ruby/misc/lldb_cruby.py

752 lines
32 KiB
Python
Raw Normal View History

#!/usr/bin/env python
#coding: utf-8
#
# Usage: run `command script import -r misc/lldb_cruby.py` on LLDB
#
# Test: misc/test_lldb_cruby.rb
#
from __future__ import print_function
import lldb
import os
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
import inspect
import sys
import shlex
import platform
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
import glob
from lldb_rb.constants import *
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
# BEGIN FUNCTION STYLE DECLS
# This will be refactored to use class style decls in the misc/commands dir
Mostly recover a Ruby stack trace from a core file Update the lldb script so it can mostly recover a Ruby stack trace from a core file. It's still missing line numbers and dealing with CFUNCs, but you use it like this: ``` (lldb) rbbt ec rb_control_frame_t TYPE 0x7f6fd6555fa0 EVAL ./bootstraptest/runner.rb error!! 0x7f6fd6555f68 METHOD ./bootstraptest/runner.rb main 0x7f6fd6555f30 METHOD ./bootstraptest/runner.rb in_temporary_working_directory 0x7f6fd6555ef8 METHOD /home/aaron/git/ruby/lib/tmpdir.rb mktmpdir 0x7f6fd6555ec0 BLOCK ./bootstraptest/runner.rb block in in_temporary_working_directory 0x7f6fd6555e88 CFUNC 0x7f6fd6555e50 BLOCK ./bootstraptest/runner.rb block (2 levels) in in_temporary_working_directory 0x7f6fd6555e18 BLOCK ./bootstraptest/runner.rb block in main 0x7f6fd6555de0 METHOD ./bootstraptest/runner.rb exec_test 0x7f6fd6555da8 CFUNC 0x7f6fd6555d70 BLOCK ./bootstraptest/runner.rb block in exec_test 0x7f6fd6555d38 CFUNC 0x7f6fd6555d00 TOP /home/aaron/git/ruby/bootstraptest/test_insns.rb error!! 0x7f6fd6555cc8 CFUNC 0x7f6fd6555c90 BLOCK /home/aaron/git/ruby/bootstraptest/test_insns.rb block in <top (required)> 0x7f6fd6555c58 METHOD ./bootstraptest/runner.rb assert_equal 0x7f6fd6555c20 METHOD ./bootstraptest/runner.rb assert_check 0x7f6fd6555be8 METHOD ./bootstraptest/runner.rb show_progress 0x7f6fd6555bb0 METHOD ./bootstraptest/runner.rb with_stderr 0x7f6fd6555b78 BLOCK ./bootstraptest/runner.rb block in show_progress 0x7f6fd6555b40 BLOCK ./bootstraptest/runner.rb block in assert_check 0x7f6fd6555b08 METHOD ./bootstraptest/runner.rb get_result_string 0x7f6fd6555ad0 METHOD ./bootstraptest/runner.rb make_srcfile 0x7f6fd6555a98 CFUNC 0x7f6fd6555a60 BLOCK ./bootstraptest/runner.rb block in make_srcfile ``` Getting the main execution context is difficult (it is stored in a thread local) so for now you must supply an ec and this will make a backtrace
2020-10-14 16:41:07 -07:00
class BackTrace:
VM_FRAME_MAGIC_METHOD = 0x11110001
VM_FRAME_MAGIC_BLOCK = 0x22220001
VM_FRAME_MAGIC_CLASS = 0x33330001
VM_FRAME_MAGIC_TOP = 0x44440001
VM_FRAME_MAGIC_CFUNC = 0x55550001
VM_FRAME_MAGIC_IFUNC = 0x66660001
VM_FRAME_MAGIC_EVAL = 0x77770001
VM_FRAME_MAGIC_RESCUE = 0x78880001
VM_FRAME_MAGIC_DUMMY = 0x79990001
VM_FRAME_MAGIC_MASK = 0x7fff0001
VM_FRAME_MAGIC_NAME = {
VM_FRAME_MAGIC_TOP: "TOP",
VM_FRAME_MAGIC_METHOD: "METHOD",
VM_FRAME_MAGIC_CLASS: "CLASS",
VM_FRAME_MAGIC_BLOCK: "BLOCK",
VM_FRAME_MAGIC_CFUNC: "CFUNC",
VM_FRAME_MAGIC_IFUNC: "IFUNC",
VM_FRAME_MAGIC_EVAL: "EVAL",
VM_FRAME_MAGIC_RESCUE: "RESCUE",
0: "-----"
}
def __init__(self, debugger, command, result, internal_dict):
self.debugger = debugger
self.command = command
self.result = result
self.target = debugger.GetSelectedTarget()
self.process = self.target.GetProcess()
self.thread = self.process.GetSelectedThread()
self.frame = self.thread.GetSelectedFrame()
self.tRString = self.target.FindFirstType("struct RString").GetPointerType()
self.tRArray = self.target.FindFirstType("struct RArray").GetPointerType()
rb_cft_len = len("rb_control_frame_t")
method_type_length = sorted(map(len, self.VM_FRAME_MAGIC_NAME.values()), reverse=True)[0]
# cfp address, method type, function name
self.fmt = "%%-%ds %%-%ds %%s" % (rb_cft_len, method_type_length)
def vm_frame_magic(self, cfp):
ep = cfp.GetValueForExpressionPath("->ep")
frame_type = ep.GetChildAtIndex(0).GetValueAsUnsigned() & self.VM_FRAME_MAGIC_MASK
return self.VM_FRAME_MAGIC_NAME.get(frame_type, "(none)")
def rb_iseq_path_str(self, iseq):
tRBasic = self.target.FindFirstType("struct RBasic").GetPointerType()
pathobj = iseq.GetValueForExpressionPath("->body->location.pathobj")
pathobj = pathobj.Cast(tRBasic)
flags = pathobj.GetValueForExpressionPath("->flags").GetValueAsUnsigned()
flType = flags & RUBY_T_MASK
if flType == RUBY_T_ARRAY:
pathobj = pathobj.Cast(self.tRArray)
if flags & RUBY_FL_USER1:
len = ((flags & (RUBY_FL_USER3|RUBY_FL_USER4|RUBY_FL_USER5|RUBY_FL_USER6|RUBY_FL_USER7|RUBY_FL_USER8|RUBY_FL_USER9)) >> (RUBY_FL_USHIFT+3))
