packages/gperftools
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Package init

Browse Source
This commit is contained in:
overweight 2019-09-30 10:51:34 -04:00
commit 3713cee362
17 changed files with 1221 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

64
Avoid-static-initialization-of-pprof-path-for-symbol.patch Normal file
View File

@ -0,0 +1,64 @@
From 893bff51bcf220b724a812d340d878b5fb8ce911 Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 26 Aug 2018 11:35:44 -0700
Subject: [PATCH 23/39] Avoid static initialization of pprof path for
symbolization.
This is one of the things that chrome's fork fixes, but with c++11 we
can do it even nicer. Proposed fix is to use c++11 local static
variable to ensure that pprof path is initialized once on as-needed
basis.
---
src/symbolize.cc | 20 +++++++++++---------
1 file changed, 11 insertions(+), 9 deletions(-)
diff --git a/src/symbolize.cc b/src/symbolize.cc
index 88609ff..4c71010 100755
--- a/src/symbolize.cc
+++ b/src/symbolize.cc
@@ -67,15 +67,17 @@
using std::string;
using tcmalloc::DumpProcSelfMaps; // from sysinfo.h
-
-DEFINE_string(symbolize_pprof,
- EnvToString("PPROF_PATH", "pprof"),
- "Path to pprof to call for reporting function names.");
-
-// heap_profile_table_pprof may be referenced after destructors are
+// pprof may be used after destructors are
// called (since that's when leak-checking is done), so we make
// a more-permanent copy that won't ever get destroyed.
-static string* g_pprof_path = new string(FLAGS_symbolize_pprof);
+static char* get_pprof_path() {
+ static char* result = ([] () {
+ string pprof_string = EnvToString("PPROF_PATH", "pprof");
+ return strdup(pprof_string.c_str());
+ })();
+
+ return result;
+}
// Returns NULL if we're on an OS where we can't get the invocation name.
// Using a static var is ok because we're not called from a thread.
@@ -144,7 +146,7 @@ int SymbolTable::Symbolize() {
PrintError("Cannot figure out the name of this executable (argv0)");
return 0;
}
- if (access(g_pprof_path->c_str(), R_OK) != 0) {
+ if (access(get_pprof_path(), R_OK) != 0) {
PrintError("Cannot find 'pprof' (is PPROF_PATH set correctly?)");
return 0;
}
@@ -206,7 +208,7 @@ int SymbolTable::Symbolize() {
unsetenv("HEAPPROFILE");
unsetenv("HEAPCHECK");
unsetenv("PERFTOOLS_VERBOSE");
- execlp(g_pprof_path->c_str(), g_pprof_path->c_str(),
+ execlp(get_pprof_path(), get_pprof_path(),
"--symbols", argv0, NULL);
_exit(3); // if execvp fails, it's bad news for us
}
--
1.8.3.1

70
Disable-large-allocation-report-by-default.patch Normal file
View File

@ -0,0 +1,70 @@
From 99d4b81d3fcedb7c912acbd813573a89d32ade22 Mon Sep 17 00:00:00 2001
From: Junhao Li <streaver91@gmail.com>
Date: Sun, 20 May 2018 13:45:32 -0400
Subject: [PATCH] Disable large allocation report by default
Fixes issue #360.
[alkondratenko@gmail.com: adjusted commit message a bit]
[alkondratenko@gmail.com: adjusted configure help message]
Signed-off-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
---
configure.ac | 9 +++++++++
src/tcmalloc.cc | 4 ++++
2 files changed, 13 insertions(+)
diff --git a/configure.ac b/configure.ac
index fa03f1a..f874e19 100644
--- a/configure.ac
+++ b/configure.ac
@@ -663,6 +663,15 @@ AS_IF([test "x$enable_generic_dynamic_tls" = xyes],
[AC_DEFINE([ENABLE_GENERIC_DYNAMIC_TLS], 1,
[Use the generic dynamic TLS model])])
+# Disable large allocation report by default.
+AC_ARG_ENABLE([large-alloc-report],
+ [AS_HELP_STRING([--enable-large-alloc-report],
+ [report very large allocations to stderr])],
+ [enable_large_alloc_report="$enableval"],
+ [enable_large_alloc_report=no])
+AS_IF([test "x$enable_large_alloc_report" = xyes],
+ [AC_DEFINE([ENABLE_LARGE_ALLOC_REPORT], 1, [report large allocation])])
+
# Write generated configuration file
AC_CONFIG_FILES([Makefile
src/gperftools/tcmalloc.h src/windows/gperftools/tcmalloc.h])
diff --git a/src/tcmalloc.cc b/src/tcmalloc.cc
index 7b18ddb..48d4530 100644
--- a/src/tcmalloc.cc
+++ b/src/tcmalloc.cc
@@ -1272,9 +1272,11 @@ void* handle_oom(malloc_fn retry_fn,
// Copy of FLAGS_tcmalloc_large_alloc_report_threshold with
// automatic increases factored in.
+#ifdef ENABLE_LARGE_ALLOC_REPORT
static int64_t large_alloc_threshold =
(kPageSize > FLAGS_tcmalloc_large_alloc_report_threshold
? kPageSize : FLAGS_tcmalloc_large_alloc_report_threshold);
+#endif
static void ReportLargeAlloc(Length num_pages, void* result) {
StackTrace stack;
@@ -1295,6 +1297,7 @@ static void ReportLargeAlloc(Length num_pages, void* result) {
// Must be called with the page lock held.
inline bool should_report_large(Length num_pages) {
+#ifdef ENABLE_LARGE_ALLOC_REPORT
const int64 threshold = large_alloc_threshold;
if (threshold > 0 && num_pages >= (threshold >> kPageShift)) {
// Increase the threshold by 1/8 every time we generate a report.
@@ -1303,6 +1306,7 @@ inline bool should_report_large(Length num_pages) {
? threshold + threshold/8 : 8ll<<30);
return true;
}
+#endif
return false;
}
--
1.8.3.1

