| 1 |
.TH STAB 8 "31 October 2011" iproute2 Linux |
| 2 |
. |
| 3 |
.SH NAME |
| 4 |
tc\-stab \- Generic size table manipulations |
| 5 |
. |
| 6 |
.SH SYNOPSIS |
| 7 |
.nf |
| 8 |
tc qdisc add ... stab |
| 9 |
.RS 4 |
| 10 |
[ \fBmtu\fR BYTES ] [ \fBtsize\fR SLOTS ] |
| 11 |
[ \fBmpu\fR BYTES ] [ \fBoverhead\fR BYTES ] |
| 12 |
[ \fBlinklayer\fR { adsl | atm | ethernet } ] ... |
| 13 |
.RE |
| 14 |
.fi |
| 15 |
|
| 16 |
.SH OPTIONS |
| 17 |
For the description of BYTES \- please refer to the \fBUNITS\fR |
| 18 |
section of \fBtc\fR(8). |
| 19 |
|
| 20 |
.IP \fBmtu\fR 4 |
| 21 |
.br |
| 22 |
maximum packet size we create size table for, assumed 2048 if not specified explicitly |
| 23 |
.IP \fBtsize\fR |
| 24 |
.br |
| 25 |
required table size, assumed 512 if not specified explicitly |
| 26 |
.IP \fBmpu\fR |
| 27 |
.br |
| 28 |
minimum packet size used in computations |
| 29 |
.IP \fBoverhead\fR |
| 30 |
.br |
| 31 |
per\-packet size overhead (can be negative) used in computations |
| 32 |
.IP \fBlinklayer\fR |
| 33 |
.br |
| 34 |
required linklayer specification. |
| 35 |
.PP |
| 36 |
. |
| 37 |
.SH DESCRIPTION |
| 38 |
. |
| 39 |
Size tables allow manipulation of packet sizes, as seen by the whole scheduler |
| 40 |
framework (of course, the actual packet size remains the same). Adjusted packet |
| 41 |
size is calculated only once \- when a qdisc enqueues the packet. Initial root |
| 42 |
enqueue initializes it to the real packet's size. |
| 43 |
|
| 44 |
Each qdisc can use a different size table, but the adjusted size is stored in |
| 45 |
an area shared by whole qdisc hierarchy attached to the interface. The effect is |
| 46 |
that if you have such a setup, the last qdisc with a stab in a chain "wins". For |
| 47 |
example, consider HFSC with simple pfifo attached to one of its leaf classes. |
| 48 |
If that pfifo qdisc has stab defined, it will override lengths calculated |
| 49 |
during HFSC's enqueue; and in turn, whenever HFSC tries to dequeue a packet, it |
| 50 |
will use a potentially invalid size in its calculations. Normal setups will |
| 51 |
usually include stab defined only on root qdisc, but further overriding gives |
| 52 |
extra flexibility for less usual setups. |
| 53 |
|
| 54 |
The initial size table is calculated by \fBtc\fR tool using \fBmtu\fR and |
| 55 |
\fBtsize\fR parameters. The algorithm sets each slot's size to the smallest |
| 56 |
power of 2 value, so the whole \fBmtu\fR is covered by the size table. Neither |
| 57 |
\fBtsize\fR, nor \fBmtu\fR have to be power of 2 value, so the size |
| 58 |
table will usually support more than is required by \fBmtu\fR. |
| 59 |
|
| 60 |
For example, with \fBmtu\fR\~=\~1500 and \fBtsize\fR\~=\~128, a table with 128 |
| 61 |
slots will be created, where slot 0 will correspond to sizes 0\-16, slot 1 to |
| 62 |
17\~\-\~32, \&..., slot 127 to 2033\~\-\~2048. Sizes assigned to each slot |
| 63 |
depend on \fBlinklayer\fR parameter. |
| 64 |
|
| 65 |
Stab calculation is also safe for an unusual case, when a size assigned to a |
| 66 |
slot would be larger than 2^16\-1 (you will lose the accuracy though). |
| 67 |
|
| 68 |
During the kernel part of packet size adjustment, \fBoverhead\fR will be added |
| 69 |
to original size, and then slot will be calculated. If the size would cause |
| 70 |
overflow, more than 1 slot will be used to get the final size. This of course |
| 71 |
will affect accuracy, but it's only a guard against unusual situations. |
| 72 |
|
| 73 |
Currently there are two methods of creating values stored in the size table \- |
| 74 |
ethernet and atm (adsl): |
| 75 |
|
| 76 |
.IP ethernet 4 |
| 77 |
.br |
| 78 |
This is basically 1\-1 mapping, so following our example from above |
| 79 |
(disregarding \fBmpu\fR for a moment) slot 0 would have 8, slot 1 would have 16 |
| 80 |
and so on, up to slot 127 with 2048. Note, that \fBmpu\fR\~>\~0 must be |
| 81 |
specified, and slots that would get less than specified by \fBmpu\fR will get |
| 82 |
\fBmpu\fR instead. If you don't specify \fBmpu\fR, the size table will not be |
