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
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"
"http://www.w3.org/TR/REC-html40/loose.dtd">
<html>
<head>
<meta content="text/html; charset=iso-8859-1" http-equiv="Content-Type">
<title>Porting UML to a new architecture</title>
</head>
<body alink="#FF0000" vlink="#55188A" link="#0000EF" bgcolor="#FFFFFF" text="#000099">
<table border="0">
<tr align="left">
<td valign="top">
<table border="0">
<tr align="left"><td valign="top" >
<img src="http://user-mode-linux.sourceforge.net/uml-small.png" height="171" width="120"/>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/index.html">Site Home Page</a></font>
<br>
<font size="-1"><a href="http://uml.harlowhill.com">The UML Wiki</a></font>
<br>
<font size="-1"><a href="http://usermodelinux.org">UML Community Site</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/roadmap.html">The UML roadmap</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/uses.html">What it's good for</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/case-studies.html">Case Studies</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/kernel.html">Kernel Capabilities</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/dl-sf.html">Downloading it</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/run.html">Running it</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/compile.html">Compiling</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/install.html">Installation</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/skas.html">Skas Mode</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/patches.html">Incremental Patches</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/tests.html">Test Suite</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/devanon.html">Host memory use</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/fs_making.html">Building filesystems</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/faq.html">Troubles</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/contrib.html">User Contributions</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/links.html">Related Links</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/todo.html">The ToDo list</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/projects.html">Projects</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/diary.html">Diary</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/thanks.html">Thanks</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/contacts.html">Contacts</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Tutorials<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/UserModeLinux-HOWTO.html">The HOWTO (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/UserModeLinux-HOWTO.txt">The HOWTO (text)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/hostfs.html">Host file access</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/input.html">Device inputs</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/shared_fs.html">Sharing filesystems</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/fs.html">Creating filesystems</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/resize.html">Resizing filesystems</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/networking.html">Virtual Networking</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/mconsole.html">Management Console</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/debugging.html">Kernel Debugging</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/honeypots.html">UML Honeypots</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/gprof.html">gprof and gcov</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/xtut.html">Running X</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/trouble.html">Diagnosing problems</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/config.html">Configuration</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slack_readme.html">Installing Slackware</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/arch-port.html">Porting UML</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/iomem.html">IO memory emulation</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/2G-2G.html">UML on 2G/2G hosts</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/lksct/index.html">Adding a UML system call</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/nesting.html">Running nested UMLs</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">How you can help<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/help-gen.html">Overview</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/help-doc.html">Documentation</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/help-userspace.html">Utilities</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/help-kernel-v1.html">Kernel bugs</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/projects.html">Kernel projects</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Screenshots<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/net.html">A virtual network</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/x.html">An X session</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Transcripts<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/login.html">A login session</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/debug-session.html">A debugging session</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slackinst.html">Slackware installation</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Reference<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/switches.html">Kernel switches</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slack_readme.html">Slackware README</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Papers<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/als2000/index.html">ALS 2000 paper (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/als2000.tex">ALS 2000 paper (TeX)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/als2000/slides.html">ALS 2000 slides</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/lca2001/lca.html">LCA 2001 slides</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/ols2001/index.html">OLS 2001 paper (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/ols2001.tex">OLS 2001 paper (TeX)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/als2001/index.html">ALS 2001 paper (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/als2001.tex">ALS 2001 paper (TeX)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/ists2002/umlsec.htm">UML security (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/lca2002/lca2002.htm">LCA 2002 (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/wvu2002/wvu2002.htm">WVU 2002 (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/ists_rt/ists_rt.htm">Security Roundtable (html)</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/ols2002/ols2002.html">OLS 2002 slides</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/slides/lwe2005/LWE2005.html">LWE 2005 slides</a></font>
</td></tr>
<tr align="left"><td valign="top" bgcolor="#e0e0e0">Fun and Games<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/cgi-bin/hangman">Kernel Hangman</a></font>
<br>
<font size="-1"><a href="http://user-mode-linux.sourceforge.net/sdotm.html">Disaster of the Month</a></font>
</td></tr>
</table>
</td>
<td valign="top" align="left">
<center>
<h3>Porting UML to a new architecture</h3>
</center>
Even though UML is running on a host Linux, which insulates it from
the underlying platform to a great extent, some details of the hardware
still leak through and make porting UML to Linux on a new architecture
more than a simple rebuild.
