On Thu, 4 Nov 2021 10:04:32 -0700
Beau Belgrave <beaub@xxxxxxxxxxxxxxxxxxx> wrote:
> Pass iterator through to probes to allow copying data directly to the
> probe buffers instead of taking multiple copies. Enables eBPF user and
> raw iterator types out to programs for no-copy scenarios.
>
> Signed-off-by: Beau Belgrave <beaub@xxxxxxxxxxxxxxxxxxx>
> ---
> kernel/trace/trace_events_user.c | 97 +++++++++++++++++++++++---------
> 1 file changed, 69 insertions(+), 28 deletions(-)
>
> diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
> index b5fe0550b489..d50118b9630a 100644
> --- a/kernel/trace/trace_events_user.c
> +++ b/kernel/trace/trace_events_user.c
> @@ -39,6 +39,10 @@
> #define MAX_EVENT_DESC 512
> #define EVENT_NAME(user_event) ((user_event)->tracepoint.name)
>
> +#define MAX_BPF_COPY_SIZE PAGE_SIZE
> +#define MAX_STACK_BPF_DATA 512
> +#define copy_nofault copy_from_iter_nocache
> +
> static char *register_page_data;
>
> static DEFINE_MUTEX(reg_mutex);
> @@ -63,8 +67,7 @@ struct user_event_refs {
> struct user_event *events[];
> };
>
> -typedef void (*user_event_func_t) (struct user_event *user,
> - void *data, u32 datalen,
> +typedef void (*user_event_func_t) (struct user_event *user, struct iov_iter *i,
> void *tpdata);
>
> static int user_event_parse(char *name, char *args, char *flags,
> @@ -491,7 +494,7 @@ static struct user_event *find_user_event(char *name, u32 *outkey)
> /*
> * Writes the user supplied payload out to a trace file.
> */
> -static void user_event_ftrace(struct user_event *user, void *data, u32 datalen,
> +static void user_event_ftrace(struct user_event *user, struct iov_iter *i,
> void *tpdata)
> {
> struct trace_event_file *file;
> @@ -506,41 +509,82 @@ static void user_event_ftrace(struct user_event *user, void *data, u32 datalen,
> return;
>
> entry = trace_event_buffer_reserve(&event_buffer, file,
> - sizeof(*entry) + datalen);
> + sizeof(*entry) + i->count);
>
> if (unlikely(!entry))
> return;
>
> - memcpy(entry + 1, data, datalen);
> + if (unlikely(!copy_nofault(entry + 1, i->count, i)))
Need:
__trace_event_discard_commit(event_buffer.buffer, event_buffer.event);
Because the trace_event_buffer_reserve() will not only allocate space on
the ring buffer, but may also disable preemption.
-- Steve
> + return;
>
> trace_event_buffer_commit(&event_buffer);
> }
>
> #ifdef CONFIG_PERF_EVENTS
|