mirror/binutils-gdb
2
0
Fork 0
mirror of https://sourceware.org/git/binutils-gdb.git synced 2024-12-15 04:31:49 +08:00
Code Issues Packages Projects Releases Wiki Activity
4ec8aa9e94
binutils-gdb/gdb/std-operator.def
Andrew Burgess 6f0dabd46d gdb/x86: handle stap probe arguments in xmm registers 
On x86 machines with xmm register, and with recent versions of
systemtap (and gcc?), it can occur that stap probe arguments will be
placed into xmm registers.

I notice this happening on a current Fedora Rawhide install with the
following package versions installed:

  $ rpm -q glibc systemtap gcc
  glibc-2.35.9000-10.fc37.x86_64
  systemtap-4.7~pre16468670g9f253544-1.fc37.x86_64
  gcc-12.0.1-0.12.fc37.x86_64

If I check the probe data in libc, I see this:

  $ readelf -n /lib64/libc.so.6
  ...
  stapsdt              0x0000004d       NT_STAPSDT (SystemTap probe descriptors)
    Provider: libc
    Name: pthread_start
    Location: 0x0000000000090ac3, Base: 0x00000000001c65c4, Semaphore: 0x0000000000000000
    Arguments: 8@%xmm1 8@1600(%rbx) 8@1608(%rbx)
  stapsdt              0x00000050       NT_STAPSDT (SystemTap probe descriptors)
    Provider: libc
    Name: pthread_create
    Location: 0x00000000000912f1, Base: 0x00000000001c65c4, Semaphore: 0x0000000000000000
    Arguments: 8@%xmm1 8@%r13 8@8(%rsp) 8@16(%rsp)
  ...

Notice that for both of these probes, the first argument is a uint64_t
stored in the xmm1 register.

Unfortunately, if I try to use this probe within GDB, then I can't
view the first argument.  Here's an example session:

  $ gdb $(which gdb)
  (gdb) start
  ...
  (gdb) info probes stap  libc pthread_create
  ...
  (gdb) break *0x00007ffff729e2f1       # Use address of probe.
  (gdb) continue
  ...
  (gdb) p $_probe_arg0
  Invalid cast.

What's going wrong?  If I re-run my session, but this time use 'set
debug stap-expression 1', this is what I see:

  (gdb) set debug stap-expression 1
  (gdb) p $_probe_arg0
  Operation: UNOP_CAST
   Operation: OP_REGISTER
    String: xmm1
   Type: uint64_t
  Operation: UNOP_CAST
   Operation: OP_REGISTER
    String: r13
   Type: uint64_t
  Operation: UNOP_CAST
   Operation: UNOP_IND
    Operation: UNOP_CAST
     Operation: BINOP_ADD
      Operation: OP_LONG
       Type: long
       Constant: 0x0000000000000008
      Operation: OP_REGISTER
       String: rsp
     Type: uint64_t *
   Type: uint64_t
  Operation: UNOP_CAST
   Operation: UNOP_IND
    Operation: UNOP_CAST
     Operation: BINOP_ADD
      Operation: OP_LONG
       Type: long
       Constant: 0x0000000000000010
      Operation: OP_REGISTER
       String: rsp
     Type: uint64_t *
   Type: uint64_t
  Invalid cast.
  (gdb)

The important bit is this:

  Operation: UNOP_CAST
   Operation: OP_REGISTER
    String: xmm1
   Type: uint64_t

Which is where we cast the xmm1 register to uint64_t.  And the final
piece of the puzzle is:

  (gdb) ptype $xmm1
  type = union vec128 {
      v8bf16 v8_bfloat16;
      v4f v4_float;
      v2d v2_double;
      v16i8 v16_int8;
      v8i16 v8_int16;
      v4i32 v4_int32;
      v2i64 v2_int64;
      uint128_t uint128;
  }

So, we are attempting to cast a union type to a scalar type, which is
not supporting in C/C++, and as a consequence GDB's expression
evaluator throws an error when we attempt to do this.

