From 2096fa4bf9b288072134879e1426ac31e4c5c47b Mon Sep 17 00:00:00 2001 From: Francis Ricci Date: Wed, 19 Apr 2017 14:00:35 +0000 Subject: [PATCH] Move valid caller-pc checks out of platform-specific checks Summary: ProcessPlatformSpecificAllocations for linux leak sanitizer iterated over memory chunks and ran two checks concurrently: 1) Ensured the pc was valid 2) Checked whether it was a linker allocation All platforms will need the valid pc check, so it is moved out of the platform- specific file. To prevent code and logic duplication, the linker allocation check is moved as well, with the name of the linker supplied by the platform-specific module. In cases where we don't need to check for linker allocations (ie Darwin), this name will be a nullptr, and we'll only run the caller pc checks. Reviewers: kubamracek, alekseyshl, kcc Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D32130 llvm-svn: 300690 --- compiler-rt/lib/lsan/lsan_common.cc | 70 ++++++++++++++++++++++++++++++- compiler-rt/lib/lsan/lsan_common.h | 4 ++ compiler-rt/lib/lsan/lsan_common_linux.cc | 65 +--------------------------- compiler-rt/lib/lsan/lsan_common_mac.cc | 2 + 4 files changed, 77 insertions(+), 64 deletions(-) diff --git a/compiler-rt/lib/lsan/lsan_common.cc b/compiler-rt/lib/lsan/lsan_common.cc index e7eea68..9d65918 100644 --- a/compiler-rt/lib/lsan/lsan_common.cc +++ b/compiler-rt/lib/lsan/lsan_common.cc @@ -356,6 +356,72 @@ static void CollectIgnoredCb(uptr chunk, void *arg) { } } +static uptr GetCallerPC(u32 stack_id, StackDepotReverseMap *map) { + CHECK(stack_id); + StackTrace stack = map->Get(stack_id); + // The top frame is our malloc/calloc/etc. The next frame is the caller. + if (stack.size >= 2) + return stack.trace[1]; + return 0; +} + +struct InvalidPCParam { + Frontier *frontier; + StackDepotReverseMap *stack_depot_reverse_map; + bool skip_linker_allocations; +}; + +// ForEachChunk callback. If the caller pc is invalid or is within the linker, +// mark as reachable. Called by ProcessPlatformSpecificAllocations. +static void MarkInvalidPCCb(uptr chunk, void *arg) { + CHECK(arg); + InvalidPCParam *param = reinterpret_cast(arg); + chunk = GetUserBegin(chunk); + LsanMetadata m(chunk); + if (m.allocated() && m.tag() != kReachable && m.tag() != kIgnored) { + u32 stack_id = m.stack_trace_id(); + uptr caller_pc = 0; + if (stack_id > 0) + caller_pc = GetCallerPC(stack_id, param->stack_depot_reverse_map); + // If caller_pc is unknown, this chunk may be allocated in a coroutine. Mark + // it as reachable, as we can't properly report its allocation stack anyway. + if (caller_pc == 0 || (param->skip_linker_allocations && + GetLinker()->containsAddress(caller_pc))) { + m.set_tag(kReachable); + param->frontier->push_back(chunk); + } + } +} + +// On Linux, handles dynamically allocated TLS blocks by treating all chunks +// allocated from ld-linux.so as reachable. +// Dynamic TLS blocks contain the TLS variables of dynamically loaded modules. +// They are allocated with a __libc_memalign() call in allocate_and_init() +// (elf/dl-tls.c). Glibc won't tell us the address ranges occupied by those +// blocks, but we can make sure they come from our own allocator by intercepting +// __libc_memalign(). On top of that, there is no easy way to reach them. Their +// addresses are stored in a dynamically allocated array (the DTV) which is +// referenced from the static TLS. Unfortunately, we can't just rely on the DTV +// being reachable from the static TLS, and the dynamic TLS being reachable from +// the DTV. This is because the initial DTV is allocated before our interception +// mechanism kicks in, and thus we don't recognize it as allocated memory. We +// can't special-case it either, since we don't know its size. +// Our solution is to include in the root set all allocations made from +// ld-linux.so (which is where allocate_and_init() is implemented). This is +// guaranteed to include all dynamic TLS blocks (and possibly other allocations +// which we don't care about). +// On all other platforms, this simply checks to ensure that the caller pc is +// valid before reporting chunks as leaked. +void ProcessPC(Frontier *frontier) { + StackDepotReverseMap stack_depot_reverse_map; + InvalidPCParam arg; + arg.frontier = frontier; + arg.stack_depot_reverse_map = &stack_depot_reverse_map; + arg.skip_linker_allocations = + flags()->use_tls && flags()->use_ld_allocations && GetLinker() != nullptr; + ForEachChunk(MarkInvalidPCCb, &arg); +} + // Sets the appropriate tag on each chunk. static void ClassifyAllChunks(SuspendedThreadsList const &suspended_threads) { // Holds the flood fill frontier. @@ -367,11 +433,13 @@ static void ClassifyAllChunks(SuspendedThreadsList const &suspended_threads) { ProcessRootRegions(&frontier); FloodFillTag(&frontier, kReachable); + CHECK_EQ(0, frontier.size()); + ProcessPC(&frontier); + // The check here is relatively expensive, so we do this in a separate flood // fill. That way we can skip the check for chunks that are reachable // otherwise. LOG_POINTERS("Processing platform-specific allocations.