forked from blackmagic-debug/blackmagic
-
Notifications
You must be signed in to change notification settings - Fork 0
/
UsingSWO
255 lines (209 loc) · 10.5 KB
/
UsingSWO
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
SWO is a datastream that comes out of a single pin when the debug interface
is in SWD mode. It can be encoded either using NRZ (UART) or RZ (Manchester)
formats. The pin is a dedicated one that would be used for TDO when the
debug interface is in JTAG mode. On the STM32 it's port PB3.
When in NRZ mode the SWO data rate that comes out of the chip _must_ match
the rate that the debugger expects. By default on BMP the baudrate is
2.25MBps but that can be changed as an optional parameter to the monitor
traceswo command, like this;
monitor traceswo 115200
....would set the swo output at the low speed of 115kbps.
We are constrained on maximum input speed by both the capabilities of the
BMP STM32F103 USART and the ability to get the packets back out over the USB
link. The UART baudrate is set by b=(72x10^6)/(16*d)...so for d=1 that means
a maximum speed of 4.5Mbps. For continious streaming that turns out to be
_too_ fast for the USB link, so the next available option is the 2.25Mbps
that we use. ....you can safely use the 4.5Mbps setting if your debug data
is bursty, or if you're using a different CPU to the STM32F103 as your BMP
host, but you potentially run the risk of losing packets if you have long
runs of sending which the usb cannot flush in time (there's a 12K buffer, so
the it is a pretty long run before it becomes a problem).
Note that the baudrate equation means there are only certain speeds
available. The highest half dozen are;
SWO uses USART1(stlink) USART2(swlink)
1 4.50 Mbps 2.25 Mbps
2 2.25 Mbps 1.125 Mbps
3 1.50 Mbps 0.75 Mbps
4 1.125 Mbps 0.5635 Mbps
5 0.900 Mbps 0.45 Mbps
6 0.750 Mbps 0.375 Mbps
...the USART will cope with some timing slip, but it's advisible to stay as
close to these values as you can. As the speed comes down the spread between
each valid value so mis-timing is less of an issue. The 'monitor traceswo
<x>' command will automatically find the closest divisor to the value you
set for the speed, so be aware the error could be significant.
Depending on what you're using to wake up SWO on the target side, you may
need code to get it into the correct mode and emitting data. You can do that
via gdb direct memory accesses, or from program code.
An example for a STM32F103 for the UART (NRZ) data format that we use;
/* STM32 specific configuration to enable the TRACESWO IO pin */
RCC->APB2ENR |= RCC_APB2ENR_AFIOEN;
AFIO->MAPR |= (2 << 24); // Disable JTAG to release TRACESWO
DBGMCU->CR |= DBGMCU_CR_TRACE_IOEN; // Enable IO trace pins
TPI->ACPR = 31; // Output bits at 72000000/(31+1)=2.25MHz.
TPI->SPPR = 2; // Use Async mode (1 for RZ/Manchester)
TPI-FFCR = 0; // Disable formatter
/* Configure instrumentation trace macroblock */
ITM->LAR = 0xC5ACCE55;
ITM->TCR = 1 << ITM_TCR_TraceBusID_Pos | ITM_TCR_SYNCENA_Msk |
ITM_TCR_ITMENA_Msk;
ITM->TER = 0xFFFFFFFF; // Enable all stimulus ports
Code for the STM32L476 might look like:
#define BAUDRATE 115200
DBGMCU->CR |= DBGMCU_CR_TRACE_IOEN; /* Enable IO pins for Async trace */
uint32_t divisor, clk_frequency;
clk_frequency = NutGetCpuClock();
divisor = clk_frequency / BAUDRATE;
divisor--;
TPI->CSPSR = 1; /* port size = 1 bit */
TPI->ACPR = divisor;
TPI->SPPR = 2; /*Use Async mode pin protocol */
TPI->FFCR = 0x00; /* Bypass the TPIU formatter and send output directly*/
/* Configure Trace Port Interface Unit */
CoreDebug->DEMCR |= CoreDebug_DEMCR_TRCENA_Msk; // Enable access to registers
DWT->CTRL = 0x400003FE; // DWT needs to provide sync for ITM
ITM->LAR = 0xC5ACCE55; // Allow access to the Control Register
ITM->TPR = 0x0000000F; // Trace access privilege from user level code, please
ITM->TCR = 0x0001000D; // ITM_TCR_TraceBusID_Msk | ITM_TCR_DWTENA_Msk | ITM_TCR_SYNCENA_Msk | ITM_TCR_ITMENA_Msk
ITM->TER = 1; // Only Enable stimulus port 1
while(1) {
for (uint32_t i = 'A'; i <= 'Z'; i++) {
ITM_SendChar(i);
NutSleep(1);
}
}
If you're using RZ mode (e.g. on a genuine BMP) then you will need the trace
output speed to be quite a lot lower...in the order of 200kHz by means of
changing the divisor to something like 359. That's because the STM32F103
doesn't have a dedicated RZ decoder so it all has to be done in
software. The advantage of RZ is that the probe can adapt to the speed of
the target, so you don't have to set the speed on the probe in the monitor
traceswo command, and it will be tolerant of different speeds.
The SWO data appears on USB Interface 5, Endpoint 5.
