1		function tracer guts
2		====================
3		By Mike Frysinger
4
5Introduction
6------------
7
8Here we will cover the architecture pieces that the common function tracing
9code relies on for proper functioning.  Things are broken down into increasing
10complexity so that you can start simple and at least get basic functionality.
11
12Note that this focuses on architecture implementation details only.  If you
13want more explanation of a feature in terms of common code, review the common
14ftrace.txt file.
15
16Ideally, everyone who wishes to retain performance while supporting tracing in
17their kernel should make it all the way to dynamic ftrace support.
18
19
20Prerequisites
21-------------
22
23Ftrace relies on these features being implemented:
24 STACKTRACE_SUPPORT - implement save_stack_trace()
25 TRACE_IRQFLAGS_SUPPORT - implement include/asm/irqflags.h
26
27
28HAVE_FUNCTION_TRACER
29--------------------
30
31You will need to implement the mcount and the ftrace_stub functions.
32
33The exact mcount symbol name will depend on your toolchain.  Some call it
34"mcount", "_mcount", or even "__mcount".  You can probably figure it out by
35running something like:
36	$ echo 'main(){}' | gcc -x c -S -o - - -pg | grep mcount
37	        call    mcount
38We'll make the assumption below that the symbol is "mcount" just to keep things
39nice and simple in the examples.
40
41Keep in mind that the ABI that is in effect inside of the mcount function is
42*highly* architecture/toolchain specific.  We cannot help you in this regard,
43sorry.  Dig up some old documentation and/or find someone more familiar than
44you to bang ideas off of.  Typically, register usage (argument/scratch/etc...)
45is a major issue at this point, especially in relation to the location of the
46mcount call (before/after function prologue).  You might also want to look at
47how glibc has implemented the mcount function for your architecture.  It might
48be (semi-)relevant.
49
50The mcount function should check the function pointer ftrace_trace_function
51to see if it is set to ftrace_stub.  If it is, there is nothing for you to do,
52so return immediately.  If it isn't, then call that function in the same way
53the mcount function normally calls __mcount_internal -- the first argument is
54the "frompc" while the second argument is the "selfpc" (adjusted to remove the
55size of the mcount call that is embedded in the function).
56
57For example, if the function foo() calls bar(), when the bar() function calls
58mcount(), the arguments mcount() will pass to the tracer are:
59	"frompc" - the address bar() will use to return to foo()
60	"selfpc" - the address bar() (with mcount() size adjustment)
61
62Also keep in mind that this mcount function will be called *a lot*, so
63optimizing for the default case of no tracer will help the smooth running of
64your system when tracing is disabled.  So the start of the mcount function is
65typically the bare minimum with checking things before returning.  That also
66means the code flow should usually be kept linear (i.e. no branching in the nop
67case).  This is of course an optimization and not a hard requirement.
68
69Here is some pseudo code that should help (these functions should actually be
70implemented in assembly):
71
72void ftrace_stub(void)
73{
74	return;
75}
76
77void mcount(void)
78{
79	/* save any bare state needed in order to do initial checking */
80
81	extern void (*ftrace_trace_function)(unsigned long, unsigned long);
82	if (ftrace_trace_function != ftrace_stub)
83		goto do_trace;
84
85	/* restore any bare state */
86
87	return;
88
89do_trace:
90
91	/* save all state needed by the ABI (see paragraph above) */
92
93	unsigned long frompc = ...;
94	unsigned long selfpc = <return address> - MCOUNT_INSN_SIZE;
95	ftrace_trace_function(frompc, selfpc);
96
97	/* restore all state needed by the ABI */
98}
99
100Don't forget to export mcount for modules !
101extern void mcount(void);
102EXPORT_SYMBOL(mcount);
103
104
105HAVE_FUNCTION_GRAPH_TRACER
106--------------------------
107
108Deep breath ... time to do some real work.  Here you will need to update the
109mcount function to check ftrace graph function pointers, as well as implement
110some functions to save (hijack) and restore the return address.
111
112The mcount function should check the function pointers ftrace_graph_return
113(compare to ftrace_stub) and ftrace_graph_entry (compare to
114ftrace_graph_entry_stub).  If either of those is not set to the relevant stub
115function, call the arch-specific function ftrace_graph_caller which in turn
116calls the arch-specific function prepare_ftrace_return.  Neither of these
117function names is strictly required, but you should use them anyway to stay
118consistent across the architecture ports -- easier to compare & contrast
119things.
120
121The arguments to prepare_ftrace_return are slightly different than what are
122passed to ftrace_trace_function.  The second argument "selfpc" is the same,
123but the first argument should be a pointer to the "frompc".  Typically this is
124located on the stack.  This allows the function to hijack the return address
125temporarily to have it point to the arch-specific function return_to_handler.
126That function will simply call the common ftrace_return_to_handler function and
127that will return the original return address with which you can return to the
128original call site.
