This solves an assertion failure when a remote provides a target
description which only refers to floating point registers by their
hardware name (e.g. f0), rather than their ABI name (e.g. ft0). GDB
assumed that should the floating point register feature be presented,
it would contain a register called ft0.
The floating point length is now instead determined by searching for
the same register, but looking for any of its aliases.
gdb/ChangeLog:
* riscv-tdep.c (riscv_gdbarch_init): Support determining flen from
target descriptions using exclusively floating point register name
aliases.
Some testsuite cases (gdb.cp/nsalias.exp for example) construct dwarf2
debug info for fake functions to test that this debug info is handled
correctly.
We currently get an error trying to read from an invalid address while
creating breakpoints for these fake functions.
Other targets allow creating breakpoints on invalid addresses, and
only error when GDB actually tries to insert the breakpoints.
In order to make RISC-V behave in the same way as other targets, this
commit makes the failure to read memory during breakpoint creation
non-fatal, we then expect to see a failure when GDB tries to insert
the breakpoint, just like other targets.
Tested with a riscv64-linux native testsuite run.
gdb/ChangeLog:
* riscv-tdep.c (riscv_breakpoint_kind_from_pc): Hanndle case where
code read might fail, assume 4-byte breakpoint in that case.
Make use of the type_align function and remove riscv_type_alignment as
it is no longer needed. I tested this against a number of RV32 and
RV64 targets, and I also ran the tests with an assertion in place
checking that the old riscv_type_alignment function gives the same
answer as the common type_align function - it does, and all the tests
still pass.
gdb/ChangeLog:
* riscv-tdep.c (riscv_type_align): New function.
(riscv_type_alignment): Delete.
(riscv_arg_location): Use 'type_align'.
(riscv_gdbarch_init): Register riscv_type_align gdbarch function.
This commit resolves a large number of failures in the test script
gdb.base/infcall-nested-structs.exp which were caused by GDB (for
RISC-V) incorrectly handling empty C++ structures when preparing
arguments for a dummy call, or collecting a return value.
The issue is further complicated in that there was a bug in GCC, such
that in some cases GCC would generate incorrect code when passing a
small structure that contained empty sub-structures. This was fixed
in GCC trunk on 5-March-2019, so in order to see the best results with
this patch you'll need a recent version of GCC.
Anything that used to work should continue to work after this patch,
regardless of GCC version being used.
The fix in this commit is that GDB now pays more attention to the
offset of fields within a structure when preparing arguments as in C++
an empty structure has a non-zero size, this is an example:
struct s1 { struct s2 { } empty; int f; };
We previously assumed that 'f' was at offset 0 inside type 's1',
however this is not the case in C++ as 's2' has size 1, and with
alignment 'f' is likely at some even bigger offset inside 's1'.
gdb/ChangeLog:
* riscv-tdep.c (riscv_call_arg_complex_float): Fix offset of first
component to 0.
(riscv_struct_info::riscv_struct_info): Initialise m_offsets
member.
(riscv_struct_info::analyse): New implementation using new
analyse_inner member function.
(riscv_struct_info::field_offset): New member function.
(riscv_struct_info::m_offsets): New member variable.
(riscv_struct_info::analyse_inner): New private member function,
takes the old implementation of riscv_struct_info::analyse but
extended to track field offsets.
(riscv_call_arg_struct): Update the struct folding special cases
to handle cases where empty C++ structs, which are non-zero
length, are found.
(riscv_arg_location): Initialise the length of each location, a
non-zero length now indicates the location is in use.
(riscv_push_dummy_call): Allow for the first location having a
non-zero offset when setting up arguments.
(riscv_return_value): Likewise, but for return values.
This rewrites gdb's TRY/CATCH to plain C++ try/catch. The patch was
largely written by script, though one change (to a comment in
common-exceptions.h) was reverted by hand.
gdb/ChangeLog
2019-04-08 Tom Tromey <tom@tromey.com>
* xml-support.c: Use C++ exception handling.
* x86-linux-nat.c: Use C++ exception handling.
* windows-nat.c: Use C++ exception handling.
* varobj.c: Use C++ exception handling.
* value.c: Use C++ exception handling.
* valprint.c: Use C++ exception handling.
* valops.c: Use C++ exception handling.
* unittests/parse-connection-spec-selftests.c: Use C++ exception
handling.
* unittests/cli-utils-selftests.c: Use C++ exception handling.
* typeprint.c: Use C++ exception handling.
* tui/tui.c: Use C++ exception handling.
* tracefile-tfile.c: Use C++ exception handling.
* top.c: Use C++ exception handling.
* thread.c: Use C++ exception handling.
* target.c: Use C++ exception handling.
* symmisc.c: Use C++ exception handling.
* symfile-mem.c: Use C++ exception handling.
* stack.c: Use C++ exception handling.
* sparc64-linux-tdep.c: Use C++ exception handling.
* solib.c: Use C++ exception handling.
* solib-svr4.c: Use C++ exception handling.
* solib-spu.c: Use C++ exception handling.
* solib-frv.c: Use C++ exception handling.
* solib-dsbt.c: Use C++ exception handling.
