--- /dev/null 2016-08-24 15:41:39.598575000 -0400 +++ new/src/cpu/arm/vm/frame_arm.cpp 2016-12-13 12:49:41.009483736 -0500 @@ -0,0 +1,655 @@ +/* + * Copyright (c) 2008, 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 + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + * + */ + +#include "precompiled.hpp" +#include "interpreter/interpreter.hpp" +#include "memory/resourceArea.hpp" +#include "oops/markOop.hpp" +#include "oops/method.hpp" +#include "oops/oop.inline.hpp" +#include "runtime/frame.inline.hpp" +#include "runtime/handles.inline.hpp" +#include "runtime/javaCalls.hpp" +#include "runtime/monitorChunk.hpp" +#include "runtime/signature.hpp" +#include "runtime/stubCodeGenerator.hpp" +#include "runtime/stubRoutines.hpp" +#include "vmreg_arm.inline.hpp" +#ifdef COMPILER1 +#include "c1/c1_Runtime1.hpp" +#include "runtime/vframeArray.hpp" +#endif +#include "prims/methodHandles.hpp" + +#ifdef ASSERT +void RegisterMap::check_location_valid() { +} +#endif + + +// Profiling/safepoint support + +bool frame::safe_for_sender(JavaThread *thread) { + address sp = (address)_sp; + address fp = (address)_fp; + address unextended_sp = (address)_unextended_sp; + + static size_t stack_guard_size = os::uses_stack_guard_pages() ? + (JavaThread::stack_red_zone_size() + JavaThread::stack_yellow_zone_size()) : 0; + size_t usable_stack_size = thread->stack_size() - stack_guard_size; + + // sp must be within the usable part of the stack (not in guards) + bool sp_safe = (sp != NULL && + (sp <= thread->stack_base()) && + (sp >= thread->stack_base() - usable_stack_size)); + + if (!sp_safe) { + return false; + } + + bool unextended_sp_safe = (unextended_sp != NULL && + (unextended_sp <= thread->stack_base()) && + (unextended_sp >= sp)); + if (!unextended_sp_safe) { + return false; + } + + // We know sp/unextended_sp are safe. Only fp is questionable here. + + bool fp_safe = (fp != NULL && + (fp <= thread->stack_base()) && + fp >= sp); + + if (_cb != NULL ) { + + // First check if frame is complete and tester is reliable + // Unfortunately we can only check frame complete for runtime stubs and nmethod + // other generic buffer blobs are more problematic so we just assume they are + // ok. adapter blobs never have a frame complete and are never ok. + + if (!_cb->is_frame_complete_at(_pc)) { + if (_cb->is_compiled() || _cb->is_adapter_blob() || _cb->is_runtime_stub()) { + return false; + } + } + + // Could just be some random pointer within the codeBlob + if (!_cb->code_contains(_pc)) { + return false; + } + + // Entry frame checks + if (is_entry_frame()) { + // an entry frame must have a valid fp. + return fp_safe && is_entry_frame_valid(thread); + } + + intptr_t* sender_sp = NULL; + address sender_pc = NULL; + + if (is_interpreted_frame()) { + // fp must be safe + if (!fp_safe) { + return false; + } + + sender_pc = (address) this->fp()[return_addr_offset]; + sender_sp = (intptr_t*) addr_at(sender_sp_offset); + + } else { + // must be some sort of compiled/runtime frame + // fp does not have to be safe (although it could be check for c1?) + + sender_sp = _unextended_sp + _cb->frame_size(); + // Is sender_sp safe? + if ((address)sender_sp >= thread->stack_base()) { + return false; + } + // With our calling conventions, the return_address should + // end up being the word on the stack + sender_pc = (address) *(sender_sp - sender_sp_offset + return_addr_offset); + } + + // We must always be able to find a recognizable pc + CodeBlob* sender_blob = CodeCache::find_blob_unsafe(sender_pc); + if (sender_pc == NULL || sender_blob == NULL) { + return false; + } + + + // If the potential sender is the interpreter then we can do some more checking + if (Interpreter::contains(sender_pc)) { + + // FP is always saved in a recognizable place in any code we generate. However + // only if the sender is interpreted/call_stub (c1 too?) are we certain that the saved FP + // is really a frame pointer. + + intptr_t *saved_fp = (intptr_t*)*(sender_sp - frame::sender_sp_offset + link_offset); + bool saved_fp_safe = ((address)saved_fp <= thread->stack_base()) && (saved_fp > sender_sp); + + if (!saved_fp_safe) { + return false; + } + + // construct the