The first approach I considered for solving this problem was to try
and make use of gdbarch_stap_adjust_register.  We already have a
gdbarch method (gdbarch_stap_adjust_register) that allows us to tweak
the name of the register that we access.  Currently only x86
architectures use this to transform things like ax to eax in some
cases.

I wondered, what if we change gdbarch_stap_adjust_register to do more
than just change the register names?  What if this method instead
became gdbarch_stap_read_register.  This new method would return a
operation_up, and would take the register name, and the type we are
trying to read from the register, and return the operation that
actually reads the register.

The default implementation of this method would just use
user_reg_map_name_to_regnum, and then create a register_operation,
like we already do in stap_parse_register_operand.  But, for x86
architectures this method would fist possibly adjust the register
name, then do the default action to read the register.  Finally, for
x86 this method would spot when we were accessing an xmm register,
and, based on the type being pulled from the register, would extract
the correct field from the union.

The benefit of this approach is that it would work with the expression
types that GDB currently supports.  The draw back would be that this
approach would not be very generic.  We'd need code to handle each
sub-field size with an xmm register.  If other architectures started
using vector registers for probe arguments, those architectures would
have to create their own gdbarch_stap_read_register method.  And
finally, the type of the xmm registers comes from the type defined in
the target description, there's a risk that GDB might end up
hard-coding the names of type sub-fields, then if a target uses a
different target description, with different field names for xmm
registers, the stap probes would stop working.

And so, based on all the above draw backs, I rejected this first
approach.

My second plan involves adding a new expression type to GDB called
unop_extract_operation.  This new expression takes a value and a type,
during evaluation the value contents are fetched, and then a new value
is extracted from the value contents (based on type).  This is similar
to the following C expression:

  result_value = *((output_type *) &input_value);

Obviously we can't actually build this expression in this case, as the
input_value is in a register, but hopefully the above makes it clearer
what I'm trying to do.

The benefit of the new expression approach is that this code can be
shared across all architectures, and it doesn't care about sub-field
names within the union type.

The draw-backs that I see are potential future problems if arguments
are not stored within the least significant bytes of the register.
However if/when that becomes an issue we can adapt the
gdbarch_stap_read_register approach to allow architectures to control
how a value is extracted.

For testing, I've extended the existing gdb.base/stap-probe.exp test
to include a function that tries to force an argument into an xmm
register.  Obviously, that will only work on a x86 target, so I've
guarded the new function with an appropriate GCC define.  In the exp
script we use readelf to check if the probe exists, and is using the
xmm register.

If the probe doesn't exist then the associated tests are skipped.

If the probe exists, put isn't using the xmm register (which will
depend on systemtap/gcc versions), then again, the tests are skipped.

Otherwise, we can run the test.  I think the cost of running readelf
is pretty low, so I don't feel too bad making all the non-xmm targets
running this step.

I found that on a Fedora 35 install, with these packages installed, I
was able to run this test and have the probe argument be placed in an
xmm register:

  $ rpm -q systemtap gcc glibc
  systemtap-4.6-4.fc35.x86_64
  gcc-11.2.1-9.fc35.x86_64
  glibc-2.34-7.fc35.x86_64

Finally, as this patch adds a new operation type, then I need to
consider how to generate an agent expression for the new operation
type.

I have kicked the can down the road a bit on this.  In the function
stap_parse_register_operand, I only create a unop_extract_operation in
the case where the register type is non-scalar, this means that in
most cases I don't need to worry about generating an agent expression
at all.

In the xmm register case, when an unop_extract_operation will be
created, I have sketched out how the agent expression could be
handled, however, this code is currently not reached.  When we try to
generate the agent expression to place the xmm register on the stack,
GDB hits this error:

  (gdb) trace -probe-stap test:xmmreg
  Tracepoint 1 at 0x401166
  (gdb) actions
  Enter actions for tracepoint 1, one per line.
  End with a line saying just "end".
  >collect $_probe_arg0
  Value not scalar: cannot be an rvalue.

