memprof 0.2.7 → 0.2.9

data/ext/tramp.c

@@ -0,0 +1,143 @@
+#include <ruby.h>
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+
+#include <sysexits.h>
+#include <sys/mman.h>
+#include <err.h>
+#include <assert.h>
+
+#include "arch.h"
+#include "bin_api.h"
+#include "util.h"
+
+#define FREELIST_INLINES 3
+
+/*
+ * trampoline specific stuff
+ */
+static struct tramp_st2_entry *tramp_table;
+static size_t tramp_size;
+
+/*
+ * inline trampoline specific stuff
+ */
+static size_t inline_tramp_size;
+static struct inline_tramp_st2_entry *inline_tramp_table;
+
+extern struct memprof_config memprof_config;
+
+void
+create_tramp_table()
+{
+  size_t i;
+  void *region, *ent, *inline_ent;
+  size_t tramp_sz = 0, inline_tramp_sz = 0;
+
+  ent = arch_get_st2_tramp(&tramp_sz);
+  inline_ent = arch_get_inline_st2_tramp(&inline_tramp_sz);
+  assert(ent && inline_ent);
+
+  region = bin_allocate_page();
+  if (region == MAP_FAILED)
+    errx(EX_SOFTWARE, "Failed to allocate memory for stage 1 trampolines.");
+
+  tramp_table = region;
+  inline_tramp_table = region + memprof_config.pagesize / 2;
+
+  for (i = 0; i < (memprof_config.pagesize / 2) / tramp_sz; i++) {
+    memcpy(tramp_table + i, ent, tramp_sz);
+  }
+
+  for (i = 0; i < (memprof_config.pagesize / 2) / inline_tramp_sz; i++) {
+    memcpy(inline_tramp_table + i, inline_ent, inline_tramp_sz);
+  }
+}
+
+static void
+hook_freelist(int entry, void *tramp)
+{
+  size_t sizes[FREELIST_INLINES];
+  void *freelist_inliners[FREELIST_INLINES];
+  void *freelist = NULL;
+  unsigned char *byte = NULL;
+  int i = 0, tramps_completed = 0;
+
+  assert(memprof_config.gc_sweep != NULL);
+  assert(memprof_config.finalize_list != NULL);
+  assert(memprof_config.rb_gc_force_recycle != NULL);
+  assert(memprof_config.freelist != NULL);
+  assert(memprof_config.gc_sweep_size > 0);
+  assert(memprof_config.finalize_list_size > 0);
+  assert(memprof_config.rb_gc_force_recycle_size > 0);
+
+  freelist_inliners[0] = memprof_config.gc_sweep;
+  freelist_inliners[1] = memprof_config.finalize_list;
+  freelist_inliners[2] = memprof_config.rb_gc_force_recycle;
+  sizes[0] = memprof_config.gc_sweep_size;
+  sizes[1] = memprof_config.finalize_list_size;
+  sizes[2] = memprof_config.rb_gc_force_recycle_size;
+
+  freelist = memprof_config.freelist;
+
+  /* start the search for places to insert the inline tramp */
+  byte = freelist_inliners[i];
+
+  while (i < FREELIST_INLINES) {
+    if (arch_insert_inline_st2_tramp(byte, freelist, tramp,
+        &inline_tramp_table[entry]) == 0) {
+      /* insert occurred, so increment internal counters for the tramp table */
+      entry++;
+      inline_tramp_size++;
+
+      /*
+       * add_freelist() only gets inlined *ONCE* into any of the 3 functions
+       * that we're scanning, so move on to the next 'inliner' after we
+       * tramp the first instruction we find.  REE's gc_sweep has 2 calls,
+       * but this gets optimized into a single inlining and a jmp to it.
+       * older patchlevels of 1.8.7 don't have an add_freelist(), but the
+       * instruction should be the same.
+       */
+      tramps_completed++;
+      i++;
+      byte = freelist_inliners[i];
+      continue;
+    }
+
+    /* if we've looked at all the bytes in this function... */
+    if (((void *)byte - freelist_inliners[i]) >= sizes[i]) {
+      /* move on to the next function */
+      i++;
+      byte = freelist_inliners[i];
+    }
+    byte++;
+  }
+
+  if (tramps_completed != 3)
+    errx(EX_SOFTWARE, "Inline add_freelist tramp insertion failed! "
+         "Only inserted %d tramps.", tramps_completed);
+}
+
+void
+insert_tramp(const char *trampee, void *tramp)
+{
+  void *trampee_addr = bin_find_symbol(trampee, NULL);
+  int inline_ent = inline_tramp_size;
+
+  if (trampee_addr == NULL) {
+    if (strcmp("add_freelist", trampee) == 0) {
+      /* XXX super hack */
+      inline_tramp_size++;
+      hook_freelist(inline_ent, tramp /* freelist_tramp() */);
+    } else {
+      errx(EX_SOFTWARE, "Failed to locate required symbol %s", trampee);
+    }
+  } else {
+    tramp_table[tramp_size].addr = tramp;
+    if (bin_update_image(trampee, &tramp_table[tramp_size]) != 0)
+      errx(EX_SOFTWARE, "Failed to insert tramp for %s", trampee);
+    tramp_size++;
+  }
+}

