8141491: Unaligned memory access in Bits.c

Introduce alignment-safe Copy::conjoint_swap and j.i.m.Unsafe.copySwapMemory

Reviewed-by: jrose, dholmes, psandoz
This commit is contained in:
Mikael Vidstedt 2016-02-10 15:20:38 -08:00
parent 3fc0fa269a
commit 4e53f48812
4 changed files with 230 additions and 4 deletions

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2000, 2016, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
@ -660,6 +660,36 @@ UNSAFE_ENTRY(void, Unsafe_CopyMemory(JNIEnv *env, jobject unsafe, jobject srcObj
Copy::conjoint_memory_atomic(src, dst, sz);
UNSAFE_END
// This function is a leaf since if the source and destination are both in native memory
// the copy may potentially be very large, and we don't want to disable GC if we can avoid it.
// If either source or destination (or both) are on the heap, the function will enter VM using
// JVM_ENTRY_FROM_LEAF
JVM_LEAF(void, Unsafe_CopySwapMemory0(JNIEnv *env, jobject unsafe, jobject srcObj, jlong srcOffset, jobject dstObj, jlong dstOffset, jlong size, jlong elemSize)) {
UnsafeWrapper("Unsafe_CopySwapMemory0");
size_t sz = (size_t)size;
size_t esz = (size_t)elemSize;
if (srcObj == NULL && dstObj == NULL) {
// Both src & dst are in native memory
address src = (address)srcOffset;
address dst = (address)dstOffset;
Copy::conjoint_swap(src, dst, sz, esz);
} else {
// At least one of src/dst are on heap, transition to VM to access raw pointers
JVM_ENTRY_FROM_LEAF(env, void, Unsafe_CopySwapMemory0) {
oop srcp = JNIHandles::resolve(srcObj);
oop dstp = JNIHandles::resolve(dstObj);
address src = (address)index_oop_from_field_offset_long(srcp, srcOffset);
address dst = (address)index_oop_from_field_offset_long(dstp, dstOffset);
Copy::conjoint_swap(src, dst, sz, esz);
} JVM_END
}
} JVM_END
////// Random queries
@ -1363,6 +1393,7 @@ static JNINativeMethod jdk_internal_misc_Unsafe_methods[] = {
{CC "getLoadAverage", CC "([DI)I", FN_PTR(Unsafe_Loadavg)},
{CC "copyMemory", CC "(" OBJ "J" OBJ "JJ)V", FN_PTR(Unsafe_CopyMemory)},
{CC "copySwapMemory0", CC "(" OBJ "J" OBJ "JJJ)V", FN_PTR(Unsafe_CopySwapMemory0)},
{CC "setMemory", CC "(" OBJ "JJB)V", FN_PTR(Unsafe_SetMemory)},
{CC "defineAnonymousClass", CC "(" DAC_Args ")" CLS, FN_PTR(Unsafe_DefineAnonymousClass)},

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
@ -417,6 +417,14 @@ class RuntimeHistogramElement : public HistogramElement {
os::verify_stack_alignment(); \
/* begin of body */
#define VM_ENTRY_BASE_FROM_LEAF(result_type, header, thread) \
TRACE_CALL(result_type, header) \
debug_only(ResetNoHandleMark __rnhm;) \
HandleMarkCleaner __hm(thread); \
Thread* THREAD = thread; \
os::verify_stack_alignment(); \
/* begin of body */
// ENTRY routines may lock, GC and throw exceptions
@ -584,6 +592,14 @@ extern "C" { \
VM_LEAF_BASE(result_type, header)
#define JVM_ENTRY_FROM_LEAF(env, result_type, header) \
{ { \
JavaThread* thread=JavaThread::thread_from_jni_environment(env); \
ThreadInVMfromNative __tiv(thread); \
debug_only(VMNativeEntryWrapper __vew;) \
VM_ENTRY_BASE_FROM_LEAF(result_type, header, thread)
#define JVM_END } }
#endif // SHARE_VM_RUNTIME_INTERFACESUPPORT_HPP

