell_log = cell_log;
- r.mpu = mpu;
+ tc_calc_ratespec(&r, rtab, r.rate, cell_log, pktsize, mpu, atm,
overhead);
}
if (ewma_log < 0)
ewma_log = TC_CBQ_DEF_EWMA;
@@ -464,8 +476,12 @@ static int cbq_print_opt(struct qdisc_ut
fprintf(f, "rate %s ", buf);
if (show_details) {
fprintf(f, "cell %ub ", 1<<r->cell_log);
- if (r->mpu)
- fprintf(f, "mpu %ub ", r->mpu);
+ if (r->mpu & 0xff)
+ fprintf(f, "mpu %ub ", (__u8)r->mpu);
+ if ((r->mpu >> 8))
+ fprintf(f, "overhead %db ", (__s8)(r->mpu >>
8));
+ if (r->feature & 0x0001)
+ fprintf(f, "atm ");
}
}
if (lss && lss->flags) {
--- iproute-2.6.22/tc/q_htb.c.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/q_htb.c 2007-07-18 22:32:14.140392445 +0300
@@ -34,14 +34,14 @@ static void explain(void)
" default minor id of class to which unclassified packets are
sent {0}\n"
" r2q DRR quantums are computed as rate in Bps/r2q {10}\n"
" debug string of 16 numbers each 0-3 {0}\n\n"
- "... class add ... htb rate R1 [burst B1] [mpu B] [overhead
O]\n"
+ "... class add ... htb rate R1 [burst B1] [mpu B] [overhead O]
[atm]\n"
" [prio P] [slot S] [pslot PS]\n"
" [ceil R2] [cburst B2] [mtu MTU] [quantum
Q]\n"
" rate rate allocated to this class (class can still
borrow)\n"
" burst max bytes burst which can be accumulated during idle
period {computed}\n"
" mpu minimum packet size used in rate computations\n"
" overhead per-packet size overhead used in rate computations\n"
-
+ " atm include atm cell tax in rate computations\n"
" ceil definite upper class rate (no borrows) {rate}\n"
" cburst burst but for ceil {computed}\n"
" mtu max packet size we create rate map for {1600}\n"
@@ -107,8 +107,10 @@ static int htb_parse_class_opt(struct qd
__u32 rtab[256],ctab[256];
unsigned buffer=0,cbuffer=0;
int cell_log=-1,ccell_log = -1;
- unsigned mtu, mpu;
- unsigned char mpu8 = 0, overhead = 0;
+ unsigned mtu;
+ __u8 mpu8=0;
+ __s8 overhead=0;
+ int atm=0;
struct rtattr *tail;
memset(&opt, 0, sizeof(opt)); mtu = 1600; /* eth packet len */
@@ -132,9 +134,11 @@ static int htb_parse_class_opt(struct qd
}
} else if (matches(*argv, "overhead") == 0) {
NEXT_ARG();
- if (get_u8(&overhead, *argv, 10)) {
+ if (get_s8(&overhead, *argv, 10)) {
explain1("overhead"); return -1;
}
+ } else if (matches(*argv, "atm") == 0) {
+ atm = 1;
} else if (matches(*argv, "quantum") == 0) {
NEXT_ARG();
if (get_u32(&opt.quantum, *argv, 10)) {
@@ -206,23 +210,12 @@ static int htb_parse_class_opt(struct qd
if (!buffer) buffer = opt.rate.rate / get_hz() + mtu;
if (!cbuffer) cbuffer = opt.ceil.rate / get_hz() + mtu;
-/* encode overhead and mpu, 8 bits each, into lower 16 bits */
- mpu = (unsigned)mpu8 | (unsigned)overhead << 8;
- opt.ceil.mpu = mpu; opt.rate.mpu = mpu;
+ /* encode overhead and mpu, 8 bits each, into lower 16 bits */
+ tc_calc_ratespec(&opt.rate, rtab, opt.rate.rate, cell_log, mtu, mpu8,
atm, overhead);
+ tc_calc_ratespec(&opt.ceil, ctab, opt.ceil.rate, cell_log, mtu, mpu8,
atm, overhead);
- if ((cell_log = tc_calc_rtable(opt.rate.rate, rtab, cell_log, mtu,
mpu)) < 0) {
- fprintf(stderr, "htb: failed to calculate rate table.\n");
- return -1;
- }
opt.buffer = tc_calc_xmittime(opt.rate.rate, buffer);
- opt.rate.cell_log = cell_log;
-
- if ((ccell_log = tc_calc_rtable(opt.ceil.rate, ctab, cell_log, mtu,
mpu)) < 0) {
- fprintf(stderr, "htb: failed to calculate ceil rate table.\n");
- return -1;
- }
opt.cbuffer = tc_calc_xmittime(opt.ceil.rate, cbuffer);
- opt.ceil.cell_log = ccell_log;
