Skip to contentSkip to conversation

fix tracing for modules

RFC PATCH Superseded Gerd Hoffmann 3 participants 34 messages Jan 21, 2021 – Apr 15, 2021 Arm/Amlogic Meson SoC support

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover fix tracing for modules

First version that actually works. Only qxl covered for this RfC, other modules will follow once the basics are hashed out.

v3:
- handle initialization of modular tracepoints.

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

More context:
https://bugzilla.redhat.com/show_bug.cgi?id=1898700
https://bugzilla.redhat.com/show_bug.cgi?id=1869642

take care,
Gerd

Gerd Hoffmann (8):
meson: add trace_events_config[]
meson: move up hw subdir (specifically before trace subdir)
meson: add module_trace & module_trace_src
meson: move qxl trace events to separate file
trace: iter init tweaks
trace: add trace_event_iter_init_group trace/simple: pass iter to st_write_event_mapping trace/simple: add st_init_group

 trace/control.h             | 30 ++++++++++++++---
 trace/simple.h              |  1 +
 hw/display/qxl-render.c     |  1 +
 hw/display/qxl.c            |  1 +
 monitor/misc.c              |  4 +--
 trace/control-target.c      |  2 +-
 trace/control.c             | 39 ++++++++++++++++-----
 trace/qmp.c                 |  6 ++--
 trace/simple.c              | 22 +++++++++---
 hw/display/meson.build      |  5 +++
 hw/display/trace-events     | 67 -------------------------------------
 hw/display/trace-events-qxl | 66 ++++++++++++++++++++++++++++++++++++
 meson.build                 |  7 ++--
 trace/meson.build           | 37 +++++++++++++++-----
 14 files changed, 187 insertions(+), 101 deletions(-)
 create mode 100644 hw/display/trace-events-qxl
Signature · 1 line
2.29.2
#1 lore raw

PATCH v3 1/8 meson: add trace_events_config[]

It's an array of dicts, where each dict holds the configuration for one trace-events file. For now just fill it from trace_events_subdirs.

---
 meson.build       |  1 +
 trace/meson.build | 21 ++++++++++++++++-----
 2 files changed, 17 insertions(+), 5 deletions(-)
meson.build +1 -0
index af2bc8974111..1f3d48b53a06 100644
@@ -1677,6 +1677,7 @@ target_softmmu_arch = {}
 
 # TODO: add each directory to the subdirs from its own meson.build, once
 # we have those
+trace_events_config = []
 trace_events_subdirs = [
   'accel/kvm',
   'accel/tcg',
trace/meson.build +16 -5
index a0be8f9b0db9..3a2b39dd6291 100644
@@ -2,12 +2,23 @@
 specific_ss.add(files('control-target.c'))
 
 trace_events_files = []
-foreach dir : [ '.' ] + trace_events_subdirs
-  trace_events_file = meson.source_root() / dir / 'trace-events'
+
+trace_events_config += {
+  'file'  : meson.source_root() / 'trace-events',
+  'group' : 'root',
+}
+foreach dir : trace_events_subdirs
+  trace_events_config += {
+    'file'  : meson.source_root() / dir / 'trace-events',
+    'group' : dir.underscorify(),
+  }
+endforeach
+
+foreach c : trace_events_config
+  trace_events_file = c.get('file')
   trace_events_files += [ trace_events_file ]
-  group_name = dir == '.' ? 'root' : dir.underscorify()
-  group = '--group=' + group_name
-  fmt = '@0@-' + group_name + '.@1@'
+  group = '--group=' + c.get('group')
+  fmt = '@0@-' + c.get('group') + '.@1@'
 
   trace_h = custom_target(fmt.format('trace', 'h'),
                           output: fmt.format('trace', 'h'),
Signature · 1 line
2.29.2
Reviewed-by on v3 1/8
#2 lore raw

Re: [PATCH v3 1/8] meson: add trace_events_config[]

Quoted 8 lines · Gerd Hoffmann

It's an array of dicts, where each dict holds the configuration for one trace-events file. For now just fill it from trace_events_subdirs.

---
 meson.build       |  1 +
 trace/meson.build | 21 ++++++++++++++++-----
 2 files changed, 17 insertions(+), 5 deletions(-)
#3 lore raw

PATCH v3 2/8 meson: move up hw subdir (specifically before trace subdir)

In reply to Gerd Hoffmann

Needed so trace/meson.build can see
stuff done in hw/*/meson.build.

---
 meson.build | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
meson.build +2 -1
index 1f3d48b53a06..7462a50b4c36 100644
@@ -1777,6 +1777,8 @@ if 'CONFIG_VHOST_USER' in config_host
   vhost_user = libvhost_user.get_variable('vhost_user_dep')
 endif
 
+subdir('hw')
+
 subdir('qapi')
 subdir('qobject')
 subdir('stubs')
@@ -1863,7 +1865,6 @@ subdir('migration')
 subdir('monitor')
 subdir('net')
 subdir('replay')
-subdir('hw')
 subdir('accel')
 subdir('plugins')
 subdir('bsd-user')
Signature · 1 line
2.29.2
#4 lore raw

Re: [PATCH v3 2/8] meson: move up hw subdir (specifically before trace subdir)

Quoted 13 lines · Gerd Hoffmann

Needed so trace/meson.build can see
stuff done in hw/*/meson.build.

---
 meson.build | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
meson.build +0 -0
index 1f3d48b53a06..7462a50b4c36 100644

meson.build if 'CONFIG_VHOST_USER' in config_host · line 1782

Quoted 6 lines
meson.build +2 -0
@@ -1777,6 +1777,8 @@ if 'CONFIG_VHOST_USER' in config_host
   vhost_user = libvhost_user.get_variable('vhost_user_dep')
 endif
 
+subdir('hw')
+
 subdir('qapi')

Please add a comment into the meson.build file explaining the ordering requirement. This will prevent people accidentally breaking this in the future.

