Re: [Qemu-devel] [PATCH V4 2/7] CAN bus support to connect bust to Linux host SocketCAN interface.

[Stefan Hajnoczi]

On Fri, Jan 19, 2018 at 10:37:22AM -0300, Philippe Mathieu-Daudé wrote:
> On 01/19/2018 05:51 AM, Pavel Pisa wrote:
> > On Tuesday 16 of January 2018 01:12:09 Philippe Mathieu-Daudé wrote:
> >> On 01/15/2018 06:29 PM, Pavel Pisa wrote:
> > But if the second format is preferred then I update the patch.
> > 
> >>> Trace events seems as nice feature. But simple text printf
> >>> like output has been enough till now for CAN testing.
> >>> There is no debugging output in production build.
> >>> So I would add tracing infrastructure later if needed.
> >>
> >> They are as useful as console printf, but less invasive and more
> >> powerful: you can use a much precise timing, select which set of events
> >> to display without having to recompile.
> >> The default backend behaves as console printf.
> >>
> >> You should try them :)
> > 
> > I have tried them on 
> > 
> >   pci_update_mappings_del
> >   pci_update_mappings_add
> >   pci_cfg_write
> > 
> > and they work great. They would be nice for SJA1000
> > register accesses, PCI boards configuration etc.
> > I am not sure how to use them for CAN messages
> > which has a variable length data field.
> 
> Yes, I hit this problem with variable length data in the SD Bus.
> 
> Let's ask Stefan what is the best approach (he said "Tracing is not
> great for dumping large amounts of data")

Yes, tracing isn't pcap.  It's not designed for dumping payloads -
especially when the length is more than 64 bytes or so.

I suggest writing out a pcap file so wireshark or tcpdump can be used to
analyze it.  The code is pretty simple, see net/dump.c.

Stefan

signature.asc
Description: PGP signature