196
Drop-not-very-portable-and-not-very-useful-unwind-be.patch Normal file
View File

@ -0,0 +1,196 @@
From 630c9ce77822212a4ce3ee92afabbec0eea0f94b Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 7 Oct 2018 08:17:04 -0700
Subject: [PATCH] Drop not very portable and not very useful unwind benchmark.
---
Makefile.am | 8 ----
benchmark/getcontext_light.cc | 60 -----------------------------
benchmark/unwind_bench.cc | 89 -------------------------------------------
3 files changed, 157 deletions(-)
delete mode 100644 benchmark/getcontext_light.cc
delete mode 100644 benchmark/unwind_bench.cc
diff --git a/Makefile.am b/Makefile.am
index c79227a..1f52d3d 100755
--- a/Makefile.am
+++ b/Makefile.am
@@ -881,14 +881,6 @@ malloc_bench_shared_full_CXXFLAGS = $(PTHREAD_CFLAGS) $(AM_CXXFLAGS) $(NO_BUILTI
malloc_bench_shared_full_LDFLAGS = $(PTHREAD_CFLAGS) $(TCMALLOC_FLAGS)
malloc_bench_shared_full_LDADD = librun_benchmark.la libtcmalloc.la $(PTHREAD_LIBS)
-if !OSX
-noinst_PROGRAMS += unwind_bench
-unwind_bench_SOURCES = benchmark/unwind_bench.cc benchmark/getcontext_light.cc
-unwind_bench_CXXFLAGS = $(PTHREAD_CFLAGS) $(AM_CXXFLAGS) $(NO_BUILTIN_CXXFLAGS)
-unwind_bench_LDFLAGS = $(PTHREAD_CFLAGS) $(TCMALLOC_FLAGS)
-unwind_bench_LDADD = librun_benchmark.la libtcmalloc.la $(PTHREAD_LIBS)
-endif !OSX
-
endif WITH_HEAP_PROFILER_OR_CHECKER
binary_trees_SOURCES = benchmark/binary_trees.cc
diff --git a/benchmark/getcontext_light.cc b/benchmark/getcontext_light.cc
deleted file mode 100644
index e8ff8b3..0000000
--- a/benchmark/getcontext_light.cc
+++ /dev/null
@@ -1,60 +0,0 @@
-#include <ucontext.h>
-#include <stddef.h>
-
-extern "C" void getcontext_light(ucontext_t *ctx);
-
-// clang's built-in asm cannot handle .set directives
-#if defined(__GNUC__) && !defined(__llvm__) && defined(__x86_64) && defined(_LP64)
-
-#define R(r) offsetof(ucontext_t, uc_mcontext.gregs[r])
-
-static __attribute__((used))
-void getcontext_tramp(ucontext_t *ctx) {
- __asm__ __volatile__(".set oRBX, %c0\n"
- ".set oRBP, %c1\n"
- ".set oR12, %c2\n"
- ".set oR13, %c3\n"
- ".set oR14, %c4\n"
- ".set oR15, %c5\n"
- ".set oRIP, %c6\n"
- ".set oRSP, %c7\n"
- : :
- "p" (R(REG_RBX)),
- "p" (R(REG_RBP)),
- "p" (R(REG_R12)),
- "p" (R(REG_R13)),
- "p" (R(REG_R14)),
- "p" (R(REG_R15)),
- "p" (R(REG_RIP)),
- "p" (R(REG_RSP)));
- getcontext_light(ctx);
-}
-
-__asm__(".pushsection .text; .globl getcontext_light\n"
- ".type getcontext_light, @function\n"
-"getcontext_light:\n"
- "\t.cfi_startproc\n"
- "\tmovq %rbx, oRBX(%rdi)\n"
- "\tmovq %rbp, oRBP(%rdi)\n"
- "\tmovq %r12, oR12(%rdi)\n"
- "\tmovq %r13, oR13(%rdi)\n"
- "\tmovq %r14, oR14(%rdi)\n"
- "\tmovq %r15, oR14(%rdi)\n"
-
- "\tmovq (%rsp), %rcx\n"
- "\tmovq %rcx, oRIP(%rdi)\n"
- "\tleaq 8(%rsp), %rcx\n" /* Exclude the return address. */
- "\tmovq %rcx, oRSP(%rdi)\n"
- "\tret\n"
- ".cfi_endproc\n"
- ".size getcontext_light, .-getcontext_light\n"
- ".popsection\n"
- );
-
-#else
-
-extern "C" void getcontext_light(ucontext_t *ctx) {
- getcontext(ctx);
-}
-
-#endif
diff --git a/benchmark/unwind_bench.cc b/benchmark/unwind_bench.cc
deleted file mode 100644
index 9c4e2f8..0000000
--- a/benchmark/unwind_bench.cc
+++ /dev/null
@@ -1,89 +0,0 @@
-// -*- Mode: C++; c-basic-offset: 2; indent-tabs-mode: nil -*-
-#include "config.h"
-
-#include "base/basictypes.h"
-#include "gperftools/stacktrace.h"
-
-#include <stdlib.h>
-#include <stdio.h>
-#include <stdint.h>
-#include <sys/ucontext.h>
-#if HAVE_LIBUNWIND_H
-#include <libunwind.h>
-#endif
-
-#include "run_benchmark.h"
-
-extern "C" void getcontext_light(ucontext_t *ctx);
-
-#define MAX_FRAMES 1024
-static void *frames[MAX_FRAMES];
-
-enum measure_mode {
- MODE_NOOP,
- MODE_WITH_CONTEXT,
- MODE_WITHOUT_CONTEXT
-};
-
-static int ATTRIBUTE_NOINLINE measure_unwind(int maxlevel, int mode) {
- int n;
-
- if (mode == MODE_NOOP)
- return 0;
-
- if (mode == MODE_WITH_CONTEXT) {
- ucontext_t uc;
- getcontext_light(&uc);
- n = GetStackTraceWithContext(frames, MAX_FRAMES, 0, &uc);
- } else {
- n = GetStackTrace(frames, MAX_FRAMES, 0);
- }
- if (n < maxlevel) {
- fprintf(stderr, "Expected at least %d frames, got %d\n", maxlevel, n);
- abort();
- }
- return 0;
-}
-
-static int ATTRIBUTE_NOINLINE frame_forcer(int rv) {
- return rv;
-}
-
-static int ATTRIBUTE_NOINLINE f1(int level, int maxlevel, int mode) {
- if (level == maxlevel)
- return frame_forcer(measure_unwind(maxlevel, mode));
- return frame_forcer(f1(level + 1, maxlevel, mode));
-}
-
-static void bench_unwind_no_op(long iterations, uintptr_t param) {
- do {
- f1(0, param, MODE_NOOP);
- iterations -= param;
- } while (iterations > 0);
-}
-
-static void bench_unwind_context(long iterations, uintptr_t param) {
- do {
- f1(0, param, MODE_WITH_CONTEXT);
- iterations -= param;
- } while (iterations > 0);
-}
-
-static void bench_unwind_no_context(long iterations, uintptr_t param) {
- do {
- f1(0, param, MODE_WITHOUT_CONTEXT);
- iterations -= param;
- } while (iterations > 0);
-}
-
-int main(void) {
- report_benchmark("unwind_no_op", bench_unwind_no_op, 100);
- report_benchmark("unwind_context", bench_unwind_context, 100);
- report_benchmark("unwind_no_context", bench_unwind_no_context, 100);
-
-//// TODO: somehow this fails at linking step. Figure out why this is missing
-// #if HAVE_LIBUNWIND_H
-// unw_set_caching_policy(unw_local_addr_space, UNW_CACHE_PER_THREAD);
-// #endif
- return 0;
-}
--
1.8.3.1