#5 lore raw

PATCH v3 3/8 meson: add module_trace & module_trace_src

In reply to Gerd Hoffmann

module_trace is a dict which keeps track of the trace source files for a module.

module_trace_src collects the trace source files for a given trace-events file, which then either added to the source set or to a new module_trace dict depending on whenever they are for a module or core qemu.

---
 meson.build       |  3 ++-
 trace/meson.build | 16 ++++++++++++----
 2 files changed, 14 insertions(+), 5 deletions(-)
meson.build +2 -1
index 7462a50b4c36..a073fcd301c5 100644
@@ -1890,7 +1890,8 @@ foreach d, list : modules
   foreach m, module_ss : list
     if enable_modules and targetos != 'windows'
       module_ss = module_ss.apply(config_all, strict: false)
-      sl = static_library(d + '-' + m, [genh, module_ss.sources()],
+      module_trace_src = module_trace.get(d + '-' + m, [])
+      sl = static_library(d + '-' + m, [genh, module_ss.sources(), module_trace_src],
                           dependencies: [modulecommon, module_ss.dependencies()], pic: true)
       if d == 'block'
         block_mods += sl
trace/meson.build +12 -4
index 3a2b39dd6291..843b472ba943 100644
@@ -2,6 +2,7 @@
 specific_ss.add(files('control-target.c'))
 
 trace_events_files = []
+module_trace = {}
 
 trace_events_config += {
   'file'  : meson.source_root() / 'trace-events',
@@ -19,6 +20,8 @@ foreach c : trace_events_config
   trace_events_files += [ trace_events_file ]
   group = '--group=' + c.get('group')
   fmt = '@0@-' + c.get('group') + '.@1@'
+  mod = c.get('module', '')
+  module_trace_src = []
 
   trace_h = custom_target(fmt.format('trace', 'h'),
                           output: fmt.format('trace', 'h'),
@@ -34,10 +37,10 @@ foreach c : trace_events_config
                                 output: fmt.format('trace-ust', 'h'),
                                 input: trace_events_file,
                                 command: [ tracetool, group, '--format=ust-events-h', '@INPUT@', '@OUTPUT@' ])
-    trace_ss.add(trace_ust_h, lttng, urcubp)
+    module_trace_src += [ trace_ust_h, lttng, urcubp ]
     genh += trace_ust_h
   endif
-  trace_ss.add(trace_h, trace_c)
+  module_trace_src += [ trace_h, trace_c ]
   if 'CONFIG_TRACE_DTRACE' in config_host
     trace_dtrace = custom_target(fmt.format('trace-dtrace', 'dtrace'),
                                  output: fmt.format('trace-dtrace', 'dtrace'),
@@ -47,17 +50,22 @@ foreach c : trace_events_config
                                    output: fmt.format('trace-dtrace', 'h'),
                                    input: trace_dtrace,
                                    command: [ 'dtrace', '-DSTAP_SDT_V2', '-o', '@OUTPUT@', '-h', '-s', '@INPUT@' ])
-    trace_ss.add(trace_dtrace_h)
+    module_trace_src += trace_dtrace_h
     if host_machine.system() != 'darwin'
       trace_dtrace_o = custom_target(fmt.format('trace-dtrace', 'o'),
                                      output: fmt.format('trace-dtrace', 'o'),
                                      input: trace_dtrace,
                                      command: [ 'dtrace', '-DSTAP_SDT_V2', '-o', '@OUTPUT@', '-G', '-s', '@INPUT@' ])
-      trace_ss.add(trace_dtrace_o)
+      module_trace_src += trace_dtrace_o
     endif
 
     genh += trace_dtrace_h
   endif
+  if enable_modules and mod != ''
+    module_trace += { mod : module_trace_src }
+  else
+    trace_ss.add(module_trace_src)
+  endif
 endforeach
 
 trace_events_all = custom_target('trace-events-all',
Signature · 1 line
2.29.2
Reviewed-by on v3 3/8
#6 lore raw

Re: [PATCH v3 3/8] meson: add module_trace & module_trace_src

Quoted 7 lines · Gerd Hoffmann

module_trace is a dict which keeps track of the trace source files for a module.

module_trace_src collects the trace source files for a given trace-events file, which then either added to the source set or to a new module_trace dict depending on whenever they are for a module or core qemu.

Quoted 5 lines
---
 meson.build       |  3 ++-
 trace/meson.build | 16 ++++++++++++----
 2 files changed, 14 insertions(+), 5 deletions(-)
#7 lore raw

PATCH v3 4/8 meson: move qxl trace events to separate file

In reply to Gerd Hoffmann

Move qxl trace events to separate trace-events-qxl file.

---
 hw/display/qxl-render.c     |  1 +
 hw/display/qxl.c            |  1 +
 hw/display/meson.build      |  5 +++
 hw/display/trace-events     | 67 -------------------------------------
 hw/display/trace-events-qxl | 66 ++++++++++++++++++++++++++++++++++++
 5 files changed, 73 insertions(+), 67 deletions(-)
 create mode 100644 hw/display/trace-events-qxl
hw/display/qxl-render.c +1 -0
index 3ce2e57b8feb..cc4862e26eb6 100644
@@ -23,6 +23,7 @@
 #include "qxl.h"
 #include "sysemu/runstate.h"
 #include "trace.h"
+#include "trace/trace-hw_display_qxl.h"
 
 static void qxl_blit(PCIQXLDevice *qxl, QXLRect *rect)
 {
hw/display/qxl.c +1 -0
index 431c1070967a..4e8d1bb8d77b 100644
@@ -33,6 +33,7 @@
 #include "migration/blocker.h"
 #include "migration/vmstate.h"
 #include "trace.h"
+#include "trace/trace-hw_display_qxl.h"
 
 #include "qxl.h"
 
hw/display/meson.build +5 -0
index 9d79e3951d9e..945c05ddbe42 100644
@@ -42,6 +42,11 @@ if config_all_devices.has_key('CONFIG_QXL')
   qxl_ss = ss.source_set()
   qxl_ss.add(when: 'CONFIG_QXL', if_true: [files('qxl.c', 'qxl-logger.c', 'qxl-render.c'),
                                            pixman, spice])
+  trace_events_config += {
+    'file'   : meson.source_root() / 'hw' / 'display' / 'trace-events-qxl',
+    'group'  : 'hw_display_qxl',
+    'module' : 'hw-display-qxl',
+  }
   hw_display_modules += {'qxl': qxl_ss}
 endif
 
hw/display/trace-events +0 -67

73 lines not on this page · raw message

hw/display/trace-events-qxl new file +66 -0

66 lines not on this page · raw message

Signature · 1 line
2.29.2
Reviewed-by on v3 4/8
#8 lore raw

Re: [PATCH v3 4/8] meson: move qxl trace events to separate file

Quoted 11 lines · Gerd Hoffmann

Move qxl trace events to separate trace-events-qxl file.

