From: Serhei Makarov <serhei@serhei.io>
Date: Thu, 18 Dec 2025 21:04:34 +0000 (-0500)
Subject: crackle: icebreaker commit pre hook up unwinder
X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=66070e5b366a2ff864f77407b1642e6524f8432b;p=thirdparty%2Felfutils.git

crackle: icebreaker commit pre hook up unwinder
---

diff --git a/src/stackprof.cxx b/src/stackprof.cxx
index 619f9fb5..0fad4906 100644
--- a/src/stackprof.cxx
+++ b/src/stackprof.cxx
@@ -90,7 +90,8 @@ public:
   
   ~PerfReader();
 
-  void process_some(); // run briefly, relay decoded perf_events to consumer  
+  void process_some(); // run briefly, relay decoded perf_events to consumer
+  int n_sample_regs() { return this->sample_regs_count; }
 };
 
 
@@ -163,6 +164,7 @@ public:
     junk.pid = getpid();
     usc->process(& junk);
   }
+  void process(const perf_event_header* sample); // handle process lifecycle events; relay unwound call stack events to a consumer 
 };
 
 
@@ -435,6 +437,40 @@ main (int argc, char *argv[])
 }
 
 
+////////////////////////////////////////////////////////////////////////
+// perf_events data format
+
+class PerfReader;
+
+/* TODO: Make sure the config above matches PERF_SAMPLE_TYPE */
+#define PERF_SAMPLE_TYPE (PERF_SAMPLE_IP | PERF_SAMPLE_TID | PERF_SAMPLE_TIME \
+			  | PERF_SAMPLE_REGS_USER | PERF_SAMPLE_STACK_USER)
+typedef struct
+{
+  struct perf_event_header header;
+  uint64_t ip;
+  uint32_t pid, tid;
+  uint64_t time;
+  uint64_t abi;
+  uint64_t regs[]; /* variable size */
+  /* uint64_t size; */
+  /* char data[]; -- variable size */
+} PerfSample;
+
+uint64_t
+perf_sample_get_size (PerfReader *reader, PerfSample *sample)
+{
+  int nregs = reader->n_sample_regs();
+  return *(uint64_t *)(sample->regs + nregs);
+}
+
+char *
+perf_sample_get_data (PerfReader *reader, PerfSample *sample)
+{
+  int nregs = reader->n_sample_regs();
+  return (char *)(sample->regs + (nregs + 1));
+}
+
 ////////////////////////////////////////////////////////////////////////
 // perf reader
 
@@ -445,7 +481,7 @@ PerfReader::PerfReader(perf_event_attr* attr, PerfConsumer* consumer, int pid)
   this->mmap_size = this->page_size * (this->page_count + 1); // total mmap size, incl header page
   this->event_wraparound_temp.resize(this->mmap_size); // NB: never resize this object again!
   
-  Ebl *default_ebl = ebl_openbackend_machine(EM_X86_64); // XXX
+  Ebl *default_ebl = ebl_openbackend_machine(EM_X86_64); /* TODO: Generalize to architectures beyond x86. */
   this->sample_regs_user = ebl_perf_frame_regs_mask (default_ebl);
   this->sample_regs_count = bitset<64>(this->sample_regs_user).count();
   
@@ -518,6 +554,23 @@ uint64_t millis_monotonic()
 }
 
 
+
+static inline uint64_t
+ring_buffer_read_head(volatile struct perf_event_mmap_page *base)
+{
+  uint64_t head = base->data_head;
+  asm volatile("" ::: "memory"); // memory fence
+  return head;
+}
+
+static inline void
+ring_buffer_write_tail(volatile struct perf_event_mmap_page *base,
+		       uint64_t tail)
+{
+  asm volatile("" ::: "memory"); // memory fence
+  base->data_tail = tail;
+}
+
 void PerfReader::process_some()
 {
   if (! this->enabled)
@@ -544,8 +597,7 @@ void PerfReader::process_some()
         if (this->pollfds[i].revents & POLLIN) // found an fd with fresh yummy events
           {
             perf_event_mmap_page *header = perf_headers[i];
-            uint64_t data_head = header->data_head;
-            asm volatile("" ::: "memory"); // memory fence
+	    uint64_t data_head = ring_buffer_read_head(header);
             uint64_t data_tail = header->data_tail; 
             uint8_t *base = ((uint8_t *) header) + this->page_size;
             struct perf_event_header *ehdr;
@@ -558,7 +610,8 @@ void PerfReader::process_some()
                 if (verbose > 3)
                   clog << "perf head=" << (void*) data_head
                        << " tail=" << (void*) data_tail
-                       << " ehdr=" << (void*) ehdr << " size=" << ehdr_size << endl;
+                       << " ehdr=" << (void*) ehdr
+		       << " size=" << setbase(10) << ehdr_size << setbase(16) << endl;
                 
                 if (((uint8_t *)ehdr) + ehdr_size > base + ring_buffer_size) // mmap region wraparound?
                   {
@@ -576,8 +629,7 @@ void PerfReader::process_some()
                 data_tail += ehdr_size;
               }
 
-            asm volatile("" ::: "memory"); // memory fence
-            header->data_tail = data_tail;
+	    ring_buffer_write_tail(header, data_tail);
           }
     }
 }
@@ -747,11 +799,13 @@ void StatsPerfConsumer::process(const perf_event_header* ehdr)
   this->event_type_counts[ehdr->type] ++;
 }
 
-
+void PerfConsumerUnwinder::process(const perf_event_header* ehdr)
+{
+}
 
 
 ////////////////////////////////////////////////////////////////////////
-// unwind consumers // gprof
+// UNWIND consumers // gprof
 
 
 UnwindStatsConsumer::~UnwindStatsConsumer()