<p>
The major aspects of the hardware that show through are
<ul>
<li>
register names used by ptrace
</li>
<li>
organization of the process address space
</li>
</ul>
<p>
This page will describe how to port UML to a new architecture. It
will acquire new material as we learn more about how to do it. At
this point, this is based on what we learned from the ppc port, which
is the only real port of UML that's been done so far. The i386 port
doesn't really count since that was part of the overall development of
UML rather than a separate porting effort.
<p>
Below, there are references to $(SUBARCH). This is the make variable
which holds the value of the host architecture in the UML build. On
Intel boxes, it's "i386" and on PowerPC boxes, it's "ppc".
<a name="Overview"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Overview</font>
</b>
</td>
</tr>
</table>
<blockquote head="Overview">
UML is split between architecture-independent code and headers which
are found under arch/um in
<ul>
<li>
kernel
</li>
<li>
drivers
</li>
<li>
fs
</li>
<li>
ptproxy
</li>
<li>
include
</li>
</ul>
and the architecture-dependent code and definitions under arch/um in
<ul>
<li>
Makefile-*
</li>
<li>
sys-*
</li>
<li>
include/sysdep-*
</li>
</ul>
Each '*' is the name of an architecture, so the i386-specific code is
under arch/um/sys-i386 and the ppc-specific code is under arch/um/sys-ppc.
<p>
</blockquote>
<a name="The host"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">The host</font>
</b>
</td>
</tr>
</table>
<blockquote head="The host">
Not all architectures can currently run UML. The potential problem is
the ability of ptrace to change system call numbers. i386 couldn't
until I got the change into 2.3.22 and 2.2.15, ppc could, and IA64 and
mips can't. I don't know about the other arches.
<p>
This is necessary because it's critical to UML's ability to virtualize
system calls. Process system calls must be nullified in the host, and
this is done by converting them into getpid.
<p>
So, before starting to work on your new port of UML, make sure ptrace
is fully functional. <a href="examples/ptrace_test.c">This
little program</a> starts a child, which makes calls to getpid and
prints them out, while the parent is converting the getpid calls to
getppid. The parent prints out its own pid, while the child prints out
what it thinks is its own pid, and they should be the same. So, if
your machine is able to run UML, you will see output like this:
<tt>
<pre>
<font size="-1">
Parent pid = 3246
getpid() returned 3246
getpid() returned 3246
</font>
</pre>
</tt>
If not, you will likely get errors from ptrace. Less likely is
different pids being printed out from the two processes. If either
happens, then you need to figure out how to remove that restriction
from the host Linux.
<p>
Note that when you compile ptrace.c, you will need to change the
references to ORIG_EAX, which contains the system call number, to
whatever is appropriate for your architecture.
</blockquote>
<a name="Address space layout"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Address space layout</font>
</b>
</td>
</tr>
</table>
<blockquote head="Address space layout">
Before delving into the code, you need to do some high-level
conceptual thinking about how to organize the address space of a UML
process. UML maps its executable, physical memory, and kernel virtual
memory into the address space of each of its processes. You need to
decide where to put each of these so as to minimize the likelihood of
a process trying to allocate that memory for its own use.
<p>
The only arch hook at this point is where in the address space the UML
binary is going to load. The other addresses are still hard-coded
because they happen to work for both i386 and ppc. UML puts its own
memory in the area starting at 0x5000000 and process stacks 4M below
its own process stack. These choices may not work on all
architectures, so feel free to generalize them. To locate a likely
area on your arch, staring at /proc/<pid>/maps of various processes
on the host has been the technique so far.
</blockquote>
<a name="Architecture Makefile"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Architecture Makefile</font>
</b>
</td>
</tr>
</table>
<blockquote head="Architecture Makefile">
You need to create arch/um/Makefile-$(SUBARCH), which contains the
following definitions:
<ul>
<li>
START_ADDR - The address where the UML executable will load in
memory. This address must be chosen so that it won't conflict with
any memory that a UML process is going to want to use. The i386
definition is
<tt>
<pre>
<font size="-1">START_ADDR = 0x10000000</font>
</pre>
</tt>
</li>
<li>
ARCH_CFLAGS - Anything that needs to be added to CFLAGS goes here.