Mostly recover a Ruby stack trace from a core file Update the lldb script so it can mostly recover a Ruby stack trace from a core file. It's still missing line numbers and dealing with CFUNCs, but you use it like this: ``` (lldb) rbbt ec rb_control_frame_t TYPE 0x7f6fd6555fa0 EVAL ./bootstraptest/runner.rb error!! 0x7f6fd6555f68 METHOD ./bootstraptest/runner.rb main 0x7f6fd6555f30 METHOD ./bootstraptest/runner.rb in_temporary_working_directory 0x7f6fd6555ef8 METHOD /home/aaron/git/ruby/lib/tmpdir.rb mktmpdir 0x7f6fd6555ec0 BLOCK ./bootstraptest/runner.rb block in in_temporary_working_directory 0x7f6fd6555e88 CFUNC 0x7f6fd6555e50 BLOCK ./bootstraptest/runner.rb block (2 levels) in in_temporary_working_directory 0x7f6fd6555e18 BLOCK ./bootstraptest/runner.rb block in main 0x7f6fd6555de0 METHOD ./bootstraptest/runner.rb exec_test 0x7f6fd6555da8 CFUNC 0x7f6fd6555d70 BLOCK ./bootstraptest/runner.rb block in exec_test 0x7f6fd6555d38 CFUNC 0x7f6fd6555d00 TOP /home/aaron/git/ruby/bootstraptest/test_insns.rb error!! 0x7f6fd6555cc8 CFUNC 0x7f6fd6555c90 BLOCK /home/aaron/git/ruby/bootstraptest/test_insns.rb block in <top (required)> 0x7f6fd6555c58 METHOD ./bootstraptest/runner.rb assert_equal 0x7f6fd6555c20 METHOD ./bootstraptest/runner.rb assert_check 0x7f6fd6555be8 METHOD ./bootstraptest/runner.rb show_progress 0x7f6fd6555bb0 METHOD ./bootstraptest/runner.rb with_stderr 0x7f6fd6555b78 BLOCK ./bootstraptest/runner.rb block in show_progress 0x7f6fd6555b40 BLOCK ./bootstraptest/runner.rb block in assert_check 0x7f6fd6555b08 METHOD ./bootstraptest/runner.rb get_result_string 0x7f6fd6555ad0 METHOD ./bootstraptest/runner.rb make_srcfile 0x7f6fd6555a98 CFUNC 0x7f6fd6555a60 BLOCK ./bootstraptest/runner.rb block in make_srcfile ``` Getting the main execution context is difficult (it is stored in a thread local) so for now you must supply an ec and this will make a backtrace
2020-10-14 16:41:07 -07:00
ptr = pathobj.GetValueForExpressionPath("->as.ary")
else:
len = pathobj.GetValueForExpressionPath("->as.heap.len").GetValueAsSigned()
ptr = pathobj.GetValueForExpressionPath("->as.heap.ptr")
pathobj = ptr.GetChildAtIndex(0)
pathobj = pathobj.Cast(self.tRString)
ptr, len = string2cstr(pathobj)
err = lldb.SBError()
path = self.target.process.ReadMemory(ptr, len, err)
if err.Success():
return path.decode("utf-8")
else:
return "unknown"
def dump_iseq_frame(self, cfp, iseq):
m = self.vm_frame_magic(cfp)
if iseq.GetValueAsUnsigned():
iseq_label = iseq.GetValueForExpressionPath("->body->location.label")
path = self.rb_iseq_path_str(iseq)
ptr, len = string2cstr(iseq_label.Cast(self.tRString))
err = lldb.SBError()
iseq_name = self.target.process.ReadMemory(ptr, len, err)
if err.Success():
iseq_name = iseq_name.decode("utf-8")
else:
iseq_name = "error!!"
else:
print("No iseq", file=self.result)
print(self.fmt % (("%0#12x" % cfp.GetAddress().GetLoadAddress(self.target)), m, "%s %s" % (path, iseq_name)), file=self.result)
def dump_cfunc_frame(self, cfp):
print(self.fmt % ("%0#12x" % (cfp.GetAddress().GetLoadAddress(self.target)), "CFUNC", ""), file=self.result)
def print_bt(self, ec):
tRbExecutionContext_t = self.target.FindFirstType("rb_execution_context_t")
ec = ec.Cast(tRbExecutionContext_t.GetPointerType())
vm_stack = ec.GetValueForExpressionPath("->vm_stack")
vm_stack_size = ec.GetValueForExpressionPath("->vm_stack_size")
last_cfp_frame = ec.GetValueForExpressionPath("->cfp")
cfp_type_p = last_cfp_frame.GetType()
stack_top = vm_stack.GetValueAsUnsigned() + (
vm_stack_size.GetValueAsUnsigned() * vm_stack.GetType().GetByteSize())
cfp_frame_size = cfp_type_p.GetPointeeType().GetByteSize()
start_cfp = stack_top
# Skip dummy frames
start_cfp -= cfp_frame_size
start_cfp -= cfp_frame_size
last_cfp = last_cfp_frame.GetValueAsUnsigned()
size = ((start_cfp - last_cfp) / cfp_frame_size) + 1
print(self.fmt % ("rb_control_frame_t", "TYPE", ""), file=self.result)
curr_addr = start_cfp
while curr_addr >= last_cfp:
cfp = self.target.CreateValueFromAddress("cfp", lldb.SBAddress(curr_addr, self.target), cfp_type_p.GetPointeeType())
ep = cfp.GetValueForExpressionPath("->ep")
iseq = cfp.GetValueForExpressionPath("->iseq")
frame_type = ep.GetChildAtIndex(0).GetValueAsUnsigned() & self.VM_FRAME_MAGIC_MASK
if iseq.GetValueAsUnsigned():
pc = cfp.GetValueForExpressionPath("->pc")
if pc.GetValueAsUnsigned():
self.dump_iseq_frame(cfp, iseq)
else:
if frame_type == self.VM_FRAME_MAGIC_CFUNC:
self.dump_cfunc_frame(cfp)
curr_addr -= cfp_frame_size
def lldb_init(debugger):
target = debugger.GetSelectedTarget()
global SIZEOF_VALUE
SIZEOF_VALUE = target.FindFirstType("VALUE").GetByteSize()
value_types = []
g = globals()
imemo_types = target.FindFirstType('enum imemo_type')
2022-07-06 13:20:53 -07:00
enum_members = imemo_types.GetEnumMembers()
2022-07-06 13:20:53 -07:00
for i in range(enum_members.GetSize()):
member = enum_members.GetTypeEnumMemberAtIndex(i)
g[member.GetName()] = member.GetValueAsUnsigned()
for enum in target.FindFirstGlobalVariable('ruby_dummy_gdb_enums'):
enum = enum.GetType()
members = enum.GetEnumMembers()
for i in range(0, members.GetSize()):
member = members.GetTypeEnumMemberAtIndex(i)
name = member.GetName()
value = member.GetValueAsUnsigned()
g[name] = value
if name.startswith('RUBY_T_'):
value_types.append(name)
g['value_types'] = value_types
def string2cstr(rstring):
"""Returns the pointer to the C-string in the given String object"""
if rstring.TypeIsPointerType():
rstring = rstring.Dereference()
flags = rstring.GetValueForExpressionPath(".basic->flags").unsigned
if flags & RUBY_T_MASK != RUBY_T_STRING:
raise TypeError("not a string")
if flags & RUBY_FL_USER1:
cptr = int(rstring.GetValueForExpressionPath(".as.heap.ptr").value, 0)
clen = int(rstring.GetValueForExpressionPath(".as.heap.len").value, 0)
else:
cptr = int(rstring.GetValueForExpressionPath(".as.embed.ary").location, 0)
2022-01-06 09:03:45 -05:00
clen = int(rstring.GetValueForExpressionPath(".as.embed.len").value, 0)
return cptr, clen
def output_string(debugger, result, rstring):
cptr, clen = string2cstr(rstring)
expr = "print *(const char (*)[%d])%0#x" % (clen, cptr)
append_command_output(debugger, expr, result)
def fixnum_p(x):
return x & RUBY_FIXNUM_FLAG != 0
def flonum_p(x):
