Re: [PATCH bpf-next v2 2/5] nbd: trace sending nbd requests

[Matt Mullins <mmullins@fb.com>]

On Sat, 2019-04-06 at 20:10 -0400, Josef Bacik wrote:
> On Fri, Apr 05, 2019 at 04:55:03PM -0700, Matt Mullins wrote:
> > This adds a tracepoint that can both observe the nbd request being sent
> > to the server, as well as modify that request , e.g., setting a flag in
> > the request that will cause the server to collect detailed tracing data.
> > 
> > The struct request * being handled is included to permit correlation
> > with the block tracepoints.
> > 
> > Signed-off-by: Matt Mullins <mmullins@fb.com>
> > ---
> >  MAINTAINERS                |  1 +
> >  drivers/block/nbd.c        |  5 ++++
> >  include/trace/events/nbd.h | 56 ++++++++++++++++++++++++++++++++++++++
> >  3 files changed, 62 insertions(+)
> >  create mode 100644 include/trace/events/nbd.h
> > 
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index e17ebf70b548..6db583d2b0ea 100644
> > --- a/MAINTAINERS
> > +++ b/MAINTAINERS
> > @@ -10737,6 +10737,7 @@ L:	linux-block@vger.kernel.org
> >  L:	nbd@other.debian.org
> >  F:	Documentation/blockdev/nbd.txt
> >  F:	drivers/block/nbd.c
> > +F:	include/trace/events/nbd.h
> >  F:	include/uapi/linux/nbd.h
> >  
> >  NETWORK DROP MONITOR
> > diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> > index 90ba9f4c03f3..7393d04d255c 100644
> > --- a/drivers/block/nbd.c
> > +++ b/drivers/block/nbd.c
> > @@ -44,6 +44,9 @@
> >  #include <linux/nbd-netlink.h>
> >  #include <net/genetlink.h>
> >  
> > +#define CREATE_TRACE_POINTS
> > +#include <trace/events/nbd.h>
> > +
> >  static DEFINE_IDR(nbd_index_idr);
> >  static DEFINE_MUTEX(nbd_index_mutex);
> >  static int nbd_total_devices = 0;
> > @@ -526,6 +529,8 @@ static int nbd_send_cmd(struct nbd_device *nbd, struct nbd_cmd *cmd, int index)
> >  	handle = nbd_cmd_handle(cmd);
> >  	memcpy(request.handle, &handle, sizeof(handle));
> >  
> > +	trace_nbd_send_request(&request, nbd->index, blk_mq_rq_from_pdu(cmd));
> > +
> 
> Just use the handle, not the pointer.  Thanks,

Assuming you're talking about the struct request *, we are tracing
things through from blk_mq_start_request, which emits tracepoints like

DEFINE_EVENT(block_rq, block_rq_issue,
	TP_PROTO(struct request_queue *q, struct request *rq),

We need at least one of the tracepoints to observe a correlation
between NBD handle (in this case, embedded in the struct nbd_request)
and the struct request.  That said, I don't actually mind if they're
the hashed kind of pointers, as long as the rest of the block
tracepoints do the same...