101
Fix-incompatible-aliasing-warnings.patch Normal file
View File

@ -0,0 +1,101 @@
From 71c8cedacafe700e662c008b65f5064b23426070 Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 5 Aug 2018 18:54:01 -0700
Subject: [PATCH 19/39] Fix incompatible aliasing warnings
We aliased functions with different signatures and gcc now correctly
gives warning for that. Originally gcc 5 same code merging feature
caused us to alias more than necessary, but I am not able to reproduce
this problem anymore. So we're now aliasing only compatible functions.
---
src/libc_override_gcc_and_weak.h | 8 ++++----
src/tcmalloc.cc | 16 ----------------
2 files changed, 4 insertions(+), 20 deletions(-)
diff --git a/src/libc_override_gcc_and_weak.h b/src/libc_override_gcc_and_weak.h
index 6875164..bb99b69 100644
--- a/src/libc_override_gcc_and_weak.h
+++ b/src/libc_override_gcc_and_weak.h
@@ -126,9 +126,9 @@ void operator delete[](void *p, size_t size) CPP_NOTHROW
#else /* !ENABLE_SIZED_DELETE && !ENABLE_DYN_SIZED_DELETE */
void operator delete(void *p, size_t size) CPP_NOTHROW
- ALIAS(tc_delete);
+ ALIAS(tc_delete_sized);
void operator delete[](void *p, size_t size) CPP_NOTHROW
- ALIAS(tc_deletearray);
+ ALIAS(tc_deletearray_sized);
#endif /* !ENABLE_SIZED_DELETE && !ENABLE_DYN_SIZED_DELETE */
@@ -197,9 +197,9 @@ void operator delete[](void *p, size_t size, std::align_val_t al) CPP_NOTHROW
#else /* defined(ENABLE_DYN_SIZED_DELETE) */
void operator delete(void *p, size_t size, std::align_val_t al) CPP_NOTHROW
- ALIAS(tc_delete);
+ ALIAS(tc_delete_sized_aligned);
void operator delete[](void *p, size_t size, std::align_val_t al) CPP_NOTHROW
- ALIAS(tc_deletearray);
+ ALIAS(tc_deletearray_sized_aligned);
#endif /* defined(ENABLE_DYN_SIZED_DELETE) */
diff --git a/src/tcmalloc.cc b/src/tcmalloc.cc
index 37c1440..b320b30 100644
--- a/src/tcmalloc.cc
+++ b/src/tcmalloc.cc
@@ -1998,9 +1998,6 @@ TC_ALIAS(tc_free);
// (via ::operator delete(ptr, nothrow)).
// But it's really the same as normal delete, so we just do the same thing.
extern "C" PERFTOOLS_DLL_DECL void tc_delete_nothrow(void* p, const std::nothrow_t&) PERFTOOLS_NOTHROW
-#ifdef TC_ALIAS
-TC_ALIAS(tc_free);
-#else
{
if (PREDICT_FALSE(!base::internal::delete_hooks_.empty())) {
tcmalloc::invoke_hooks_and_free(p);
@@ -2008,7 +2005,6 @@ TC_ALIAS(tc_free);
}
do_free(p);
}
-#endif
extern "C" PERFTOOLS_DLL_DECL void* tc_newarray(size_t size)
#ifdef TC_ALIAS
@@ -2080,33 +2076,21 @@ extern "C" PERFTOOLS_DLL_DECL void* tc_new_aligned_nothrow(size_t size, std::ali
}
extern "C" PERFTOOLS_DLL_DECL void tc_delete_aligned(void* p, std::align_val_t) PERFTOOLS_NOTHROW
-#ifdef TC_ALIAS
-TC_ALIAS(tc_delete);
-#else
{
free_fast_path(p);
}
-#endif
// There is no easy way to obtain the actual size used by do_memalign to allocate aligned storage, so for now
// just ignore the size. It might get useful in the future.
extern "C" PERFTOOLS_DLL_DECL void tc_delete_sized_aligned(void* p, size_t size, std::align_val_t align) PERFTOOLS_NOTHROW
-#ifdef TC_ALIAS
-TC_ALIAS(tc_delete);
-#else
{
free_fast_path(p);
}
-#endif
extern "C" PERFTOOLS_DLL_DECL void tc_delete_aligned_nothrow(void* p, std::align_val_t, const std::nothrow_t&) PERFTOOLS_NOTHROW
-#ifdef TC_ALIAS
-TC_ALIAS(tc_delete);
-#else
{
free_fast_path(p);
}
-#endif
extern "C" PERFTOOLS_DLL_DECL void* tc_newarray_aligned(size_t size, std::align_val_t align)
#ifdef TC_ALIAS
--
1.8.3.1