return (x&RUBY_FLONUM_MASK) == RUBY_FLONUM_FLAG
def static_sym_p(x):
return (x&~(~0<<RUBY_SPECIAL_SHIFT)) == RUBY_SYMBOL_FLAG
def append_command_output(debugger, command, result):
output1 = result.GetOutput()
debugger.GetCommandInterpreter().HandleCommand(command, result)
output2 = result.GetOutput()
result.Clear()
result.write(output1)
result.write(output2)
def lldb_rp(debugger, command, result, internal_dict):
if not ('RUBY_Qfalse' in globals()):
lldb_init(debugger)
target = debugger.GetSelectedTarget()
process = target.GetProcess()
thread = process.GetSelectedThread()
frame = thread.GetSelectedFrame()
if frame.IsValid():
val = frame.EvaluateExpression(command)
else:
val = target.EvaluateExpression(command)
error = val.GetError()
if error.Fail():
print(error, file=result)
return
lldb_inspect(debugger, target, result, val)
def lldb_inspect(debugger, target, result, val):
num = val.GetValueAsSigned()
if num == RUBY_Qfalse:
print('false', file=result)
elif num == RUBY_Qtrue:
print('true', file=result)
elif num == RUBY_Qnil:
print('nil', file=result)
elif num == RUBY_Qundef:
print('undef', file=result)
elif fixnum_p(num):
print(num >> 1, file=result)
elif flonum_p(num):
append_command_output(debugger, "print rb_float_value(%0#x)" % val.GetValueAsUnsigned(), result)
elif static_sym_p(num):
if num < 128:
print("T_SYMBOL: %c" % num, file=result)
else:
print("T_SYMBOL: (%x)" % num, file=result)
append_command_output(debugger, "p rb_id2name(%0#x)" % (num >> 8), result)
elif num & RUBY_IMMEDIATE_MASK:
print('immediate(%x)' % num, file=result)
else:
tRBasic = target.FindFirstType("struct RBasic").GetPointerType()
tRValue = target.FindFirstType("struct RVALUE")
val = val.Cast(tRBasic)
flags = val.GetValueForExpressionPath("->flags").GetValueAsUnsigned()
flaginfo = ""
page = get_page(lldb, target, val)
page_type = target.FindFirstType("struct heap_page").GetPointerType()
page.Cast(page_type)
dump_bits(target, result, page, val.GetValueAsUnsigned())
if (flags & RUBY_FL_PROMOTED) == RUBY_FL_PROMOTED:
flaginfo += "[PROMOTED] "
2019-05-09 12:27:44 -07:00
if (flags & RUBY_FL_FREEZE) == RUBY_FL_FREEZE:
flaginfo += "[FROZEN] "
flType = flags & RUBY_T_MASK
if flType == RUBY_T_NONE:
print('T_NONE: %s%s' % (flaginfo, val.Dereference()), file=result)
elif flType == RUBY_T_NIL:
print('T_NIL: %s%s' % (flaginfo, val.Dereference()), file=result)
elif flType == RUBY_T_OBJECT:
result.write('T_OBJECT: %s' % flaginfo)
append_command_output(debugger, "print *(struct RObject*)%0#x" % val.GetValueAsUnsigned(), result)
elif flType == RUBY_T_CLASS or flType == RUBY_T_MODULE or flType == RUBY_T_ICLASS:
result.write('T_%s: %s' % ('CLASS' if flType == RUBY_T_CLASS else 'MODULE' if flType == RUBY_T_MODULE else 'ICLASS', flaginfo))
append_command_output(debugger, "print *(struct RClass*)%0#x" % val.GetValueAsUnsigned(), result)
tRClass = target.FindFirstType("struct RClass")
if not val.Cast(tRClass).GetChildMemberWithName("ptr").IsValid():
append_command_output(debugger, "print *(struct rb_classext_struct*)%0#x" % (val.GetValueAsUnsigned() + tRClass.GetByteSize()), result)
elif flType == RUBY_T_STRING:
result.write('T_STRING: %s' % flaginfo)
encidx = ((flags & RUBY_ENCODING_MASK)>>RUBY_ENCODING_SHIFT)
encname = target.FindFirstType("enum ruby_preserved_encindex").GetEnumMembers().GetTypeEnumMemberAtIndex(encidx).GetName()
if encname is not None:
result.write('[%s] ' % encname[14:])
else:
result.write('[enc=%d] ' % encidx)
tRString = target.FindFirstType("struct RString").GetPointerType()
ptr, len = string2cstr(val.Cast(tRString))
if len == 0:
result.write("(empty)\n")
else:
append_command_output(debugger, "print *(const char (*)[%d])%0#x" % (len, ptr), result)
elif flType == RUBY_T_SYMBOL:
result.write('T_SYMBOL: %s' % flaginfo)
tRSymbol = target.FindFirstType("struct RSymbol").GetPointerType()
val = val.Cast(tRSymbol)
append_command_output(debugger, 'print (ID)%0#x ' % val.GetValueForExpressionPath("->id").GetValueAsUnsigned(), result)
tRString = target.FindFirstType("struct RString").GetPointerType()
output_string(debugger, result, val.GetValueForExpressionPath("->fstr").Cast(tRString))
elif flType == RUBY_T_ARRAY:
tRArray = target.FindFirstType("struct RArray").GetPointerType()
val = val.Cast(tRArray)
if flags & RUBY_FL_USER1:
len = ((flags & (RUBY_FL_USER3|RUBY_FL_USER4|RUBY_FL_USER5|RUBY_FL_USER6|RUBY_FL_USER7|RUBY_FL_USER8|RUBY_FL_USER9)) >> (RUBY_FL_USHIFT+3))
ptr = val.GetValueForExpressionPath("->as.ary")
else:
len = val.GetValueForExpressionPath("->as.heap.len").GetValueAsSigned()
ptr = val.GetValueForExpressionPath("->as.heap.ptr")
result.write("T_ARRAY: %slen=%d" % (flaginfo, len))
if flags & RUBY_FL_USER1:
result.write(" (embed)")
elif flags & RUBY_FL_USER2:
shared = val.GetValueForExpressionPath("->as.heap.aux.shared").GetValueAsUnsigned()
result.write(" (shared) shared=%016x" % shared)
else:
capa = val.GetValueForExpressionPath("->as.heap.aux.capa").GetValueAsSigned()
result.write(" (ownership) capa=%d" % capa)
if len == 0:
result.write(" {(empty)}\n")
else:
result.write("\n")
if ptr.GetValueAsSigned() == 0:
append_command_output(debugger, "expression -fx -- ((struct RArray*)%0#x)->as.ary" % val.GetValueAsUnsigned(), result)
else:
append_command_output(debugger, "expression -Z %d -fx -- (const VALUE*)%0#x" % (len, ptr.GetValueAsUnsigned()), result)
elif flType == RUBY_T_HASH:
result.write("T_HASH: %s" % flaginfo)