tail = NLMSG_TAIL(n);
addattr_l(n, 1024, TCA_OPTIONS, NULL, 0);
@@ -267,12 +260,16 @@ static int htb_print_opt(struct qdisc_ut
sprint_size(buffer, b1),
1<<hopt->rate.cell_log,
sprint_size(hopt->rate.mpu&0xFF, b2),
- sprint_size((hopt->rate.mpu>>8)&0xFF, b3));
+ sprint_size((__s8)(hopt->rate.mpu>>8), b3));
+ if (hopt->rate.feature & 0x0001)
+ fprintf(f, "atm ");
fprintf(f, "cburst %s/%u mpu %s overhead %s ",
sprint_size(cbuffer, b1),
1<<hopt->ceil.cell_log,
sprint_size(hopt->ceil.mpu&0xFF, b2),
- sprint_size((hopt->ceil.mpu>>8)&0xFF, b3));
+ sprint_size((__s8)(hopt->ceil.mpu>>8), b3));
+ if (hopt->ceil.feature & 0x0001)
+ fprintf(f, "atm ");
fprintf(f, "level %d ", (int)hopt->level);
} else {
fprintf(f, "burst %s ", sprint_size(buffer, b1));
--- iproute-2.6.22/tc/q_tbf.c.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/q_tbf.c 2007-07-18 22:38:36.990372046 +0300
@@ -26,7 +26,7 @@
static void explain(void)
{
fprintf(stderr, "Usage: ... tbf limit BYTES burst BYTES[/BYTES] rate
KBPS [ mtu BYTES[/BYTES] ]\n");
- fprintf(stderr, " [ peakrate KBPS ] [ latency TIME ]\n");
+ fprintf(stderr, " [ peakrate KBPS ] [ latency TIME ] [
overhead OVERHEAD ] [ atm ]\n");
}
static void explain1(char *arg)
@@ -43,7 +43,10 @@ static int tbf_parse_opt(struct qdisc_ut
struct tc_tbf_qopt opt;
__u32 rtab[256];
__u32 ptab[256];
- unsigned buffer=0, mtu=0, mpu=0, latency=0;
+ unsigned buffer=0, mtu=0, latency=0;
+ __u8 mpu=0;
+ __s8 overhead=0;
+ int atm=0;
int Rcell_log=-1, Pcell_log = -1;
struct rtattr *tail;
@@ -103,7 +106,7 @@ static int tbf_parse_opt(struct qdisc_ut
fprintf(stderr, "Double \"mpu\" spec\n");
return -1;
}
- if (get_size(&mpu, *argv)) {
+ if (get_u8(&mpu, *argv, 10)) {
explain1("mpu");
return -1;
}
@@ -119,6 +122,20 @@ static int tbf_parse_opt(struct qdisc_ut
return -1;
}
ok++;
+ } else if (strcmp(*argv, "overhead") == 0) {
+ NEXT_ARG();
+ if (overhead) {
+ fprintf(stderr, "Double \"overhead\" spec\n");
+ return -1;
+ }
+ if (get_s8(&overhead, *argv, 10)) {
+ explain1("overhead");
+ return -1;
+ }
+ ok++;
+ } else if (strcmp(*argv, "atm") == 0) {
+ atm = 1;
+ ok++;
} else if (matches(*argv, "peakrate") == 0) {
NEXT_ARG();
if (opt.peakrate.rate) {
@@ -170,21 +187,11 @@ static int tbf_parse_opt(struct qdisc_ut
opt.limit = lim;
}
- if ((Rcell_log = tc_calc_rtable(opt.rate.rate, rtab, Rcell_log, mtu,
mpu)) < 0) {
- fprintf(stderr, "TBF: failed to calculate rate table.\n");
- return -1;
- }
+ tc_calc_ratespec(&opt.rate, rtab, opt.rate.rate, Rcell_log, mtu, mpu,
atm, overhead);
opt.buffer = tc_calc_xmittime(opt.rate.rate, buffer);
- opt.rate.cell_log = Rcell_log;
- opt.rate.mpu = mpu;
if (opt.peakrate.rate) {
- if ((Pcell_log = tc_calc_rtable(opt.peakrate.rate, ptab,
Pcell_log, mtu, mpu)) < 0) {
- fprintf(stderr, "TBF: failed to calculate peak rate
table.\n");
- return -1;
- }
+ tc_calc_ratespec(&opt.peakrate, ptab, opt.peakrate.rate,
Pcell_log, mtu, mpu, atm, overhead);
opt.mtu = tc_calc_xmittime(opt.peakrate.rate, mtu);
- opt.peakrate.cell_log = Pcell_log;
- opt.peakrate.mpu = mpu;
}
tail = NLMSG_TAIL(n);
@@ -220,8 +227,12 @@ static int tbf_print_opt(struct qdisc_ut
fprintf(f, "rate %s ", sprint_rate(qopt->rate.rate, b1));
buffer = tc_calc_xmitsize(qopt->rate.rate, qopt->buffer);
if (show_details) {
- fprintf(f, "burst %s/%u mpu %s ", sprint_size(buffer, b1),
- 1<<qopt->rate.cell_log, sprint_size(qopt->rate.mpu,
b2));
+ fprintf(f, "burst %s/%u mpu %s overhead %d ",