81
Fix-mmap-region-iteration-while-no-regions-are-recor.patch Normal file
View File

@ -0,0 +1,81 @@
From 1de76671d478e848f0dd210e00df43bfa463ca9f Mon Sep 17 00:00:00 2001
From: Gabriel Marin <gmx@chromium.org>
Date: Thu, 4 Oct 2018 22:52:35 +0000
Subject: [PATCH 31/39] Fix mmap region iteration while no regions are
recorded.
If no mmap regions are recorded, iteration failed since the RegionSet
(std::set) object is not initialized.
Original CL https://codereview.chromium.org/14769008
Reviewed-on: https://chromium-review.googlesource.com/c/1130807
---
src/memory_region_map.cc | 19 +++++++++++++------
src/memory_region_map.h | 3 +++
2 files changed, 16 insertions(+), 6 deletions(-)
mode change 100755 => 100644 src/memory_region_map.cc
diff --git a/src/memory_region_map.cc b/src/memory_region_map.cc
old mode 100755
new mode 100644
index 841d6f3..06b6fb0
--- a/src/memory_region_map.cc
+++ b/src/memory_region_map.cc
@@ -234,6 +234,9 @@ void MemoryRegionMap::Init(int max_stack_depth, bool use_buckets) {
memset(bucket_table_, 0, table_bytes);
num_buckets_ = 0;
}
+ if (regions_ == NULL) { // init regions_
+ InitRegionSetLocked();
+ }
Unlock();
RAW_VLOG(10, "MemoryRegionMap Init done");
}
@@ -536,6 +539,15 @@ void MemoryRegionMap::RestoreSavedBucketsLocked() {
}
}
+inline void MemoryRegionMap::InitRegionSetLocked() {
+ RAW_VLOG(12, "Initializing region set");
+ regions_ = regions_rep.region_set();
+ recursive_insert = true;
+ new (regions_) RegionSet();
+ HandleSavedRegionsLocked(&DoInsertRegionLocked);
+ recursive_insert = false;
+}
+
inline void MemoryRegionMap::InsertRegionLocked(const Region& region) {
RAW_CHECK(LockIsHeld(), "should be held (by this thread)");
// We can be called recursively, because RegionSet constructor
@@ -556,12 +568,7 @@ inline void MemoryRegionMap::InsertRegionLocked(const Region& region) {
saved_regions[saved_regions_count++] = region;
} else { // not a recusrive call
if (regions_ == NULL) { // init regions_
- RAW_VLOG(12, "Initializing region set");
- regions_ = regions_rep.region_set();
- recursive_insert = true;
- new(regions_) RegionSet();
- HandleSavedRegionsLocked(&DoInsertRegionLocked);
- recursive_insert = false;
+ InitRegionSetLocked();
}
recursive_insert = true;
// Do the actual insertion work to put new regions into regions_:
diff --git a/src/memory_region_map.h b/src/memory_region_map.h
index ec388e1..f774994 100644
--- a/src/memory_region_map.h
+++ b/src/memory_region_map.h
@@ -362,6 +362,9 @@ class MemoryRegionMap {
// table where all buckets eventually should be.
static void RestoreSavedBucketsLocked();
+ // Initialize RegionSet regions_.
+ inline static void InitRegionSetLocked();
+
// Wrapper around DoInsertRegionLocked
// that handles the case of recursive allocator calls.
inline static void InsertRegionLocked(const Region& region);
--
1.8.3.1

194
Fix-page_heap_test-flakiness.patch Normal file
View File

@ -0,0 +1,194 @@
From f2bca77aed83107197abadee0652c39e2c4c73d3 Mon Sep 17 00:00:00 2001
From: Brian Silverman <brian@peloton-tech.com>
Date: Thu, 9 Aug 2018 16:33:39 -0700
Subject: [PATCH 21/39] Fix page_heap_test flakiness
After 71fa9f87 (use 2-level page map for 48-bit addresses), there is a
potential for allocs of additional leaves in the middle of the test.
With ASLR enabled on amd64 (without TCMALLOC_SMALL_BUT_SLOW), this
results in the test failing sometimes (< 1% of the time). Instead, make
sure all those leaves are allocated before deciding on the baseline heap
usage in the test.
---
src/tests/page_heap_test.cc | 79 ++++++++++++++++++++++++++++++++-------------
1 file changed, 56 insertions(+), 23 deletions(-)
diff --git a/src/tests/page_heap_test.cc b/src/tests/page_heap_test.cc
index e82a1da..3caacc0 100644
--- a/src/tests/page_heap_test.cc
+++ b/src/tests/page_heap_test.cc
@@ -6,9 +6,13 @@
// be found in the LICENSE file.
#include "config_for_unittests.h"
+
+#include <stdio.h>
+
+#include <memory>
+
#include "page_heap.h"
#include "system-alloc.h"
-#include <stdio.h>
#include "base/logging.h"
#include "common.h"
@@ -39,33 +43,63 @@ static void CheckStats(const tcmalloc::PageHeap* ph,
}
static void TestPageHeap_Stats() {
- tcmalloc::PageHeap* ph = new tcmalloc::PageHeap();
+ std::unique_ptr<tcmalloc::PageHeap> ph(new tcmalloc::PageHeap());
// Empty page heap
- CheckStats(ph, 0, 0, 0);
+ CheckStats(ph.get(), 0, 0, 0);
// Allocate a span 's1'
tcmalloc::Span* s1 = ph->New(256);
- CheckStats(ph, 256, 0, 0);
+ CheckStats(ph.get(), 256, 0, 0);
// Split span 's1' into 's1', 's2'. Delete 's2'
tcmalloc::Span* s2 = ph->Split(s1, 128);
ph->Delete(s2);
- CheckStats(ph, 256, 128, 0);
+ CheckStats(ph.get(), 256, 128, 0);
// Unmap deleted span 's2'
ph->ReleaseAtLeastNPages(1);
- CheckStats(ph, 256, 0, 128);
+ CheckStats(ph.get(), 256, 0, 128);
// Delete span 's1'
ph->Delete(s1);
- CheckStats(ph, 256, 128, 128);
+ CheckStats(ph.get(), 256, 128, 128);
+}
- delete ph;
+// The number of kMaxPages-sized Spans we will allocate and free during the
+// tests.
+// We will also do twice this many kMaxPages/2-sized ones.
+static constexpr int kNumberMaxPagesSpans = 10;
+
+// Allocates all the last-level page tables we will need. Doing this before
+// calculating the base heap usage is necessary, because otherwise if any of
+// these are allocated during the main test it will throw the heap usage
+// calculations off and cause the test to fail.
+static void AllocateAllPageTables() {
+ // Make a separate PageHeap from the main test so the test can start without
+ // any pages in the lists.
+ std::unique_ptr<tcmalloc::PageHeap> ph(new tcmalloc::PageHeap());
+ tcmalloc::Span *spans[kNumberMaxPagesSpans * 2];
+ for (int i = 0; i < kNumberMaxPagesSpans; ++i) {
+ spans[i] = ph->New(kMaxPages);
+ EXPECT_NE(spans[i], NULL);
+ }
+ for (int i = 0; i < kNumberMaxPagesSpans; ++i) {
+ ph->Delete(spans[i]);
+ }
+ for (int i = 0; i < kNumberMaxPagesSpans * 2; ++i) {
+ spans[i] = ph->New(kMaxPages >> 1);
+ EXPECT_NE(spans[i], NULL);
+ }
+ for (int i = 0; i < kNumberMaxPagesSpans * 2; ++i) {
+ ph->Delete(spans[i]);
+ }
}
static void TestPageHeap_Limit() {
- tcmalloc::PageHeap* ph = new tcmalloc::PageHeap();
+ AllocateAllPageTables();
+
+ std::unique_ptr<tcmalloc::PageHeap> ph(new tcmalloc::PageHeap());
CHECK_EQ(kMaxPages, 1 << (20 - kPageShift));
@@ -77,25 +111,26 @@ static void TestPageHeap_Limit() {
while((s = ph->New(kMaxPages)) == NULL) {
FLAGS_tcmalloc_heap_limit_mb++;
}
- FLAGS_tcmalloc_heap_limit_mb += 9;
+ FLAGS_tcmalloc_heap_limit_mb += kNumberMaxPagesSpans - 1;
ph->Delete(s);
// We are [10, 11) mb from the limit now.
}
// Test AllocLarge and GrowHeap first:
{
- tcmalloc::Span * spans[10];
- for (int i=0; i<10; ++i) {
+ tcmalloc::Span * spans[kNumberMaxPagesSpans];
+ for (int i=0; i<kNumberMaxPagesSpans; ++i) {
spans[i] = ph->New(kMaxPages);
EXPECT_NE(spans[i], NULL);
}
EXPECT_EQ(ph->New(kMaxPages), NULL);
- for (int i=0; i<10; i += 2) {
+ for (int i=0; i<kNumberMaxPagesSpans; i += 2) {
ph->Delete(spans[i]);
}
- tcmalloc::Span *defragmented = ph->New(5 * kMaxPages);
+ tcmalloc::Span *defragmented =
+ ph->New(kNumberMaxPagesSpans / 2 * kMaxPages);
if (HaveSystemRelease) {
// EnsureLimit should release deleted normal spans
@@ -109,15 +144,15 @@ static void TestPageHeap_Limit() {
EXPECT_TRUE(ph->CheckExpensive());
}
- for (int i=1; i<10; i += 2) {
+ for (int i=1; i<kNumberMaxPagesSpans; i += 2) {
ph->Delete(spans[i]);
}
}
// Once again, testing small lists this time (twice smaller spans):
{
- tcmalloc::Span * spans[20];
- for (int i=0; i<20; ++i) {
+ tcmalloc::Span * spans[kNumberMaxPagesSpans * 2];
+ for (int i=0; i<kNumberMaxPagesSpans * 2; ++i) {
spans[i] = ph->New(kMaxPages >> 1);
EXPECT_NE(spans[i], NULL);
}
@@ -125,12 +160,12 @@ static void TestPageHeap_Limit() {
tcmalloc::Span * lastHalf = ph->New(kMaxPages >> 1);
EXPECT_EQ(ph->New(kMaxPages >> 1), NULL);
- for (int i=0; i<20; i += 2) {
+ for (int i=0; i<kNumberMaxPagesSpans * 2; i += 2) {
ph->Delete(spans[i]);
}
- for(Length len = kMaxPages >> 2; len < 5 * kMaxPages; len = len << 1)
- {
+ for (Length len = kMaxPages >> 2;
+ len < kNumberMaxPagesSpans / 2 * kMaxPages; len = len << 1) {
if(len <= kMaxPages >> 1 || HaveSystemRelease) {
tcmalloc::Span *s = ph->New(len);
EXPECT_NE(s, NULL);
@@ -140,7 +175,7 @@ static void TestPageHeap_Limit() {
EXPECT_TRUE(ph->CheckExpensive());
- for (int i=1; i<20; i += 2) {
+ for (int i=1; i<kNumberMaxPagesSpans * 2; i += 2) {
ph->Delete(spans[i]);
}
@@ -148,8 +183,6 @@ static void TestPageHeap_Limit() {
ph->Delete(lastHalf);
}
}
-
- delete ph;
}
} // namespace
--
1.8.3.1

