xref: /netbsd-src/external/gpl3/gdb.old/dist/sim/README-HACKING (revision 22ebeae4b2252475e0ebe332f69734639cb946ea)
1This is a loose collection of notes for people hacking on simulators.
2If this document gets big enough it can be prettied up then.
3
4Contents
5
6- The "common" directory
7- Common Makefile Support
8- TAGS support
9- Generating "configure" files
10- C Language Assumptions
11- "dump" commands under gdb
12
13The "common" directory
14======================
15
16The common directory contains:
17
18- common documentation files (e.g. run.1, and maybe in time .texi files)
19- common source files (e.g. run.c)
20- common Makefile fragment and configury (e.g. Make-common.in, aclocal.m4).
21
22In addition "common" contains portions of the system call support
23(e.g. callback.c, target-newlib-*.c).
24
25Common Makefile Support
26=======================
27
28A common configuration framework is available for simulators that want
29to use it.  The common framework exists to remove a lot of duplication
30in configure.ac and Makefile.in, and it also provides a foundation for
31enhancing the simulators uniformly (e.g. the more they share in common
32the easier a feature added to one is added to all).
33
34The configure.ac of a simulator using the common framework should look like:
35
36--- snip ---
37dnl Process this file with autoconf to produce a configure script.
38AC_INIT(Makefile.in)
39AC_CONFIG_MACRO_DIRS([../common ../.. ../../config])
40
41... target specific additions ...
42
43SIM_AC_OUTPUT
44--- snip ---
45
46SIM_AC_OUTPUT:
47
48- creates the symbolic links defined in sim_link_{files,links}
49- creates config.h
50- creates the Makefile
51
52The Makefile.in of a simulator using the common framework should look like:
53
54--- snip ---
55# Makefile for blah ...
56# Copyright blah ...
57
58## COMMON_PRE_CONFIG_FRAG
59
60# These variables are given default values in COMMON_PRE_CONFIG_FRAG.
61# We override the ones we need to here.
62# Not all of these need to be mentioned, only the necessary ones.
63# In fact it is better to *not* mention ones if the value is the default.
64
65# List of object files, less common parts.
66SIM_OBJS =
67# List of extra dependencies.
68# Generally this consists of simulator specific files included by sim-main.h.
69SIM_EXTRA_DEPS =
70# List of flags to always pass to $(CC).
71SIM_EXTRA_CFLAGS =
72# List of extra libraries to link with.
73SIM_EXTRA_LIBS =
74# Dependency of `clean' to clean any extra files.
75SIM_EXTRA_CLEAN =
76
77## COMMON_POST_CONFIG_FRAG
78
79# Rules need to build $(SIM_OBJS), plus whatever else the target wants.
80
81... target specific rules ...
82--- snip ---
83
84COMMON_{PRE,POST}_CONFIG_FRAG are markers for SIM_AC_OUTPUT to tell it
85where to insert the two pieces of common/Make-common.in.
86The resulting Makefile is created by doing autoconf substitions on
87both the target's Makefile.in and Make-common.in, and inserting
88the two pieces of Make-common.in into the target's Makefile.in at
89COMMON_{PRE,POST}_CONFIG_FRAG.
90
91Note that SIM_EXTRA_{INSTALL,CLEAN} could be removed and "::" targets
92could be used instead.  However, it's not clear yet whether "::" targets
93are portable enough.
94
95TAGS support
96============
97
98Many files generate program symbols at compile time.
99Such symbols can't be found with grep nor do they normally appear in
100the TAGS file.  To get around this, source files can add the comment
101
102/* TAGS: foo1 foo2 */
103
104where foo1, foo2 are program symbols.  Symbols found in such comments
105are greppable and appear in the TAGS file.
106
107Generating "configure" files
108============================
109
110For targets using the common framework, "configure" can be generated
111by running `autoconf'.