This is because GDB doesn't currently support placing non-scalar types
on the agent expression evaluation stack.  Solving this is clearly
related to the original problem, but feels a bit like a second
problem.  I'd like to get feedback on whether my approach to solving
the original problem is acceptable or not before I start looking at
how to handle xmm registers within agent expressions.
2022-03-21 14:39:14 +00:00

403 lines
14 KiB
Modula-2
Raw Blame History

/* Standard language operator definitions for GDB, the GNU debugger.
Copyright (C) 1986-2022 Free Software Foundation, Inc.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program 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 for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
/* Used when it's necessary to pass an opcode which will be ignored,
or to catch uninitialized values. */
OP (OP_NULL)
/* BINOP_... operate on two values computed by following subexpressions,
replacing them by one result value. They take no immediate arguments. */
OP (BINOP_ADD) /* + */
OP (BINOP_SUB) /* - */
OP (BINOP_MUL) /* * */
OP (BINOP_DIV) /* / */
OP (BINOP_REM) /* % */
OP (BINOP_MOD) /* mod (Knuth 1.2.4) */
OP (BINOP_LSH) /* << */
OP (BINOP_RSH) /* >> */
OP (BINOP_LOGICAL_AND) /* && */
OP (BINOP_LOGICAL_OR) /* || */
OP (BINOP_BITWISE_AND) /* & */
OP (BINOP_BITWISE_IOR) /* | */
OP (BINOP_BITWISE_XOR) /* ^ */
OP (BINOP_EQUAL) /* == */
OP (BINOP_NOTEQUAL) /* != */
OP (BINOP_LESS) /* < */
OP (BINOP_GTR) /* > */
OP (BINOP_LEQ) /* <= */
OP (BINOP_GEQ) /* >= */
OP (BINOP_REPEAT) /* @ */
OP (BINOP_ASSIGN) /* = */
OP (BINOP_COMMA) /* , */
OP (BINOP_SUBSCRIPT) /* x[y] */
OP (BINOP_EXP) /* Exponentiation */
/* C++. */
OP (BINOP_MIN) /* <? */
OP (BINOP_MAX) /* >? */
/* STRUCTOP_MEMBER is used for pointer-to-member constructs.
X . * Y translates into X STRUCTOP_MEMBER Y. */
OP (STRUCTOP_MEMBER)
/* STRUCTOP_MPTR is used for pointer-to-member constructs
when X is a pointer instead of an aggregate. */
OP (STRUCTOP_MPTR)
/* TYPE_INSTANCE is used when the user specifies a specific
type instantiation for overloaded methods/functions.
The format is:
TYPE_INSTANCE num_types type0 ... typeN num_types TYPE_INSTANCE. */
OP (TYPE_INSTANCE)
/* end of C++. */
/* For Modula-2 integer division DIV. */
OP (BINOP_INTDIV)
/* +=, -=, *=, and so on. The following exp_element is another opcode,
a BINOP_, saying how to modify. Then comes another BINOP_ASSIGN_MODIFY,
making three exp_elements in total. */
OP (BINOP_ASSIGN_MODIFY)
/* Modula-2 standard (binary) procedures. */
OP (BINOP_VAL)
/* Concatenate two operands, such as character strings or bitstrings.
If the first operand is a integer expression, then it means concatenate
the second operand with itself that many times. */
OP (BINOP_CONCAT)
/* Operates on three values computed by following subexpressions. */
OP (TERNOP_COND) /* ?: */
/* A sub-string/sub-array. Ada syntax: OP1(OP2..OP3). Return
elements OP2 through OP3 of OP1. */
OP (TERNOP_SLICE)
/* Multidimensional subscript operator, such as Modula-2 x[a,b,...].
The dimensionality is encoded in the operator, like the number of
function arguments in OP_FUNCALL, I.E. <OP><dimension><OP>.
The value of the first following subexpression is subscripted
by each of the next following subexpressions, one per dimension. */
OP (MULTI_SUBSCRIPT)
/* The OP_... series take immediate following arguments.
After the arguments come another OP_... (the same one)
so that the grouping can be recognized from the end. */