append_command_output(debugger, "p *(struct RHash *) %0#x" % val.GetValueAsUnsigned(), result)
elif flType == RUBY_T_BIGNUM:
tRBignum = target.FindFirstType("struct RBignum").GetPointerType()
val = val.Cast(tRBignum)
sign = '+' if (flags & RUBY_FL_USER1) != 0 else '-'
if flags & RUBY_FL_USER2:
len = ((flags & (RUBY_FL_USER3|RUBY_FL_USER4|RUBY_FL_USER5)) >> (RUBY_FL_USHIFT+3))
print("T_BIGNUM: sign=%s len=%d (embed)" % (sign, len), file=result)
append_command_output(debugger, "print ((struct RBignum *) %0#x)->as.ary" % val.GetValueAsUnsigned(), result)
else:
len = val.GetValueForExpressionPath("->as.heap.len").GetValueAsSigned()
print("T_BIGNUM: sign=%s len=%d" % (sign, len), file=result)
print(val.Dereference(), file=result)
append_command_output(debugger, "expression -Z %x -fx -- (const BDIGIT*)((struct RBignum*)%d)->as.heap.digits" % (len, val.GetValueAsUnsigned()), result)
# append_command_output(debugger, "x ((struct RBignum *) %0#x)->as.heap.digits / %d" % (val.GetValueAsUnsigned(), len), result)
elif flType == RUBY_T_FLOAT:
append_command_output(debugger, "print ((struct RFloat *)%d)->float_value" % val.GetValueAsUnsigned(), result)
elif flType == RUBY_T_RATIONAL:
tRRational = target.FindFirstType("struct RRational").GetPointerType()
val = val.Cast(tRRational)
lldb_inspect(debugger, target, result, val.GetValueForExpressionPath("->num"))
output = result.GetOutput()
result.Clear()
result.write("(Rational) " + output.rstrip() + " / ")
lldb_inspect(debugger, target, result, val.GetValueForExpressionPath("->den"))
elif flType == RUBY_T_COMPLEX:
tRComplex = target.FindFirstType("struct RComplex").GetPointerType()
val = val.Cast(tRComplex)
lldb_inspect(debugger, target, result, val.GetValueForExpressionPath("->real"))
real = result.GetOutput().rstrip()
result.Clear()
lldb_inspect(debugger, target, result, val.GetValueForExpressionPath("->imag"))
imag = result.GetOutput().rstrip()
result.Clear()
if not imag.startswith("-"):
imag = "+" + imag
print("(Complex) " + real + imag + "i", file=result)
elif flType == RUBY_T_REGEXP:
tRRegex = target.FindFirstType("struct RRegexp").GetPointerType()
val = val.Cast(tRRegex)
print("(Regex) ->src {", file=result)
lldb_inspect(debugger, target, result, val.GetValueForExpressionPath("->src"))
print("}", file=result)
elif flType == RUBY_T_DATA:
tRTypedData = target.FindFirstType("struct RTypedData").GetPointerType()
val = val.Cast(tRTypedData)
flag = val.GetValueForExpressionPath("->typed_flag")
if flag.GetValueAsUnsigned() == 1:
print("T_DATA: %s" % val.GetValueForExpressionPath("->type->wrap_struct_name"), file=result)
append_command_output(debugger, "p *(struct RTypedData *) %0#x" % val.GetValueAsUnsigned(), result)
else:
print("T_DATA:", file=result)
append_command_output(debugger, "p *(struct RData *) %0#x" % val.GetValueAsUnsigned(), result)
elif flType == RUBY_T_NODE:
tRTypedData = target.FindFirstType("struct RNode").GetPointerType()
nd_type = (flags & RUBY_NODE_TYPEMASK) >> RUBY_NODE_TYPESHIFT
append_command_output(debugger, "p (node_type) %d" % nd_type, result)
val = val.Cast(tRTypedData)
append_command_output(debugger, "p *(struct RNode *) %0#x" % val.GetValueAsUnsigned(), result)
2020-05-07 14:19:08 -07:00
elif flType == RUBY_T_MOVED:
tRTypedData = target.FindFirstType("struct RMoved").GetPointerType()
val = val.Cast(tRTypedData)
append_command_output(debugger, "p *(struct RMoved *) %0#x" % val.GetValueAsUnsigned(), result)
2020-09-02 16:42:14 -07:00
elif flType == RUBY_T_MATCH:
tRTypedData = target.FindFirstType("struct RMatch").GetPointerType()
val = val.Cast(tRTypedData)
append_command_output(debugger, "p *(struct RMatch *) %0#x" % val.GetValueAsUnsigned(), result)
elif flType == RUBY_T_IMEMO:
# I'm not sure how to get IMEMO_MASK out of lldb. It's not in globals()
imemo_type = (flags >> RUBY_FL_USHIFT) & 0x0F # IMEMO_MASK
This commit implements the Object Shapes technique in CRuby. Object Shapes is used for accessing instance variables and representing the "frozenness" of objects. Object instances have a "shape" and the shape represents some attributes of the object (currently which instance variables are set and the "frozenness"). Shapes form a tree data structure, and when a new instance variable is set on an object, that object "transitions" to a new shape in the shape tree. Each shape has an ID that is used for caching. The shape structure is independent of class, so objects of different types can have the same shape. For example: ```ruby class Foo def initialize # Starts with shape id 0 @a = 1 # transitions to shape id 1 @b = 1 # transitions to shape id 2 end end class Bar def initialize # Starts with shape id 0 @a = 1 # transitions to shape id 1 @b = 1 # transitions to shape id 2 end end foo = Foo.new # `foo` has shape id 2 bar = Bar.new # `bar` has shape id 2 ``` Both `foo` and `bar` instances have the same shape because they both set instance variables of the same name in the same order. This technique can help to improve inline cache hits as well as generate more efficient machine code in JIT compilers. This commit also adds some methods for debugging shapes on objects. See `RubyVM::Shape` for more details. For more context on Object Shapes, see [Feature: #18776] Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org> Co-Authored-By: Eileen M. Uchitelle <eileencodes@gmail.com> Co-Authored-By: John Hawthorn <john@hawthorn.email>
2022-09-23 13:54:42 -04:00
print("T_IMEMO: ", file=result)