\n"); - CHECK_EQ(0, frontier.size()); ProcessPlatformSpecificAllocations(&frontier); FloodFillTag(&frontier, kReachable); diff --git a/compiler-rt/lib/lsan/lsan_common.h b/compiler-rt/lib/lsan/lsan_common.h index 0bb575b..88d5f1c 100644 --- a/compiler-rt/lib/lsan/lsan_common.h +++ b/compiler-rt/lib/lsan/lsan_common.h @@ -212,6 +212,10 @@ uptr PointsIntoChunk(void *p); uptr GetUserBegin(uptr chunk); // Helper for __lsan_ignore_object(). IgnoreObjectResult IgnoreObjectLocked(const void *p); + +// Return the linker module, if valid for the platform. +LoadedModule *GetLinker(); + // Wrapper for chunk metadata operations. class LsanMetadata { public: diff --git a/compiler-rt/lib/lsan/lsan_common_linux.cc b/compiler-rt/lib/lsan/lsan_common_linux.cc index 0d1e998..fadd026 100644 --- a/compiler-rt/lib/lsan/lsan_common_linux.cc +++ b/compiler-rt/lib/lsan/lsan_common_linux.cc @@ -89,70 +89,9 @@ void ProcessGlobalRegions(Frontier *frontier) { dl_iterate_phdr(ProcessGlobalRegionsCallback, frontier); } -static uptr GetCallerPC(u32 stack_id, StackDepotReverseMap *map) { - CHECK(stack_id); - StackTrace stack = map->Get(stack_id); - // The top frame is our malloc/calloc/etc. The next frame is the caller. - if (stack.size >= 2) - return stack.trace[1]; - return 0; -} +LoadedModule *GetLinker() { return linker; } -struct ProcessPlatformAllocParam { - Frontier *frontier; - StackDepotReverseMap *stack_depot_reverse_map; - bool skip_linker_allocations; -}; - -// ForEachChunk callback. Identifies unreachable chunks which must be treated as -// reachable. Marks them as reachable and adds them to the frontier. -static void ProcessPlatformSpecificAllocationsCb(uptr chunk, void *arg) { - CHECK(arg); - ProcessPlatformAllocParam *param = - reinterpret_cast(arg); - chunk = GetUserBegin(chunk); - LsanMetadata m(chunk); - if (m.allocated() && m.tag() != kReachable && m.tag() != kIgnored) { - u32 stack_id = m.stack_trace_id(); - uptr caller_pc = 0; - if (stack_id > 0) - caller_pc = GetCallerPC(stack_id, param->stack_depot_reverse_map); - // If caller_pc is unknown, this chunk may be allocated in a coroutine. Mark - // it as reachable, as we can't properly report its allocation stack anyway. - if (caller_pc == 0 || (param->skip_linker_allocations && - linker->containsAddress(caller_pc))) { - m.set_tag(kReachable); - param->frontier->push_back(chunk); - } - } -} - -// Handles dynamically allocated TLS blocks by treating all chunks allocated -// from ld-linux.so as reachable. -// Dynamic TLS blocks contain the TLS variables of dynamically loaded modules. -// They are allocated with a __libc_memalign() call in allocate_and_init() -// (elf/dl-tls.c). Glibc won't tell us the address ranges occupied by those -// blocks, but we can make sure they come from our own allocator by intercepting -// __libc_memalign(). On top of that, there is no easy way to reach them. Their -// addresses are stored in a dynamically allocated array (the DTV) which is -// referenced from the static TLS. Unfortunately, we can't just rely on the DTV -// being reachable from the static TLS, and the dynamic TLS being reachable from -// the DTV. This is because the initial DTV is allocated before our interception -// mechanism kicks in, and thus we don't recognize it as allocated memory. We -// can't special-case it either, since we don't know its size. -// Our solution is to include in the root set all allocations made from -// ld-linux.so (which is where allocate_and_init() is implemented). This is -// guaranteed to include all dynamic TLS blocks (and possibly other allocations -// which we don't care about). -void ProcessPlatformSpecificAllocations(Frontier *frontier) { - StackDepotReverseMap stack_depot_reverse_map; - ProcessPlatformAllocParam arg; - arg.frontier = frontier; - arg.stack_depot_reverse_map = &stack_depot_reverse_map; - arg.skip_linker_allocations = - flags()->use_tls && flags()->use_ld_allocations && linker != nullptr; - ForEachChunk(ProcessPlatformSpecificAllocationsCb, &arg); -} +void ProcessPlatformSpecificAllocations(Frontier *frontier) {} struct DoStopTheWorldParam { StopTheWorldCallback callback; diff --git a/compiler-rt/lib/lsan/lsan_common_mac.cc b/compiler-rt/lib/lsan/lsan_common_mac.cc index dfb5ab4..20ac389 100644 --- a/compiler-rt/lib/lsan/lsan_common_mac.cc +++ b/compiler-rt/lib/lsan/lsan_common_mac.cc @@ -87,6 +87,8 @@ void SetCurrentThread(u32 tid) { get_tls_val(true)->current_thread_id = tid; } AllocatorCache *GetAllocatorCache() { return &get_tls_val(true)->cache; } +LoadedModule *GetLinker() { return nullptr; } + // Required on Linux for initialization of TLS behavior, but should not be // required on Darwin. void InitializePlatformSpecificModules() { -- 2.7.4