* selftest-arch.c: Use C++ exception handling.
* s390-tdep.c: Use C++ exception handling.
* rust-lang.c: Use C++ exception handling.
* rust-exp.y: Use C++ exception handling.
* rs6000-tdep.c: Use C++ exception handling.
* rs6000-aix-tdep.c: Use C++ exception handling.
* riscv-tdep.c: Use C++ exception handling.
* remote.c: Use C++ exception handling.
* remote-fileio.c: Use C++ exception handling.
* record-full.c: Use C++ exception handling.
* record-btrace.c: Use C++ exception handling.
* python/python.c: Use C++ exception handling.
* python/py-value.c: Use C++ exception handling.
* python/py-utils.c: Use C++ exception handling.
* python/py-unwind.c: Use C++ exception handling.
* python/py-type.c: Use C++ exception handling.
* python/py-symbol.c: Use C++ exception handling.
* python/py-record.c: Use C++ exception handling.
* python/py-record-btrace.c: Use C++ exception handling.
* python/py-progspace.c: Use C++ exception handling.
* python/py-prettyprint.c: Use C++ exception handling.
* python/py-param.c: Use C++ exception handling.
* python/py-objfile.c: Use C++ exception handling.
* python/py-linetable.c: Use C++ exception handling.
* python/py-lazy-string.c: Use C++ exception handling.
* python/py-infthread.c: Use C++ exception handling.
* python/py-inferior.c: Use C++ exception handling.
* python/py-gdb-readline.c: Use C++ exception handling.
* python/py-framefilter.c: Use C++ exception handling.
* python/py-frame.c: Use C++ exception handling.
* python/py-finishbreakpoint.c: Use C++ exception handling.
* python/py-cmd.c: Use C++ exception handling.
* python/py-breakpoint.c: Use C++ exception handling.
* python/py-arch.c: Use C++ exception handling.
* printcmd.c: Use C++ exception handling.
* ppc-linux-tdep.c: Use C++ exception handling.
* parse.c: Use C++ exception handling.
* p-valprint.c: Use C++ exception handling.
* objc-lang.c: Use C++ exception handling.
* mi/mi-main.c: Use C++ exception handling.
* mi/mi-interp.c: Use C++ exception handling.
* mi/mi-cmd-stack.c: Use C++ exception handling.
* mi/mi-cmd-break.c: Use C++ exception handling.
* main.c: Use C++ exception handling.
* linux-thread-db.c: Use C++ exception handling.
* linux-tdep.c: Use C++ exception handling.
* linux-nat.c: Use C++ exception handling.
* linux-fork.c: Use C++ exception handling.
* linespec.c: Use C++ exception handling.
* language.c: Use C++ exception handling.
* jit.c: Use C++ exception handling.
* infrun.c: Use C++ exception handling.
* infcmd.c: Use C++ exception handling.
* infcall.c: Use C++ exception handling.
* inf-loop.c: Use C++ exception handling.
* i386-tdep.c: Use C++ exception handling.
* i386-linux-tdep.c: Use C++ exception handling.
* guile/scm-value.c: Use C++ exception handling.
* guile/scm-type.c: Use C++ exception handling.
* guile/scm-symtab.c: Use C++ exception handling.
* guile/scm-symbol.c: Use C++ exception handling.
* guile/scm-pretty-print.c: Use C++ exception handling.
* guile/scm-ports.c: Use C++ exception handling.
* guile/scm-param.c: Use C++ exception handling.
* guile/scm-math.c: Use C++ exception handling.
* guile/scm-lazy-string.c: Use C++ exception handling.
* guile/scm-frame.c: Use C++ exception handling.
* guile/scm-disasm.c: Use C++ exception handling.
* guile/scm-cmd.c: Use C++ exception handling.
* guile/scm-breakpoint.c: Use C++ exception handling.
* guile/scm-block.c: Use C++ exception handling.
* guile/guile-internal.h: Use C++ exception handling.
* gnu-v3-abi.c: Use C++ exception handling.
* gdbtypes.c: Use C++ exception handling.
* frame.c: Use C++ exception handling.
* frame-unwind.c: Use C++ exception handling.
* fbsd-tdep.c: Use C++ exception handling.
* f-valprint.c: Use C++ exception handling.
* exec.c: Use C++ exception handling.
* event-top.c: Use C++ exception handling.
* event-loop.c: Use C++ exception handling.
* eval.c: Use C++ exception handling.
* dwarf2read.c: Use C++ exception handling.
* dwarf2loc.c: Use C++ exception handling.
* dwarf2-frame.c: Use C++ exception handling.
* dwarf2-frame-tailcall.c: Use C++ exception handling.
* dwarf-index-write.c: Use C++ exception handling.
* dwarf-index-cache.c: Use C++ exception handling.
* dtrace-probe.c: Use C++ exception handling.
* disasm-selftests.c: Use C++ exception handling.
* darwin-nat.c: Use C++ exception handling.
* cp-valprint.c: Use C++ exception handling.
* cp-support.c: Use C++ exception handling.
* cp-abi.c: Use C++ exception handling.
* corelow.c: Use C++ exception handling.
* completer.c: Use C++ exception handling.