| 83 |
created at all (it wouldn't make any difference), although any \fBoverhead\fR |
| 84 |
value will be respected during calculations. |
| 85 |
.IP "atm, adsl" |
| 86 |
.br |
| 87 |
ATM linklayer consists of 53 byte cells, where each of them provides 48 bytes |
| 88 |
for payload. Also all the cells must be fully utilized, thus the last one is |
| 89 |
padded if/as necessary. |
| 90 |
|
| 91 |
When the size table is calculated, adjusted size that fits properly into lowest |
| 92 |
amount of cells is assigned to a slot. For example, a 100 byte long packet |
| 93 |
requires three 48\-byte payloads, so the final size would require 3 ATM cells |
| 94 |
\- 159 bytes. |
| 95 |
|
| 96 |
For ATM size tables, 16\~bytes sized slots are perfectly enough. The default |
| 97 |
values of \fBmtu\fR and \fBtsize\fR create 4\~bytes sized slots. |
| 98 |
.PP |
| 99 |
. |
| 100 |
.SH "TYPICAL OVERHEADS" |
| 101 |
The following values are typical for different adsl scenarios (based on |
| 102 |
\fB[1]\fR and \fB[2]\fR): |
| 103 |
|
| 104 |
.nf |
| 105 |
LLC based: |
| 106 |
.RS 4 |
| 107 |
PPPoA \- 14 (PPP \- 2, ATM \- 12) |
| 108 |
PPPoE \- 40+ (PPPoE \- 8, ATM \- 18, ethernet 14, possibly FCS \- 4+padding) |
| 109 |
Bridged \- 32 (ATM \- 18, ethernet 14, possibly FCS \- 4+padding) |
| 110 |
IPoA \- 16 (ATM \- 16) |
| 111 |
.RE |
| 112 |
|
| 113 |
VC Mux based: |
| 114 |
.RS 4 |
| 115 |
PPPoA \- 10 (PPP \- 2, ATM \- 8) |
| 116 |
PPPoE \- 32+ (PPPoE \- 8, ATM \- 10, ethernet 14, possibly FCS \- 4+padding) |
| 117 |
Bridged \- 24+ (ATM \- 10, ethernet 14, possibly FCS \- 4+padding) |
| 118 |
IPoA \- 8 (ATM \- 8) |
| 119 |
.RE |
| 120 |
.fi |
| 121 |
There are a few important things regarding the above overheads: |
| 122 |
. |
| 123 |
.IP \(bu 4 |
| 124 |
IPoA in LLC case requires SNAP, instead of LLC\-NLPID (see rfc2684) \- this is |
| 125 |
the reason why it actually takes more space than PPPoA. |
| 126 |
.IP \(bu |
| 127 |
In rare cases, FCS might be preserved on protocols that include Ethernet frames |
| 128 |
(Bridged and PPPoE). In such situation, any Ethernet specific padding |
| 129 |
guaranteeing 64 bytes long frame size has to be included as well (see RFC2684). |
| 130 |
In the other words, it also guarantees that any packet you send will take |
| 131 |
minimum 2 atm cells. You should set \fBmpu\fR accordingly for that. |
| 132 |
.IP \(bu |
| 133 |
When the size table is consulted, and you're shaping traffic for the sake of |
| 134 |
another modem/router, an Ethernet header (without padding) will already be added |
| 135 |
to initial packet's length. You should compensate for that by subtracting 14 |
| 136 |
from the above overheads in this case. If you're shaping directly on the router |
| 137 |
(for example, with speedtouch usb modem) using ppp daemon, you're using raw ip |
| 138 |
interface without underlying layer2, so nothing will be added. |
| 139 |
|
| 140 |
For more thorough explanations, please see \fB[1]\fR and \fB[2]\fR. |
| 141 |
. |
| 142 |
.SH "ETHERNET CARDS CONSIDERATIONS" |
| 143 |
. |
| 144 |
It's often forgotten that modern network cards (even cheap ones on desktop |
| 145 |
motherboards) and/or their drivers often support different offloading |
| 146 |
mechanisms. In the context of traffic shaping, 'tso' and 'gso' might cause |
| 147 |
undesirable effects, due to massive TCP segments being considered during |
| 148 |
traffic shaping (including stab calculations). For slow uplink interfaces, |
| 149 |
it's good to use \fBethtool\fR to turn off offloading features. |
| 150 |
. |
| 151 |
.SH "SEE ALSO" |
| 152 |
. |
| 153 |
\fBtc\fR(8), \fBtc\-hfsc\fR(7), \fBtc\-hfsc\fR(8), |
| 154 |
.br |
| 155 |
\fB[1]\fR http://ace\-host.stuart.id.au/russell/files/tc/tc\-atm/ |
| 156 |
.br |
| 157 |
\fB[2]\fR http://www.faqs.org/rfcs/rfc2684.html |
| 158 |
|
| 159 |
Please direct bugreports and patches to: <[email protected]> |
| 160 |
. |
| 161 |
.SH "AUTHOR" |
| 162 |
. |
| 163 |
Manpage created by Michal Soltys ([email protected]) |