Both the i386 and ppc ports use this to turn off definitions that
would pull hardware-specific code into the kernel. The ppc definition
is
<tt>
<pre>
<font size="-1">ARCH_CFLAGS = -U__powerpc__ -D__UM_PPC__</font>
</pre>
</tt>
</li>
<li>
ELF_SUBARCH - This is the name of the ELF object format for the
architecture. On i386, it's 'i386', but on ppc, it's not 'ppc' (it's
'powerpc'). The i386 definition is
<tt>
<pre>
<font size="-1">ELF_SUBARCH = $(SUBARCH)</font>
</pre>
</tt>
</li>
</ul>
</blockquote>
<a name="include/asm-um"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">include/asm-um</font>
</b>
</td>
</tr>
</table>
<blockquote head="include/asm-um">
Every architecture needs to provide a set of headers to the generic
kernel. These are located in include/asm-$(ARCH). UML mostly uses
the headers of the underlying architecture. It does this by creating
a symlink from include/asm-um/arch to include/asm-$(SUBARCH). Most of
UML's headers then just include "asm/arch/header.h". As an example,
this is rwlock.h
<tt>
<pre>
<font size="-1">
#ifndef __UM_RWLOCK_H
#define __UM_RWLOCK_H
#include "asm/arch/rwlock.h"
#endif
</font>
</pre>
</tt>
Almost all of UML's headers look exactly like that. Some of the rest
are architecture-independent headers private to UML. You don't need to
worry about these.
<p>
Some headers include the underlying arch headers, but also need to
change them in some way. For example, the UML ptrace-generic.h includes
asm/arch/ptrace.h because it needs some definitions from there.
However, there are things it doesn't want because it needs to define
its own, such as struct pt_regs. So, the underlying architecture's
struct pt_regs is renamed by doing the following
<tt>
<pre>
<font size="-1">
#define pt_regs pt_regs_subarch
#include "asm/arch/ptrace.h"
#undef pt_regs
</font>
</pre>
</tt>
This changes the name of the underlying architecture's pt_regs to
struct pt_regs_subarch, allowing UML to define its own struct
pt_regs. This practice of taking most of the contents other
architecture's header and defining the unwanted bits away is useful,
but it also causes problems which porter have to deal with. For
example, the ppc ptrace.h includes "asm/system.h", which includes the
UML system.h. Since the UML system.h contains references to pt_regs
and it's being included by a header that has had pt_regs renamed to
pt_regs_subarch, the UML system.h references are similarly renamed.
This causes conflicts against files which expect references to
pt_regs. There have been several attempts to update UML/ppc and this
problem has stymied them. It's far from an insolvable problem, but it
involves staring at confusing sequences of includes, figuring out
what's happening, and how to fix it.
<p>
There are also a few headers which are archtecture-dependent.
<ul>
<li>
archparam-$(SUBARCH).h
<br>
This is a header for miscellaneous architecture-dependent
definitions. In a lot of cases, a mostly-generic header can get its
non-generic definitions from a separate header. In these cases, the
definitions are put in archparam-$(SUBARCH).h and that header includes
asm/archparam.h, which is a symbolic link to archparam-$(SUBARCH).h.
<p>
The i386 archparam mostly includes definitions for elf.h, such as
platform-specific register initializatins. The ppc archparam is
similar, adding a couple of definitions for hardirq.h and a couple of
other headers.
</li>
<li>
processor-$(SUBARCH).h
<br>
This header exists because UML/ppc needs a little logic to choose
between including one header and another.
</li>
<li>
ptrace-$(SUBARCH).h
<br>
This defines architecture-dependent access macros into struct
pt_regs. These macros aren't used in generic code, but you may need
some when you define ELF_PLAT_INIT and a few other things. Those
definitions should go here.
</li>
<li>
sigcontext-$(SUBARCH).h
<br>
This is similar to processor-$(SUBARCH).h. It exists so that that the
occurrences of pt_regs in asm-$(SUBARCH)/sigcontext.h can be defined
out of the way.
</li>
<li>
system-$(SUBARCH).h
<br>
This header exists so that UML/ppc can define the ppc _switch_to out
of the way. The i386 version doesn't do anything.
