Re: [PATCH v2 4/4] net/colo: Match is-enabled probe to tracepoint

[Roman Bolshakov <r.bolshakov@yadro.com>]

On Tue, Jul 21, 2020 at 03:06:57PM +0100, Daniel P. Berrangé wrote:
> On Sat, Jul 18, 2020 at 05:58:56PM +0000, Zhang, Chen wrote:
> > 
> > 
> > > -----Original Message-----
> > > From: Roman Bolshakov <r.bolshakov@yadro.com>
> > > Sent: Friday, July 17, 2020 5:35 PM
> > > To: qemu-devel@nongnu.org
> > > Cc: Daniel P. Berrangé <berrange@redhat.com>; Stefan Hajnoczi
> > > <stefanha@redhat.com>; Cameron Esfahani <dirty@apple.com>; Roman
> > > Bolshakov <r.bolshakov@yadro.com>; Philippe Mathieu-Daudé
> > > <philmd@redhat.com>; Zhang, Chen <chen.zhang@intel.com>; Li Zhijian
> > > <lizhijian@cn.fujitsu.com>; Jason Wang <jasowang@redhat.com>
> > > Subject: [PATCH v2 4/4] net/colo: Match is-enabled probe to tracepoint
> > > 
> > > Build of QEMU with dtrace fails on macOS:
> > > 
> > >   LINK    x86_64-softmmu/qemu-system-x86_64
> > > error: probe colo_compare_miscompare doesn't exist
> > > error: Could not register probes
> > > ld: error creating dtrace DOF section for architecture x86_64
> > > 
> > > The reason of the error is explained by Adam Leventhal [1]:
> > > 
> > >   Note that is-enabled probes don't have the stability magic so I'm not
> > >   sure how things would work if only is-enabled probes were used.
> > > 
> > > net/colo code uses is-enabled probes to determine if other probes should be
> > > used but colo_compare_miscompare itself is not used explicitly.
> > > Linker doesn't include the symbol and build fails.
> > > 
> > > The issue can be resolved if is-enabled probe matches the actual trace point
> > > that is used inside the test. Packet dump toggle is replaced with a compile-
> > > time conditional definition.
> > > 
> > > 1. http://markmail.org/message/6grq2ygr5nwdwsnb
> > > 
> > > Fixes: f4b618360e ("colo-compare: add TCP, UDP, ICMP packet comparison")
> > > Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
> > > Cc: Cameron Esfahani <dirty@apple.com>
> > > Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
> > > ---
> > >  net/colo-compare.c    | 42 ++++++++++++++++++++++--------------------
> > >  net/filter-rewriter.c | 10 ++++++++--
> > >  net/trace-events      |  2 --
> > >  3 files changed, 30 insertions(+), 24 deletions(-)
> 
> 
> > > (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE))
> > > {
> > > +    if (trace_event_get_state_backends(TRACE_COLO_COMPARE_IP_INFO))
> > > {
> > >          char pri_ip_src[20], pri_ip_dst[20], sec_ip_src[20], sec_ip_dst[20];
> > > 
> > >          strcpy(pri_ip_src, inet_ntoa(ppkt->ip->ip_src)); @@ -492,12 +494,12
> > > @@ sec:
> > >          g_queue_push_head(&conn->primary_list, ppkt);
> > >          g_queue_push_head(&conn->secondary_list, spkt);
> > > 
> > > -        if
> > > (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE))
> > > {
> > > -            qemu_hexdump((char *)ppkt->data, stderr,
> > > -                        "colo-compare ppkt", ppkt->size);
> > > -            qemu_hexdump((char *)spkt->data, stderr,
> > > -                        "colo-compare spkt", spkt->size);
> > > -        }
> > > +#ifdef DEBUG_COLO_PACKETS
> > > +        qemu_hexdump((char *)ppkt->data, stderr,
> > > +                     "colo-compare ppkt", ppkt->size);
> > > +        qemu_hexdump((char *)spkt->data, stderr,
> > > +                     "colo-compare spkt", spkt->size); #endif
> > > 
> > >          colo_compare_inconsistency_notify(s);
> > >      }
> > > @@ -533,12 +535,12 @@ static int colo_packet_compare_udp(Packet *spkt,
> > > Packet *ppkt)
> > >                                      ppkt->size - offset)) {
> > >          trace_colo_compare_udp_miscompare("primary pkt size", ppkt->size);
> > >          trace_colo_compare_udp_miscompare("Secondary pkt size", spkt-
> > > >size);
> > > -        if
> > > (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE))
> > > {
> > > -            qemu_hexdump((char *)ppkt->data, stderr, "colo-compare pri pkt",
> > > -                         ppkt->size);
> > > -            qemu_hexdump((char *)spkt->data, stderr, "colo-compare sec pkt",
> > > -                         spkt->size);
> > > -        }
> > > +#ifdef DEBUG_COLO_PACKETS
> > > +        qemu_hexdump((char *)ppkt->data, stderr, "colo-compare pri pkt",
> > > +                     ppkt->size);
> > > +        qemu_hexdump((char *)spkt->data, stderr, "colo-compare sec pkt",
> > > +                     spkt->size);
> > > +#endif
> > 
> > Hi Roman,
> > 
> > I think change the " trace_event_get_state_backends()" to
> > "trace_colo_compare_main("Dump packet hex: ")" is a better choice here.
> > It will keep the original code logic and avoid the problem here.
> 
> That may workaround the immediate bug, but this is still a misuse of the
> tracing code. Use of any trace point should only trigger actions in the
> trace infrastructure.
> 
> If I'm using dtrace backend to monitor events I don't want to see QEMU
> dumping stuff to stderr. Anything written to stderr is going to trigger
> disk I/O writing to the VM's logfile, and is also liable to trigger rate
> limiting which can impact the guest performance.
> 

Hi Daniel, Chen, Stefan,

So, what do we want to do about the series? Do we have an agreement? Is
the patch okay or I should make a change?

BTW. I've found that Apple added trace probes to Hypervisor.framework in
Big Sur and there're fbt probes in AppleHV.kext. Addition of dtrace on
macOS helps to find performance or functional issues. (I'm using the
series in my private branches for debugging).

Thanks,
Roman