* compile/compile-object-run.c: Use C++ exception handling.
* compile/compile-object-load.c: Use C++ exception handling.
* compile/compile-cplus-symbols.c: Use C++ exception handling.
* compile/compile-c-symbols.c: Use C++ exception handling.
* common/selftest.c: Use C++ exception handling.
* common/new-op.c: Use C++ exception handling.
* cli/cli-script.c: Use C++ exception handling.
* cli/cli-interp.c: Use C++ exception handling.
* cli/cli-cmds.c: Use C++ exception handling.
* c-varobj.c: Use C++ exception handling.
* btrace.c: Use C++ exception handling.
* breakpoint.c: Use C++ exception handling.
* break-catch-throw.c: Use C++ exception handling.
* arch-utils.c: Use C++ exception handling.
* amd64-tdep.c: Use C++ exception handling.
* ada-valprint.c: Use C++ exception handling.
* ada-typeprint.c: Use C++ exception handling.
* ada-lang.c: Use C++ exception handling.
* aarch64-tdep.c: Use C++ exception handling.
gdb/gdbserver/ChangeLog
2019-04-08 Tom Tromey <tom@tromey.com>
* server.c: Use C++ exception handling.
* linux-low.c: Use C++ exception handling.
* gdbreplay.c: Use C++ exception handling.
This series is revisit of Siddhesh Poyarekar's patch from back in
2012. The last status on the patch is in the following gdb-patches
thread:
https://sourceware.org/ml/gdb-patches/2012-08/msg00562.html
It appears that Tom approved the patch, but Jan had some issues
with a compiler error that made the test fail on -m32 test runs.
He wrote up a hand-tweaked .S file to deal with it. Siddesh said
he would update tests. Then nothing.
Siddesh and Jan have both moved on since.
The patch originally required a large precursor patch to work.
I have whittled this down to/rewritten the bare minimum, and this
first patch is the result, changing the type of TYPE_LENGTH
to ULONGEST from unsigned int.
The majority of the changes involve changing printf format
strings to use %s and pulongest instead of %d.
gdb/ChangeLog:
* ada-lang.c (ada_template_to_fixed_record_type_1): Use
%s/pulongest for TYPE_LENGTH instead of %d in format
strings.
* ada-typerint.c (ada_print_type): Likewise.
* amd64-windows-tdep.c (amd64_windows_store_arg_in_reg): Likewise.
* compile/compile-c-support.c (generate_register_struct): Likewise.
* gdbtypes.c (recursive_dump_type): Likewise.
* gdbtypes.h (struct type) <length>: Change type to ULONGEST.
* m2-typeprint.c (m2_array): Use %s/pulongest for TYPE_LENGTH
instead of %d in format strings.
* riscv-tdep.c (riscv_type_alignment): Cast second argument
to std::min to ULONGEST.
* symmisc.c (print_symbol): Use %s/pulongest for TYPE_LENGTH
instead of %d in format strings.
* tracepoint.c (info_scope_command): Likewise.
* typeprint.c (print_offset_data::update)
(print_offset_data::finish): Likewise.
* xtensa-tdep.c (xtensa_store_return_value)
(xtensa_push_dummy_call): Likewise.
Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.
This change has been tested with no regressions.
gdb/ChangeLog:
* gdb/riscv-tdep.c (riscv_dummy_id): Delete.
(riscv_unwind_pc): Delete.
(riscv_unwind_sp): Delete.
(riscv_gdbarch_init): Don't register deleted functions with
gdbarch.
Using the gdb.ada/call_pn.exp testcase, and running it by hand on
riscv64-elf, we get the following error:
(gdb) call pn(55)
Could not compute alignment of type
The problem occurs because the parameter's type is a TYPE_CODE_RANGE,
and that type code is not handled by riscv_type_alignment. So this patch
fixes the issue by handling TYPE_CODE_RANGE the same way we handle other
integral types.
gdb/ChangeLog:
* riscv-rdep.c (riscv_type_alignment): Handle TYPE_CODE_RANGE.
Tested on riscv64-elf using AdaCore's testsuite.
s0 is listed as both an int register name and an FP register name. The FP reg
name is wrong. This looks like a simple editting error, and has an easy fix.
Tested with riscv64-linux build and check, with no regressions.
gdb/
* riscv-tdep.c (riscv_freg_feature): Drop s0 name from f8.
The goal of this commit is to allow RV64 binaries compiled for the 'F'
extension to run on a target that supports both the 'F' and 'D'
extensions.
The 'D' extension depends on the 'F' extension and chapter 9 of the
RISC-V ISA manual implies that running a program compiled for 'F' on
a 'D' target should be fine.
To support this the gdbarch now holds two feature sets, one represents
the features that are present on the target, and one represents the
features requested in the ELF flags.
The existing error checks are relaxed slightly to allow binaries
compiled for 32-bit 'F' extension to run on targets with the 64-bit
'D' extension.
A new set of functions called riscv_abi_{xlen,flen} are added to
compliment the existing riscv_isa_{xlen,flen}, and some callers to the
isa functions now call the abi functions when that is appropriate.