potential sender + + frame sender(sender_sp, saved_fp, sender_pc); + + return sender.is_interpreted_frame_valid(thread); + } + + if (sender_blob->is_zombie() || sender_blob->is_unloaded()) { + return false; + } + + // Could just be some random pointer within the codeBlob + if (!sender_blob->code_contains(sender_pc)) { + return false; + } + + // We should never be able to see an adapter if the current frame is something from code cache + if (sender_blob->is_adapter_blob()) { + return false; + } + + // Could be the call_stub + if (StubRoutines::returns_to_call_stub(sender_pc)) { + intptr_t *saved_fp = (intptr_t*)*(sender_sp - frame::sender_sp_offset + link_offset); + bool saved_fp_safe = ((address)saved_fp <= thread->stack_base()) && (saved_fp >= sender_sp); + + if (!saved_fp_safe) { + return false; + } + + // construct the potential sender + + frame sender(sender_sp, saved_fp, sender_pc); + + // Validate the JavaCallWrapper an entry frame must have + address jcw = (address)sender.entry_frame_call_wrapper(); + + bool jcw_safe = (jcw <= thread->stack_base()) && (jcw > (address)sender.fp()); + + return jcw_safe; + } + + // If the frame size is 0 something (or less) is bad because every nmethod has a non-zero frame size + // because the return address counts against the callee's frame. + + if (sender_blob->frame_size() <= 0) { + assert(!sender_blob->is_compiled(), "should count return address at least"); + return false; + } + + // We should never be able to see anything here except an nmethod. If something in the + // code cache (current frame) is called by an entity within the code cache that entity + // should not be anything but the call stub (already covered), the interpreter (already covered) + // or an nmethod. + + if (!sender_blob->is_compiled()) { + return false; + } + + // Could put some more validation for the potential non-interpreted sender + // frame we'd create by calling sender if I could think of any. Wait for next crash in forte... + + // One idea is seeing if the sender_pc we have is one that we'd expect to call to current cb + + // We've validated the potential sender that would be created + return true; + } + + // Must be native-compiled frame. Since sender will try and use fp to find + // linkages it must be safe + + if (!fp_safe) { + return false; + } + + // Will the pc we fetch be non-zero (which we'll find at the oldest frame) + + if ((address) this->fp()[return_addr_offset] == NULL) return false; + + + // could try and do some more potential verification of native frame if we could think of some... + + return true; +} + + +void frame::patch_pc(Thread* thread, address pc) { + address* pc_addr = &((address *)sp())[-sender_sp_offset+return_addr_offset]; + if (TracePcPatching) { + tty->print_cr("patch_pc at address" INTPTR_FORMAT " [" INTPTR_FORMAT " -> " INTPTR_FORMAT "] ", + p2i(pc_addr), p2i(*pc_addr), p2i(pc)); + } + *pc_addr = pc; + _cb = CodeCache::find_blob(pc); + address original_pc = CompiledMethod::get_deopt_original_pc(this); + if (original_pc != NULL) { + assert(original_pc == _pc, "expected original PC to be stored before patching"); + _deopt_state = is_deoptimized; + // leave _pc as is + } else { + _deopt_state = not_deoptimized; + _pc = pc; + } +} + +bool frame::is_interpreted_frame() const { + return Interpreter::contains(pc()); +} + +int frame::frame_size(RegisterMap* map) const { + frame sender = this->sender(map); + return sender.sp() - sp(); +} + +intptr_t* frame::entry_frame_argument_at(int offset) const { + assert(is_entry_frame(), "entry frame expected"); + // convert offset to index to deal with tsi + int index = (Interpreter::expr_offset_in_bytes(offset)/wordSize); + // Entry frame's arguments are always in relation to unextended_sp() + return &unextended_sp()[index]; +} + +// sender_sp +intptr_t* frame::interpreter_frame_sender_sp() const { + assert(is_interpreted_frame(), "interpreted frame expected"); + return (intptr_t*) at(interpreter_frame_sender_sp_offset); +} + +void frame::set_interpreter_frame_sender_sp(intptr_t* sender_sp) { + assert(is_interpreted_frame(), "interpreted frame expected"); + ptr_at_put(interpreter_frame_sender_sp_offset, (intptr_t) sender_sp); +} + + +// monitor