Re: [PATCH] Documentation: ftrace-uses: Change pt_regs to ftrace_regs

From: Steven Rostedt
Date: Mon Jul 29 2024 - 11:27:03 EST


On Sat, 27 Jul 2024 19:45:54 -0300
Marcos Paulo de Souza <mpdesouza@xxxxxxxx> wrote:

> Since commit d19ad0775dcd ("ftrace: Have the callbacks receive a struct
> ftrace_regs instead of pt_regs") the callback function receives a
> ftrace_regs argument, and not a pt_regs anymore. Change the
> documentation to reflect the reality.
>

Thanks, but this should probably add how to use ftrace_regs as well.

For example, if you need the pt_regs() you use ftrace_get_regs(fregs).
That is, if you have a callback:

void callback_func(unsigned long ip, unsigned long parent_ip,
struct ftrace_ops *op, struct ftrace_regs *fregs)
{
struct pt_regs *regs = ftrace_regs(fregs);

then if the function was passed a full set of regs, then regs would be
the pt_regs, like usual. But if the function does not have it, then
regs would be NULL. So it is safe to do:

if (!regs)
return;

// safely play with real regs

There should also be a mention of how if HAVE_DYNAMIC_FTRACE_WITH_ARGS
is set, you can then get to the arguments of the function (or at least
the words of the arguments) with ftrace_regs_get_argument(). Basically,
if we are updating this doc to reflect the changes for using
ftrace_regs, it should be fully changed, not partially.

-- Steve


> Signed-off-by: Marcos Paulo de Souza <mpdesouza@xxxxxxxx>
> ---
> Documentation/trace/ftrace-uses.rst | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/trace/ftrace-uses.rst b/Documentation/trace/ftrace-uses.rst
> index e198854ace79..05620714d99f 100644
> --- a/Documentation/trace/ftrace-uses.rst
> +++ b/Documentation/trace/ftrace-uses.rst
> @@ -87,7 +87,7 @@ The prototype of the callback function is as follows (as of v4.14):
> .. code-block:: c
>
> void callback_func(unsigned long ip, unsigned long parent_ip,
> - struct ftrace_ops *op, struct pt_regs *regs);
> + struct ftrace_ops *op, struct ftrace_regs *regs);
>
> @ip
> This is the instruction pointer of the function that is being traced.
> @@ -104,7 +104,7 @@ The prototype of the callback function is as follows (as of v4.14):
> @regs
> If the FTRACE_OPS_FL_SAVE_REGS or FTRACE_OPS_FL_SAVE_REGS_IF_SUPPORTED
> flags are set in the ftrace_ops structure, then this will be pointing
> - to the pt_regs structure like it would be if an breakpoint was placed
> + to the ftrace_regs structure like it would be if an breakpoint was placed
> at the start of the function where ftrace was tracing. Otherwise it
> either contains garbage, or NULL.
>
> @@ -169,10 +169,10 @@ Some of the flags are used for internal infrastructure of ftrace, but the
> ones that users should be aware of are the following:
>
> FTRACE_OPS_FL_SAVE_REGS
> - If the callback requires reading or modifying the pt_regs
> + If the callback requires reading or modifying the ftrace_regs
> passed to the callback, then it must set this flag. Registering
> a ftrace_ops with this flag set on an architecture that does not
> - support passing of pt_regs to the callback will fail.
> + support passing of ftrace_regs to the callback will fail.
>
> FTRACE_OPS_FL_SAVE_REGS_IF_SUPPORTED
> Similar to SAVE_REGS but the registering of a
> @@ -199,7 +199,7 @@ FTRACE_OPS_FL_IPMODIFY
> Requires FTRACE_OPS_FL_SAVE_REGS set. If the callback is to "hijack"
> the traced function (have another function called instead of the
> traced function), it requires setting this flag. This is what live
> - kernel patches uses. Without this flag the pt_regs->ip can not be
> + kernel patches uses. Without this flag the ftrace_regs->ip can not be
> modified.
>
> Note, only one ftrace_ops with FTRACE_OPS_FL_IPMODIFY set may be
>
> ---
> base-commit: 256abd8e550ce977b728be79a74e1729438b4948
> change-id: 20240727-ftrace-docs-cb-args-963ef5676a81
>
> Best regards,