129
130Here is the updated mcount pseudo code:
131void mcount(void)
132{
133...
134	if (ftrace_trace_function != ftrace_stub)
135		goto do_trace;
136
137+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
138+	extern void (*ftrace_graph_return)(...);
139+	extern void (*ftrace_graph_entry)(...);
140+	if (ftrace_graph_return != ftrace_stub ||
141+	    ftrace_graph_entry != ftrace_graph_entry_stub)
142+		ftrace_graph_caller();
143+#endif
144
145	/* restore any bare state */
146...
147
148Here is the pseudo code for the new ftrace_graph_caller assembly function:
149#ifdef CONFIG_FUNCTION_GRAPH_TRACER
150void ftrace_graph_caller(void)
151{
152	/* save all state needed by the ABI */
153
154	unsigned long *frompc = &...;
155	unsigned long selfpc = <return address> - MCOUNT_INSN_SIZE;
156	/* passing frame pointer up is optional -- see below */
157	prepare_ftrace_return(frompc, selfpc, frame_pointer);
158
159	/* restore all state needed by the ABI */
160}
161#endif
162
163For information on how to implement prepare_ftrace_return(), simply look at the
164x86 version (the frame pointer passing is optional; see the next section for
165more information).  The only architecture-specific piece in it is the setup of
166the fault recovery table (the asm(...) code).  The rest should be the same
167across architectures.
168
169Here is the pseudo code for the new return_to_handler assembly function.  Note
170that the ABI that applies here is different from what applies to the mcount
171code.  Since you are returning from a function (after the epilogue), you might
172be able to skimp on things saved/restored (usually just registers used to pass
173return values).
174
175#ifdef CONFIG_FUNCTION_GRAPH_TRACER
176void return_to_handler(void)
177{
178	/* save all state needed by the ABI (see paragraph above) */
179
180	void (*original_return_point)(void) = ftrace_return_to_handler();
181
182	/* restore all state needed by the ABI */
183
184	/* this is usually either a return or a jump */
185	original_return_point();
186}
187#endif
188
189
190HAVE_FUNCTION_GRAPH_FP_TEST
191---------------------------
192
193An arch may pass in a unique value (frame pointer) to both the entering and
194exiting of a function.  On exit, the value is compared and if it does not
195match, then it will panic the kernel.  This is largely a sanity check for bad
196code generation with gcc.  If gcc for your port sanely updates the frame
197pointer under different optimization levels, then ignore this option.
198
199However, adding support for it isn't terribly difficult.  In your assembly code
200that calls prepare_ftrace_return(), pass the frame pointer as the 3rd argument.
201Then in the C version of that function, do what the x86 port does and pass it
202along to ftrace_push_return_trace() instead of a stub value of 0.
203
204Similarly, when you call ftrace_return_to_handler(), pass it the frame pointer.
205
206
207HAVE_FTRACE_NMI_ENTER
208---------------------
209
210If you can't trace NMI functions, then skip this option.
211
212<details to be filled>
213
214
215HAVE_SYSCALL_TRACEPOINTS
216------------------------
217
218You need very few things to get the syscalls tracing in an arch.
219
220- Support HAVE_ARCH_TRACEHOOK (see arch/Kconfig).
221- Have a NR_syscalls variable in <asm/unistd.h> that provides the number
222  of syscalls supported by the arch.
223- Support the TIF_SYSCALL_TRACEPOINT thread flags.
224- Put the trace_sys_enter() and trace_sys_exit() tracepoints calls from ptrace
225  in the ptrace syscalls tracing path.
226- If the system call table on this arch is more complicated than a simple array
227  of addresses of the system calls, implement an arch_syscall_addr to return
228  the address of a given system call.
229- If the symbol names of the system calls do not match the function names on
230  this arch, define ARCH_HAS_SYSCALL_MATCH_SYM_NAME in asm/ftrace.h and
231  implement arch_syscall_match_sym_name with the appropriate logic to return
232  true if the function name corresponds with the symbol name.
233- Tag this arch as HAVE_SYSCALL_TRACEPOINTS.
234
235
236HAVE_FTRACE_MCOUNT_RECORD
237-------------------------
238
239See scripts/recordmcount.pl for more info.  Just fill in the arch-specific
240details for how to locate the addresses of mcount call sites via objdump.
241This option doesn't make much sense without also implementing dynamic ftrace.
242
243
244HAVE_DYNAMIC_FTRACE
245-------------------
246
247You will first need HAVE_FTRACE_MCOUNT_RECORD and HAVE_FUNCTION_TRACER, so
248scroll your reader back up if you got over eager.