data/ext/tramp.h

@@ -0,0 +1,23 @@
+#if !defined(TRAMP__)
+#define TRAMP__
+
+/*
+ * create_tramp_table - create the trampoline tables.
+ */
+void
+create_tramp_table();
+
+/*
+ * insert_tramp - insert a trampoline.
+ *
+ * Given:
+ *  - trampee: function in which we want to install the trampoline.
+ *  - tramp:   pointer to the function to be called from the trampoline.
+ *
+ * This function is responsible for installing the requested trampoline
+ * at the location of "trampee".  This results in tramp() being called
+ * whenever trampee() is executed.
+ */
+void
+insert_tramp(const char *trampee, void *tramp);
+#endif

data/ext/util.c

@@ -0,0 +1,72 @@
+/* This is the CRC function used by GNU. Stripped executables may contain a
+ * section .gnu_debuglink which holds the name of an elf object with debug
+ * information and a checksum.
+ */
+#include <stdlib.h>
+/*   !!!! DO NOT MODIFY THIS FUNCTION !!!!
+ *   TODO create specs for this!
+ */
+unsigned long
+gnu_debuglink_crc32(unsigned long crc, unsigned char *buf, size_t len)
+{
+  static const unsigned int crc32_table[256] = {
+    0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419,
+    0x706af48f, 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4,
+    0xe0d5e91e, 0x97d2d988, 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07,
+    0x90bf1d91, 0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de,
+    0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 0x136c9856,
+    0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9,
+    0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4,
+    0xa2677172, 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b,
+    0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3,
+    0x45df5c75, 0xdcd60dcf, 0xabd13d59, 0x26d930ac, 0x51de003a,
+    0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423, 0xcfba9599,
+    0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
+    0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190,
+    0x01db7106, 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f,
+    0x9fbfe4a5, 0xe8b8d433, 0x7807c9a2, 0x0f00f934, 0x9609a88e,
+    0xe10e9818, 0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01,
+    0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 0x6c0695ed,
+    0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950,
+    0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3,
+    0xfbd44c65, 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2,
+    0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a,
+    0x346ed9fc, 0xad678846, 0xda60b8d0, 0x44042d73, 0x33031de5,
+    0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa, 0xbe0b1010,
+    0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
+    0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17,
+    0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6,
+    0x03b6e20c, 0x74b1d29a, 0xead54739, 0x9dd277af, 0x04db2615,
+    0x73dc1683, 0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8,
+    0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 0xf00f9344,
+    0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb,
+    0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a,
+    0x67dd4acc, 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5,
+    0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1,
+    0xa6bc5767, 0x3fb506dd, 0x48b2364b, 0xd80d2bda, 0xaf0a1b4c,
+    0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55, 0x316e8eef,
+    0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
+    0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe,
+    0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31,
+    0x2cd99e8b, 0x5bdeae1d, 0x9b64c2b0, 0xec63f226, 0x756aa39c,
+    0x026d930a, 0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713,
+    0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 0x92d28e9b,
+    0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242,
+    0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1,
+    0x18b74777, 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c,
+    0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 0xa00ae278,
+    0xd70dd2ee, 0x4e048354, 0x3903b3c2, 0xa7672661, 0xd06016f7,
+    0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc, 0x40df0b66,
+    0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
+    0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605,
+    0xcdd70693, 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8,
+    0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b,
+    0x2d02ef8d
+  };
+  unsigned char *end;
+
+  crc = ~crc & 0xffffffff;
+  for (end = buf + len; buf < end; ++buf)
+    crc = crc32_table[(crc ^ *buf) & 0xff] ^ (crc >> 8);
+  return ~crc & 0xffffffff;
+}