append_command_output(debugger, "p (enum imemo_type) %d" % imemo_type, result)
This commit implements the Object Shapes technique in CRuby. Object Shapes is used for accessing instance variables and representing the "frozenness" of objects. Object instances have a "shape" and the shape represents some attributes of the object (currently which instance variables are set and the "frozenness"). Shapes form a tree data structure, and when a new instance variable is set on an object, that object "transitions" to a new shape in the shape tree. Each shape has an ID that is used for caching. The shape structure is independent of class, so objects of different types can have the same shape. For example: ```ruby class Foo def initialize # Starts with shape id 0 @a = 1 # transitions to shape id 1 @b = 1 # transitions to shape id 2 end end class Bar def initialize # Starts with shape id 0 @a = 1 # transitions to shape id 1 @b = 1 # transitions to shape id 2 end end foo = Foo.new # `foo` has shape id 2 bar = Bar.new # `bar` has shape id 2 ``` Both `foo` and `bar` instances have the same shape because they both set instance variables of the same name in the same order. This technique can help to improve inline cache hits as well as generate more efficient machine code in JIT compilers. This commit also adds some methods for debugging shapes on objects. See `RubyVM::Shape` for more details. For more context on Object Shapes, see [Feature: #18776] Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org> Co-Authored-By: Eileen M. Uchitelle <eileencodes@gmail.com> Co-Authored-By: John Hawthorn <john@hawthorn.email>
2022-09-23 13:54:42 -04:00
if imemo_type == imemo_shape:
append_command_output(debugger, "p *(rb_shape_t *) %0#x" % val.GetValueAsUnsigned(), result)
else:
append_command_output(debugger, "p *(struct MEMO *) %0#x" % val.GetValueAsUnsigned(), result)
2022-06-21 16:12:27 -04:00
elif flType == RUBY_T_STRUCT:
tRTypedData = target.FindFirstType("struct RStruct").GetPointerType()
val = val.Cast(tRTypedData)
append_command_output(debugger, "p *(struct RStruct *) %0#x" % val.GetValueAsUnsigned(), result)
2020-08-27 09:00:14 -07:00
elif flType == RUBY_T_ZOMBIE:
tRZombie = target.FindFirstType("struct RZombie").GetPointerType()
val = val.Cast(tRZombie)
append_command_output(debugger, "p *(struct RZombie *) %0#x" % val.GetValueAsUnsigned(), result)
else:
print("Not-handled type %0#x" % flType, file=result)
print(val, file=result)
def count_objects(debugger, command, ctx, result, internal_dict):
objspace = ctx.frame.EvaluateExpression("ruby_current_vm->objspace")
num_pages = objspace.GetValueForExpressionPath(".heap_pages.allocated_pages").unsigned
counts = {}
total = 0
for t in range(0x00, RUBY_T_MASK+1):
counts[t] = 0
for i in range(0, num_pages):
print("\rcounting... %d/%d" % (i, num_pages), end="")
page = objspace.GetValueForExpressionPath('.heap_pages.sorted[%d]' % i)
p = page.GetChildMemberWithName('start')
num_slots = page.GetChildMemberWithName('total_slots').unsigned
for j in range(0, num_slots):
obj = p.GetValueForExpressionPath('[%d]' % j)
flags = obj.GetValueForExpressionPath('.as.basic.flags').unsigned
obj_type = flags & RUBY_T_MASK
counts[obj_type] += 1
total += num_slots
print("\rTOTAL: %d, FREE: %d" % (total, counts[0x00]))
for sym in value_types:
print("%s: %d" % (sym, counts[globals()[sym]]))
def stack_dump_raw(debugger, command, ctx, result, internal_dict):
ctx.frame.EvaluateExpression("rb_vmdebug_stack_dump_raw_current()")
def check_bits(page, bitmap_name, bitmap_index, bitmap_bit, v):
bits = page.GetChildMemberWithName(bitmap_name)
plane = bits.GetChildAtIndex(bitmap_index).GetValueAsUnsigned()
if (plane & bitmap_bit) != 0:
return v
else:
return ' '
def heap_page_body(debugger, command, ctx, result, internal_dict):
target = debugger.GetSelectedTarget()
process = target.GetProcess()
thread = process.GetSelectedThread()
frame = thread.GetSelectedFrame()
val = frame.EvaluateExpression(command)
page = get_page_body(lldb, target, val)
print("Page body address: ", page.GetAddress(), file=result)
print(page, file=result)
def get_page_body(lldb, target, val):
tHeapPageBody = target.FindFirstType("struct heap_page_body")
addr = val.GetValueAsUnsigned()
page_addr = addr & ~(HEAP_PAGE_ALIGN_MASK)
address = lldb.SBAddress(page_addr, target)
return target.CreateValueFromAddress("page", address, tHeapPageBody)
def get_page(lldb, target, val):
body = get_page_body(lldb, target, val)
return body.GetValueForExpressionPath("->header.page")
def dump_node(debugger, command, ctx, result, internal_dict):
args = shlex.split(command)
if not args:
return
node = args[0]
dump = ctx.frame.EvaluateExpression("(struct RString*)rb_parser_dump_tree((NODE*)(%s), 0)" % node)
output_string(ctx, result, dump)