---
 hw/display/qxl-render.c     |  1 +
 hw/display/qxl.c            |  1 +
 hw/display/meson.build      |  5 +++
 hw/display/trace-events     | 67 -------------------------------------
 hw/display/trace-events-qxl | 66 ++++++++++++++++++++++++++++++++++++
 5 files changed, 73 insertions(+), 67 deletions(-)
 create mode 100644 hw/display/trace-events-qxl
#9 lore raw

Re: [PATCH v3 4/8] meson: move qxl trace events to separate file

In reply to Gerd Hoffmann
Quoted 17 lines · Gerd Hoffmann

Move qxl trace events to separate trace-events-qxl file.

---
 hw/display/qxl-render.c     |  1 +
 hw/display/qxl.c            |  1 +
 hw/display/meson.build      |  5 +++
 hw/display/trace-events     | 67 -------------------------------------
 hw/display/trace-events-qxl | 66 ++++++++++++++++++++++++++++++++++++
 5 files changed, 73 insertions(+), 67 deletions(-)
 create mode 100644 hw/display/trace-events-qxl
hw/display/qxl-render.c +0 -0
index 3ce2e57b8feb..cc4862e26eb6 100644

hw/display/qxl-render.c line 26

Quoted 4 lines
hw/display/qxl-render.c +1 -0
@@ -23,6 +23,7 @@
 #include "qxl.h"
 #include "sysemu/runstate.h"
 #include "trace.h"
+#include "trace/trace-hw_display_qxl.h"

I'm not a fan of this, as it feels like we're setting up a completely different naming convention from the one that already exists.

The previous "trace.h" header is relative to the source directory. ie, "trace.h" is pulling in "hw/display/trace.h".

IMHO, we should be generating 'trace-qxl.h', alongside the existing trace.h, so we can use a bare "trace-qxl.h" include that refers to "hw/display/trace-qxl.h".

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#10 lore raw

PATCH v3 5/8 trace: iter init tweaks

In reply to Gerd Hoffmann

Rename trace_event_iter_init() to trace_event_iter_init_pattern(), add trace_event_iter_init_all() for interating over all events.

---
 trace/control.h        | 17 +++++++++++++----
 monitor/misc.c         |  4 ++--
 trace/control-target.c |  2 +-
 trace/control.c        | 16 +++++++++++-----
 trace/qmp.c            |  6 +++---
 trace/simple.c         |  2 +-
 6 files changed, 31 insertions(+), 16 deletions(-)
trace/control.h +13 -4
index 9522a7b318e2..ce40bd040574 100644
@@ -20,15 +20,24 @@ typedef struct TraceEventIter {
 
 
 /**
- * trace_event_iter_init:
+ * trace_event_iter_init_all:
  * @iter: the event iterator struct
- * @pattern: optional pattern to filter events on name
  *
  * Initialize the event iterator struct @iter,
- * optionally using @pattern to filter out events
+ * for all events.
+ */
+void trace_event_iter_init_all(TraceEventIter *iter);
+
+/**
+ * trace_event_iter_init_pattern:
+ * @iter: the event iterator struct
+ * @pattern: pattern to filter events on name
+ *
+ * Initialize the event iterator struct @iter,
+ * using @pattern to filter out events
  * with non-matching names.
  */
-void trace_event_iter_init(TraceEventIter *iter, const char *pattern);
+void trace_event_iter_init_pattern(TraceEventIter *iter, const char *pattern);
 
 /**
  * trace_event_iter_next:
monitor/misc.c +2 -2
index a7650ed74702..d62d5d25aa2e 100644
@@ -2025,7 +2025,7 @@ void info_trace_events_completion(ReadLineState *rs, int nb_args, const char *st
         TraceEventIter iter;
         TraceEvent *ev;
         char *pattern = g_strdup_printf("%s*", str);
-        trace_event_iter_init(&iter, pattern);
+        trace_event_iter_init_pattern(&iter, pattern);
         while ((ev = trace_event_iter_next(&iter)) != NULL) {
             readline_add_completion(rs, trace_event_get_name(ev));
         }
@@ -2043,7 +2043,7 @@ void trace_event_completion(ReadLineState *rs, int nb_args, const char *str)
         TraceEventIter iter;
         TraceEvent *ev;
         char *pattern = g_strdup_printf("%s*", str);
-        trace_event_iter_init(&iter, pattern);
+        trace_event_iter_init_pattern(&iter, pattern);
         while ((ev = trace_event_iter_next(&iter)) != NULL) {
             readline_add_completion(rs, trace_event_get_name(ev));
         }
trace/control-target.c +1 -1
index e293eeed7c00..8418673c18cf 100644
@@ -127,7 +127,7 @@ void trace_init_vcpu(CPUState *vcpu)
 {
     TraceEventIter iter;
     TraceEvent *ev;
-    trace_event_iter_init(&iter, NULL);
+    trace_event_iter_init_all(&iter);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         if (trace_event_is_vcpu(ev) &&
             trace_event_get_state_static(ev) &&
trace/control.c +11 -5
index cd04dd4e0c1c..bd5d79aacd94 100644
@@ -90,7 +90,7 @@ TraceEvent *trace_event_name(const char *name)
 