elements + +BasicObjectLock* frame::interpreter_frame_monitor_begin() const { + return (BasicObjectLock*) addr_at(interpreter_frame_monitor_block_bottom_offset); +} + +BasicObjectLock* frame::interpreter_frame_monitor_end() const { + BasicObjectLock* result = (BasicObjectLock*) *addr_at(interpreter_frame_monitor_block_top_offset); + // make sure the pointer points inside the frame + assert((intptr_t) fp() > (intptr_t) result, "result must < than frame pointer"); + assert((intptr_t) sp() <= (intptr_t) result, "result must >= than stack pointer"); + return result; +} + +void frame::interpreter_frame_set_monitor_end(BasicObjectLock* value) { + *((BasicObjectLock**)addr_at(interpreter_frame_monitor_block_top_offset)) = value; +} + +#ifdef AARCH64 + +// Used by template based interpreter deoptimization +void frame::interpreter_frame_set_stack_top(intptr_t* stack_top) { + *((intptr_t**)addr_at(interpreter_frame_stack_top_offset)) = stack_top; +} + +// Used by template based interpreter deoptimization +void frame::interpreter_frame_set_extended_sp(intptr_t* sp) { + *((intptr_t**)addr_at(interpreter_frame_extended_sp_offset)) = sp; +} + +#else + +// Used by template based interpreter deoptimization +void frame::interpreter_frame_set_last_sp(intptr_t* sp) { + *((intptr_t**)addr_at(interpreter_frame_last_sp_offset)) = sp; +} + +#endif // AARCH64 + +frame frame::sender_for_entry_frame(RegisterMap* map) const { + assert(map != NULL, "map must be set"); + // Java frame called from C; skip all C frames and return top C + // frame of that chunk as the sender + JavaFrameAnchor* jfa = entry_frame_call_wrapper()->anchor(); + assert(!entry_frame_is_first(), "next Java fp must be non zero"); + assert(jfa->last_Java_sp() > sp(), "must be above this frame on stack"); + map->clear(); + assert(map->include_argument_oops(), "should be set by clear"); +#ifdef AARCH64 + assert (jfa->last_Java_pc() != NULL, "pc should be stored"); + frame fr(jfa->last_Java_sp(), jfa->last_Java_fp(), jfa->last_Java_pc()); + return fr; +#else + if (jfa->last_Java_pc() != NULL) { + frame fr(jfa->last_Java_sp(), jfa->last_Java_fp(), jfa->last_Java_pc()); + return fr; + } + frame fr(jfa->last_Java_sp(), jfa->last_Java_fp()); + return fr; +#endif // AARCH64 +} + +//------------------------------------------------------------------------------ +// frame::verify_deopt_original_pc +// +// Verifies the calculated original PC of a deoptimization PC for the +// given unextended SP. The unextended SP might also be the saved SP +// for MethodHandle call sites. +#ifdef ASSERT +void frame::verify_deopt_original_pc(CompiledMethod* nm, intptr_t* unextended_sp, bool is_method_handle_return) { + frame fr; + + // This is ugly but it's better than to change {get,set}_original_pc + // to take an SP value as argument. And it's only a debugging + // method anyway. + fr._unextended_sp = unextended_sp; + + address original_pc = nm->get_original_pc(&fr); + assert(nm->insts_contains(original_pc), "original PC must be in nmethod"); + assert(nm->is_method_handle_return(original_pc) == is_method_handle_return, "must be"); +} +#endif + +//------------------------------------------------------------------------------ +// frame::adjust_unextended_sp +void frame::adjust_unextended_sp() { + // same as on x86 + + // If we are returning to a compiled MethodHandle call site, the + // saved_fp will in fact be a saved value of the unextended SP. The + // simplest way to tell whether we are returning to such a call site + // is as follows: + + CompiledMethod* sender_cm = (_cb == NULL) ? NULL : _cb->as_compiled_method_or_null(); + if (sender_cm != NULL) { + // If the sender PC is a deoptimization point, get the original + // PC. For MethodHandle call site the unextended_sp is stored in + // saved_fp. + if (sender_cm->is_deopt_mh_entry(_pc)) { + DEBUG_ONLY(verify_deopt_mh_original_pc(sender_cm, _fp)); + _unextended_sp = _fp; + } + else if (sender_cm->is_deopt_entry(_pc)) { + DEBUG_ONLY(verify_deopt_original_pc(sender_cm, _unextended_sp)); + } + else if (sender_cm->is_method_handle_return(_pc)) { + _unextended_sp = _fp; + } + } +} + +//------------------------------------------------------------------------------ +// frame::update_map_with_saved_link +void frame::update_map_with_saved_link(RegisterMap* map, intptr_t** link_addr) { + // see x86 for comments + map->set_location(FP->as_VMReg(), (address) link_addr); +#ifdef AARCH64 + // also adjust a high part of register + map->set_location(FP->as_VMReg()->next(), (address) link_addr); +#endif // AARCH64 +} + +frame frame::sender_for_interpreter_frame(RegisterMap* map) const { + // SP is the raw SP from the sender after adapter or interpreter + // extension. + intptr_t* sender_sp = this->sender_sp(); + + // This is the sp before any possible extension (adapter/locals). + intptr_t* unextended_sp = interpreter_frame_sender_sp(); + +#ifdef COMPILER2 + if (map->update_map()) { + update_map_with_saved_link(map, (intptr_t**) addr_at(link_offset)); + } +#endif // COMPILER2 + + return frame(sender_sp, unextended_sp, link(), sender_pc()); +} + +frame frame::sender_for_compiled_frame(RegisterMap* map) const { + assert(map != NULL, "map must be set"); + + // frame owned by optimizing compiler + assert(_cb->frame_size() >= 0, "must have non-zero frame size"); + intptr_t* sender_sp = unextended_sp() + _cb->frame_size(); + intptr_t* unextended_sp = sender_sp; + + address sender_pc = (address) *(sender_sp - sender_sp_offset + return_addr_offset); + + // This is the saved value of FP which may or may not really be an FP. + // It is only an FP if the sender is an interpreter frame (or C1?). + intptr_t** saved_fp_addr = (intptr_t**) (sender_sp - sender_sp_offset + link_offset); + + if (map->update_map()) { + // Tell GC to use argument oopmaps for some runtime stubs that need it. + // For C1, the runtime stub might not have oop maps, so set this flag + // outside of update_register_map. + map->set_include_argument_oops(_cb->caller_must_gc_arguments(map->thread())); + if (_cb->oop_maps() != NULL) { + OopMapSet::update_register_map(this, map); + } + + // Since the prolog does the save and restore of FP there is no oopmap + // for it so we must fill in its location as if there was an oopmap entry + // since if our caller was compiled code there could be live jvm state in it. + update_map_with_saved_link(map, saved_fp_addr); + } + + assert(sender_sp != sp(), "must have changed"); + return frame(sender_sp, unextended_sp, *saved_fp_addr, sender_pc); +} + +frame frame::sender(RegisterMap* map) const { + // Default is we done have to follow them. The sender_for_xxx will + // update it accordingly + map->set_include_argument_oops(false); + + if (is_entry_frame()) return sender_for_entry_frame(map); + if (is_interpreted_frame()) return sender_for_interpreter_frame(map); + assert(_cb == CodeCache::find_blob(pc()),"Must be the same"); + + if (_cb != NULL) { + return sender_for_compiled_frame(map); + } + + assert(false, "should not be called for a C frame"); + return frame(); +} + +bool frame::is_interpreted_frame_valid(JavaThread* thread) const { + assert(is_interpreted_frame(), "Not an interpreted frame"); + // These are reasonable sanity checks + if (fp() == 0 || (intptr_t(fp()) & (wordSize-1)) != 0) { + return false; + } + if (sp() == 0 || (intptr_t(sp()) & (wordSize-1)) != 0) { + return false; + } + if (fp() + interpreter_frame_initial_sp_offset < sp()) { + return false; + } + // These are hacks to keep us out of trouble. + // The problem with these is that they mask other problems + if (fp() <= sp()) { // this attempts to deal with unsigned comparison above + return false; + } + // do some validation of frame elements + + // first the method + + Method* m = *interpreter_frame_method_addr(); + + // validate the method we'd find in this potential sender + if (!m->is_valid_method()) return false; + + // stack frames shouldn't be much larger than max_stack elements + + if (fp() - sp() > 1024 + m->max_stack()*Interpreter::stackElementSize) { + return false; + } + + // validate bci/bcp + + address bcp = interpreter_frame_bcp(); + if (m->validate_bci_from_bcp(bcp) < 0) { + return false; + } + + // validate ConstantPoolCache* + ConstantPoolCache* cp = *interpreter_frame_cache_addr(); + if (cp == NULL || !cp->is_metaspace_object()) return false; + + // validate locals + + address