</li>
<li>
ptrace-generic.h
<br>
ptrace-generic.h isn't architecture-dependent, but it does put some
requirements on arch/um/include/sysdep/ptrace.h, which you will meet
below. It defines its struct pt_regs as
<tt>
<pre>
<font size="-1">
struct pt_regs {
struct uml_pt_regs regs;
};
</font>
</pre>
</tt>
and it's up to the architecture to define struct uml_pt_regs. This is
done this way because some UML userspace code needs to refer to
register values. So, uml_pt_regs is the structure that is safe for
userspace code to look at, which pt_regs is for kernel code. Access
macros for pt_regs therefore just call the equivalent access macro for
uml_pt_regs, like
<tt>
<pre>
<font size="-1">
#define PT_REGS_IP(r) UPT_IP(&(r)->regs)
</font>
</pre>
</tt>
and these should be defined in arch/um/include/sysdep/ptrace.h, which
should be able to be included in both user and kernel code. This
means it can't include either libc headers or kernel headers, but may
include safe UML headers.
</li>
</ul>
</blockquote>
<a name="Architecture headers"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Architecture headers</font>
</b>
</td>
</tr>
</table>
<blockquote head="Architecture headers">
There are three headers, which go in arch/um/include/sysdep-$(SUBARCH),
which need to be written, and each needs to
contain a certain set of definitions.
<p>
<b>frame.h</b>
<p>
One of the first things that UML does when it boots is it creates and
saves a signal frame. This will be used when it delivers signals to
its processes. It will be copied onto the process stack and the data
in the original frame, like the signal number, sigcontext contents,
and restorer will be replaced. UML knows how to replace this stuff
because it figured out where in the stack frame it is. And it did
that by having the signal handler record the address of the signal,
the sigcontext structure, and a few other things.
<p>
This is done in a mostly architecture-independent way, but a little
help is needed from architecture-specific code. That code goes into
frame.h.
<p>
What's needed here are definitions of:
<tt>
<pre>
<font size="-1">
static inline unsigned long frame_restorer(void)
</font>
</pre>
</tt>
This returns the location on the stack of the signal state restorer.
On i386, this is the return address, which is next to the frame
pointer, so the i386 definition is
<tt>
<pre>
<font size="-1">
static inline unsigned long frame_restorer(void)
{
unsigned long *fp;
fp = __builtin_frame_address(0);
return((unsigned long) (fp + 1));
}
</font>
</pre>
</tt>
<tt>
<pre>
<font size="-1">
static inline unsigned long frame_sp(void)
</font>
</pre>
</tt>
This returns the value of the stack pointer when the signal handler is
first entered. Note that this is not necessarily the same value as
when the signal handler code is executing because it may have been
adjusted for local variables.
<p>
On i386, this is four bytes more than the frame pointer, so its
definition is
<tt>
<pre>
<font size="-1">
static inline unsigned long frame_sp(void)
{
unsigned long *fp;
fp = __builtin_frame_address(0);
return((unsigned long) (fp + 1));
}
</font>
</pre>
</tt>
In addition, there may be a need for the architecture to save more
information from the signal frame. There are two pairs of structures
and procedures which allow you to do this. The first pair are
expected to record raw addresses from the frame:
<tt>
<pre>
<font size="-1">
struct arch_frame_data_raw {
...
};
static inline void setup_arch_frame_raw(struct arch_frame_data_raw *data,
struct sigcontext *sc)
</font>
</pre>
</tt>
The arch_frame_data_raw may contain anything you want and
setup_arch_frame_raw is expected to fill it in. Both may be empty.