112
113To regenerate the configure files for all targets using the common framework:
114
115	$  cd devo/sim
116	$  make -f Makefile.in SHELL=/bin/sh autoconf-common
117
118To add a change-log entry to the ChangeLog file for each updated
119directory (WARNING - check the modified new-ChangeLog files before
120renaming):
121
122	$  make -f Makefile.in SHELL=/bin/sh autoconf-changelog
123	$  more */new-ChangeLog
124	$  make -f Makefile.in SHELL=/bin/sh autoconf-install
125
126In a similar vein, both the configure and config.in files can be
127updated using the sequence:
128
129	$  cd devo/sim
130	$  make -f Makefile.in SHELL=/bin/sh autoheader-common
131	$  make -f Makefile.in SHELL=/bin/sh autoheader-changelog
132	$  more */new-ChangeLog
133	$  make -f Makefile.in SHELL=/bin/sh autoheader-install
134
135To add the entries to an alternative ChangeLog file, use:
136
137	$  make ChangeLog=MyChangeLog ....
138
139
140C Language Assumptions
141======================
142
143An ISO C11 compiler is required, as is an ISO C standard library.
144
145"dump" commands under gdb
146=========================
147
148gdbinit.in contains the following
149
150define dump
151set sim_debug_dump ()
152end
153
154Simulators that define the sim_debug_dump function can then have their
155internal state pretty printed from gdb.
156
157FIXME: This can obviously be made more elaborate.  As needed it will be.
158
159Rebuilding target-newlib-* files
160================================
161
162Checkout a copy of the SIM and LIBGLOSS modules (Unless you've already
163got one to hand):
164
165	$  mkdir /tmp/$$
166	$  cd /tmp/$$
167	$  cvs checkout sim-no-testsuite libgloss-no-testsuite newlib-no-testsuite
168
169Configure things for an arbitrary simulator target (d10v is used here for
170convenience):
171
172	$  mkdir /tmp/$$/build
173	$  cd /tmp/$$/build
174	$  /tmp/$$/devo/configure --target=d10v-elf
175
176In the sim/ directory rebuild the headers:
177
178	$  cd sim/
179	$  make nltvals
180
181If the target uses the common syscall table (libgloss/syscall.h), then you're
182all set!  If the target has a custom syscall table, you need to declare it:
183
184	devo/sim/common/gennltvals.py
185
186		Add your new processor target (you'll need to grub
187		around to find where your syscall.h lives).
188
189	devo/sim/<processor>/*.[ch]
190
191		Include target-newlib-syscall.h instead of syscall.h.
192
193Tracing
194=======
195
196For ports based on CGEN, tracing instrumentation should largely be for free,
197so we will cover the basic non-CGEN setup here.  The assumption is that your
198target is using the common autoconf macros and so the build system already
199includes the sim-trace configure flag.
200
201The full tracing API is covered in sim-trace.h, so this section is an overview.
202
203Before calling any trace function, you should make a call to the trace_prefix()
204function.  This is usually done in the main sim_engine_run() loop before
205simulating the next instruction.  You should make this call before every
206simulated insn.  You can probably copy & paste this:
207  if (TRACE_ANY_P (cpu))
208    trace_prefix (sd, cpu, NULL_CIA, oldpc, TRACE_LINENUM_P (cpu), NULL, 0, "");
209
210You will then need to instrument your simulator code with calls to the
211trace_generic() function with the appropriate trace index.  Typically, this
212will take a form similar to the above snippet.  So to trace instructions, you
213would use something like:
214  if (TRACE_INSN_P (cpu))
215    trace_generic (sd, cpu, TRACE_INSN_IDX, "NOP;");
216
217The exact output format is up to you.  See the trace index enum in sim-trace.h
218to see the different tracing info available.
219
220To utilize the tracing features at runtime, simply use the --trace-xxx flags.
221  run --trace-insn ./some-program
222
223Profiling
224=========
225
226Similar to the tracing section, this is merely an overview for non-CGEN based
227ports.  The full API may be found in sim-profile.h.  Its API is also similar
228to the tracing API.
229
230Note that unlike the tracing command line options, in addition to the profile
231flags, you have to use the --verbose option to view the summary report after
232execution.  Tracing output is displayed on the fly, but the profile output is
233only summarized.