/* OP_LONG is followed by a type pointer in the next exp_element
and the long constant value in the following exp_element.
Then comes another OP_LONG.
Thus, the operation occupies four exp_elements. */
OP (OP_LONG)
/* OP_FLOAT is similar but takes a floating-point constant encoded in
the target format for the given type instead of a long. */
OP (OP_FLOAT)
/* OP_VAR_VALUE takes one struct block * in the following element,
and one struct symbol * in the following exp_element, followed
by another OP_VAR_VALUE, making four exp_elements. If the
block is non-NULL, evaluate the symbol relative to the
innermost frame executing in that block; if the block is NULL
use the selected frame. */
OP (OP_VAR_VALUE)
/* OP_VAR_ENTRY_VALUE takes one struct symbol * in the following element,
followed by another OP_VAR_ENTRY_VALUE, making three exp_elements.
somename@entry may mean parameter value as present at the entry of the
current function. Implemented via DW_OP_entry_value. */
OP (OP_VAR_ENTRY_VALUE)
/* OP_VAR_MSYM_VALUE takes one struct objfile * in the following
element, and one struct minimal_symbol * in the following
exp_element, followed by another OP_VAR_MSYM_VALUE, making four
exp_elements. */
OP (OP_VAR_MSYM_VALUE)
/* OP_LAST is followed by an integer in the next exp_element.
The integer is zero for the last value printed,
or it is the absolute number of a history element.
With another OP_LAST at the end, this makes three exp_elements. */
OP (OP_LAST)
/* OP_REGISTER is followed by a string in the next exp_element.
This is the name of a register to fetch. */
OP (OP_REGISTER)
/* OP_INTERNALVAR is followed by an internalvar ptr in the next
exp_element. With another OP_INTERNALVAR at the end, this
makes three exp_elements. */
OP (OP_INTERNALVAR)
/* OP_FUNCALL is followed by an integer in the next exp_element.
The integer is the number of args to the function call.
That many plus one values from following subexpressions
are used, the first one being the function.
The integer is followed by a repeat of OP_FUNCALL,
making three exp_elements. */
OP (OP_FUNCALL)
/* OP_OBJC_MSGCALL is followed by a string in the next exp_element
and then an integer. The string is the selector string. The
integer is the number of arguments to the message call. That
many plus one values are used, the first one being the object
pointer. This is an Objective C message. */
OP (OP_OBJC_MSGCALL)
/* OP_COMPLEX takes a type in the following element, followed by another
OP_COMPLEX, making three exp_elements. It is followed by two double
args, and converts them into a complex number of the given type. */
OP (OP_COMPLEX)
/* OP_STRING represents a string constant.
Its format is the same as that of a STRUCTOP, but the string
data is just made into a string constant when the operation
is executed. */
OP (OP_STRING)
/* OP_ARRAY creates an array constant out of the following subexpressions.
It is followed by two exp_elements, the first containing an integer
that is the lower bound of the array and the second containing another
integer that is the upper bound of the array. The second integer is
followed by a repeat of OP_ARRAY, making four exp_elements total.
The bounds are used to compute the number of following subexpressions
to consume, as well as setting the bounds in the created array constant.
The type of the elements is taken from the type of the first subexp,
and they must all match. */
OP (OP_ARRAY)
/* UNOP_EXTRACT takes a value and a type, like a cast, but, instead of
casting the value to the given type, a new value (of the given