sprint_size(buffer, b1),
+ 1<<qopt->rate.cell_log,
+ sprint_size(qopt->rate.mpu & 0xFF, b2),
+ (__s8)(qopt->rate.mpu >> 8));
+ if (qopt->rate.feature & 0x0001)
+ fprintf(f, "atm ");
} else {
fprintf(f, "burst %s ", sprint_size(buffer, b1));
}
@@ -232,8 +243,12 @@ static int tbf_print_opt(struct qdisc_ut
if (qopt->mtu || qopt->peakrate.mpu) {
mtu = tc_calc_xmitsize(qopt->peakrate.rate, qopt->mtu);
if (show_details) {
- fprintf(f, "mtu %s/%u mpu %s ",
sprint_size(mtu, b1),
- 1<<qopt->peakrate.cell_log,
sprint_size(qopt->peakrate.mpu, b2));
+ fprintf(f, "mtu %s/%u mpu %s overhead %d ",
sprint_size(mtu, b1),
+ 1<<qopt->peakrate.cell_log,
+ sprint_size(qopt->peakrate.mpu & 0xFF,
b2),
+ (__s8)(qopt->peakrate.mpu >> 8));
+ if (qopt->peakrate.feature & 0x0001)
+ fprintf(f, "atm ");
} else {
fprintf(f, "minburst %s ", sprint_size(mtu,
b1));
}
--- iproute-2.6.22/tc/tc_core.c.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/tc_core.c 2007-07-18 22:46:04.381040136 +0300
@@ -66,33 +66,124 @@ unsigned tc_calc_xmitsize(unsigned rate,
}
/*
- rtab[pkt_len>>cell_log] = pkt_xmit_time
+ * Calculate the ATM cell overhead. ATM sends each packet in 48 byte
+ * chunks, the last chunk being padded if necessary. Each chunk carries
+ * an additional 5 byte overhead - the ATM header.
*/
-int tc_calc_rtable(unsigned bps, __u32 *rtab, int cell_log, unsigned mtu,
- unsigned mpu)
+static int tc_align_to_cells(int size)
+{
+ int cells;
+
+ cells = size / ATM_CELL_PAYLOAD;
+ if (size % ATM_CELL_PAYLOAD != 0)
+ cells++;
+ return cells * ATM_CELL_SIZE;
+}
+
+/*
+ * The number this function calculates is subtle. Ignore it and just believe
+ * it works if you have a choice, otherwise ..
+ *
+ * If there we are calculating the ATM cell overhead the kernel calculations
+ * will be out sometimes if the range of packet sizes spanned by one
+ * rate table element crosses an ATM cell boundary. Consider these three
+ * senarios:
+ * (a) the packet is sent across the ATM link without addition
+ * overheads the kernel doesn't know about, and
+ * (b) a packet that has 1 byte of additional overhead the kernel
+ * doesn't know about. Here
+ * (c) a packet that has 2 bytes of additional overhead the
+ * kernel doesn't know about.
+ * The table below presents what happens. Each row is for a single rate
+ * table element. The "Sizes" column shows what packet sizes the rate table
+ * element will be used for. This packet size includes the "unknown to
+ * kernel" overhead, but does not include overhead incurred by breaking the
+ * packet up into ATM cells. This ATM cell overhead consists of the 5 byte
+ * header per ATM cell, plus the padding in the last cell. The "ATM" column
+ * shows how many bytes are actually sent across the ATM link, ie it does
+ * include the ATM cell overhead.
+ *
+ * RateTable Entry Sizes(a) ATM(a) Sizes(b) ATM(b) Sizes(c) ATM(c)
+ * ratetable[0] 0..7 53 1..8 53 2..9 53
+ * ratetable[1] 8..15 53 9..16 53 2..17 53
+ * ratetable[2] 16..23 53 17..24 53 18..25 53
+ * ratetable[3] 24..31 53 25..32 53 26..33 53
+ * ratetable[4] 32..39 53 33..40 53 34..41 53
+ * ratetable[5] 40..47 53 41..48 53 42..49 53,106
+ * ratetable[6] 48..55 53,106 49..56 106 50..57 106
+ *
+ * For senario (a), the ratetable[6] entry covers two cases: one were a single
+ * ATM cell is needed to transmit the data, and one where two ATM cells are
+ * required. It can't be right for both. Unfortunately the error is large.