31
Fix-uninitialized-memory-use-in-sampler_test.patch Normal file
View File

@ -0,0 +1,31 @@
From 0b588e7490eac4ef37b59c9265f1f3f05f4bb27c Mon Sep 17 00:00:00 2001
From: Brian Silverman <brian@peloton-tech.com>
Date: Tue, 31 Jul 2018 15:53:47 -0700
Subject: [PATCH 05/39] Fix uninitialized memory use in sampler_test
Sampler's documentation states the following:
C++03 requires that types stored in TLS be POD. As a result, you must
initialize these members to {0, 0, false} before using this class!
However, the test code wasn't doing that. MemorySanitizer and
UndefinedBehaviorSanitizer both failed because of it.
---
src/tests/sampler_test.cc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/tests/sampler_test.cc b/src/tests/sampler_test.cc
index e0d24d4..9a55b70 100755
--- a/src/tests/sampler_test.cc
+++ b/src/tests/sampler_test.cc
@@ -361,7 +361,7 @@ double StandardDeviationsErrorInSample(
}
TEST(Sampler, LargeAndSmallAllocs_CombinedTest) {
- tcmalloc::Sampler sampler;
+ tcmalloc::Sampler sampler{0, 0, false};
sampler.Init(1);
int counter_big = 0;
int counter_small = 0;
--
1.8.3.1

79
Format-and-fix-out-of-bound-access-in-CpuProfilerSwi.patch Normal file
View File

@ -0,0 +1,79 @@
From 8dd304035848e780372a5a4bb1afac2b2d20bc43 Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 5 Aug 2018 19:52:11 -0700
Subject: [PATCH 18/39] Format and fix out of bound access in CpuProfilerSwitch
GCC was giving warning on snprintf and it hinted at base_profile_name
and full_profile_name not being long enough.
Fix is to ensure base_profile_name is long enough for PATH_MAX and
that full_profile_name is enough longer to fit extra chars.
---
src/profiler.cc | 49 +++++++++++++++++++++++--------------------------
1 file changed, 23 insertions(+), 26 deletions(-)
diff --git a/src/profiler.cc b/src/profiler.cc
index f4f5990..3bd0ed9 100644
--- a/src/profiler.cc
+++ b/src/profiler.cc
@@ -144,34 +144,31 @@ class CpuProfiler {
// number is defined in the environment variable CPUPROFILESIGNAL.
static void CpuProfilerSwitch(int signal_number)
{
- bool static started = false;
- static unsigned profile_count = 0;
- static char base_profile_name[1024] = "\0";
-
- if (base_profile_name[0] == '\0') {
- if (!GetUniquePathFromEnv("CPUPROFILE", base_profile_name)) {
- RAW_LOG(FATAL,"Cpu profiler switch is registered but no CPUPROFILE is defined");
- return;
- }
- }
- if (!started)
- {
- char full_profile_name[1024];
-
- snprintf(full_profile_name, sizeof(full_profile_name), "%s.%u",
- base_profile_name, profile_count++);
-
- if(!ProfilerStart(full_profile_name))
- {
- RAW_LOG(FATAL, "Can't turn on cpu profiling for '%s': %s\n",
- full_profile_name, strerror(errno));
- }
+ static unsigned profile_count;
+ static char base_profile_name[PATH_MAX];
+ static bool started = false;
+
+ if (base_profile_name[0] == '\0') {
+ if (!GetUniquePathFromEnv("CPUPROFILE", base_profile_name)) {
+ RAW_LOG(FATAL,"Cpu profiler switch is registered but no CPUPROFILE is defined");
+ return;
}
- else
- {
- ProfilerStop();
+ }
+
+ if (!started) {
+ char full_profile_name[PATH_MAX + 16];
+
+ snprintf(full_profile_name, sizeof(full_profile_name), "%s.%u",
+ base_profile_name, profile_count++);
+
+ if(!ProfilerStart(full_profile_name)) {
+ RAW_LOG(FATAL, "Can't turn on cpu profiling for '%s': %s\n",
+ full_profile_name, strerror(errno));
}
- started = !started;
+ } else {
+ ProfilerStop();
+ }
+ started = !started;
}
// Profile data structure singleton: Constructor will check to see if
--
1.8.3.1