<p>
i386 needs to know the size of the floating point state that's on the
signal frame, so these save the address of the beginning of the sigcontext
structure, where the floating point state ends:
<tt>
<pre>
<font size="-1">
struct arch_frame_data_raw {
unsigned long sc_end;
};
static inline void setup_arch_frame_raw(struct arch_frame_data_raw *data,
struct sigcontext *sc)
{
data->sc_end = (unsigned long) sc;
data->sc_end += sizeof(*sc);
}
</font>
</pre>
</tt>
Then a similar structure and function pair is used to process the raw
addresses into something that's usable later. The i386 code assumes
that the floating point state runs from the top of the stack (which is
alone on its own page, so the top of the stack is the end of the page)
to the start of the sigcontext structure:
<tt>
<pre>
<font size="-1">
struct arch_frame_data {
int fpstate_size;
};
static inline void setup_arch_frame(struct arch_frame_data_raw *in,
struct arch_frame_data *out)
{
unsigned long fpstate_start = in->sc_end;
fpstate_start &= ~PAGE_MASK;
out->fpstate_size = PAGE_SIZE - fpstate_start;
}
</font>
</pre>
</tt>
<p>
<b>ptrace.h</b>
<p>
ptrace.h deals with the machine's register set. It defines the
following:
<tt>
<pre>
<font size="-1">
struct uml_pt_regs
</font>
</pre>
</tt>
which should contain a system call number, a set of system call
arguments, a flag saying whether the kernel was entered from userspace
or kernelspace, and a pointer to the sigcontext structure on the
current stack. The i386 definition is
<tt>
<pre>
<font size="-1">
struct uml_pt_regs {
unsigned long args[6];
long syscall;
int is_user;
void *sc;
};
</font>
</pre>
</tt>
which may in fact be architecture-independent. The only thing that
may need changing is the size of args[].
<tt>
<pre>
<font size="-1">
EMPTY_UML_PT_REGS
</font>
</pre>
</tt>
which is an initializer for uml_pt_regs. The i386 definition is
<tt>
<pre>
<font size="-1">
#define EMPTY_UML_PT_REGS { \
syscall : -1, \
args : { [0 ... 5] = 0 }, \
is_user : 0, \
sc : NULL }
</font>
</pre>
</tt>
<tt>
<pre>
<font size="-1">
UPT_REG(regs, reg)
</font>
</pre>
</tt>
which returns the value of the appropriate register from the saved
sigcontext.
<tt>
<pre>
<font size="-1">
UPT_SET(regs, reg, val)
</font>
</pre>
</tt>
which sets the value of the appropriate register in the saved
sigcontext to whatever value is passed in. The i386 definitions of
these are big switch statements
<tt>
<pre>
<font size="-1">
#define UPT_REG(regs, reg) \
({ unsigned long val; \
switch(reg){ \
case EIP: val = UPT_IP(regs); break; \
case UESP: val = UPT_SP(regs); break; \
case EAX: val = UPT_EAX(regs); break; \
case EBX: val = UPT_EBX(regs); break; \
case ECX: val = UPT_ECX(regs); break; \
case EDX: val = UPT_EDX(regs); break; \
case ESI: val = UPT_ESI(regs); break; \
case EDI: val = UPT_EDI(regs); break; \
case EBP: val = UPT_EBP(regs); break; \
case ORIG_EAX: val = UPT_ORIG_EAX(regs); break; \
case CS: val = UPT_CS(regs); break; \
case SS: val = UPT_SS(regs); break; \
case DS: val = UPT_DS(regs); break; \
case ES: val = UPT_ES(regs); break; \
case FS: val = UPT_FS(regs); break; \
case GS: val = UPT_GS(regs); break; \
case EFL: val = UPT_EFLAGS(regs); break; \
default : \
panic("Bad register in UPT_REG : %d\n", reg); \
val = -1; \
} \
val; \
})
#define UPT_SET(regs, reg, val) \
do { \
switch(reg){ \
case EIP: UPT_IP(regs) = val; break; \
case UESP: UPT_SP(regs) = val; break; \
case EAX: UPT_EAX(regs) = val; break; \
case EBX: UPT_EBX(regs) = val; break; \
case ECX: UPT_ECX(regs) = val; break; \
case EDX: UPT_EDX(regs) = val; break; \
case ESI: UPT_ESI(regs) = val; break; \
case EDI: UPT_EDI(regs) = val; break; \
case EBP: UPT_EBP(regs) = val; break; \
case ORIG_EAX: UPT_ORIG_EAX(regs) = val; break; \
case CS: UPT_CS(regs) = val; break; \
case SS: UPT_SS(regs) = val; break; \
case DS: UPT_DS(regs) = val; break; \
case ES: UPT_ES(regs) = val; break; \
case FS: UPT_FS(regs) = val; break; \
case GS: UPT_GS(regs) = val; break; \
case EFL: UPT_EFLAGS(regs) = val; break; \
default : \
panic("Bad register in UPT_SET : %d\n", reg); \
break; \
} \
} while (0)
</font>
</pre>
</tt>
In addition to whatever macros you call from any additional PT_REGS_*
macros you define, there are a few that you definitely need
equivalents to. These will generally call into sigcontext macros
since they need to modify the current sigcontext.