+ * The same problem arises in senario (c) for ratetable[5]. The problem
+ * doesn't happen for senario (b), because the boundary between rate table
+ * entries happens to match the boundary between ATM cells.
+ *
+ * What we would like to do is ensure that ratetable boundaries always match
+ * the ATM cells. If we do this the error goes away. The solution is to make
+ * the kernel add a small bias to the packet size. (Small because the bias
+ * will always be smaller than cell_log.) Adding this small bias will in
+ * effect slide the ratetable along a bit, so the boundaries match. The code
+ * below calculates that bias. Provided the MTU is less than 4092, doing
+ * this can always eliminate the error.
+ *
+ * Old kernels won't add this bias, so they will have the error described above
+ * in most cases. In the worst case senario, considering all possible ATM cell
+ * sizes (1..48), for 7 of these sizes the old kernel will calculate the rate
+ * wrongly - ie, be out by 53 bytes.
+ */
+static int tc_calc_cell_align(int atm_cell_tax, char overhead, int cell_log)
+{
+ int cell_size;
+
+ if (!atm_cell_tax)
+ return 0;
+ cell_size = 1 << cell_log;
+ return (overhead + cell_size - 2) % cell_size - cell_size + 1;
+}
+
+/*
+ * A constructor for a tc_ratespec.
+ */
+void tc_calc_ratespec(struct tc_ratespec* spec, __u32* rtab, unsigned bps,
+ int cell_log, unsigned mtu, unsigned char mpu,
+ int atm_cell_tax, char overhead)
{
int i;
- unsigned overhead = (mpu >> 8) & 0xFF;
- mpu = mpu & 0xFF;
if (mtu == 0)
mtu = 2047;
+ /* rtab[pkt_len>>cell_log] = pkt_xmit_time */
if (cell_log < 0) {
cell_log = 0;
while ((mtu>>cell_log) > 255)
cell_log++;
}
+
for (i=0; i<256; i++) {
- unsigned sz = (i<<cell_log);
- if (overhead)
- sz += overhead;
+ /*
+ * sz is the length of packet we will use for this ratetable
+ * entry. The time taken to send a packet of this length will
+ * be used for all packet lengths this ratetable entry applies
+ * to. As underestimating how long it will take to transmit a
+ * packet is a worse error than overestimating it, the longest
+ * packet this rate table entry applies to is used.
+ */
+ int sz = ((i+1)<<cell_log) - 1 + overhead;
if (sz < mpu)
sz = mpu;
+ if (atm_cell_tax)
+ sz = tc_align_to_cells(sz);
rtab[i] = tc_calc_xmittime(bps, sz);
}
- return cell_log;
+
+ spec->cell_align = tc_calc_cell_align(atm_cell_tax, overhead, cell_log);
+ spec->cell_log = cell_log;
+ spec->feature = 0x8000 | (atm_cell_tax ? 1 : 0);
+ spec->mpu = mpu | (unsigned)(overhead << 8);
+ spec->rate = bps;
}
int tc_core_init()
--- iproute-2.6.22/tc/tc_core.h.bak 2007-07-11 04:34:14.000000000 +0300
+++ iproute-2.6.22/tc/tc_core.h 2007-07-18 22:47:53.289179306 +0300
@@ -6,6 +6,9 @@
#define TIME_UNITS_PER_SEC 1000000
+#define ATM_CELL_SIZE 53
+#define ATM_CELL_PAYLOAD 48
+
int tc_core_time2big(long time);
long tc_core_time2tick(long time);
long tc_core_tick2time(long tick);
@@ -13,7 +16,9 @@ long tc_core_time2ktime(long time);
long tc_core_ktime2time(long ktime);
unsigned tc_calc_xmittime(unsigned rate, unsigned size);
unsigned tc_calc_xmitsize(unsigned rate, unsigned ticks);
-int tc_calc_rtable(unsigned bps, __u32 *rtab, int cell_log, unsigned mtu,
unsigned mpu);
+void tc_calc_ratespec(struct tc_ratespec* spec, __u32* rtab, unsigned bps,
+ int cell_log, unsigned mtu, unsigned char mpu,
+ int atm_cell_tax, char overhead);
int tc_setup_estimator(unsigned A, unsigned time_const, struct tc_estimator
*est);