71
Use-initial-exec-TLS-model-for-all-thread-local-vari.patch Normal file
View File

@ -0,0 +1,71 @@
From 1e36ed705548ae2a5170bb00dfee0954b8ef76de Mon Sep 17 00:00:00 2001
From: Romain Geissler <romain.geissler@amadeus.com>
Date: Tue, 29 Jan 2019 18:31:19 +0000
Subject: [PATCH 38/39] Use initial exec TLS model for all thread local
variables from thread_cache.cc
This avoids a deadlock when a library which is being dlopen'ed creates
as part of its static constructors a thread which quickly need to call
malloc. We are still in the dlopen call (so with some internal glibc
mutex taken) when the thread executes code and later needs to call
malloc which in term calls tls_get_addr_tail, which wait for the dlopen
mutex to be unlocked. If later the dlopen'ing thread also calls malloc
as part of its constructors, we are in a deadlock.
Fix is similar to
https://github.com/gperftools/gperftools/commit/7852eeb75b9375cf52a7da01be044da6e915dd08
Stack of the dlopening thread:
#0 0x00007fd5406ca93c in __lll_lock_wait () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
#1 0x00007fd5406c45a5 in pthread_mutex_lock () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
... proprietary code in the stack
#9 0x00007fd5074f0367 in __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at src/ClientImpl.cpp:15
#10 0x00007fd5074f06d7 in _GLOBAL__sub_I_ClientImpl.cpp(void) () at src/ClientImpl.cpp:85
#11 0x00007fd50757aa46 in __do_global_ctors_aux ()
#12 0x00007fd5073e985f in _init () from ...
#13 0x00007fd53bf9dec8 in ?? () from ...
#14 0x00007fd54d637a5d in call_init.part () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#15 0x00007fd54d637bab in _dl_init () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#16 0x00007fd54d63c160 in dl_open_worker () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#17 0x00007fd54d637944 in _dl_catch_error () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#18 0x00007fd54d63b7d9 in _dl_open () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#19 0x00007fd54d61f2b9 in dlopen_doit () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
#20 0x00007fd54d637944 in _dl_catch_error () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#21 0x00007fd54d61f889 in _dlerror_run () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
#22 0x00007fd54d61f351 in dlopen@@GLIBC_2.2.5 () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
Stack of the newly created thread calling tls_get_addr_tail:
#0 0x00007fd5406ca93c in __lll_lock_wait () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
#1 0x00007fd5406c4622 in pthread_mutex_lock () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
#2 0x00007fd54d63a2ed in tls_get_addr_tail () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
#3 0x00007fd53fee877d in tcmalloc::ThreadCache::CreateCacheIfNecessary () at src/thread_cache.cc:344
#4 0x00007fd53fecb4ab in tcmalloc::ThreadCache::GetCache () at src/thread_cache.h:437
#5 0x00007fd53fefeccb in (anonymous namespace)::do_malloc (size=56) at src/tcmalloc.cc:1354
#6 tcmalloc::do_allocate_full<tcmalloc::cpp_throw_oom> (size=56) at src/tcmalloc.cc:1762
#7 tcmalloc::allocate_full_cpp_throw_oom (size=56) at src/tcmalloc.cc:1776
#8 0x00007fd53ff01b80 in tcmalloc::dispatch_allocate_full<tcmalloc::cpp_throw_oom> (size=56) at src/tcmalloc.cc:1785
#9 malloc_fast_path<tcmalloc::cpp_throw_oom> (size=56) at src/tcmalloc.cc:1845
#10 tc_new (size=56) at src/tcmalloc.cc:1980
... proprietary code in the stack
#26 0x00007fd5406c1ef4 in start_thread () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
#27 0x00007fd5403ba01d in clone () from /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libc.so.6
---
src/thread_cache.cc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/thread_cache.cc b/src/thread_cache.cc
index 6d2f832..97f3ffe 100644
--- a/src/thread_cache.cc
+++ b/src/thread_cache.cc
@@ -334,7 +334,7 @@ ThreadCache* ThreadCache::CreateCacheIfNecessary() {
bool seach_condition = true;
#ifdef HAVE_TLS
- static __thread ThreadCache** current_heap_ptr;
+ static __thread ThreadCache** current_heap_ptr ATTR_INITIAL_EXEC;
if (tsd_inited_) {
// In most common case we're avoiding expensive linear search
// through all heaps (see below). Working TLS enables faster
--
1.8.3.1

29
Use-standard-conforming-alignof-in-debugallocation.c.patch Normal file
View File

@ -0,0 +1,29 @@
From c41688bf20186723367b560ceb539b2330951ddb Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 5 Aug 2018 20:39:07 -0700
Subject: [PATCH 20/39] Use standard-conforming alignof in debugallocation.cc
Clang was giving warning that alignof(<variable>) is extension. So
we're now doing alignof of decltype.
---
src/debugallocation.cc | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/debugallocation.cc b/src/debugallocation.cc
index 7c438f2..39dea5a 100644
--- a/src/debugallocation.cc
+++ b/src/debugallocation.cc
@@ -1152,8 +1152,8 @@ static union {
REGISTER_MODULE_INITIALIZER(debugallocation, {
#if (__cplusplus >= 201103L)
- COMPILE_ASSERT(alignof(debug_malloc_implementation_space) >= alignof(DebugMallocImplementation),
- debug_malloc_implementation_space_is_not_properly_aligned);
+ static_assert(alignof(decltype(debug_malloc_implementation_space)) >= alignof(DebugMallocImplementation),
+ "DebugMallocImplementation is expected to need just word alignment");
#endif
// Either we or valgrind will control memory management. We
// register our extension if we're the winner. Otherwise let
--
1.8.3.1

26
benchmark-use-angle-brackets-to-include-ucontext.h.patch Normal file
View File