<ul>
<li>
UPT_SET_SYSCALL_RETURN
<br>
This sets the system call return value
</li>
<li>
UPT_RESTART_SYSCALL
<br>
This does whatever is necessary to make sure that the current system
call will restart when userspace is entered. Backing up the IP so
that it points at the system call instruction is probably enough.
</li>
<li>
UPT_ORIG_SYSCALL
<br>
This is the original location of the system call number. i386 moves
it from EAX to ORIG_EAX, so it refers to EAX. This is used when
restarting a system call to restore the registers to their original
values.
</li>
<li>
UPT_SYSCALL_NR
<br>
This pulls the system call number from the uml_pt_regs. On i386, it
comes directly from the uml_pt_regs structure. On other
architectures, it may make sense to get it from the sigcontext.
</li>
<li>
UPT_SYSCALL_RET
<br>
This returns the system call return value.
</li>
</ul>
<p>
<b>ptrace_user.h</b>
<p>
This file defines a set of access macros into the hosts's pt_regs
structure. This is purely a userspace file which is used by parts of
UML which use ptrace to pull the process registers from the host
kernel and need to interpret them. These are the definitions that
should be here, and they should be implemented in terms of register
definitions in the host <asm/ptrace.h>.
<ul>
<li>
PT_SYSCALL_NR
</li>
<li>
PT_SYSCALL_NR_OFFSET
<br>
The pt_regs index and ptrace offset of the system call number
</li>
<li>
PT_SYSCALL_ARG1_OFFSET
</li>
<li>
PT_SYSCALL_ARG2_OFFSET
</li>
<li>
PT_SYSCALL_ARG3_OFFSET
</li>
<li>
PT_SYSCALL_ARG4_OFFSET
</li>
<li>
PT_SYSCALL_ARG5_OFFSET
<br>
The offsets of the system call arguments
</li>
<li>
PT_SYSCALL_RET_OFFSET
<br>
The offset of the system call return value
</li>
<li>
PT_IP
</li>
<li>
PT_IP_OFFSET
<br>
The pt_regs index and offset of the IP
</li>
<li>
PT_SP
<br>
The index of the stack pointer
</li>
<li>
FRAME_SIZE
</li>
<li>
FRAME_SIZE_OFFSET
<br>
If the host pt_regs doesn't define FRAME_SIZE, set it to the number of
general purpose registers. Set FRAME_SIZE_OFFSET to the maximum
offset for PTRACE_GETREGS.
</li>
<li>
FP_FRAME_SIZE
</li>
<li>
FPX_FRAME_SIZE
<br>
These are the number of floating point registers and extended floating
point registers, respectively. The second is likely x86-specific. If
you don't define UM_HAVE_GETFPREGS/UM_HAVE_SETFPREGS or
UM_HAVE_GETFPXREGS/UM_HAVE_SETFPXREGS (see below), you can leave the
corresponding _FRAME_SIZE undefined.
</li>
<li>
UM_HAVE_GETREGS
</li>
<li>
UM_HAVE_SETREGS
</li>
<li>
UM_HAVE_GETFPREGS
</li>
<li>
UM_HAVE_SETFPREGS
</li>
<li>
UM_HAVE_GETFPXREGS
</li>
<li>
UM_HAVE_SETFPXREGS
<br>
These should be defined if the architecture defines PTRACE_GETREGS,
PTRACE_SETREGS, PTRACE_GETFPREGS, PTRACE_SETFPREGS, PTRACE_GETFPXREGS,
PTRACE_,SETFPXREGS respectively.
</li>
</ul>
<p>
<b>sigcontext.h</b>
<p>
sigcontext.h defines a few sigcontext-related macros. Some of them
are accessed through the PT_REGS and UPT_REGS macros, so their
details are up to you. You will have to define similar things at the
very least, so here's what i386 defines
<ul>
<li>
SC_RESTART_SYSCALL
<br>
Does whatever IP fiddling is needed to cause the current system call
to restart when userspace is re-entered. On i386, this just subtracts
2 from the IP since a system call instruction is 2 bytes long.
</li>
<li>
SC_SET_SYSCALL_RETURN
<br>
Sets the system call return value. On i386, this just sets %eax. On
ppc, it does a little more than that.
