Chuck,
I forgot to add you to the Cc of the series, but you can download it from
here:
https://patchwork.kernel.org/series/604304/mbox/
-- Steve
On Mon, 10 Jan 2022 19:17:03 -0500
Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> Chuck Lever reported that adding the following:
>
> #define field_sockaddr(field, len) __dynamic_array(u8, field, len)
> #define assign_sockaddr(dest, src, len) memcpy(__get_dynamic_array(dest), src, len)
> #define __get_sockaddr(field) ((struct sockaddr *)__get_dynamic_array(field))
>
> TRACE_EVENT(sockaddr_example,
> TP_PROTO(
> const struct sockaddr *sap,
> size_t salen
> ),
> TP_ARGS(sap, salen),
> TP_STRUCT__entry(
> __field_sockaddr(addr, salen)
> ),
> TP_fast_assign(
> __assign_sockaddr(addr, sap, salen);
> ),
> TP_printk("addr=%pIS", __get_sockaddr(addr))
> );
>
> Causes trace-cmd to report:
>
> "addr=ARG TYPE NOT FIELD BUT 7"
>
> Which is not only unwanted, but rather unhelpful.
>
> Fix it to allow %pIS to work with dynamic arrays, and also allow all fields to
> still work if they are typecasted.
>
> Link: https://lore.kernel.org/all/164182978641.8391.8277203495236105391.stgit@xxxxxxxxxxxxxxxxxxxxxxx/
>
> Steven Rostedt (2):
> libtraceevent: Do not fail field parsing if field has typecast
> libtraceevent: Allow ipsa arg to use dynamic arrays
>
> src/event-parse.c | 70 ++++++++++++++++++++++++++++++++++++-----------
> 1 file changed, 54 insertions(+), 16 deletions(-)
>
|