     TraceEventIter iter;
     TraceEvent *ev;
-    trace_event_iter_init(&iter, NULL);
+    trace_event_iter_init_all(&iter);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         if (strcmp(trace_event_get_name(ev), name) == 0) {
             return ev;
@@ -99,10 +99,16 @@ TraceEvent *trace_event_name(const char *name)
     return NULL;
 }
 
-void trace_event_iter_init(TraceEventIter *iter, const char *pattern)
+void trace_event_iter_init_all(TraceEventIter *iter)
 {
     iter->event = 0;
     iter->group = 0;
+    iter->pattern = NULL;
+}
+
+void trace_event_iter_init_pattern(TraceEventIter *iter, const char *pattern)
+{
+    trace_event_iter_init_all(iter);
     iter->pattern = pattern;
 }
 
@@ -129,7 +135,7 @@ void trace_list_events(FILE *f)
 {
     TraceEventIter iter;
     TraceEvent *ev;
-    trace_event_iter_init(&iter, NULL);
+    trace_event_iter_init_all(&iter);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         fprintf(f, "%s\n", trace_event_get_name(ev));
     }
@@ -149,7 +155,7 @@ static void do_trace_enable_events(const char *line_buf)
     TraceEvent *ev;
     bool is_pattern = trace_event_is_pattern(line_ptr);
 
-    trace_event_iter_init(&iter, line_ptr);
+    trace_event_iter_init_pattern(&iter, line_ptr);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         if (!trace_event_get_state_static(ev)) {
             if (!is_pattern) {
@@ -257,7 +263,7 @@ void trace_fini_vcpu(CPUState *vcpu)
 
     trace_guest_cpu_exit(vcpu);
 
-    trace_event_iter_init(&iter, NULL);
+    trace_event_iter_init_all(&iter);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         if (trace_event_is_vcpu(ev) &&
             trace_event_get_state_static(ev) &&
trace/qmp.c +3 -3
index 85f81e47cc6d..3b4f4702b4f0 100644
@@ -55,7 +55,7 @@ static bool check_events(bool has_vcpu, bool ignore_unavailable, bool is_pattern
         /* error for unavailable events */
         TraceEventIter iter;
         TraceEvent *ev;
-        trace_event_iter_init(&iter, name);
+        trace_event_iter_init_pattern(&iter, name);
         while ((ev = trace_event_iter_next(&iter)) != NULL) {
             if (!ignore_unavailable && !trace_event_get_state_static(ev)) {
                 error_setg(errp, "event \"%s\" is disabled", trace_event_get_name(ev));
@@ -90,7 +90,7 @@ TraceEventInfoList *qmp_trace_event_get_state(const char *name,
     }
 
     /* Get states (all errors checked above) */
-    trace_event_iter_init(&iter, name);
+    trace_event_iter_init_pattern(&iter, name);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         TraceEventInfo *value;
         bool is_vcpu = trace_event_is_vcpu(ev);
@@ -153,7 +153,7 @@ void qmp_trace_event_set_state(const char *name, bool enable,
     }
 
     /* Apply changes (all errors checked above) */
-    trace_event_iter_init(&iter, name);
+    trace_event_iter_init_pattern(&iter, name);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         if (!trace_event_get_state_static(ev) ||
             (has_vcpu && !trace_event_is_vcpu(ev))) {
trace/simple.c +1 -1
index 9cd2ed1fb3f4..97b6f85168e7 100644
@@ -286,7 +286,7 @@ static int st_write_event_mapping(void)
     TraceEventIter iter;
     TraceEvent *ev;
 
-    trace_event_iter_init(&iter, NULL);
+    trace_event_iter_init_all(&iter);
     while ((ev = trace_event_iter_next(&iter)) != NULL) {
         uint64_t id = trace_event_get_id(ev);
         const char *name = trace_event_get_name(ev);
Signature · 1 line
2.29.2
Reviewed-by on v3 5/8
#11 lore raw

Re: [PATCH v3 5/8] trace: iter init tweaks

Quoted 4 lines · Gerd Hoffmann

Rename trace_event_iter_init() to trace_event_iter_init_pattern(), add trace_event_iter_init_all() for interating over all events.

Quoted 8 lines
---
 trace/control.h        | 17 +++++++++++++----
 monitor/misc.c         |  4 ++--
 trace/control-target.c |  2 +-
 trace/control.c        | 16 +++++++++++-----
 trace/qmp.c            |  6 +++---
 trace/simple.c         |  2 +-
 6 files changed, 31 insertions(+), 16 deletions(-)
#12 lore raw

PATCH v3 6/8 trace: add trace_event_iter_init_group

In reply to Gerd Hoffmann

This allows to interate over an event group.

---
 trace/control.h | 13 +++++++++++++
 trace/control.c | 19 ++++++++++++++++---
 2 files changed, 29 insertions(+), 3 deletions(-)
trace/control.h +13 -0
index ce40bd040574..23b8393b297e 100644
@@ -13,8 +13,11 @@
 #include "event-internal.h"
 
 typedef struct TraceEventIter {
+    /* iter state */
     size_t event;
     size_t group;
+    /* filter conditions */
+    size_t group_id;
     const char *pattern;
 } TraceEventIter;
 
@@ -39,6 +42,16 @@ void trace_event_iter_init_all(TraceEventIter *iter);
  */
 void trace_event_iter_init_pattern(TraceEventIter *iter, const char *pattern);
 
+/**
+ * trace_event_iter_init_group:
+ * @iter: the event iterator struct
+ * @group_id: group_id to filter events by group.
+ *
+ * Initialize the event iterator struct @iter,
+ * using @group_id to filter for events in the group.
+ */
+void trace_event_iter_init_group(TraceEventIter *iter, size_t group_id);
+
 /**
  * trace_event_iter_next:
  * @iter: the event iterator struct
trace/control.c +16 -3
index bd5d79aacd94..f1cc880b7cd1 100644
@@ -103,6 +103,7 @@ void trace_event_iter_init_all(TraceEventIter *iter)
 {
     iter->event = 0;
     iter->group = 0;
+    iter->group_id = -1;
     iter->pattern = NULL;
 }
 