</li>
<li>
SC_FAULT_ADDR
</li>
<li>
SC_FAULT_WRITE
<br>
These two are called from generic UML code, so you have to implement
these as described. On a segfault, they pick out from a sigcontext
the fault address and whether the fault was on a write.
</li>
<li>
SEGV_IS_FIXABLE
<br>
This evaluates to non-zero if a segfault is one that can be fixed by
mapping in a page or changing page protections. If not, then it
returns zero, and the faulting process will simply be segfaulted.
This is called from generic UML code.
</li>
<li>
SC_START_SYSCALL
<br>
This is a general hook that's called at the start of a system call
before ptrace gets to see it. On x86, strace expects %eax to contain
-ENOSYS when it sees a system call entry. So, this macro sets that.
On other architectures, this may do nothing. This is called from
generic UML code.
</li>
<li>
void sc_to_regs(struct uml_pt_regs *regs, struct sigcontext *sc,
unsigned long syscall)
<br>
This is called at the beginning of a system call to fill in a pt_regs
structure from the sigcontext. It parses the system call from the
sigcontext and fills in the system call number and the arguments in
the pt_regs. This is called from generic UML code.
</li>
</ul>
<p>
<p>
<b>syscalls.h</b>
<p>
syscalls.h defines any architecture-specific system calls. It does so
by defining
<ul>
<li>
ARCH_SYSCALLS
<br> which is a set of array element initializers which will be
included in the initialization of the system call table. The i386
ARCH_SYSCALLS looks in part like this
<tt>
<pre>
<font size="-1">
#define ARCH_SYSCALLS \
[ __NR_mmap ] = old_mmap_i386, \
[ __NR_select ] = old_select, \
[ __NR_vm86old ] = sys_ni_syscall, \
[ __NR_modify_ldt ] = sys_modify_ldt, \
...
[ 222 ] = sys_ni_syscall,
</font>
</pre>
</tt>
</li>
</ul>
Also, you must define
<ul>
<li>
LAST_ARCH_SYSCALL
<br>
to be the last initialized element defined by ARCH_SYSCALLS.
This is used to fill the end of the system call table properly.
</li>
</ul>
In addition, syscalls.h defines
<ul>
<li>
EXECUTE_SYSCALL(syscall, regs)
<br>
which calls into a system call entry. The i386 pt_regs has been
arranged so that it can just be dumped on the stack and the right
thing will happen
<tt>
<pre>
<font size="-1">
#define EXECUTE_SYSCALL(syscall, regs) (*sys_call_table[syscall])(*regs);
</font>
</pre>
</tt>
</li>
</ul>
</blockquote>
<a name="Port implementation"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Port implementation</font>
</b>
</td>
</tr>
</table>
<blockquote head="Port implementation">
The actual implementation of the port is contained in sys-$(SUBARCH).
You have complete freedom in this directory, except that when it is
built, it must produce an object file named sys.o which contains all
the code required by the generic kernel.
<p>
Here is a list of the files used by the existing ports, along with
what they define.
<p>
<b>ptrace.c</b>
<tt>
<pre>
<font size="-1">
int putreg(struct task_struct *child, unsigned long regno, unsigned long value)
</font>
</pre>
</tt>
This does any needed validity checking on the register and the value,
and assigns the value to the appropriate register in
child->thread.process_regs. If it fails, it returns -EIO.
<p>
This may be changed in the future so that it is provided with just the
register set rather than the whole task structure.
<tt>
<pre>
<font size="-1">
unsigned long getreg(struct task_struct *child, unsigned long regno)
</font>
</pre>
</tt>
getreg fetches the value of the requested register from
child->thread.process_regs, doing any required masking of registers
which don't use all their bits. This may also be changed to take the
register set rather than the task structure.
<p>
<b>ptrace_user.c</b>
<p>
Linux doesn't implement PTRACE_SETREGS and PTRACE_GETREGS on all
architectures. This file contains definitions of ptrace_getregs and
ptrace_setregs to hide this difference from the generic code.