data/ext/util.h

@@ -0,0 +1,40 @@
+#if !defined(__util_h__)
+#define __util_h__
+
+#if defined(_MEMPROF_DEBUG)
+#define dbg_printf(...) fprintf(stderr, __VA_ARGS__)
+#else
+#define dbg_printf(...)
+#endif
+
+#define ASSERT_ON_COMPILE(pred) \
+      switch(0){case 0:case pred:;}
+
+struct memprof_config {
+  void *gc_sweep;
+  size_t gc_sweep_size;
+  void *finalize_list;
+  size_t finalize_list_size;
+  void *rb_gc_force_recycle;
+  size_t rb_gc_force_recycle_size;
+  void *freelist;
+  void *classname;
+  void *add_freelist;
+  void *rb_mark_table_add_filename;
+  void *heaps;
+  void *heaps_used;
+  size_t sizeof_RVALUE;
+  size_t sizeof_heaps_slot;
+  int offset_heaps_slot_limit;
+  int offset_heaps_slot_slot;
+  size_t pagesize;
+};
+
+/* This is the CRC function used by GNU. Stripped executables may contain a
+ * section .gnu_debuglink which holds the name of an elf object with debug
+ * information and a checksum.
+ */
+unsigned long
+gnu_debuglink_crc32 (unsigned long crc, unsigned char *buf, size_t len);
+
+#endif

data/ext/x86_64.c

@@ -6,6 +6,7 @@
 
 #include "arch.h"
 #include "x86_gen.h"
+#include "util.h"
 
 /*
  * inline_st1_tramp - inline stage 1 trampoline
@@ -124,8 +125,7 @@ arch_insert_inline_st2_tramp(void *addr, void *marker, void *trampoline, void *t
   struct inline_st1_base *base = addr;
   struct inline_tramp_st2_entry *entry = table_entry;
 
-  /* TODO make this a compile time assert */
-  assert(sizeof(struct inline_st1_base) ==
+  ASSERT_ON_COMPILE(sizeof(struct inline_st1_base) ==
          sizeof(struct inline_st1_tramp));
 
   if (!arch_check_ins(base))

data/lib/memprof/usr2.rb

@@ -3,6 +3,7 @@ Memprof.start
 trap('USR2'){
   pid = Process.pid
   fork{
+    GC.start
     Memprof.dump_all("/tmp/memprof-#{pid}-#{Time.now.to_i}.json")
     exit!
   }

data/memprof.gemspec

@@ -1,12 +1,12 @@
 spec = Gem::Specification.new do |s|
   s.name = 'memprof'
-  s.version = '0.2.7'
-  s.date = '2010-03-13'
+  s.version = '0.2.9'
+  s.date = '2010-03-15'
   s.summary = 'Ruby Memory Profiler'
   s.description = "Ruby memory profiler similar to bleak_house, but without patches to the Ruby VM"
   s.homepage = "http://github.com/ice799/memprof"
   s.has_rdoc = false
-  s.authors = ["Joe Damato", "Aman Gupta", "Jake Douglas"]
+  s.authors = ["Joe Damato", "Aman Gupta", "Jake Douglas", "Rob Benson"]
   s.email = ["joe@memprof.com", "aman@memprof.com", "jake@memprof.com"]
   s.extensions = "ext/extconf.rb"
   s.files = `git ls-files`.split

metadata

@@ -1,17 +1,23 @@
 --- !ruby/object:Gem::Specification 
 name: memprof
 version: !ruby/object:Gem::Version 
-  version: 0.2.7
+  prerelease: false
+  segments: 
+  - 0
+  - 2
+  - 9
+  version: 0.2.9
 platform: ruby
 authors: 
 - Joe Damato
 - Aman Gupta
 - Jake Douglas
+- Rob Benson
 autorequire: 
 bindir: bin
 cert_chain: []
 
-date: 2010-03-13 00:00:00 -08:00
+date: 2010-03-15 00:00:00 -07:00
 default_executable: 
 dependencies: []
 
@@ -41,6 +47,10 @@ files:
 - ext/src/libdwarf-20091118.tar.gz
 - ext/src/libelf-0.8.13.tar.gz
 - ext/src/yajl-1.0.9.tar.gz
+- ext/tramp.c
+- ext/tramp.h
+- ext/util.c
+- ext/util.h
 - ext/x86_64.c
 - ext/x86_64.h
 - ext/x86_gen.h
@@ -61,18 +71,20 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements: 
   - - ">="
     - !ruby/object:Gem::Version 
+      segments: 
+      - 0
       version: "0"
-  version: 
 required_rubygems_version: !ruby/object:Gem::Requirement 
   requirements: 
   - - ">="
     - !ruby/object:Gem::Version 
+      segments: 
+      - 0
       version: "0"
-  version: 
 requirements: []
 
 rubyforge_project: 
-rubygems_version: 1.3.5
+rubygems_version: 1.3.6
 signing_key: 
 specification_version: 3
 summary: Ruby Memory Profiler