type) is extracted from the contents of the old value, starting
from the least significant byte.
It is invalid for the given type to be larger than the type of the
given value. */
OP (UNOP_EXTRACT)
/* UNOP_CAST is followed by a type pointer in the next exp_element.
With another UNOP_CAST at the end, this makes three exp_elements.
It casts the value of the following subexpression. */
OP (UNOP_CAST)
/* Like UNOP_CAST, but the type is a subexpression. */
OP (UNOP_CAST_TYPE)
/* The C++ dynamic_cast operator. */
OP (UNOP_DYNAMIC_CAST)
/* The C++ reinterpret_cast operator. */
OP (UNOP_REINTERPRET_CAST)
/* UNOP_MEMVAL is followed by a type pointer in the next exp_element
With another UNOP_MEMVAL at the end, this makes three exp_elements.
It casts the contents of the word addressed by the value of the
following subexpression. */
OP (UNOP_MEMVAL)
/* Like UNOP_MEMVAL, but the type is supplied as a subexpression. */
OP (UNOP_MEMVAL_TYPE)
/* UNOP_... operate on one value from a following subexpression
and replace it with a result. They take no immediate arguments. */
OP (UNOP_NEG) /* Unary - */
OP (UNOP_LOGICAL_NOT) /* Unary ! */
OP (UNOP_COMPLEMENT) /* Unary ~ */
OP (UNOP_IND) /* Unary * */
OP (UNOP_ADDR) /* Unary & */
OP (UNOP_PREINCREMENT) /* ++ before an expression */
OP (UNOP_POSTINCREMENT) /* ++ after an expression */
OP (UNOP_PREDECREMENT) /* -- before an expression */
OP (UNOP_POSTDECREMENT) /* -- after an expression */
OP (UNOP_SIZEOF) /* Unary sizeof (followed by expression) */
OP (UNOP_ALIGNOF) /* Unary alignof (followed by expression) */
OP (UNOP_PLUS) /* Unary plus */
OP (UNOP_ABS)
OP (UNOP_HIGH)
OP (OP_BOOL) /* Modula-2 builtin BOOLEAN type */
/* STRUCTOP_... operate on a value from a following subexpression
by extracting a structure component specified by a string
that appears in the following exp_elements (as many as needed).
STRUCTOP_STRUCT is used for "." and STRUCTOP_PTR for "->".
They differ only in the error message given in case the value is
not suitable or the structure component specified is not found.
The length of the string follows the opcode, followed by
BYTES_TO_EXP_ELEM(length) elements containing the data of the
string, followed by the length again and the opcode again. */
OP (STRUCTOP_STRUCT)
OP (STRUCTOP_PTR)
/* Anonymous field access, e.g. "foo.3". Used in Rust. */
OP (STRUCTOP_ANONYMOUS)
/* C++: OP_THIS is just a placeholder for the class instance variable.
It just comes in a tight (OP_THIS, OP_THIS) pair. */
OP (OP_THIS)
/* Objective C: "@selector" pseudo-operator. */
OP (OP_OBJC_SELECTOR)
/* OP_SCOPE surrounds a type name and a field name. The type
name is encoded as one element, but the field name stays as
a string, which, of course, is variable length. */
OP (OP_SCOPE)
/* OP_FUNC_STATIC_VAR refers to a function local static variable. The
function is taken from the following subexpression. The length of
the variable name as a string follows the opcode, followed by
BYTES_TO_EXP_ELEM(length) elements containing the data of the
string, followed by the length again and the opcode again.
Note this is used by C++, but not C. The C parser handles local
static variables in the parser directly. Also, this is only used
in C++ if the function/method name is not quoted, like e.g.:
p S:method()::var
p S:method() const::var
If the function/method is quoted like instead:
p 'S:method() const'::var
then the C-specific handling directly in the parser takes over (see
block/variable productions).