Mostly recover a Ruby stack trace from a core file Update the lldb script so it can mostly recover a Ruby stack trace from a core file. It's still missing line numbers and dealing with CFUNCs, but you use it like this: ``` (lldb) rbbt ec rb_control_frame_t TYPE 0x7f6fd6555fa0 EVAL ./bootstraptest/runner.rb error!! 0x7f6fd6555f68 METHOD ./bootstraptest/runner.rb main 0x7f6fd6555f30 METHOD ./bootstraptest/runner.rb in_temporary_working_directory 0x7f6fd6555ef8 METHOD /home/aaron/git/ruby/lib/tmpdir.rb mktmpdir 0x7f6fd6555ec0 BLOCK ./bootstraptest/runner.rb block in in_temporary_working_directory 0x7f6fd6555e88 CFUNC 0x7f6fd6555e50 BLOCK ./bootstraptest/runner.rb block (2 levels) in in_temporary_working_directory 0x7f6fd6555e18 BLOCK ./bootstraptest/runner.rb block in main 0x7f6fd6555de0 METHOD ./bootstraptest/runner.rb exec_test 0x7f6fd6555da8 CFUNC 0x7f6fd6555d70 BLOCK ./bootstraptest/runner.rb block in exec_test 0x7f6fd6555d38 CFUNC 0x7f6fd6555d00 TOP /home/aaron/git/ruby/bootstraptest/test_insns.rb error!! 0x7f6fd6555cc8 CFUNC 0x7f6fd6555c90 BLOCK /home/aaron/git/ruby/bootstraptest/test_insns.rb block in <top (required)> 0x7f6fd6555c58 METHOD ./bootstraptest/runner.rb assert_equal 0x7f6fd6555c20 METHOD ./bootstraptest/runner.rb assert_check 0x7f6fd6555be8 METHOD ./bootstraptest/runner.rb show_progress 0x7f6fd6555bb0 METHOD ./bootstraptest/runner.rb with_stderr 0x7f6fd6555b78 BLOCK ./bootstraptest/runner.rb block in show_progress 0x7f6fd6555b40 BLOCK ./bootstraptest/runner.rb block in assert_check 0x7f6fd6555b08 METHOD ./bootstraptest/runner.rb get_result_string 0x7f6fd6555ad0 METHOD ./bootstraptest/runner.rb make_srcfile 0x7f6fd6555a98 CFUNC 0x7f6fd6555a60 BLOCK ./bootstraptest/runner.rb block in make_srcfile ``` Getting the main execution context is difficult (it is stored in a thread local) so for now you must supply an ec and this will make a backtrace
2020-10-14 16:41:07 -07:00
def rb_backtrace(debugger, command, result, internal_dict):
if not ('RUBY_Qfalse' in globals()):
lldb_init(debugger)
Mostly recover a Ruby stack trace from a core file Update the lldb script so it can mostly recover a Ruby stack trace from a core file. It's still missing line numbers and dealing with CFUNCs, but you use it like this: ``` (lldb) rbbt ec rb_control_frame_t TYPE 0x7f6fd6555fa0 EVAL ./bootstraptest/runner.rb error!! 0x7f6fd6555f68 METHOD ./bootstraptest/runner.rb main 0x7f6fd6555f30 METHOD ./bootstraptest/runner.rb in_temporary_working_directory 0x7f6fd6555ef8 METHOD /home/aaron/git/ruby/lib/tmpdir.rb mktmpdir 0x7f6fd6555ec0 BLOCK ./bootstraptest/runner.rb block in in_temporary_working_directory 0x7f6fd6555e88 CFUNC 0x7f6fd6555e50 BLOCK ./bootstraptest/runner.rb block (2 levels) in in_temporary_working_directory 0x7f6fd6555e18 BLOCK ./bootstraptest/runner.rb block in main 0x7f6fd6555de0 METHOD ./bootstraptest/runner.rb exec_test 0x7f6fd6555da8 CFUNC 0x7f6fd6555d70 BLOCK ./bootstraptest/runner.rb block in exec_test 0x7f6fd6555d38 CFUNC 0x7f6fd6555d00 TOP /home/aaron/git/ruby/bootstraptest/test_insns.rb error!! 0x7f6fd6555cc8 CFUNC 0x7f6fd6555c90 BLOCK /home/aaron/git/ruby/bootstraptest/test_insns.rb block in <top (required)> 0x7f6fd6555c58 METHOD ./bootstraptest/runner.rb assert_equal 0x7f6fd6555c20 METHOD ./bootstraptest/runner.rb assert_check 0x7f6fd6555be8 METHOD ./bootstraptest/runner.rb show_progress 0x7f6fd6555bb0 METHOD ./bootstraptest/runner.rb with_stderr 0x7f6fd6555b78 BLOCK ./bootstraptest/runner.rb block in show_progress 0x7f6fd6555b40 BLOCK ./bootstraptest/runner.rb block in assert_check 0x7f6fd6555b08 METHOD ./bootstraptest/runner.rb get_result_string 0x7f6fd6555ad0 METHOD ./bootstraptest/runner.rb make_srcfile 0x7f6fd6555a98 CFUNC 0x7f6fd6555a60 BLOCK ./bootstraptest/runner.rb block in make_srcfile ``` Getting the main execution context is difficult (it is stored in a thread local) so for now you must supply an ec and this will make a backtrace
2020-10-14 16:41:07 -07:00
bt = BackTrace(debugger, command, result, internal_dict)
frame = bt.frame
if command:
if frame.IsValid():
val = frame.EvaluateExpression(command)
else:
val = target.EvaluateExpression(command)
error = val.GetError()
if error.Fail():
print >> result, error
return
else:
print("Need an EC for now")
bt.print_bt(val)
def dump_bits(target, result, page, object_address, end = "\n"):
tRValue = target.FindFirstType("struct RVALUE")
tUintPtr = target.FindFirstType("uintptr_t") # bits_t
num_in_page = (object_address & HEAP_PAGE_ALIGN_MASK) // tRValue.GetByteSize();
bits_bitlength = tUintPtr.GetByteSize() * 8
bitmap_index = num_in_page // bits_bitlength
bitmap_offset = num_in_page & (bits_bitlength - 1)
bitmap_bit = 1 << bitmap_offset
print("bits: [%s%s%s%s%s]" % (
check_bits(page, "uncollectible_bits", bitmap_index, bitmap_bit, "L"),
check_bits(page, "mark_bits", bitmap_index, bitmap_bit, "M"),
check_bits(page, "pinned_bits", bitmap_index, bitmap_bit, "P"),
check_bits(page, "marking_bits", bitmap_index, bitmap_bit, "R"),
check_bits(page, "wb_unprotected_bits", bitmap_index, bitmap_bit, "U"),
), end=end, file=result)
class HeapPageIter:
def __init__(self, page, target):
self.page = page
self.target = target
self.start = page.GetChildMemberWithName('start').GetValueAsUnsigned();
self.num_slots = page.GetChildMemberWithName('total_slots').unsigned
self.slot_size = page.GetChildMemberWithName('size_pool').GetChildMemberWithName('slot_size').unsigned
self.counter = 0
self.tRBasic = target.FindFirstType("struct RBasic")
self.tRValue = target.FindFirstType("struct RVALUE")
def is_valid(self):
heap_page_header_size = self.target.FindFirstType("struct heap_page_header").GetByteSize()
rvalue_size = self.slot_size
heap_page_obj_limit = int((HEAP_PAGE_SIZE - heap_page_header_size) / self.slot_size)
return (heap_page_obj_limit - 1) <= self.num_slots <= heap_page_obj_limit
def __iter__(self):
return self
def __next__(self):
if self.counter < self.num_slots:
obj_addr_i = self.start + (self.counter * self.slot_size)
obj_addr = lldb.SBAddress(obj_addr_i, self.target)
slot_info = (self.counter, obj_addr_i, self.target.CreateValueFromAddress("object", obj_addr, self.tRBasic))