@ -0,0 +1,26 @@
From 3af509d4f91aa1fae3f8caeb17b158f20e9b5e59 Mon Sep 17 00:00:00 2001
From: Gabriel Marin <gmx@chromium.org>
Date: Tue, 17 Jul 2018 18:28:01 -0700
Subject: [PATCH 13/39] benchmark: use angle brackets to include ucontext.h
Using quotes for a system header file fails a presubmit check in Chromium.
---
benchmark/unwind_bench.cc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/benchmark/unwind_bench.cc b/benchmark/unwind_bench.cc
index f919fc0..9c4e2f8 100644
--- a/benchmark/unwind_bench.cc
+++ b/benchmark/unwind_bench.cc
@@ -7,7 +7,7 @@
#include <stdlib.h>
#include <stdio.h>
#include <stdint.h>
-#include "sys/ucontext.h"
+#include <sys/ucontext.h>
#if HAVE_LIBUNWIND_H
#include <libunwind.h>
#endif
--
1.8.3.1

30
don-t-try-to-mark-rsp-as-clobbered-in-linux-syscall-.patch Normal file
View File

@ -0,0 +1,30 @@
From 9e5b1628737c67b4587f937164572774592978c4 Mon Sep 17 00:00:00 2001
From: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Date: Sun, 10 Feb 2019 12:46:18 -0800
Subject: [PATCH 39/39] don't try to mark rsp as clobbered in linux syscall
support
rsp is not actually clobbered by that code and later gccs actually (correctly) bark
at it.
Fixed issue #1076.
---
src/base/linux_syscall_support.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/base/linux_syscall_support.h b/src/base/linux_syscall_support.h
index 13aa415..199061a 100644
--- a/src/base/linux_syscall_support.h
+++ b/src/base/linux_syscall_support.h
@@ -1485,7 +1485,7 @@ struct kernel_stat {
"d"(LSS_SYSCALL_ARG(parent_tidptr)),
"r"(LSS_SYSCALL_ARG(newtls)),
"r"(LSS_SYSCALL_ARG(child_tidptr))
- : "rsp", "memory", "r8", "r10", "r11", "rcx");
+ : "memory", "r8", "r10", "r11", "rcx");
}
LSS_RETURN(int, __res);
}
--
1.8.3.1

32
gperftools-2.6.1-disable-generic-dynamic-tls.patch Normal file
View File

@ -0,0 +1,32 @@
diff -up ./configure.ac.dynload ./configure.ac
--- ./configure.ac.dynload 2017-08-24 15:23:01.126622099 -0400
+++ ./configure.ac 2017-08-24 15:29:46.332132222 -0400
@@ -614,6 +614,16 @@ AC_COMPILE_IFELSE(
AC_MSG_RESULT([yes])],
[AC_MSG_RESULT([no])])
+default_enable_generic_dynamic_tls=yes
+AC_ARG_ENABLE([general-dynamic-tls],
+ [AS_HELP_STRING([--disable-general-dynamic-tls],
+ [Do not use the general dynamic TLS model])],
+ [],
+ [enable_generic_dynamic_tls="$default_enable_generic_dynamic_tls"])
+AS_IF([test "x$enable_generic_dynamic_tls" = xyes],
+ [AC_DEFINE([ENABLE_GENERIC_DYNAMIC_TLS], 1,
+ [Use the generic dynamic TLS model])])
+
# Write generated configuration file
AC_CONFIG_FILES([Makefile
src/gperftools/tcmalloc.h src/windows/gperftools/tcmalloc.h])
diff -up ./src/base/basictypes.h.dynload ./src/base/basictypes.h
--- ./src/base/basictypes.h.dynload 2017-08-24 15:30:30.210996293 -0400
+++ ./src/base/basictypes.h 2017-08-24 15:31:01.889176214 -0400
@@ -200,7 +200,7 @@ struct CompileAssert {
# define ATTRIBUTE_UNUSED
#endif
-#if defined(HAVE___ATTRIBUTE__) && defined(HAVE_TLS)
+#if defined(HAVE___ATTRIBUTE__) && defined(HAVE_TLS) && defined(ENABLE_GENERIC_DYNAMIC_TLS)
#define ATTR_INITIAL_EXEC __attribute__ ((tls_model ("initial-exec")))
#else
#define ATTR_INITIAL_EXEC

BIN
gperftools-2.7.tar.gz Normal file
View File

Binary file not shown.