SWOListen
=========
A program swolisten.c is found in ./scripts which will listen to this
endpoint, decode the datastream, and output it to a set of unix fifos which
can then be used as the input to other programs (e.g. cat, or something more
sophisticated like gnuplot, octave or whatever). This program doesn't care
if the data originates from a RZ or NRZ port, or at what speed.
Note that swolisten can be used with either BMP firmware, or with a
conventional TTL serial dongle. See at the bottom of this file for
information on how to use a dongle.
The command line to build the swolisten tool is;
gcc -I /usr/local/include/libusb-1.0 -L /usr/local/lib -lusb-1.0 swolisten.c -o swolisten
For Opensuse:
gcc -I /usr/include/libusb-1.0 -lusb-1.0 swolisten.c swolisten -std=gnu99 -g -Og
...you will obviously need to change the paths to your libusb files.
Attach to BMP to your PC:
Start gdb: "arm-none-eabi-gdb"
Choose bmp as target, like:
"target extended /dev/ttyACM0(*)"
Start SWO output: "mon traceswo"
If async SWO is used, give the baudrate your device sends
out as argument. 2.25 MBaud is the default, for the STM32L476 example above
the command would be: "mon traceswo 115200(*)".
Scan the SWD "mon swdp_scan"
Attach to the device: : "attach 1"
Start the program: "r".
(*) Your milage may vary
Now start swolisten without further options.
By default the tool will create fifos for the first 32 channels in a
directory swo (which you will need to create) as follows;
>ls swo/
chan00 chan02 chan04 chan06 chan08 chan0A chan0C chan0E chan10 chan12 chan14
chan16 chan18 chan1A chan1C chan1E chan01 chan03 chan05 chan07 chan09 chan0B
chan0D chan0F chan11 chan13 chan15 chan17 chan19 chan1B chan1D chan1F
>cat swo/channel0
<<OUTPUT FROM ITM Channel 0>>
With the F103 and L476 examples above, an endless stream of
"ABCDEFGHIJKLMNOPQRSTUVWXYZ" should be seen. During reset of the target
device, no output will appear, but with release of reset output restarts.
Information about command line options can be found with the -h option.
swolisten is specifically designed to be 'hardy' to probe and target
disconnects and restarts (y'know, like you get in the real world). The
intention being to give you streams whenever it can get them. It does _not_
require gdb to be running. For the time being traceswo is not turned on by
default in the BMP to avoid possible interactions and making the overall
thing less reliable so You do need gdb to send the initial 'monitor
traceswo' to the probe, but beyond that there's no requirement for gdb to be
present.
Reliability
===========
A whole chunk of work has gone into making sure the dataflow over the SWO
link is reliable. The TL;DR is that the link _is_ reliable. There are
factors outside of our control (i.e. the USB bus you connect to) that could
potentially break the reliabilty but there's not too much we can do about
that since the SWO link is unidirectional (no opportunity for
re-transmits). The following section provides evidence for the claim that
the link is good;
A test 'mule' sends data flat out to the link at the maximum data rate of
2.25Mbps using a loop like the one below;
while (1)
{
for (uint32_t r=0; r<26; r++)
{
for (uint32_t g=0; g<31; g++)
{
ITM_SendChar('A'+r);
}
ITM_SendChar('\n');
}
}
100MB of data (more than 200MB of actual SWO packets, due to the encoding) was sent from the mule to the BMP where the
output from swolisten chan00 was cat'ted into a file;
>cat swo/chan00 > o
....this process was interrupted once the file had grown to 100MB. The first
and last lines were removed from it (these represent previously buffered
data and an incomplete packet at the point where the capture was
interrupted) and the resulting file analysed for consistency;
> sort o | uniq -c
The output was;
126462 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
126462 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB
126462 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
126462 DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD
126461 EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
126461 FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
126461 GGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG
126461 HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH
126461 IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
126461 JJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJ
126461 KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK
126461 LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
126461 MMMMMMMMMMMMMMMMMMMMMMMMMMMMMMM
126461 NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN
126461 OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
126461 PPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP
126461 QQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQ
126461 RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR
126461 SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
126461 TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT
126461 UUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU
126461 VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV
126461 WWWWWWWWWWWWWWWWWWWWWWWWWWWWWWW
126461 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
126461 YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY
126461 ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
(On inspection, the last line of recorded data was indeed a 'D' line).
Swolisten, using a TTL Serial Dongle
====================================
The NRZ data that comes out of the SWO is just UART formatted, but in a
frame. swolisten has been extended to accomodate TTL Serial Dongles that
can pick this up. Success has been had with CP2102 dongles at up to 921600
baud.
To use this mode just connect SWO to the RX pin of your dongle, and start
swolisten with parameters representing the speed and port. An example;
>./swolisten -p /dev/cu.SLAB_USBtoUART -v -b swo/ -s 921600
Any individual dongle will only support certain baudrates (Generally
multiples of 115200) so you may have to experiment to find the best
supported ones. For the CP2102 dongle 1.3824Mbps wasn't supported and
1.8432Mbps returned corrupted data.
Please email [email protected] with information about dongles you find work
well and at what speed.
Further information
===================
SWO is a wide field. Read e.g. the blogs around SWD on
http://shadetail.com/blog/swo-starting-the-steroids/
An open source program suite for SWO under active development is
https://github.com/mubes/orbuculum