self.counter += 1
return slot_info
else:
raise StopIteration
def dump_page_internal(page, target, process, thread, frame, result, debugger, highlight=None):
if not ('RUBY_Qfalse' in globals()):
lldb_init(debugger)
ruby_type_map = ruby_types(debugger)
freelist = []
fl_start = page.GetChildMemberWithName('freelist').GetValueAsUnsigned()
tRVALUE = target.FindFirstType("struct RVALUE")
while fl_start > 0:
freelist.append(fl_start)
obj_addr = lldb.SBAddress(fl_start, target)
obj = target.CreateValueFromAddress("object", obj_addr, tRVALUE)
fl_start = obj.GetChildMemberWithName("as").GetChildMemberWithName("free").GetChildMemberWithName("next").GetValueAsUnsigned()
page_iter = HeapPageIter(page, target)
if page_iter.is_valid():
for (page_index, obj_addr, obj) in page_iter:
dump_bits(target, result, page, obj_addr, end= " ")
flags = obj.GetChildMemberWithName('flags').GetValueAsUnsigned()
flType = flags & RUBY_T_MASK
flidx = ' '
if flType == RUBY_T_NONE:
try:
flidx = "%3d" % freelist.index(obj_addr)
except ValueError:
flidx = ' -1'
if flType == RUBY_T_NONE:
klass = obj.GetChildMemberWithName('klass').GetValueAsUnsigned()
result_str = "%s idx: [%3d] freelist_idx: {%s} Addr: %0#x (flags: %0#x, next: %0#x)" % (rb_type(flags, ruby_type_map), page_index, flidx, obj_addr, flags, klass)
else:
result_str = "%s idx: [%3d] freelist_idx: {%s} Addr: %0#x (flags: %0#x)" % (rb_type(flags, ruby_type_map), page_index, flidx, obj_addr, flags)
if highlight == obj_addr:
result_str = ' '.join([result_str, "<<<<<"])
print(result_str, file=result)
else:
print("%s is not a valid heap page" % page, file=result)
def dump_page(debugger, command, result, internal_dict):
target = debugger.GetSelectedTarget()
process = target.GetProcess()
thread = process.GetSelectedThread()
frame = thread.GetSelectedFrame()
tHeapPageP = target.FindFirstType("struct heap_page").GetPointerType()
page = frame.EvaluateExpression(command)
page = page.Cast(tHeapPageP)
dump_page_internal(page, target, process, thread, frame, result, debugger)
def dump_page_rvalue(debugger, command, result, internal_dict):
target = debugger.GetSelectedTarget()
process = target.GetProcess()
thread = process.GetSelectedThread()
frame = thread.GetSelectedFrame()
val = frame.EvaluateExpression(command)
page = get_page(lldb, target, val)
page_type = target.FindFirstType("struct heap_page").GetPointerType()
page.Cast(page_type)
dump_page_internal(page, target, process, thread, frame, result, debugger, highlight=val.GetValueAsUnsigned())
def rb_type(flags, ruby_types):
flType = flags & RUBY_T_MASK
return "%-10s" % (ruby_types.get(flType, ("%0#x" % flType)))
def ruby_types(debugger):
target = debugger.GetSelectedTarget()
types = {}
for enum in target.FindFirstGlobalVariable('ruby_dummy_gdb_enums'):
enum = enum.GetType()
members = enum.GetEnumMembers()
for i in range(0, members.GetSize()):
member = members.GetTypeEnumMemberAtIndex(i)
name = member.GetName()
value = member.GetValueAsUnsigned()
if name.startswith('RUBY_T_'):
types[value] = name.replace('RUBY_', '')
return types
def rb_ary_entry(target, ary, idx, result):
tRArray = target.FindFirstType("struct RArray").GetPointerType()
ary = ary.Cast(tRArray)
flags = ary.GetValueForExpressionPath("->flags").GetValueAsUnsigned()
if flags & RUBY_FL_USER1:
ptr = ary.GetValueForExpressionPath("->as.ary")
else:
ptr = ary.GetValueForExpressionPath("->as.heap.ptr")
ptr_addr = ptr.GetValueAsUnsigned() + (idx * ptr.GetType().GetByteSize())
return target.CreateValueFromAddress("ary_entry[%d]" % idx, lldb.SBAddress(ptr_addr, target), ptr.GetType().GetPointeeType())
def rb_id_to_serial(id_val):
if id_val > tLAST_OP_ID:
return id_val >> RUBY_ID_SCOPE_SHIFT
else:
return id_val
def rb_id2str(debugger, command, result, internal_dict):
if not ('RUBY_Qfalse' in globals()):
lldb_init(debugger)
target = debugger.GetSelectedTarget()
process = target.GetProcess()
thread = process.GetSelectedThread()
frame = thread.GetSelectedFrame()
global_symbols = target.FindFirstGlobalVariable("ruby_global_symbols")
id_val = frame.EvaluateExpression(command).GetValueAsUnsigned()
num = rb_id_to_serial(id_val)
last_id = global_symbols.GetChildMemberWithName("last_id").GetValueAsUnsigned()
ID_ENTRY_SIZE = 2
ID_ENTRY_UNIT = int(target.FindFirstGlobalVariable("ID_ENTRY_UNIT").GetValue())
ids = global_symbols.GetChildMemberWithName("ids")
if (num <= last_id):
idx = num // ID_ENTRY_UNIT
ary = rb_ary_entry(target, ids, idx, result)
pos = (num % ID_ENTRY_UNIT) * ID_ENTRY_SIZE
id_str = rb_ary_entry(target, ary, pos, result)
lldb_inspect(debugger, target, result, id_str)
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
# END FUNCTION STYLE DECLS
load_dir, _ = os.path.split(os.path.realpath(__file__))
for fname in glob.glob(f"{load_dir}/lldb_rb/commands/*_command.py"):
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
_, basename = os.path.split(fname)
mname, _ = os.path.splitext(basename)
exec(f"import lldb_rb.commands.{mname}")
def __lldb_init_module(debugger, internal_dict):
