gecko-dev/xpcom/base/CountingAllocatorBase.h
Nicholas Nethercote c2306345d5 Bug 1297658 - Avoid unnecessary checking in memory reporters. r=erahm.
This patch removes checking of all the callback calls in memory reporter
CollectReport() functions, because it's not useful.

The patch also does some associated clean-up.

- Replaces some uses of nsIMemoryReporterCallback with the preferred
  nsIHandleReportCallback typedef.

- Replaces aCallback/aCb/aClosure with aHandleRepor/aData for CollectReports()
  parameter names, for consistency.

- Adds MOZ_MUST_USE/[must_use] in a few places in nsIMemoryReporter.idl.

- Uses the MOZ_COLLECT_REPORT macro in all suitable places.

Overall the patch reduces code size by ~300 lines and reduces the size of
libxul by about 37 KiB on my Linux64 builds.

--HG--
extra : rebase_source : e94323614bd10463a0c5134a7276238a7ca1cf23
2016-08-24 15:23:45 +10:00

141 lines
4.2 KiB
C++

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef CountingAllocatorBase_h
#define CountingAllocatorBase_h
#include "mozilla/Assertions.h"
#include "mozilla/Atomics.h"
#include "nsIMemoryReporter.h"
namespace mozilla {
// This CRTP class handles several details of wrapping allocators and should
// be preferred to manually counting with MOZ_DEFINE_MALLOC_SIZE_OF_ON_ALLOC
// and MOZ_DEFINE_MALLOC_SIZE_OF_ON_FREE. The typical use is in a memory
// reporter for a particular third party library:
//
// class MyMemoryReporter : public CountingAllocatorBase<MyMemoryReporter>
// {
// ...
// NS_IMETHOD
// CollectReports(nsIHandleReportCallback* aHandleReport,
// nsISupports* aData, bool aAnonymize) override
// {
// MOZ_COLLECT_REPORT(
// "explicit/path/to/somewhere", KIND_HEAP, UNITS_BYTES,
// MemoryAllocated(),
// "A description of what we are reporting.");
//
// return NS_OK;
// }
// };
//
// ...somewhere later in the code...
// SetThirdPartyMemoryFunctions(MyMemoryReporter::CountingAlloc,
// MyMemoryReporter::CountingFree);
template<typename T>
class CountingAllocatorBase
{
public:
CountingAllocatorBase()
{
#ifdef DEBUG
// There must be only one instance of this class, due to |sAmount| being
// static.
static bool hasRun = false;
MOZ_ASSERT(!hasRun);
hasRun = true;
#endif
}
static size_t
MemoryAllocated()
{
return sAmount;
}
static void*
CountingMalloc(size_t size)
{
void* p = malloc(size);
sAmount += MallocSizeOfOnAlloc(p);
return p;
}
static void*
CountingCalloc(size_t nmemb, size_t size)
{
void* p = calloc(nmemb, size);
sAmount += MallocSizeOfOnAlloc(p);
return p;
}
static void*
CountingRealloc(void* p, size_t size)
{
size_t oldsize = MallocSizeOfOnFree(p);
void *pnew = realloc(p, size);
if (pnew) {
size_t newsize = MallocSizeOfOnAlloc(pnew);
sAmount += newsize - oldsize;
} else if (size == 0) {
// We asked for a 0-sized (re)allocation of some existing pointer
// and received NULL in return. 0-sized allocations are permitted
// to either return NULL or to allocate a unique object per call (!).
// For a malloc implementation that chooses the second strategy,
// that allocation may fail (unlikely, but possible).
//
// Given a NULL return value and an allocation size of 0, then, we
// don't know if that means the original pointer was freed or if
// the allocation of the unique object failed. If the original
// pointer was freed, then we have nothing to do here. If the
// allocation of the unique object failed, the original pointer is
// still valid and we ought to undo the decrement from above.
// However, we have no way of knowing how the underlying realloc
// implementation is behaving. Assuming that the original pointer
// was freed is the safest course of action. We do, however, need
// to note that we freed memory.
sAmount -= oldsize;
} else {
// realloc failed. The amount allocated hasn't changed.
}
return pnew;
}
// Some library code expects that realloc(x, 0) will free x, which is not
// the behavior of the version of jemalloc we're using, so this wrapped
// version of realloc is needed.
static void*
CountingFreeingRealloc(void* p, size_t size)
{
if (size == 0) {
CountingFree(p);
return nullptr;
}
return CountingRealloc(p, size);
}
static void
CountingFree(void* p)
{
sAmount -= MallocSizeOfOnFree(p);
free(p);
}
private:
// |sAmount| can be (implicitly) accessed by multiple threads, so it
// must be thread-safe.
static Atomic<size_t> sAmount;
MOZ_DEFINE_MALLOC_SIZE_OF_ON_ALLOC(MallocSizeOfOnAlloc)
MOZ_DEFINE_MALLOC_SIZE_OF_ON_FREE(MallocSizeOfOnFree)
};
} // namespace mozilla
#endif // CountingAllocatorBase_h