Also, if the whole function+var is quoted like this:
p 'S:method() const::var'
then the whole quoted expression is interpreted as a single symbol
name and we don't use OP_FUNC_STATIC_VAR either. In that case, the
C++-specific symbol lookup routines take care of the
function-local-static search. */
OP (OP_FUNC_STATIC_VAR)
/* OP_TYPE is for parsing types, and used with the "ptype" command
so we can look up types that are qualified by scope, either with
the GDB "::" operator, or the Modula-2 '.' operator. */
OP (OP_TYPE)
/* An Objective C Foundation Class NSString constant. */
OP (OP_OBJC_NSSTRING)
/* An array range operator (in Fortran 90, for "exp:exp", "exp:",
":exp" and ":"). */
OP (OP_RANGE)
/* OP_ADL_FUNC specifies that the function is to be looked up in an
Argument Dependent manner (Koenig lookup). */
OP (OP_ADL_FUNC)
/* The typeof operator. This has one expression argument, which is
evaluated solely for its type. */
OP (OP_TYPEOF)
/* The decltype operator. This has one expression argument, which is
evaluated solely for its type. This is similar to typeof, but has
slight different semantics. */
OP (OP_DECLTYPE)
/* The typeid operator. This has one expression argument. */
OP (OP_TYPEID)
/* This is used for the Rust [expr; N] form of array construction. It
takes two expression arguments. */
OP (OP_RUST_ARRAY)
/* ================ Ada operators ================ */
/* X IN A'RANGE(N). N is an immediate operand, surrounded by
BINOP_IN_BOUNDS before and after. A is an array, X an index
value. Evaluates to true iff X is within range of the Nth
dimension (1-based) of A. (A multi-dimensional array
type is represented as array of array of ...) */
OP (BINOP_IN_BOUNDS)
/* X IN L .. U. True iff L <= X <= U. */
OP (TERNOP_IN_RANGE)
/* Ada attributes ('Foo). */
OP (OP_ATR_FIRST)
OP (OP_ATR_LAST)
OP (OP_ATR_LENGTH)
OP (OP_ATR_POS)
OP (OP_ATR_SIZE)
OP (OP_ATR_TAG)
OP (OP_ATR_VAL)
/* Ada type qualification. It is encoded as for UNOP_CAST, above,
and denotes the TYPE'(EXPR) construct. */
OP (UNOP_QUAL)
/* X IN TYPE. The `TYPE' argument is immediate, with
UNOP_IN_RANGE before and after it. True iff X is a member of
type TYPE (typically a subrange). */
OP (UNOP_IN_RANGE)
/* An aggregate. A single immediate operand, N>0, gives
the number of component specifications that follow. The
immediate operand is followed by a second OP_AGGREGATE.
Next come N component specifications. A component
specification is either an OP_OTHERS (others=>...), an
OP_CHOICES (for named associations), or other expression (for
positional aggregates only). Aggregates currently
occur only as the right sides of assignments. */
OP (OP_AGGREGATE)
/* ================ Fortran operators ================ */
/* This is EXACTLY like OP_FUNCALL but is semantically different.
In F77, array subscript expressions, substring expressions and
function calls are all exactly the same syntactically. They
may only be disambiguated at runtime. Thus this operator,
which indicates that we have found something of the form
<name> ( <stuff> ). */
OP (OP_F77_UNDETERMINED_ARGLIST)
/* Single operand builtins. */
OP (UNOP_FORTRAN_KIND)
OP (UNOP_FORTRAN_FLOOR)
OP (UNOP_FORTRAN_CEILING)
OP (UNOP_FORTRAN_ALLOCATED)
OP (UNOP_FORTRAN_RANK)
OP (UNOP_FORTRAN_SHAPE)
OP (UNOP_FORTRAN_LOC)
/* Two operand builtins. */
OP (BINOP_FORTRAN_CMPLX)
OP (BINOP_FORTRAN_MODULO)
/* Builtins that take one or two operands. */
OP (FORTRAN_LBOUND)
OP (FORTRAN_UBOUND)
OP (FORTRAN_ASSOCIATED)
OP (FORTRAN_ARRAY_SIZE)
Reference in New Issue View Git Blame Copy Permalink