On Thu, 4 Nov 2021 at 05:20, Lianjun Huang <hljhnu@xxxxxxxxx> wrote:
>
> Change-Id: I361d582f30a04040969f1774064d5d1a4b646389
> Signed-off-by: Lianjun Huang <hljhnu@xxxxxxxxx>
> ---
> fs/fuse/dev.c | 4 ++++
> include/trace/events/fuse.h | 27 +++++++++++++++++++++++++++
> 2 files changed, 31 insertions(+)
> create mode 100644 include/trace/events/fuse.h
>
> diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
> index 9d2d321bd60b..83f20799683d 100644
> --- a/fs/fuse/dev.c
> +++ b/fs/fuse/dev.c
> @@ -23,6 +23,8 @@
> #include <linux/splice.h>
> #include <linux/sched.h>
> #include <linux/freezer.h>
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/fuse.h>
>
> MODULE_ALIAS_MISCDEV(FUSE_MINOR);
> MODULE_ALIAS("devname:fuse");
> @@ -323,6 +325,7 @@ static u64 fuse_get_unique(struct fuse_iqueue *fiq)
>
> static void queue_request(struct fuse_iqueue *fiq, struct fuse_req *req)
> {
> + trace_fuse_info(req->in.h.opcode, req->in.h.unique, req->in.h.nodeid, "queue request");
I'm very much reluctant to add tracepoints to fuse at all.
Fuse protocol is traceable on the syscall interface (with strace) so I
don't see how this adds any value.
Can you please explain why you need this and why is tracing on the
userspace ABI not sufficient?
Thanks,
Miklos