Architectures which define PTRACE_SETREGS and PTRACE_GETREGS will
implement these functions as follows
<tt>
<pre>
<font size="-1">
int ptrace_getregs(long pid, struct sys_pt_regs *regs_out)
{
return(ptrace(PTRACE_GETREGS, pid, 0, regs_out));
}
int ptrace_setregs(long pid, struct sys_pt_regs *regs)
{
return(ptrace(PTRACE_SETREGS, pid, 0, regs));
}
</font>
</pre>
</tt>
Architectures which don't will implement them as loops which call
ptrace(PTRACE_GETREG, ...) or ptrace(PTRACE_SETREG, ...) for each register.
<p>
<b>semaphore.c</b>
<p>
This implements the architecture's semaphore primitives. It is highly
recommended to steal this from the underlying architecture by having
the Makefile make a link from arch/$(SUBARCH)/kernel/semaphore.c to
arch/um/sys-$(SUBARCH)/semaphore.c.
<p>
<b>checksum.c or checksum.S</b>
<p>
This implements the architecture's ip checksumming. This is
stolen from the underlying architecture in the same manner as
semaphore.c.
<p>
<b>sigcontext.c</b>
<p>
This defines a few sigcontext-related functions
<ul>
<li>
int sc_size(void *data)
<br>
How big is a sigcontext? On x86, this takes the floating point state
into account as well as just the sigcontext structure itself.
</li>
<li>
int copy_sc_to_user(void *to_ptr, void *from_ptr, void *data)
<br>
Copies a sigcontext to a process stack. It must use copy_to_user, not
memcopy.
</li>
<li>
int copy_sc_from_user(void *to_ptr, void *from_ptr, void *data)
<br>
Copies a sigcontext from a process stack into kernel memory.
Similarly, this must use copy_from_user, not memcopy.
</li>
<li>
void sc_to_sc(void *to_ptr, void *from_ptr)
<br>
This copies a sigcontext from one kernel stack to another. It is used
during thread creation (kernel_thread(), fork(), or clone()) to
initialize a kernel stack with a signal frame that the new process can
return from.
</li>
</ul>
<p>
<b>sysrq.c</b>
<p>
This needs to define
<ul>
<li>
void show_regs(struct pt_regs *regs)
<br>
For the benefit of the SysRq driver.
</li>
</ul>
<p>
<b>Other files</b>
<p>
If Linux on your architecture defines any private system calls, you
will need to implement them here. Normally, you can take the code
from the underlying architecture, and you might get away with linking
to the files in the other architecture that implement them.
</blockquote>
<a name="Debugging the new port"/><table width="100%" bgcolor="#e0e0e0">
<tr>
<td>
<b>
<font color="black">Debugging the new port</font>
</b>
</td>
</tr>
</table>
<blockquote head="Debugging the new port">
There's no algorithm for doing this stage of the port, so I'll just
describe a number of useful tricks.
<p>
gdb is available. Use it. It's usable for any part of the kernel
after the beginning of start_kernel. If you need to debug anything
before that, the 'debugtrace' option is handle. It causes the tracing
thread to stop and wait to be attached with gdb. Then you can step
through the very early boot before start_kernel.
<p>
If you're post-mortem-ing a bug and you want to see what just happened
inside UML, there are some arrays which store some useful recent
history:
<ul>
<li>
signal_record - stores the last 1024 signals seen by the tracing
thread, including the host pid of the process getting the signal, the
time, and the IP at which the signal happened. This is a circular
buffer and the latest entry is at index signal_index - 1.
</li>
<li>
syscall_record - the same, except it stores process system calls. It
stores the system call number, the return value (and 0xdeadbeef is
stored there if it hasn't returned), the UML pid of the process, and
the time. It is indexed by syscall_index, so the most recent entry is
at index syscall_index - 1.
</li>
</ul>
These provide a decent picture of what UML has been doing lately.
Looking for unusual things here immediately before a bug happened is a
useful debugging technique. Correlating timestamps between the two
arrays is also sometimes useful.
<p>
If you have reproducable memory corruption, an extremely useful way to
track it down is to set the page that it happens on read-only and to
see what seg faults when it tries writing to that page. Obviously,
this only works if there aren't legitimate writes happening to that
page at the same time.
</blockquote>
</td>
</tr>
</table>
<center>
<font size="-1">Hosted at </font>
<a href="http://sourceforge.net">
<img src="http://sourceforge.net/sflogo.php?group_id=429" width="88" height="31" border="0" alt="SourceForge Logo">
</a>
</center>
</body>
</html>
|