In riscv_call_arg_struct two asserts are removed, these asserts no
longer make sense. The asserts were both like this:
gdb_assert (TYPE_LENGTH (ainfo->type)
<= (cinfo->flen + cinfo->xlen));
And were made in two cases, when passing structures like these:
struct {
integer field1;
float field2;
};
or,
struct {
float field1;
integer field2;
};
When running on an RV64 target which only has 32-bit float then the
integer field could be 64-bits, while if the float field is 32-bits
the overall size of the structure can be 128-bits (with 32-bits of
padding). In this case the assertion would fail, however, the code
isn't incorrect, so its safe to just remove the assertion.
This was tested by running on an RV64IMFDC target using a compiler
configured for RV64IMFC, and comparing the results with those obtained
when using a compiler configured for RV64IMFDC. The only regressions
I see (now) are in gdb.base/store.exp and are related too different
code generation choices GCC makes between the two targets.
Finally, this commit does not make any attempt to support running
binaries compiled for RV32 on an RV64 target, though nothing in here
should prevent that being supported in the future.
gdb/ChangeLog:
* arch/riscv.h (struct riscv_gdbarch_features) <hw_float_abi>:
Delete.
<operator==>: Update with for removed field.
<hash>: Likewise.
* riscv-tdep.h (struct gdbarch_tdep) <features>: Renamed to...
<isa_features>: ...this.
<abi_features>: New field.
(riscv_isa_flen): Update comment.
(riscv_abi_xlen): New declaration.
(riscv_abi_flen): New declaration.
* riscv-tdep.c (riscv_isa_xlen): Update to get answer from
isa_features.
(riscv_abi_xlen): New function.
(riscv_isa_flen): Update to get answer from isa_features.
(riscv_abi_flen): New function.
(riscv_has_fp_abi): Update to get answer from abi_features.
(riscv_call_info::riscv_call_info): Use abi xlen and flen, not isa
xlen and flen.
(riscv_call_info) <xlen, flen>: Update comment.
(riscv_call_arg_struct): Remove invalid assertions
(riscv_features_from_gdbarch_info): Update now hw_float_abi field
is removed.
(riscv_gdbarch_init): Gather isa features and abi features
separately, ensure both match on the gdbarch when reusing an old
gdbarch. Relax an error check to allow 32-bit abi float to run on
a target with 64-bit float hardware.
This commit applies all changes made after running the gdb/copyright.py
script.
Note that one file was flagged by the script, due to an invalid
copyright header
(gdb/unittests/basic_string_view/element_access/char/empty.cc).
As the file was copied from GCC's libstdc++-v3 testsuite, this commit
leaves this file untouched for the time being; a patch to fix the header
was sent to gcc-patches first.
gdb/ChangeLog:
Update copyright year range in all GDB files.
The existing code for reading and writing the return value can
overflow the passed in buffers in a couple of situations. This commit
aims to resolve these issues.
The problems were detected using valgrind, here are two examples,
first from gdb.base/structs.exp:
(gdb) p/x fun9()
==31353== Invalid write of size 8
==31353== at 0x4C34153: memmove (vg_replace_strmem.c:1270)
==31353== by 0x632EBB: memcpy (string_fortified.h:34)
==31353== by 0x632EBB: readable_regcache::raw_read(int, unsigned char*) (regcache.c:538)
==31353== by 0x659D3F: riscv_return_value(gdbarch*, value*, type*, regcache*, unsigned char*, unsigned char const*) (riscv-tdep.c:2593)
==31353== by 0x583641: get_call_return_value (infcall.c:448)
==31353== by 0x583641: call_thread_fsm_should_stop(thread_fsm*, thread_info*) (infcall.c:546)
==31353== by 0x59BBEC: fetch_inferior_event(void*) (infrun.c:3883)
==31353== by 0x53890B: check_async_event_handlers (event-loop.c:1064)
==31353== by 0x53890B: gdb_do_one_event() [clone .part.4] (event-loop.c:326)
==31353== by 0x6CA34B: wait_sync_command_done() (top.c:503)
==31353== by 0x584653: run_inferior_call (infcall.c:621)
...
And from gdb.base/call-sc.exp:
(gdb) advance fun
fun () at /gdb/gdb/testsuite/gdb.base/call-sc.c:41
41 return foo;
(gdb) finish
==1968== Invalid write of size 8
==1968== at 0x4C34153: memmove (vg_replace_strmem.c:1270)
==1968== by 0x632EBB: memcpy (string_fortified.h:34)
==1968== by 0x632EBB: readable_regcache::raw_read(int, unsigned char*) (regcache.c:538)
==1968== by 0x659D01: riscv_return_value(gdbarch*, value*, type*, regcache*, unsigned char*, unsigned char const*) (riscv-tdep.c:2576)
==1968== by 0x5891E4: get_return_value(value*, type*) (infcmd.c:1640)
==1968== by 0x5892C4: finish_command_fsm_should_stop(thread_fsm*, thread_info*) (infcmd.c:1808)
==1968== by 0x59BBEC: fetch_inferior_event(void*) (infrun.c:3883)
==1968== by 0x53890B: check_async_event_handlers (event-loop.c:1064)
==1968== by 0x53890B: gdb_do_one_event() [clone .part.4] (event-loop.c:326)
==1968== by 0x6CA34B: wait_sync_command_done() (top.c:503)
...