@@ -112,20 +113,32 @@ void trace_event_iter_init_pattern(TraceEventIter *iter, const char *pattern)
     iter->pattern = pattern;
 }
 
+void trace_event_iter_init_group(TraceEventIter *iter, size_t group_id)
+{
+    trace_event_iter_init_all(iter);
+    iter->group_id = group_id;
+}
+
 TraceEvent *trace_event_iter_next(TraceEventIter *iter)
 {
     while (iter->group < nevent_groups &&
            event_groups[iter->group].events[iter->event] != NULL) {
         TraceEvent *ev = event_groups[iter->group].events[iter->event];
+        size_t group = iter->group;
         iter->event++;
         if (event_groups[iter->group].events[iter->event] == NULL) {
             iter->event = 0;
             iter->group++;
         }
-        if (!iter->pattern ||
-            g_pattern_match_simple(iter->pattern, trace_event_get_name(ev))) {
-            return ev;
+        if (iter->pattern &&
+            !g_pattern_match_simple(iter->pattern, trace_event_get_name(ev))) {
+            continue;
         }
+        if (iter->group_id != -1 &&
+            iter->group_id != group) {
+            continue;
+        }
+        return ev;
     }
 
     return NULL;
Signature · 1 line
2.29.2
#14 lore raw

PATCH v3 7/8 trace/simple: pass iter to st_write_event_mapping

In reply to Gerd Hoffmann

Pass an iter to st_write_event_mapping, so the function can interate different things depending on how we initialize the iter.

---
 trace/simple.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
trace/simple.c +5 -5
index 97b6f85168e7..ec2156d135cb 100644
@@ -280,14 +280,12 @@ void trace_record_finish(TraceBufferRecord *rec)
     }
 }
 
-static int st_write_event_mapping(void)
+static int st_write_event_mapping(TraceEventIter *iter)
 {
     uint64_t type = TRACE_RECORD_TYPE_MAPPING;
-    TraceEventIter iter;
     TraceEvent *ev;
 
-    trace_event_iter_init_all(&iter);
-    while ((ev = trace_event_iter_next(&iter)) != NULL) {
+    while ((ev = trace_event_iter_next(iter)) != NULL) {
         uint64_t id = trace_event_get_id(ev);
         const char *name = trace_event_get_name(ev);
         uint32_t len = strlen(name);
@@ -309,6 +307,7 @@ static int st_write_event_mapping(void)
  */
 bool st_set_trace_file_enabled(bool enable)
 {
+    TraceEventIter iter;
     bool was_enabled = trace_fp;
 
     if (enable == !!trace_fp) {
@@ -333,8 +332,9 @@ bool st_set_trace_file_enabled(bool enable)
             return was_enabled;
         }
 
+        trace_event_iter_init_all(&iter);
         if (fwrite(&header, sizeof header, 1, trace_fp) != 1 ||
-            st_write_event_mapping() < 0) {
+            st_write_event_mapping(&iter) < 0) {
             fclose(trace_fp);
             trace_fp = NULL;
             return was_enabled;
Signature · 1 line
2.29.2
Reviewed-by on v3 7/8
#15 lore raw

Re: [PATCH v3 7/8] trace/simple: pass iter to st_write_event_mapping

Quoted 7 lines · Gerd Hoffmann

Pass an iter to st_write_event_mapping, so the function can interate different things depending on how we initialize the iter.

---
 trace/simple.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
#16 lore raw

PATCH v3 8/8 trace/simple: add st_init_group

In reply to Gerd Hoffmann

Add helper function and call it for each trace event group added. Makes sure that events added at module load time are initialized properly.

---
 trace/simple.h  |  1 +
 trace/control.c |  4 ++++
 trace/simple.c  | 12 ++++++++++++
 3 files changed, 17 insertions(+)
trace/simple.h +1 -0
index 26ccbc8b8ae3..ee1983ce5617 100644
@@ -15,6 +15,7 @@ void st_print_trace_file_status(void);
 bool st_set_trace_file_enabled(bool enable);
 void st_set_trace_file(const char *file);
 bool st_init(void);
+void st_init_group(size_t group);
 void st_flush_trace_buffer(void);
 
 typedef struct {
trace/control.c +4 -0
index f1cc880b7cd1..9649d0979e70 100644
@@ -81,6 +81,10 @@ void trace_event_register_group(TraceEvent **events)
     event_groups = g_renew(TraceEventGroup, event_groups, nevent_groups + 1);
     event_groups[nevent_groups].events = events;
     nevent_groups++;
+
+#ifdef CONFIG_TRACE_SIMPLE
+    st_init_group(nevent_groups - 1);
+#endif
 }
 
 
trace/simple.c +12 -0
index ec2156d135cb..ac499edee0d5 100644
@@ -422,3 +422,15 @@ bool st_init(void)
     atexit(st_flush_trace_buffer);
     return true;
 }
+
+void st_init_group(size_t group)
+{
+    TraceEventIter iter;
+
+    if (!trace_writeout_enabled) {
+        return;
+    }
+
+    trace_event_iter_init_group(&iter, group);
+    st_write_event_mapping(&iter);
+}
Signature · 1 line
2.29.2
Reviewed-by on v3 8/8
#17 lore raw

Re: [PATCH v3 8/8] trace/simple: add st_init_group

Quoted 10 lines · Gerd Hoffmann

Add helper function and call it for each trace event group added. Makes sure that events added at module load time are initialized properly.

---
 trace/simple.h  |  1 +
 trace/control.c |  4 ++++
 trace/simple.c  | 12 ++++++++++++
 3 files changed, 17 insertions(+)
#18 lore raw
In reply to Gerd Hoffmann
Quoted 5 lines · Gerd Hoffmann

First version that actually works. Only qxl covered for this RfC, other modules will follow once the basics are hashed out.

v3:
- handle initialization of modular tracepoints.

Cool, this looks promising!

Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

#19 lore raw

Hi,

Quoted 23 lines
Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

Quoted 7 lines

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

Finally found the time to look at this again...