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2006, 2016, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
@ -53,6 +53,175 @@ void Copy::conjoint_memory_atomic(void* from, void* to, size_t size) {
}
}
class CopySwap : AllStatic {
public:
/**
* Copy and byte swap elements
*
* @param src address of source
* @param dst address of destination
* @param byte_count number of bytes to copy
* @param elem_size size of the elements to copy-swap
*/
static void conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
assert(src != NULL, "address must not be NULL");
assert(dst != NULL, "address must not be NULL");
assert(elem_size == 2 || elem_size == 4 || elem_size == 8,
"incorrect element size: " SIZE_FORMAT, elem_size);
assert(is_size_aligned(byte_count, elem_size),
"byte_count " SIZE_FORMAT " must be multiple of element size " SIZE_FORMAT, byte_count, elem_size);
address src_end = src + byte_count;
if (dst <= src || dst >= src_end) {
do_conjoint_swap<RIGHT>(src, dst, byte_count, elem_size);
} else {
do_conjoint_swap<LEFT>(src, dst, byte_count, elem_size);
}
}
private:
/**
* Byte swap a 16-bit value
*/
static uint16_t byte_swap(uint16_t x) {
return (x << 8) | (x >> 8);
}
/**
* Byte swap a 32-bit value
*/
static uint32_t byte_swap(uint32_t x) {
uint16_t lo = (uint16_t)x;
uint16_t hi = (uint16_t)(x >> 16);
return ((uint32_t)byte_swap(lo) << 16) | (uint32_t)byte_swap(hi);
}
/**
* Byte swap a 64-bit value
*/
static uint64_t byte_swap(uint64_t x) {
uint32_t lo = (uint32_t)x;
uint32_t hi = (uint32_t)(x >> 32);
return ((uint64_t)byte_swap(lo) << 32) | (uint64_t)byte_swap(hi);
}
enum CopyDirection {
RIGHT, // lower -> higher address
LEFT // higher -> lower address
};
/**
* Copy and byte swap elements
*
* <T> - type of element to copy
* <D> - copy direction
* <is_src_aligned> - true if src argument is aligned to element size
* <is_dst_aligned> - true if dst argument is aligned to element size
*
* @param src address of source
* @param dst address of destination
* @param byte_count number of bytes to copy
*/
template <typename T, CopyDirection D, bool is_src_aligned, bool is_dst_aligned>
static void do_conjoint_swap(address src, address dst, size_t byte_count) {
address cur_src, cur_dst;
switch (D) {
case RIGHT:
cur_src = src;
cur_dst = dst;
break;
case LEFT:
cur_src = src + byte_count - sizeof(T);
cur_dst = dst + byte_count - sizeof(T);
break;
}
for (size_t i = 0; i < byte_count / sizeof(T); i++) {
T tmp;
if (is_src_aligned) {
tmp = *(T*)cur_src;
} else {
memcpy(&tmp, cur_src, sizeof(T));
}
tmp = byte_swap(tmp);
if (is_dst_aligned) {
*(T*)cur_dst = tmp;
} else {
memcpy(cur_dst, &tmp, sizeof(T));
}
switch (D) {
case RIGHT:
cur_src += sizeof(T);
cur_dst += sizeof(T);
break;
case LEFT:
cur_src -= sizeof(T);
cur_dst -= sizeof(T);
break;
}
}
}
/**
* Copy and byte swap elements
*
* <T> - type of element to copy
* <D> - copy direction
*
* @param src address of source
* @param dst address of destination
* @param byte_count number of bytes to copy
*/
template <typename T, CopyDirection direction>
static void do_conjoint_swap(address src, address dst, size_t byte_count) {
if (is_ptr_aligned(src, sizeof(T))) {
if (is_ptr_aligned(dst, sizeof(T))) {
do_conjoint_swap<T,direction,true,true>(src, dst, byte_count);
} else {
do_conjoint_swap<T,direction,true,false>(src, dst, byte_count);
}
} else {
if (is_ptr_aligned(dst, sizeof(T))) {
do_conjoint_swap<T,direction,false,true>(src, dst, byte_count);
} else {
do_conjoint_swap<T,direction,false,false>(src, dst, byte_count);
}
}
}
/**
* Copy and byte swap elements
*
* <D> - copy direction
*
* @param src address of source
* @param dst address of destination
* @param byte_count number of bytes to copy
* @param elem_size size of the elements to copy-swap
*/
template <CopyDirection D>
static void do_conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
switch (elem_size) {
case 2: do_conjoint_swap<uint16_t,D>(src, dst, byte_count); break;
case 4: do_conjoint_swap<uint32_t,D>(src, dst, byte_count); break;
case 8: do_conjoint_swap<uint64_t,D>(src, dst, byte_count); break;
default: guarantee(false, "do_conjoint_swap: Invalid elem_size %zd\n", elem_size);
}
}
};
void Copy::conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
CopySwap::conjoint_swap(src, dst, byte_count, elem_size);
}
// Fill bytes; larger units are filled atomically if everything is aligned.
void Copy::fill_to_memory_atomic(void* to, size_t size, jubyte value) {

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2003, 2011, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2003, 2016, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
@ -227,6 +227,16 @@ class Copy : AllStatic {
}
}
/**
* Copy and *unconditionally* byte swap elements
*
* @param src address of source
* @param dst address of destination
* @param byte_count number of bytes to copy
* @param elem_size size of the elements to copy-swap
*/
static void conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size);
// Fill methods
// Fill word-aligned words, not atomic on each word