There are a couple of problems with the existing code, that are all
related.
In riscv_call_arg_struct we incorrectly rounded up the size of a
structure argument. This is unnecessary, and caused GDB to read too
much data into the output buffer when extracting a struct return
value.
In fixing this it became clear that we were incorrectly assuming that
any value being placed in a register (or read from a register) would
always access the entire register. This is not true, for example a
9-byte struct on a 64-bit target places 8-bytes in one registers and
1-byte in a second register (assuming available registers). To handle
this I switch from using cooked_read to cooked_read_part.
Finally, when processing basic integer return value types these are
extended to xlen sized types and then passed in registers. We
currently don't handle this type expansion in riscv_return_value, but
we do in riscv_push_dummy_call. The result is that small integer
types (like char) result in a full xlen sized register being written
into the output buffer, which results in buffer overflow. To address
this issue we now create a value of the expanded type and use this
values contents buffer to hold the return value before casting the
value down to the smaller expected type.
This patch resolves all of the valgrind issues I have found so far,
and causes no regressions. Tested against RV32/64 with and without
floating point support.
gdb/ChangeLog:
* riscv-tdep.c (riscv_call_arg_struct): Don't adjust size before
assigning locations.
(riscv_return_value): Take more care not to read/write outside of
argument buffer. Cast return value between the declared type and
the abi type.
We should save and restore the floating point status registers. This
became an issue when testing 32-bit float on a target with 64-bit with
the gdb.base/callfuncs.exp test.
gdb/ChangeLog:
* riscv-tdep.c (riscv_register_reggroup_p): Save and restore fcsr,
fflags, and frm registers.
Provide a mapping between GDB's register numbers and DWARF's register
numbers. This resolves some failures that I was seeing on
gdb.base/store.exp when running on an rv64imfdc target.
gdb/ChangeLog:
* riscv-tdep.c (riscv_dwarf_reg_to_regnum): New function.
(riscv_gdbarch_init): Register new function with gdbarch.
* riscv-tdep.h: New enum to define RISC-V DWARF register numbers.
Avoid compiler errors caused by trying to print CORE_ADDR using '%ld'
format, instead convert to a string and print that instead.
gdb/ChangeLog:
* riscv-tdep.c (riscv_scan_prologue): Use plongest to format
a signed offset as a string.
* riscv-tdep.c (riscv_print_one_register_info): For MSTATUS, add
comment for SD field, and correct xlen calculation. For MISA, add
comment for MXL field, add call to register_size, and correct base
calculation.
This commit fixes some test failures in gdb.base/varargs.exp when
running on targets with floating point hardware. Floating point
unnamed (variadic) arguments should be passed in integer registers
according to the abi.
After this commit I see no failures in gdb.base/varargs.exp on 32 or
64 bit targets with floating point hardware.
gdb/ChangeLog:
* riscv-tdep.c (riscv_call_arg_scalar_float): Unnamed (variadic)
arguments are passed in integer registers.
(riscv_call_arg_complex_float): Likewise.
This fixes an ARI warning in riscv-tdep.c that whitespace before a
gdb/ChangeLog:
* riscv-tdep.c (riscv_register_name): Fix ARI warning by removing
leading whitespace before #include line.
Currently, if the target announces that it has floating point
registers in its target description then GDB assumes that the hardware
float ABI should be used. However, there's nothing stopping a user
compiling a program for the soft-float abi, and then trying to run
this on a target with hardware floating point registers.
This commit adjusts the logic that decides if GDB should use the
hardware float abi. The primary decision now is based on what the ELF
currently being executed says in its headers. If the file was
compiled for h/w float abi, then GDB uses h/w float abi, otherwise s/w
float is used.
If the current BFD is not an ELF then we don't currently have a
mechanism for figuring out if the file was compiled for float or not.
In this case we disable the h/w float abi. This shouldn't be a
problem as, right now, the RISC-V linker can only produce ELFs.
If there is NO current BFD (can this happen?) then we will enable h/w
float abi if the target has floating point hardware, otherwise, s/w
float abi is used.
This commit also adds some sanity checking that the features requested
in the BFD (xlen and flen) match the target description.
For testing I ran the testsuite on a target that returns a target
description containing both integer and floating point registers, but
used a compiler that didn't have floating point support. Before this
commit I would see failures on may tests that made inferior calls
using floating point arguments, after this commit, all of these issues
are resolved. One example from the testsuite is
gdb.base/infcall-nested-structs.exp.
gdb/ChangeLog:
* riscv-tdep.c (riscv_features_from_gdbarch_info): New function.
(riscv_find_default_target_description): Use new function to
extract feature from gdbarch_info.
(riscv_gdbarch_init): Add error checks for xlen and flen between
target description and bfd headers. Be smarter about when we
think the hardware floating point abi should be used.
Add '==' and '!=' operators for the struct riscv_gdb_features,
allowing a small simplification.
gdb/ChangeLog:
* arch/riscv.h (riscv_gdb_features::operator==): New.