So, we already have a "group". Which is basically the sub-directory of the trace-events file right now, and it seems to be mostly a build system thing. We get many small lists instead of one huge, but there seems to be no other effect. We could change that though, by giving each group an (optional?) prefix.

There also is a probe prefix, apparently used by dtrace only. Not sure how to deal with that. It prefix is qemu-<target-type>-<target-name>. Giving qemu modules its own dtrace prefix looks sensible to me. That would probably something like "qemu-module-<name>".

take care,
Gerd

#20 lore raw
Quoted 33 lines · Gerd Hoffmann

Hi,

Quoted 23 lines
Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

Quoted 7 lines

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

Quoted 7 lines

Finally found the time to look at this again...

So, we already have a "group". Which is basically the sub-directory of the trace-events file right now, and it seems to be mostly a build system thing. We get many small lists instead of one huge, but there seems to be no other effect. We could change that though, by giving each group an (optional?) prefix.

Yes. This reminds me of an idea that was mentioned at the beginning of this effort: maybe QEMU modules should always have their own directory in the source tree instead of being alongside other source files that are built into the main binary.

Quoted 4 lines

There also is a probe prefix, apparently used by dtrace only. Not sure how to deal with that. It prefix is qemu-<target-type>-<target-name>. Giving qemu modules its own dtrace prefix looks sensible to me. That would probably something like "qemu-module-<name>".

I think the DTrace prefix needs to be the same as the executable name, but I'm not sure. I also don't know how that extends to shared libraries like QEMU modules. I'm afraid you would need to investigate the DTrace prefix.

Stefan

#21 lore raw
Quoted 42 lines · Stefan Hajnoczi
Quoted 33 lines · Gerd Hoffmann

Hi,

Quoted 23 lines
Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

Quoted 7 lines

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

Quoted 7 lines

Finally found the time to look at this again...

So, we already have a "group". Which is basically the sub-directory of the trace-events file right now, and it seems to be mostly a build system thing. We get many small lists instead of one huge, but there seems to be no other effect. We could change that though, by giving each group an (optional?) prefix.

Quoted 4 lines

Yes. This reminds me of an idea that was mentioned at the beginning of this effort: maybe QEMU modules should always have their own directory in the source tree instead of being alongside other source files that are built into the main binary.

This implies that each time we modularize something, we have to move its source code. It is possible, but it wouldn't be by preferred choice.

Quoted 9 lines
Quoted 4 lines

There also is a probe prefix, apparently used by dtrace only. Not sure how to deal with that. It prefix is qemu-<target-type>-<target-name>. Giving qemu modules its own dtrace prefix looks sensible to me. That would probably something like "qemu-module-<name>".

I think the DTrace prefix needs to be the same as the executable name, but I'm not sure. I also don't know how that extends to shared libraries like QEMU modules. I'm afraid you would need to investigate the DTrace prefix.

I'm not aware of any requirement for dtrace prefix to match the executable name. We don't even have an executable called "qemu" so we're not matching even today.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#22 lore raw
Quoted 43 lines · Daniel P. Berrangé
Quoted 42 lines · Stefan Hajnoczi
Quoted 40 lines · Gerd Hoffmann

Hi,

Quoted 23 lines
Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

Quoted 7 lines

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

Finally found the time to look at this again...

So, we already have a "group". Which is basically the sub-directory of the trace-events file right now, and it seems to be mostly a build system thing. We get many small lists instead of one huge, but there seems to be no other effect. We could change that though, by giving each group an (optional?) prefix.

Quoted 8 lines
Quoted 4 lines

Yes. This reminds me of an idea that was mentioned at the beginning of this effort: maybe QEMU modules should always have their own directory in the source tree instead of being alongside other source files that are built into the main binary.

This implies that each time we modularize something, we have to move its source code. It is possible, but it wouldn't be by preferred choice.

If it bothers you then it probably bothers others too. Let's leave it.

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

There also is a probe prefix, apparently used by dtrace only. Not sure how to deal with that. It prefix is qemu-<target-type>-<target-name>. Giving qemu modules its own dtrace prefix looks sensible to me. That would probably something like "qemu-module-<name>".

Quoted 8 lines
Quoted 4 lines

I think the DTrace prefix needs to be the same as the executable name, but I'm not sure. I also don't know how that extends to shared libraries like QEMU modules. I'm afraid you would need to investigate the DTrace prefix.

I'm not aware of any requirement for dtrace prefix to match the executable name. We don't even have an executable called "qemu" so we're not matching even today.

Great, thanks for the other email reply where you showed how the SystemTap tapsetsuse the prefix!

Stefan

#23 lore raw
In reply to Gerd Hoffmann
Quoted 41 lines · Gerd Hoffmann

Hi,

Quoted 23 lines
Quoted 5 lines

TODO:
Enabling modular tracepoints via -trace cmd line doesn't work yet. Guess we need to store the list somewhere for later re-processing. Error handling is tricky, specifically the "tracepoint doesn't exist" error. Suggestions / ideas are welcome.

Two ideas:

Global trace event name list
----------------------------
Build *some* global information about all trace events, including modules, into the main QEMU binary. For example, generate an array of all trace event names so QEMU can always print an error if a non-existent trace event name is used. (This is similar to the trace-events-all file, which is a global list of all trace events.)

Module name prefixes
--------------------
Allow an optional module/group prefix like qxl:my_trace_event. When the user says:

--trace qxl:my_trace_event

Quoted 7 lines

QEMU knows that this trace event belongs to the "qxl" module/group. It will not attempt to load it until the qxl module registers itself.

If "my_trace_event" doesn't exist in the qxl module:
1. If the qxl module is not loaded we don't hit an error. Nevermind.
2. When the qxl module is loaded pending events are resolved and an error is printed.

Finally found the time to look at this again...

So, we already have a "group". Which is basically the sub-directory of the trace-events file right now, and it seems to be mostly a build system thing. We get many small lists instead of one huge, but there seems to be no other effect. We could change that though, by giving each group an (optional?) prefix.