249
250Once those are out of the way, you will need to implement:
251	- asm/ftrace.h:
252		- MCOUNT_ADDR
253		- ftrace_call_adjust()
254		- struct dyn_arch_ftrace{}
255	- asm code:
256		- mcount() (new stub)
257		- ftrace_caller()
258		- ftrace_call()
259		- ftrace_stub()
260	- C code:
261		- ftrace_dyn_arch_init()
262		- ftrace_make_nop()
263		- ftrace_make_call()
264		- ftrace_update_ftrace_func()
265
266First you will need to fill out some arch details in your asm/ftrace.h.
267
268Define MCOUNT_ADDR as the address of your mcount symbol similar to:
269	#define MCOUNT_ADDR ((unsigned long)mcount)
270Since no one else will have a decl for that function, you will need to:
271	extern void mcount(void);
272
273You will also need the helper function ftrace_call_adjust().  Most people
274will be able to stub it out like so:
275	static inline unsigned long ftrace_call_adjust(unsigned long addr)
276	{
277		return addr;
278	}
279<details to be filled>
280
281Lastly you will need the custom dyn_arch_ftrace structure.  If you need
282some extra state when runtime patching arbitrary call sites, this is the
283place.  For now though, create an empty struct:
284	struct dyn_arch_ftrace {
285		/* No extra data needed */
286	};
287
288With the header out of the way, we can fill out the assembly code.  While we
289did already create a mcount() function earlier, dynamic ftrace only wants a
290stub function.  This is because the mcount() will only be used during boot
291and then all references to it will be patched out never to return.  Instead,
292the guts of the old mcount() will be used to create a new ftrace_caller()
293function.  Because the two are hard to merge, it will most likely be a lot
294easier to have two separate definitions split up by #ifdefs.  Same goes for
295the ftrace_stub() as that will now be inlined in ftrace_caller().
296
297Before we get confused anymore, let's check out some pseudo code so you can
298implement your own stuff in assembly:
299
300void mcount(void)
301{
302	return;
303}
304
305void ftrace_caller(void)
306{
307	/* save all state needed by the ABI (see paragraph above) */
308
309	unsigned long frompc = ...;
310	unsigned long selfpc = <return address> - MCOUNT_INSN_SIZE;
311
312ftrace_call:
313	ftrace_stub(frompc, selfpc);
314
315	/* restore all state needed by the ABI */
316
317ftrace_stub:
318	return;
319}
320
321This might look a little odd at first, but keep in mind that we will be runtime
322patching multiple things.  First, only functions that we actually want to trace
323will be patched to call ftrace_caller().  Second, since we only have one tracer
324active at a time, we will patch the ftrace_caller() function itself to call the
325specific tracer in question.  That is the point of the ftrace_call label.
326
327With that in mind, let's move on to the C code that will actually be doing the
328runtime patching.  You'll need a little knowledge of your arch's opcodes in
329order to make it through the next section.
330
331Every arch has an init callback function.  If you need to do something early on
332to initialize some state, this is the time to do that.  Otherwise, this simple
333function below should be sufficient for most people:
334
335int __init ftrace_dyn_arch_init(void)
336{
337	return 0;
338}
339
340There are two functions that are used to do runtime patching of arbitrary
341functions.  The first is used to turn the mcount call site into a nop (which
342is what helps us retain runtime performance when not tracing).  The second is
343used to turn the mcount call site into a call to an arbitrary location (but
344typically that is ftracer_caller()).  See the general function definition in
345linux/ftrace.h for the functions:
346	ftrace_make_nop()
347	ftrace_make_call()
348The rec->ip value is the address of the mcount call site that was collected
349by the scripts/recordmcount.pl during build time.
350
351The last function is used to do runtime patching of the active tracer.  This
352will be modifying the assembly code at the location of the ftrace_call symbol
353inside of the ftrace_caller() function.  So you should have sufficient padding
354at that location to support the new function calls you'll be inserting.  Some
355people will be using a "call" type instruction while others will be using a
356"branch" type instruction.  Specifically, the function is:
357	ftrace_update_ftrace_func()
358
359
360HAVE_DYNAMIC_FTRACE + HAVE_FUNCTION_GRAPH_TRACER
361------------------------------------------------
362
363The function grapher needs a few tweaks in order to work with dynamic ftrace.
364Basically, you will need to:
365	- update:
366		- ftrace_caller()
367		- ftrace_graph_call()
368		- ftrace_graph_caller()
369	- implement:
370		- ftrace_enable_ftrace_graph_caller()
371		- ftrace_disable_ftrace_graph_caller()
372
373<details to be filled>
374Quick notes:
375	- add a nop stub after the ftrace_call location named ftrace_graph_call;
376	  stub needs to be large enough to support a call to ftrace_graph_caller()
377	- update ftrace_graph_caller() to work with being called by the new
378	  ftrace_caller() since some semantics may have changed
379	- ftrace_enable_ftrace_graph_caller() will runtime patch the
380	  ftrace_graph_call location with a call to ftrace_graph_caller()
381	- ftrace_disable_ftrace_graph_caller() will runtime patch the
382	  ftrace_graph_call location with nops
383