On Sat, Nov 02, 2019 at 05:49:00PM +0530, Amit Daniel Kachhap wrote:Ok sure. It makes sense.
Hi Mark,
Hi Amit,
Thanks for the review!
On 10/29/19 10:28 PM, Mark Rutland wrote:
Architectures may need to perform special initialization of ftraces/an mcount/a mcount.
callsites, and today they do so by special-casing ftrace_make_nop() when
the expected branch address is MCOUNT_ADDR. In some cases (e.g. for
patchable-function-entry), we don't have an mcount-like symbol and don't
want a synthetic MCOUNT_ADDR, but we may need to perform someSame as above.
initialization of callsites.
To make it possible to separate initialization from runtime
modification, and to handle cases without an mcount-like symbol, this
Using 'an' in both of these cases is correct, even though 'mcount'
starts with a consonant, since it's pronounced as-if it were 'emcount'.
I will leave this as-is.
ok Its fine this way too. Actually from the comment, I could not understand which one of the compiler contents this points to as in this case there are 2 nops.
patch adds an optional ftrace_init_nop() function that architectures canNit: Will it be better to write it as "@rec->ip should store the adjusted
implement, which does not pass a branch address.
Where an architecture does not provide ftrace_init_nop(), we will fall
back to the existing behaviour of calling ftrace_make_nop() with
MCOUNT_ADDR.
At the same time, ftrace_code_disable() is renamed to
ftrace_nop_initialize() to make it clearer that it is intended to
intialize a callsite into a disabled state, and is not for disabling a
callsite that has been runtime enabled. The kerneldoc description of rec
arguments is updated to cover non-mcount callsites.
Signed-off-by: Mark Rutland <mark.rutland@xxxxxxx>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
Cc: Torsten Duwe <duwe@xxxxxxx>
---
include/linux/ftrace.h | 35 ++++++++++++++++++++++++++++++++---
kernel/trace/ftrace.c | 6 +++---
2 files changed, 35 insertions(+), 6 deletions(-)
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 8a8cb3c401b2..9867d90d635e 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -499,7 +499,7 @@ static inline int ftrace_disable_ftrace_graph_caller(void) { return 0; }
/**
* ftrace_make_nop - convert code into nop
* @mod: module structure if called by module load initialization
- * @rec: the mcount call site record
+ * @rec: the call site record (e.g. mcount/fentry)
* @addr: the address that the call site should be calling
*
* This is a very sensitive operation and great care needs
@@ -520,9 +520,38 @@ static inline int ftrace_disable_ftrace_graph_caller(void) { return 0; }
extern int ftrace_make_nop(struct module *mod,
struct dyn_ftrace *rec, unsigned long addr);
+
+/**
+ * ftrace_init_nop - initialize a nop call site
+ * @mod: module structure if called by module load initialization
+ * @rec: the call site record (e.g. mcount/fentry)
+ *
+ * This is a very sensitive operation and great care needs
+ * to be taken by the arch. The operation should carefully
+ * read the location, check to see if what is read is indeed
+ * what we expect it to be, and then on success of the compare,
+ * it should write to the location.
+ *
+ * The code segment at @rec->ip should contain the contents created by
+ * the compiler
ftrace entry address of the call site" or something like that.
This was the specific wording requested by Steve, and it's trying to
describe the instructions at rec->ip, rather than the value of rec->ip,
so I think it's better to leave this as-is.
ok.
+ *Now that ftrace_init_nop is also an arch implemented function so it may be
+ * Return must be:
+ * 0 on success
+ * -EFAULT on error reading the location
+ * -EINVAL on a failed compare of the contents
+ * -EPERM on error writing to the location
+ * Any other value will be considered a failure.
+ */
+#ifndef ftrace_init_nop
+static inline int ftrace_init_nop(struct module *mod, struct dyn_ftrace *rec)
+{
+ return ftrace_make_nop(mod, rec, MCOUNT_ADDR);
+}
+#endif
+
added in Documentation/trace/ftrace-design.rst along with ftrace_make_nop.
In general also, adding some description about patchable-function-entry
in kernel Documentation will be useful.
I agree that would be a good thing, but as Steve suggests I will leave
this for subsequent rework, as I think that also implies more
rework/renaming in the core code (e.g. to more cleanly separate the
notion of a callsite from mcount specifically).
Steve, if you'd like help with that (or review), I'd be happy to help.
Thanks,
Mark.