Quoted 4 lines

There also is a probe prefix, apparently used by dtrace only. Not sure how to deal with that. It prefix is qemu-<target-type>-<target-name>. Giving qemu modules its own dtrace prefix looks sensible to me. That would probably something like "qemu-module-<name>".

The prefix is used by the systemtap backend. It lets us define friendly probes, scoped for each QEMU emulator target.

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-ppc64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-aarch64").mark("dma_map_wait")

there is nothing here that reqiures us to change the prefix for a module - we can carry on using the same systemtap probe name whether modules are used or not.

Currently we kind of unofficially have a convention that the name of a trace point should reflect its functional area. So all QXL related probes have a name prefix "qxl_".

We could make that explicit, by having the trace-events files support a group, IOW instead of

qxl_destroy_primary(int qid) "%d"
qxl_enter_vga_mode(int qid) "%d"
qxl_exit_vga_mode(int qid) "%d"

We could say that a dot separates the group from the probe name, and thus have

qxl.destroy_primary(int qid) "%d"
qxl.enter_vga_mode(int qid) "%d"
qxl.exit_vga_mode(int qid) "%d"

this would be backwards compatible, as we can turn the "." back into a "_" for all existing trace backends, except stp.

NB this is all tagential to use of modules at a build system /runtime level. I think this explicit grouping of probes would make sense for everything, to make the naming convention explicit instead of implicit.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#24 lore raw

Hi,

Quoted 6 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-ppc64").mark("dma_map_wait")

Probe qemu.system.ppc64.dma_map_wait = ...

Can I trace qemu started from build directory? Seems scripts/qemu-trace-stap doesn't support that.

Quoted 6 lines

qxl.destroy_primary(int qid) "%d"
qxl.enter_vga_mode(int qid) "%d"
qxl.exit_vga_mode(int qid) "%d"

this would be backwards compatible, as we can turn the "." back into a "_" for all existing trace backends, except stp.

Hmm, not sure I like this inconsistency. I think names should be the same no matter which trace backend is used.

take care,
Gerd

#25 lore raw
Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

Quoted 7 lines
Quoted 2 lines

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-ppc64").mark("dma_map_wait")

Probe qemu.system.ppc64.dma_map_wait = ...

Can I trace qemu started from build directory? Seems scripts/qemu-trace-stap doesn't support that.

We should really generate extra equiv .stp files just for running from the build.

Quoted 9 lines
Quoted 6 lines

qxl.destroy_primary(int qid) "%d"
qxl.enter_vga_mode(int qid) "%d"
qxl.exit_vga_mode(int qid) "%d"

this would be backwards compatible, as we can turn the "." back into a "_" for all existing trace backends, except stp.

Hmm, not sure I like this inconsistency. I think names should be the same no matter which trace backend is used.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#26 lore raw
Quoted 10 lines · Daniel P. Berrangé
Quoted 9 lines · Gerd Hoffmann

Hi,

Quoted 6 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-ppc64").mark("dma_map_wait")

Quoted 7 lines
Quoted 4 lines

Probe qemu.system.ppc64.dma_map_wait = ...

Can I trace qemu started from build directory? Seems scripts/qemu-trace-stap doesn't support that.

We should really generate extra equiv .stp files just for running from the build.

Well, "make install" with --prefix=$HOME/qemu-install fixed that for the time being.

Now I have this:

kraxel@sirius ~/qemu-install/bin# sudo ./qemu-trace-stap -v run ./qemu-system-x86_64 "qxl_soft_reset"
Using tapset dir '/home/kraxel/qemu-install/share/systemtap/tapset' for binary './qemu-system-x86_64'
Compiling script 'probe qemu.system.x86_64.log.qxl_soft_reset {}'
semantic error: unresolved function pid: identifier 'pid' at /home/kraxel/qemu-install/share/systemtap/tapset/qemu-system-x86_64-log.stp:5451:41

        source:     printf("%d@%d qxl_soft_reset %d\n", pid(), gettimeofday_ns(), qid)
                                                        ^

Pass 2: analysis failed. [man error::pass2]

Any clue why pid() isn't known?

thanks,
Gerd

#27 lore raw
Quoted 28 lines · Gerd Hoffmann
Quoted 10 lines · Daniel P. Berrangé
Quoted 9 lines · Gerd Hoffmann

Hi,

Quoted 6 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait") probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-ppc64").mark("dma_map_wait")

Quoted 7 lines
Quoted 4 lines

Probe qemu.system.ppc64.dma_map_wait = ...

Can I trace qemu started from build directory? Seems scripts/qemu-trace-stap doesn't support that.

We should really generate extra equiv .stp files just for running from the build.

Well, "make install" with --prefix=$HOME/qemu-install fixed that for the time being.

Now I have this:

kraxel@sirius ~/qemu-install/bin# sudo ./qemu-trace-stap -v run ./qemu-system-x86_64 "qxl_soft_reset"
Using tapset dir '/home/kraxel/qemu-install/share/systemtap/tapset' for binary './qemu-system-x86_64'
Compiling script 'probe qemu.system.x86_64.log.qxl_soft_reset {}'
semantic error: unresolved function pid: identifier 'pid' at /home/kraxel/qemu-install/share/systemtap/tapset/qemu-system-x86_64-log.stp:5451:41

Quoted 6 lines
        source:     printf("%d@%d qxl_soft_reset %d\n", pid(), gettimeofday_ns(), qid)
                                                        ^

Pass 2: analysis failed. [man error::pass2]

Any clue why pid() isn't known?

Hmm, strange, makes me think we have a bug causing it to not pull in global functions.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#28 lore raw

Hi,

Quoted 12 lines
Quoted 12 lines

Well, "make install" with --prefix=$HOME/qemu-install fixed that for the time being.