locals = (address) *interpreter_frame_locals_addr(); + + if (locals > thread->stack_base() || locals < (address) fp()) return false; + + // We'd have to be pretty unlucky to be mislead at this point + + return true; +} + +BasicType frame::interpreter_frame_result(oop* oop_result, jvalue* value_result) { + assert(is_interpreted_frame(), "interpreted frame expected"); + Method* method = interpreter_frame_method(); + BasicType type = method->result_type(); + + intptr_t* res_addr; + if (method->is_native()) { + // Prior to calling into the runtime to report the method_exit both of + // the possible return value registers are saved. +#ifdef AARCH64 + // Return value registers are saved into the frame + if (type == T_FLOAT || type == T_DOUBLE) { + res_addr = addr_at(interpreter_frame_fp_saved_result_offset); + } else { + res_addr = addr_at(interpreter_frame_gp_saved_result_offset); + } +#else + // Return value registers are pushed to the native stack + res_addr = (intptr_t*)sp(); +#ifdef __ABI_HARD__ + // FP result is pushed onto a stack along with integer result registers + if (type == T_FLOAT || type == T_DOUBLE) { + res_addr += 2; + } +#endif // __ABI_HARD__ +#endif // AARCH64 + } else { + res_addr = (intptr_t*)interpreter_frame_tos_address(); + } + + switch (type) { + case T_OBJECT : + case T_ARRAY : { + oop obj; + if (method->is_native()) { + obj = cast_to_oop(at(interpreter_frame_oop_temp_offset)); + } else { + obj = *(oop*)res_addr; + } + assert(obj == NULL || Universe::heap()->is_in(obj), "sanity check"); + *oop_result = obj; + break; + } + case T_BOOLEAN : value_result->z = *(jboolean*)res_addr; break; + case T_BYTE : value_result->b = *(jbyte*)res_addr; break; + case T_CHAR : value_result->c = *(jchar*)res_addr; break; + case T_SHORT : value_result->s = *(jshort*)res_addr; break; + case T_INT : value_result->i = *(jint*)res_addr; break; + case T_LONG : value_result->j = *(jlong*)res_addr; break; + case T_FLOAT : value_result->f = *(jfloat*)res_addr; break; + case T_DOUBLE : value_result->d = *(jdouble*)res_addr; break; + case T_VOID : /* Nothing to do */ break; + default : ShouldNotReachHere(); + } + + return type; +} + + +intptr_t* frame::interpreter_frame_tos_at(jint offset) const { + int index = (Interpreter::expr_offset_in_bytes(offset)/wordSize); + return &interpreter_frame_tos_address()[index]; +} + +#ifndef PRODUCT + +#define DESCRIBE_FP_OFFSET(name) \ + values.describe(frame_no, fp() + frame::name##_offset, #name) + +void frame::describe_pd(FrameValues& values, int frame_no) { + if (is_interpreted_frame()) { + DESCRIBE_FP_OFFSET(interpreter_frame_sender_sp); +#ifdef AARCH64 + DESCRIBE_FP_OFFSET(interpreter_frame_stack_top); + DESCRIBE_FP_OFFSET(interpreter_frame_extended_sp); +#else + DESCRIBE_FP_OFFSET(interpreter_frame_last_sp); +#endif // AARCH64 + DESCRIBE_FP_OFFSET(interpreter_frame_method); + DESCRIBE_FP_OFFSET(interpreter_frame_mdp); + DESCRIBE_FP_OFFSET(interpreter_frame_cache); + DESCRIBE_FP_OFFSET(interpreter_frame_locals); + DESCRIBE_FP_OFFSET(interpreter_frame_bcp); + DESCRIBE_FP_OFFSET(interpreter_frame_initial_sp); + } +} + +// This is a generic constructor which is only used by pns() in debug.cpp. +frame::frame(void* sp, void* fp, void* pc) { + init((intptr_t*)sp, (intptr_t*)fp, (address)pc); +} +#endif + +intptr_t *frame::initial_deoptimization_info() { + // used to reset the saved FP + return fp(); +} + +intptr_t* frame::real_fp() const { +#ifndef AARCH64 + if (is_entry_frame()) { + // Work-around: FP (currently) does not conform to the ABI for entry + // frames (see generate_call_stub). Might be worth fixing as another CR. + // Following code assumes (and asserts) this has not yet been fixed. + assert(frame::entry_frame_call_wrapper_offset == 0, "adjust this code"); + intptr_t* new_fp = fp(); + new_fp += 5; // saved R0,R1,R2,R4,R10 +#ifndef __SOFTFP__ + new_fp += 8*2; // saved D8..D15 +#endif + return new_fp; + } +#endif // !AARCH64 + if (_cb != NULL) { + // use the frame size if valid + int size = _cb->frame_size(); + if (size > 0) { + return unextended_sp() + size; + } + } + // else rely on fp() + assert(! is_compiled_frame(), "unknown compiled frame size"); + return fp(); +}