(riscv_gdb_features::operator!=): New.
* riscv-tdep.c (riscv_gdbarch_init): Make use of the inequality
operator.
Makes more of the interface related to fetching target descriptions
constant.
gdb/ChangeLog:
* arch/riscv.h (riscv_create_target_description): Make return type
const.
* arch/riscv.c (riscv_create_target_description): Likewise.
* riscv-tdep.c (riscv_find_default_target_description): Likewise.
This commit adds target description support for riscv.
I've used the split feature approach for specifying the architectural
features, and the CSR feature is auto-generated from the riscv-opc.h
header file.
If the target doesn't provide a suitable target description then GDB
will build one by looking at the bfd headers.
This commit does not implement target description creation for the
Linux or FreeBSD native targets, both of these will need to add
read_description methods into their respective target classes, which
probe the target features, and then call
riscv_create_target_description to build a suitable target
description. Until this is done Linux and FreeBSD will get the same
default target description based on the bfd that bare-metal targets
get.
I've only added feature descriptions for 32 and 64 bit registers, 128
bit registers (for RISC-V) are not supported in the reset of GDB yet.
This commit removes the special reading of the MISA register in order
to establish the target features, this was only used for figuring out
the f-register size, and even that wasn't done consistently. We now
rely on the target to tell us what size of registers it has (or look
in the BFD as a last resort). The result of this is that we should
now support RV64 targets with 32-bit float, though I have not
extensively tested this combination yet.
* Makefile.in (ALL_TARGET_OBS): Add arch/riscv.o.
(HFILES_NO_SRCDIR): Add arch/riscv.h.
* arch/riscv.c: New file.
* arch/riscv.h: New file.
* configure.tgt: Add cpu_obs list of riscv, move riscv-tdep.o into
this list, and add arch/riscv.o.
* features/Makefile: Add riscv features.
* features/riscv/32bit-cpu.c: New file.
* features/riscv/32bit-cpu.xml: New file.
* features/riscv/32bit-csr.c: New file.
* features/riscv/32bit-csr.xml: New file.
* features/riscv/32bit-fpu.c: New file.
* features/riscv/32bit-fpu.xml: New file.
* features/riscv/64bit-cpu.c: New file.
* features/riscv/64bit-cpu.xml: New file.
* features/riscv/64bit-csr.c: New file.
* features/riscv/64bit-csr.xml: New file.
* features/riscv/64bit-fpu.c: New file.
* features/riscv/64bit-fpu.xml: New file.
* features/riscv/rebuild-csr-xml.sh: New file.
* riscv-tdep.c: Add 'arch/riscv.h' include.
(riscv_gdb_reg_names): Delete.
(csr_reggroup): New global.
(struct riscv_register_alias): Delete.
(struct riscv_register_feature): New structure.
(riscv_register_aliases): Delete.
(riscv_xreg_feature): New global.
(riscv_freg_feature): New global.
(riscv_virtual_feature): New global.
(riscv_csr_feature): New global.
(riscv_create_csr_aliases): New function.
(riscv_read_misa_reg): Delete.
(riscv_has_feature): Delete.
(riscv_isa_xlen): Simplify, just return cached xlen.
(riscv_isa_flen): Simplify, just return cached flen.
(riscv_has_fp_abi): Update for changes in struct gdbarch_tdep.
(riscv_register_name): Update to make use of tdesc_register_name.
Look up xreg and freg names in the new globals riscv_xreg_feature
and riscv_freg_feature. Don't supply csr aliases here.
(riscv_fpreg_q_type): Delete.
(riscv_register_type): Use tdesc_register_type in almost all
cases, override the returned type in a few specific cases only.
(riscv_print_one_register_info): Handle errors reading registers.
(riscv_register_reggroup_p): Use tdesc_register_in_reggroup_p for
registers that are otherwise unknown to GDB. Also check the
csr_reggroup.
(riscv_print_registers_info): Remove assert about upper register
number, and use gdbarch_register_reggroup_p instead of
short-cutting.
(riscv_find_default_target_description): New function.
(riscv_check_tdesc_feature): New function.
(riscv_add_reggroups): New function.
(riscv_setup_register_aliases): New function.
(riscv_init_reggroups): New function.
(_initialize_riscv_tdep): Add calls to setup CSR aliases, and
setup register groups. Register new riscv debug variable.
* riscv-tdep.h: Add 'arch/riscv.h' include.
(struct gdbarch_tdep): Remove abi union, and add
riscv_gdbarch_features field. Remove cached quad floating point
type, and provide initialisation for double type field.
* target-descriptions.c (maint_print_c_tdesc_cmd): Add riscv to
the list of targets using the feature based target descriptions.
* NEWS: Mention target description support.
gdb/doc/ChangeLog:
* gdb.texinfo (Standard Target Features): Add RISC-V Features
sub-section.
For riscv64-linux target, second half of fix for
FAIL: gdb.base/gnu_vector.exp: call add_various_floatvecs
Unnamed arguments with 2*XLEN alignment are passed in aligned register pairs.
gdb/
* riscv-tdep.c (struct riscv_arg_info): New field is_unnamed.