Now I have this:

kraxel@sirius ~/qemu-install/bin# sudo ./qemu-trace-stap -v run ./qemu-system-x86_64 "qxl_soft_reset"
Using tapset dir '/home/kraxel/qemu-install/share/systemtap/tapset' for binary './qemu-system-x86_64'
Compiling script 'probe qemu.system.x86_64.log.qxl_soft_reset {}'
semantic error: unresolved function pid: identifier 'pid' at /home/kraxel/qemu-install/share/systemtap/tapset/qemu-system-x86_64-log.stp:5451:41

        source:     printf("%d@%d qxl_soft_reset %d\n", pid(), gettimeofday_ns(), qid)
                                                        ^
Quoted 6 lines
Quoted 3 lines

Pass 2: analysis failed. [man error::pass2]

Any clue why pid() isn't known?

Hmm, strange, makes me think we have a bug causing it to not pull in global functions.

Hmm. 5.1.0 fails the same way. 5.2.0 fails in a different way. Seems we had temporary breakage in 5.2.0 (was that the module thing which needed some workaround?). Given 5.1.0 fails too I suspect this is a systemtap change (/me runs fedora 33). Googling didn't found much, other than indicating pid() is used frequently in examples and tutorials.

stap-prep asked for kernel-debuginfo, but installing that (huge package with 3.5G(!) installed size) hasn't changed the situation. Guess I only actually need that if I want trace the kernel not userspace?

take care,
Gerd

#29 lore raw
In reply to Daniel P. Berrangé

Hi,

Quoted 5 lines
Quoted 2 lines

semantic error: unresolved function pid: identifier 'pid' at /home/kraxel/qemu-install/share/systemtap/tapset/qemu-system-x86_64-log.stp:5451:41

        source:     printf("%d@%d qxl_soft_reset %d\n", pid(), gettimeofday_ns(), qid)

Hmm, strange, makes me think we have a bug causing it to not pull in global functions.

Yep. qemu-trace-stap sets SYSTEMTAP_TAPSET to /home/kraxel/qemu-install/share/systemtap/tapset. Which makes stap use that exclusively and ignore /usr/share/systemtap/tapset. Which in turn makes pid() disappear because that apparently is defined somewhere in those tapsets.

IOW: It's broken for any $prefix != "/usr".

Using stap -I /home/kraxel/qemu-install/share/systemtap/tapset seems to work fine, I guess qemu-trace-stap should use that instead ...

take care,
Gerd

#30 lore raw
In reply to Daniel P. Berrangé

Hi,

Quoted 5 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait")

So, that changes with modules, we need the module name now, i.e.

    probe qemu.system.x86_64.qxl_soft_reset = \
	process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

We could repeat that in every qemu-system-$arch.stp file.

We could also have one stp file per module, with probes like this:

    probe qemu.modules.qxl_soft_reset = \
        process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

The later looks like a better fit to me, but has the drawback that the tracepoints have different names in modular and non-modular builds ...

take care,
Gerd

#31 lore raw
Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

Quoted 8 lines
Quoted 1 line

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait")

So, that changes with modules, we need the module name now, i.e.

    probe qemu.system.x86_64.qxl_soft_reset = \
	process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

We could repeat that in every qemu-system-$arch.stp file.

This would have the surprise the 'qemu.system.x86_64.qxl_soft_reset' probes will fire even for qemu-system-ppc64 / qemu-system-xxxxx etc because we've not restricted the scope as the original probe did.

If we can't fix that, then we must use the second option to avoid the surprise IMHO

Quoted 7 lines

We could also have one stp file per module, with probes like this:

    probe qemu.modules.qxl_soft_reset = \
        process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

The later looks like a better fit to me, but has the drawback that the tracepoints have different names in modular and non-modular builds ...

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#32 lore raw
Quoted 11 lines · Daniel P. Berrangé
Quoted 10 lines · Gerd Hoffmann

Hi,

Quoted 5 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait")

So, that changes with modules, we need the module name now, i.e.

Quoted 8 lines
Quoted 4 lines
    probe qemu.system.x86_64.qxl_soft_reset = \
	process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

We could repeat that in every qemu-system-$arch.stp file.

This would have the surprise the 'qemu.system.x86_64.qxl_soft_reset' probes will fire even for qemu-system-ppc64 / qemu-system-xxxxx etc because we've not restricted the scope as the original probe did.

Oh, right.

Quoted 2 lines

If we can't fix that, then we must use the second option to avoid the surprise IMHO

Yep. Got that working. Only problem is qemu-trace-stap is broken now and it seems there is no easy way out. Right now qemu-trace-stap can simply work with a constant prefix, with that change the prefix can be either qemu.system.$arch or qemu.system.modules and I suspect there is no way around listing tracepoints to figure the correct name ...

take care,
Gerd

PS: https://git.kraxel.org/cgit/qemu/log/?h=sirius/trace-modules

#33 lore raw
Quoted 23 lines · Gerd Hoffmann
Quoted 11 lines · Daniel P. Berrangé
Quoted 10 lines · Gerd Hoffmann

Hi,

Quoted 5 lines

eg a trace point "dma_map_wait" gets mapped to probes in many .stp files, once per target, because we need to match based on the executable path:

probe qemu.system.x86_64.dma_map_wait = process("/usr/libexec/qemu-system-x86_64").mark("dma_map_wait")

So, that changes with modules, we need the module name now, i.e.

Quoted 8 lines
Quoted 4 lines
    probe qemu.system.x86_64.qxl_soft_reset = \
	process("/home/kraxel/qemu-install/lib/qemu/hw-display-qxl.so").mark("qxl_soft_reset")

We could repeat that in every qemu-system-$arch.stp file.

This would have the surprise the 'qemu.system.x86_64.qxl_soft_reset' probes will fire even for qemu-system-ppc64 / qemu-system-xxxxx etc because we've not restricted the scope as the original probe did.

Oh, right.

Quoted 8 lines
Quoted 2 lines

If we can't fix that, then we must use the second option to avoid the surprise IMHO

Yep. Got that working. Only problem is qemu-trace-stap is broken now and it seems there is no easy way out. Right now qemu-trace-stap can simply work with a constant prefix, with that change the prefix can be either qemu.system.$arch or qemu.system.modules and I suspect there is no way around listing tracepoints to figure the correct name ...

Maybe just don't change the probe names, and declare that people have to tell stap to filter by pid if you want exact matches

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|