[ci-skip][Feature #18910][lldb] Provide class framework for lldb commands `lldb_cruby.py` manages lldb custom commands using functions. The file is a large list of Python functions, and an init handler to map some of the Python functions into the debugger, to enable execution of custom logic during a debugging session. Since LLDB 3.7 (September 2015) there has also been support for using python classes rather than bare functions, as long as those classes implement a specific interface. This PR Introduces some more defined structure to the LLDB helper functions by switching from the function based implementation to the class based one, and providing an auto-loading mechanism by which new functions can be loaded. The intention behind this change is to make working with the LLDB helpers easier, by reducing code duplication, providing a consistent structure and a clearer API for developers. The current function based approach has some advantages and disadvantages Advantages: - Adding new code is easy. - All the code is self contained and searchable. Disadvantages: - No visible organisation of the file contents. This means - Hard to tell which functions are utility functions and which are available to you in a debugging session - Lots of code duplication within lldb functions - Large files quickly become intimidating to work with - for example, `lldb_disasm.py` was implemented as a seperate Python module because it was easier to start with a clean slate than add significant amounts of code to `lldb_cruby.py` This PR attempts, to fix the disadvantages of the current approach and maintain, or enhance, the benefits. The new structure of a command looks like this; ``` class TestCommand(RbBaseCommand): # program is the keyword the user will type in lldb to execute this command program = "test" # help_string will be displayed in lldb when the user uses the help functions help_string = "This is a test command to show how to implement lldb commands" # call is where our command logic will be implemented def call(self, debugger, command, exe_ctx, result): pass ``` If the command fulfils the following criteria it will then be auto-loaded when an lldb session is started: - The package file must exist inside the `commands` directory and the filename must end in `_command.py` - The package must implement a class whose name ends in `Command` - The class inherits from `RbBaseCommand` or at minimum a class that shares the same interface as `RbBaseCommand` (at minimum this means defining `__init__` and `__call__`, and using `__call__` to call `call` which is defined in the subclasses). - The class must have a class variable `package` that is a String. This is the name of the command you'll call in the `lldb` debugger.
2022-07-13 13:18:03 +01:00
# Register all classes that subclass RbBaseCommand
for memname, mem in inspect.getmembers(sys.modules["lldb_rb.rb_base_command"]):
if inspect.isclass(mem):
for sclass in mem.__subclasses__():
sclass.register_lldb_command(debugger, f"{__name__}.{sclass.__module__}")
## FUNCTION INITS - These should be removed when converted to class commands
debugger.HandleCommand("command script add -f lldb_cruby.lldb_rp rp")
debugger.HandleCommand("command script add -f lldb_cruby.count_objects rb_count_objects")
debugger.HandleCommand("command script add -f lldb_cruby.stack_dump_raw SDR")
debugger.HandleCommand("command script add -f lldb_cruby.dump_node dump_node")
debugger.HandleCommand("command script add -f lldb_cruby.heap_page_body heap_page_body")
Mostly recover a Ruby stack trace from a core file Update the lldb script so it can mostly recover a Ruby stack trace from a core file. It's still missing line numbers and dealing with CFUNCs, but you use it like this: ``` (lldb) rbbt ec rb_control_frame_t TYPE 0x7f6fd6555fa0 EVAL ./bootstraptest/runner.rb error!! 0x7f6fd6555f68 METHOD ./bootstraptest/runner.rb main 0x7f6fd6555f30 METHOD ./bootstraptest/runner.rb in_temporary_working_directory 0x7f6fd6555ef8 METHOD /home/aaron/git/ruby/lib/tmpdir.rb mktmpdir 0x7f6fd6555ec0 BLOCK ./bootstraptest/runner.rb block in in_temporary_working_directory 0x7f6fd6555e88 CFUNC 0x7f6fd6555e50 BLOCK ./bootstraptest/runner.rb block (2 levels) in in_temporary_working_directory 0x7f6fd6555e18 BLOCK ./bootstraptest/runner.rb block in main 0x7f6fd6555de0 METHOD ./bootstraptest/runner.rb exec_test 0x7f6fd6555da8 CFUNC 0x7f6fd6555d70 BLOCK ./bootstraptest/runner.rb block in exec_test 0x7f6fd6555d38 CFUNC 0x7f6fd6555d00 TOP /home/aaron/git/ruby/bootstraptest/test_insns.rb error!! 0x7f6fd6555cc8 CFUNC 0x7f6fd6555c90 BLOCK /home/aaron/git/ruby/bootstraptest/test_insns.rb block in <top (required)> 0x7f6fd6555c58 METHOD ./bootstraptest/runner.rb assert_equal 0x7f6fd6555c20 METHOD ./bootstraptest/runner.rb assert_check 0x7f6fd6555be8 METHOD ./bootstraptest/runner.rb show_progress 0x7f6fd6555bb0 METHOD ./bootstraptest/runner.rb with_stderr 0x7f6fd6555b78 BLOCK ./bootstraptest/runner.rb block in show_progress 0x7f6fd6555b40 BLOCK ./bootstraptest/runner.rb block in assert_check 0x7f6fd6555b08 METHOD ./bootstraptest/runner.rb get_result_string 0x7f6fd6555ad0 METHOD ./bootstraptest/runner.rb make_srcfile 0x7f6fd6555a98 CFUNC 0x7f6fd6555a60 BLOCK ./bootstraptest/runner.rb block in make_srcfile ``` Getting the main execution context is difficult (it is stored in a thread local) so for now you must supply an ec and this will make a backtrace
2020-10-14 16:41:07 -07:00
debugger.HandleCommand("command script add -f lldb_cruby.rb_backtrace rbbt")
debugger.HandleCommand("command script add -f lldb_cruby.dump_page dump_page")
debugger.HandleCommand("command script add -f lldb_cruby.dump_page_rvalue dump_page_rvalue")
debugger.HandleCommand("command script add -f lldb_cruby.rb_id2str rb_id2str")
lldb_rb.rb_base_command.RbBaseCommand.lldb_init(debugger)
print("lldb scripts for ruby has been installed.")