234
235To profile core accesses (such as data reads/writes and insn fetches), add
236calls to PROFILE_COUNT_CORE() to your read/write functions.  So in your data
237fetch function, you'd use something like:
238  PROFILE_COUNT_CORE (cpu, target_addr, size_in_bytes, map_read);
239Then in your data write function:
240  PROFILE_COUNT_CORE (cpu, target_addr, size_in_bytes, map_write);
241And in your insn fetcher:
242  PROFILE_COUNT_CORE (cpu, target_addr, size_in_bytes, map_exec);
243
244To use the PC profiling code, you simply have to tell the system where to find
245your simulator's PC.  So in your model initialization function:
246  CPU_PC_FETCH (cpu) = function_that_fetches_the_pc;
247
248To profile branches, in every location where a branch insn is executed, call
249one of the related helpers:
250  PROFILE_BRANCH_TAKEN (cpu);
251  PROFILE_BRANCH_UNTAKEN (cpu);
252If you have stall information, you can utilize the other helpers too.
253
254Environment Simulation
255======================
256
257The simplest simulator doesn't include environment support -- it merely
258simulates the Instruction Set Architecture (ISA).  Once you're ready to move
259on to the next level, it's time to start handling the --env option.  It's
260enabled by default for all ports already.
261
262This will support for the user, virtual, and operating environments.  See the
263sim-config.h header for a more detailed description of them.  The former are
264pretty straight forward as things like exceptions (making system calls) are
265handled in the simulator.  Which is to say, an exception does not trigger an
266exception handler in the simulator target -- that is what the operating env
267is about.  See the following userspace section for more information.
268
269Userspace System Calls
270======================
271
272By default, the libgloss userspace is simulated.  That means the system call
273numbers and calling convention matches that of libgloss.  Simulating other
274userspaces (such as Linux) is pretty straightforward, but let's first focus
275on the basics.  The basic API is covered in include/sim/callback.h.
276
277When an instruction is simulated that invokes the system call method (such as
278forcing a hardware trap or exception), your simulator code should set up the
279CB_SYSCALL data structure before calling the common cb_syscall() function.
280For example:
281static int
282syscall_read_mem (host_callback *cb, struct cb_syscall *sc,
283		  unsigned long taddr, char *buf, int bytes)
284{
285  SIM_DESC sd = (SIM_DESC) sc->p1;
286  SIM_CPU *cpu = (SIM_CPU *) sc->p2;
287  return sim_core_read_buffer (sd, cpu, read_map, buf, taddr, bytes);
288}
289static int
290syscall_write_mem (host_callback *cb, struct cb_syscall *sc,
291		  unsigned long taddr, const char *buf, int bytes)
292{
293  SIM_DESC sd = (SIM_DESC) sc->p1;
294  SIM_CPU *cpu = (SIM_CPU *) sc->p2;
295  return sim_core_write_buffer (sd, cpu, write_map, buf, taddr, bytes);
296}
297void target_sim_syscall (SIM_CPU *cpu)
298{
299  SIM_DESC sd = CPU_STATE (cpu);
300  host_callback *cb = STATE_CALLBACK (sd);
301  CB_SYSCALL sc;
302
303  CB_SYSCALL_INIT (&sc);
304
305  sc.func = <fetch system call number>;
306  sc.arg1 = <fetch first system call argument>;
307  sc.arg2 = <fetch second system call argument>;
308  sc.arg3 = <fetch third system call argument>;
309  sc.arg4 = <fetch fourth system call argument>;
310  sc.p1 = (PTR) sd;
311  sc.p2 = (PTR) cpu;
312  sc.read_mem = syscall_read_mem;
313  sc.write_mem = syscall_write_mem;
314
315  cb_syscall (cb, &sc);
316
317  <store system call result from sc.result>;
318  <store system call error from sc.errcode>;
319}
320Some targets store the result and error code in different places, while others
321only store the error code when the result is an error.
322
323Keep in mind that the CB_SYS_xxx defines are normalized values with no real
324meaning with respect to the target.  They provide a unique map on the host so
325that it can parse things sanely.  For libgloss, the common/target-newlib-syscall
326file contains the target's system call numbers to the CB_SYS_xxx values.