(riscv_call_arg_scalar_int): If unnamed arg with twice xlen alignment,
then increment next_regnum if odd.
(riscv_arg_location): New arg is_unnamed. Set ainfo->is_unnamed.
(riscv_push_dummy_call): New local ftype. Call check_typedef to set
function type. Pass new arg to riscv_arg_location based on function
type.
(riscv_return_value): Pass new arg to riscv_arg_location.
For riscv64-linux target, first half of fix for
FAIL: gdb.base/gnu_vector.exp: call add_various_floatvecs
GCC gives vectors natural aligment based on total size, not element size,
bounded by the maximum supported type alignment.
gdb/
* riscv-tdep.c (BIGGEST_ALIGNMENT): New.
(riscv_type_alignment) <TYPE_CODE_ARRAY>: If TYPE_VECTOR, return min
of TYPE_LENGTH and BIGGEST_ALIGNMENT.
For riscv64-linux target, fixes
FAIL: gdb.base/gnu_vector.exp: call add_many_charvecs
Ensure that stack slots are always the same alignment as XLEN by rounding
up arg align to xlen.
gdb/
* riscv-tdep.c (riscv_call_arg_scalar_int): Use std::min when
setting len. New local align, set to max of arg align and xlen,
and pass to first riscv_assign_stack_location call.
When we connect to a remote target one of the first things GDB does is
establish a frame id. If an error is thrown while building this frame
id then GDB will disconnect from the target.
This can mean that, if the user is attempting to connect to a target
that doesn't yet have a program loaded, or the program the user is
going to load onto the target doesn't match what is already loaded, or
the target is just in some undefined state, then the very first
request for a frame id can fail (for example, by trying to load from
an invalid memory address), and GDB will disconnect. It is then
impossible for the user to connect to the target and load a new
program at all.
An example of such a session might look like this:
Reading symbols from ./gdb/testsuite/outputs/gdb.arch/riscv-reg-aliases/riscv-reg-aliases...
(gdb) target remote :37191
Remote debugging using :37191
0x0000000000000100 in ?? ()
Cannot access memory at address 0x0
(gdb) load
You can't do that when your target is `exec'
(gdb) info frame
/path/to/gdb/gdb/thread.c:93: internal-error: thread_info* inferior_thread(): Assertion `tp' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
The solution is to handle errors in riscv_frame_this_id, and leave the
this_id variable with its default value, which is the predefined
'outermost' frame.
With this fix in place, connecting to the same target now looks like
this:
(gdb) target remote :37191
Remote debugging using :37191
0x0000000000000100 in ?? ()
(gdb) info frame
Stack level 0, frame at 0x0:
pc = 0x100; saved pc = <not saved>
Outermost frame: outermost
Arglist at unknown address.
Locals at unknown address, Previous frame's sp in sp
gdb/ChangeLog:
* riscv-tdep.c (riscv_insn::decode): Update header comment.
(riscv_frame_this_id): Catch errors thrown while building the
frame cache, leave the frame id as the default, which is the outer
frame id.
Some hardware doesn't support unaligned accesses, and a bare metal target
may not have an unaligned access trap handler. So if the PC is 2-byte
aligned, then use a 2-byte breakpoint to avoid unaligned accesses.
Tested on native RV64GC Linux with gdb testsuite and cross on spike
simulator and openocd with riscv-tests/debug.
gdb/
* riscv-tdep.c (riscv_breakpoint_kind_from_pc): New local unaligned_p.
Set if pcptr if unaligned. Return 2 if unaligned_p true. Update
debugging messages.
Make riscv_isa_flen available to the linux native code, and clean up duplicate
comments.
gdb/
* riscv-tdep.c (riscv_isa_xlen): Refer to riscv-tdep.h comment.
(riscv_isa_flen): Likewise. Drop static.
* riscv-tdep.h (riscv_isa_xlen): Move riscv-tdep.c comment to here.
(riscv_isa_flen): Likewise.
When reading the MISA register, the RISC-V specification says that, if
MISA can't be found then a default value of 0 should be assumed.
As such, this patch ensures that GDB ignores errors when accessing
both the new and old locations for the MISA register.
Additionally, this patch removes an unneeded flag parameter which
didn't provide any additional functionality beyond checking the MISA
for the default value of 0.
gdb/ChangeLog:
* riscv-tdep.c (riscv_read_misa_reg): Update comment, remove
READ_P parameter, catch and ignore register access errors from
either the old or new MISA location.
(riscv_has_feature): Update call to riscv_read_misa_reg.
The recent commit:
commit 0dbfcfffe9
Date: Tue Oct 16 22:40:09 2018 +0100
gdb/riscv: Fix register access for register aliases
broke the CSR names for RISC-V, now all of the CSRs have names like,
csr0, csr1, csr2, etc. This commit restores the previous
user-friendly names.
gdb/ChangeLog:
* riscv-tdep.c (riscv_register_name): Use the user-friendly names
for CSRs.