131
gperftools.spec Normal file
View File

@ -0,0 +1,131 @@
Name: gperftools
Version: 2.7
Release: 5
Summary: high-performance malloc and performance analysis tools
License: BSD
URL: https://github.com/gperftools/gperftools
Source0: https://github.com/gperftools/gperftools/releases/download/%{name}-%{version}/%{name}-%{version}.tar.gz
Patch1: gperftools-2.6.1-disable-generic-dynamic-tls.patch
Patch2: Disable-large-allocation-report-by-default.patch
Patch3: Fix-uninitialized-memory-use-in-sampler_test.patch
Patch4: benchmark-use-angle-brackets-to-include-ucontext.h.patch
Patch5: Format-and-fix-out-of-bound-access-in-CpuProfilerSwi.patch
Patch6: Fix-incompatible-aliasing-warnings.patch
Patch7: Use-standard-conforming-alignof-in-debugallocation.c.patch
Patch8: Fix-page_heap_test-flakiness.patch
Patch9: Avoid-static-initialization-of-pprof-path-for-symbol.patch
Patch10: Fix-mmap-region-iteration-while-no-regions-are-recor.patch
Patch11: Drop-not-very-portable-and-not-very-useful-unwind-be.patch
Patch12: undef-mmap64-function.patch
Patch13: Use-initial-exec-TLS-model-for-all-thread-local-vari.patch
Patch14: don-t-try-to-mark-rsp-as-clobbered-in-linux-syscall-.patch
Patch9000: issue-1122-fix-bus-error-on-aarch64.patch
BuildRequires: autoconf automake gcc-c++ libtool libunwind-devel perl-generators
Requires: %{name}-devel = %{version}-%{release}
Requires: pprof = %{version}-%{release}
ExcludeArch: s390
%description
gperftools is a collection of a high-performance multi-threaded \
malloc() implementation, plus some pretty nifty performance analysis \
tools.
%package libs
Summary: Libraries for CPU profiler and tcmalloc
Provides: google-perftools-libs = %{version}-%{release}
Obsoletes: google-perftools-libs < 2.0
%description libs
This package contains libraries with CPU or heap profiling and thread-caching
malloc().
%package devel
Summary: Development header files
Requires: pkg-config
Requires: %{name}-libs = %{version}-%{release}
Provides: google-perftools-devel = %{version}-%{release}
Provides: pkg-config(gperftools-devel) = %{version}-%{release}
Obsoletes: google-perftools-devel < 2.0
%description devel
This package contains Development header files with CPU or heap profiling and
thread-caching malloc() which provides by google and gperftools.
%package -n pprof
Summary: Analyzes CPU and heap profiles tools
Requires: gv graphviz perl
BuildArch: noarch
Provides: pprof = %{version}-%{release}
Provides: google-perftools = %{version}-%{release}
Obsoletes: google-perftools < 2.0
%description -n pprof
pprof is an facility that analyzes CPU and heap profiles.
%prep
%autosetup -n %{name}-%{version} -p1
%build
autoreconf -vif
CFLAGS=`echo $RPM_OPT_FLAGS -fno-strict-aliasing -Wno-unused-local-typedefs -DTCMALLOC_LARGE_PAGES | sed -e 's|-fexceptions||g'`
CXXFLAGS=`echo $RPM_OPT_FLAGS -fno-strict-aliasing -Wno-unused-local-typedefs -DTCMALLOC_LARGE_PAGES | sed -e 's|-fexceptions||g'`
%configure \
%ifarch s390x aarch64
--disable-general-dynamic-tls \
%endif
--disable-dynamic-sized-delete-support \
%disable_rpath
make
%install
%make_install
%ldconfig_scriptlets libs
%files
#nothing to do
%files libs
%{_libdir}/libprofiler.so.*
%{_libdir}/libtcmalloc*.so.*
%files devel
%{_includedir}/google/*.h
%{_includedir}/gperftools/*.h
%{_libdir}/libprofiler.so
%{_libdir}/libtcmalloc*.so
%{_libdir}/pkgconfig/*.pc
%{_docdir}/%{name}/*
%{_libdir}/libprofiler.*a
%{_libdir}/libtcmalloc*.*a
%files -n pprof
%{_bindir}/*
%{_mandir}/man1/*.1.gz
%changelog
* Thu Aug 29 2019 luochunsheng <luochunsheng@huawei.com> - 2.7-5
- fix spelling errors
* Mon Jun 17 2019 Zhipeng Xie <xiezhipeng1@huawei.com> - 2.7-4
- Type:bugfix
- ID:NA
- SUG: NA
- DESC: issue-1122: fix bus error on aarch64
* Mon Apr 15 2019 lvying <lvying6@huawei.com> - 2.7-3
- Type:enhancement
- ID:NA
- SUG: NA
- DESC: peripheral package quality reinforcement: backport bugfix patches from community
* Tue Mar 5 2019 openEuler Buildteam <buildteam@openeuler.org> - 2.7-2
- Package init

42
issue-1122-fix-bus-error-on-aarch64.patch Normal file
View File

@ -0,0 +1,42 @@
From 42dbe9e2ced74ae7f0f3b70daabf634edf44d3d3 Mon Sep 17 00:00:00 2001
From: Zhipeng Xie <xiezhipeng1@huawei.com>
Date: Thu, 13 Jun 2019 23:16:43 +0800
Subject: [PATCH] issue-1122: fix bus error on aarch64
According to the ARMv8 Instruction Set Overview, among other documents,
"... if SP is used as the base register then the value of the stack
pointer prior to adding any offset must be quadword (16 byte) aligned,
or else a stack alignment exception will be generated.". So, we pass
the child_stack 16 byte aligned to sys_clone.
Signed-off-by: Zhipeng Xie <xiezhipeng1@huawei.com>
---
src/base/linuxthreads.cc | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/src/base/linuxthreads.cc b/src/base/linuxthreads.cc
index 891e70c..00b1ea3 100644
--- a/src/base/linuxthreads.cc
+++ b/src/base/linuxthreads.cc
@@ -105,6 +105,7 @@ static int local_clone (int (*fn)(void *), void *arg, ...)
#define CLONE_STACK_SIZE 4096
#endif
+#define ALIGN(x, a) (((x) + (a) - 1) & ~((a) - 1))
static int local_clone (int (*fn)(void *), void *arg, ...) {
/* Leave 4kB of gap between the callers stack and the new clone. This
* should be more than sufficient for the caller to call waitpid() until
@@ -120,7 +121,9 @@ static int local_clone (int (*fn)(void *), void *arg, ...) {
* is being debugged. This is OK and the error code will be reported
* correctly.
*/
- return sys_clone(fn, (char *)&arg - CLONE_STACK_SIZE,
+ void *child_stack = (void *)ALIGN(((unsigned long)&arg - CLONE_STACK_SIZE),
+ 16);
+ return sys_clone(fn, child_stack,
CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_UNTRACED, arg, 0, 0, 0);
}
--
1.7.12.4

44
undef-mmap64-function.patch Normal file
View File

@ -0,0 +1,44 @@
From e9ab4c53041ac62feefbbb076d326e9a77dd1567 Mon Sep 17 00:00:00 2001
From: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Date: Sun, 7 Oct 2018 18:12:03 +0200
Subject: [PATCH 34/39] undef mmap64 function
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
mmap64 can already been defined and as a result the following error is
raised:
In file included from src/malloc_hook.cc:686:0:
src/malloc_hook_mmap_linux.h: In function ‘void* mmap(void*, size_t, int, int, int, off_t)’:
src/malloc_hook_mmap_linux.h:173:18: error: redefinition of ‘void* mmap(void*, size_t, int, int, int, off_t)’
extern "C" void* mmap(void *start, size_t length, int prot, int flags,
^~~~
In file included from src/malloc_hook.cc:41:0:
src/malloc_hook_mmap_linux.h:159:18: note: ‘void* mmap(void*, size_t, int, int, int, off_t)’ previously defined here
extern "C" void* mmap64(void *start, size_t length, int prot, int flags,
^
Makefile:4874: recipe for target 'src/libtcmalloc_minimal_internal_la-malloc_hook.lo' failed
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
---
src/malloc_hook_mmap_linux.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/malloc_hook_mmap_linux.h b/src/malloc_hook_mmap_linux.h
index 0eec895..34de715 100644
--- a/src/malloc_hook_mmap_linux.h
+++ b/src/malloc_hook_mmap_linux.h
@@ -137,7 +137,8 @@ static inline void* do_mmap64(void *start, size_t length,
// malloc_hook section,
// so that MallocHook::GetCallerStackTrace can function accurately:
-// Make sure mmap doesn't get #define'd away by <sys/mman.h>
+// Make sure mmap64 and mmap doesn't get #define'd away by <sys/mman.h>
+# undef mmap64
# undef mmap
extern "C" {
--
1.8.3.1