327
328To simulate other userspace targets, you really only need to update the maps
329pointers that are part of the callback interface.  So create CB_TARGET_DEFS_MAP
330arrays for each set (system calls, errnos, open bits, etc...) and in a place
331you find useful, do something like:
332
333...
334static CB_TARGET_DEFS_MAP cb_linux_syscall_map[] = {
335# define TARGET_LINUX_SYS_open 5
336  { CB_SYS_open, TARGET_LINUX_SYS_open },
337  ...
338  { -1, -1 },
339};
340...
341  host_callback *cb = STATE_CALLBACK (sd);
342  cb->syscall_map = cb_linux_syscall_map;
343  cb->errno_map = cb_linux_errno_map;
344  cb->open_map = cb_linux_open_map;
345  cb->signal_map = cb_linux_signal_map;
346  cb->stat_map = cb_linux_stat_map;
347...
348
349Each of these cb_linux_*_map's are manually declared by the arch target.
350
351The target_sim_syscall() example above will then work unchanged (ignoring the
352system call convention) because all of the callback functions go through these
353mapping arrays.
354
355Events
356======
357
358Events are scheduled and executed on behalf of either a cpu or hardware devices.
359The API is pretty much the same and can be found in common/sim-events.h and
360common/hw-events.h.
361
362For simulator targets, you really just have to worry about the schedule and
363deschedule functions.
364
365Device Trees
366============
367
368The device tree model is based on the OpenBoot specification.  Since this is
369largely inherited from the psim code, consult the existing psim documentation
370for some in-depth details.
371	http://sourceware.org/psim/manual/
372
373Hardware Devices
374================
375
376The simplest simulator doesn't include hardware device support.  Once you're
377ready to move on to the next level, declare in your Makefile.in:
378SIM_EXTRA_HW_DEVICES = devone devtwo devthree
379
380The basic hardware API is documented in common/hw-device.h.
381
382Each device has to have a matching file name with a "dv-" prefix.  So there has
383to be a dv-devone.c, dv-devtwo.c, and dv-devthree.c files.  Further, each file
384has to have a matching hw_descriptor structure.  So the dv-devone.c file has to
385have something like:
386  const struct hw_descriptor dv_devone_descriptor[] = {
387    {"devone", devone_finish,},
388    {NULL, NULL},
389  };
390
391The "devone" string as well as the "devone_finish" function are not hard
392requirements, just common conventions.  The structure name is a hard
393requirement.
394
395The devone_finish() callback function is used to instantiate this device by
396parsing the corresponding properties in the device tree.
397
398Hardware devices typically attach address ranges to themselves.  Then when
399accesses to those addresses are made, the hardware will have its callback
400invoked.  The exact callback could be a normal I/O read/write access, as
401well as a DMA access.  This makes it easy to simulate memory mapped registers.
402
403Keep in mind that like a proper device driver, it may be instantiated many
404times over.  So any device state it needs to be maintained should be allocated
405during the finish callback and attached to the hardware device via set_hw_data.
406Any hardware functions can access this private data via the hw_data function.
407
408Ports (Interrupts / IRQs)
409=========================
410
411First, a note on terminology.  A "port" is an aspect of a hardware device that
412accepts or generates interrupts.  So devices with input ports may be the target
413of an interrupt (accept it), and/or they have output ports so that they may be
414the source of an interrupt (generate it).
415
416Each port has a symbolic name and a unique number.  These are used to identify
417the port in different contexts.  The output port name has no hard relationship
418to the input port name (same for the unique number).  The callback that accepts
419the interrupt uses the name/id of its input port, while the generator function
420uses the name/id of its output port.
421
422The device tree is used to connect the output port of a device to the input
423port of another device.  There are no limits on the number of inputs connected
424to an output, or outputs to an input, or the devices attached to the ports.
425In other words, the input port and output port could be the same device.
426
427The basics are:
428 - each hardware device declares an array of ports (hw_port_descriptor).
429   any mix of input and output ports is allowed.
430 - when setting up the device, attach the array (set_hw_ports).
431 - if the device accepts interrupts, it will have to attach a port callback
432   function (set_hw_port_event)
433 - connect ports with the device tree
434 - handle incoming interrupts with the callback
435 - generate outgoing interrupts with hw_port_event
436