When using QEMU as a RISCV simulator, hardware watchpoint events are
reported to GDB before the target memory gets written. GDB currently
expects the event to be reported after it is written. As a result of
this mismatch, upon receiving the event, GDB sees that the target
memory region has not changed, and therefore decides to ignore the
event. It therefore resumes the program's execution with a continue,
which is the start of an infinite loop between QEMU repeatedly
reporting the same watchpoint event over and over, and GDB repeatedly
ignoring it.
This patch fixes the issue by telling GDB to expect the watchpoint
event to be reported ahead of the memory region being modified.
Upon receiving the event, GDB then single-steps the program before
checking the watched memory value.
gdb/ChangeLog:
* riscv-tdep.c (riscv_gdbarch_init): Set the gdbarch's
have_nonsteppable_watchpoint attribute to 1.
Some confusion over how the register names and aliases are setup in
riscv means that we currently can't access registers through their
architectural name.
This commit fixes this issue, and moves some of the csr register
handling out of the alias handling code and deals with it separately.
This has the benefit that we can now directly access some arrays
rather than having to iterate over them.
A new test is added to ensure that register aliases now work
correctly.
gdb/ChangeLog:
* riscv-tdep.c (riscv_gdb_reg_names): Update comment, and all
register names.
(struct register_alias): Rename to...
(struct riscv_register_alias): ...this, and update comment.
(riscv_register_aliases): Update type, and alias names. Remove
CSR names from this list.
(riscv_register_name): Use riscv_gdb_reg_names for int and float
register names. Add an extra assertion.
(riscv_is_regnum_a_named_csr): New function.
(riscv_register_reggroup_p): Use riscv_is_regnum_a_named_csr.
gdb/testsuite/ChangeLog:
* gdb.arch/riscv-reg-aliases.c: New file.
* gdb.arch/riscv-reg-aliases.exp: New file.
The hardware requires that values in FP registers be NaN-boxed, so we must
extend them with 1's instead of 0's as we do for integer values.
gdb/
* riscv-tdep.c (riscv_push_dummy_call) <in_reg>: Check for value in
FP reg smaller than FP reg size, and fill with -1 instead of 0.
A 64-bit FP register can hold either a single or double float value, so
print it as both types by using a union type for FP registers. Likewise
for 128-bit regs which can also hold long double.
gdb/
* riscv-tdep.c (riscv_fpreg_d_type, riscv_fpreg_q_type): New.
(riscv_register_type): Use them.
(riscv_print_one_register_info): Handle union of floats same as float.
* riscv-tdep.h (struct gdbarch_tdep): Add riscv_fpreg_d_type and
riscv_fpreg_q_type fields.
This fixes all the straightforward -Wshadow=local warnings in gdb. A
few standard approaches are used here:
* Renaming an inner (or outer, but more commonly inner) variable;
* Lowering a declaration to avoid a clash;
* Moving a declaration into a more inner scope to avoid a clash,
including the special case of moving a declaration into a loop header.
I did not consider any of the changes in this patch to be particularly
noteworthy, though of course they should all still be examined.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* ctf.c (SET_ARRAY_FIELD): Rename "u32".
* p-valprint.c (pascal_val_print): Split inner "i" variable.
* xtensa-tdep.c (xtensa_push_dummy_call): Declare "i" in loop
header.
* xstormy16-tdep.c (xstormy16_push_dummy_call): Declare "val" in
more inner scope.
* xcoffread.c (read_xcoff_symtab): Rename inner "symbol".
* varobj.c (varobj_update): Rename inner "newobj",
"type_changed".
* valprint.c (generic_emit_char): Rename inner "buf".
* valops.c (find_overload_match): Rename inner "temp".
(value_struct_elt_for_reference): Declare "v" in more inner
scope.
* v850-tdep.c (v850_push_dummy_call): Rename "len".
* unittests/array-view-selftests.c (run_tests): Rename inner
"vec".
* tui/tui-stack.c (tui_show_frame_info): Declare "i" in loop
header.
* tracepoint.c (merge_uploaded_trace_state_variables): Declare
"tsv" in more inner scope.
(print_one_static_tracepoint_marker): Rename inner
"tuple_emitter".
* tic6x-tdep.c (tic6x_analyze_prologue): Declare "inst" lower.
(tic6x_push_dummy_call): Don't redeclare "addr".
* target-float.c: Declare "dto" lower.
* symtab.c (lookup_local_symbol): Rename inner "sym".
(find_pc_sect_line): Rename inner "pc".
* stack.c (print_frame): Don't redeclare "gdbarch".
(return_command): Rename inner "gdbarch".
* s390-tdep.c (s390_prologue_frame_unwind_cache): Renam inner
"sp".
* rust-lang.c (rust_internal_print_type): Declare "i" in loop
header.
* rs6000-tdep.c (ppc_process_record): Rename inner "addr".
* riscv-tdep.c (riscv_push_dummy_call): Declare "info" in inner
scope.
* remote.c (remote_target::update_thread_list): Don't redeclare
"tp".
(remote_target::process_initial_stop_replies): Rename inner
"thread".
(remote_target::remote_parse_stop_reply): Don't redeclare "p".
(remote_target::wait_as): Don't redeclare "stop_reply".
(remote_target::get_thread_local_address): Rename inner
"result".
(remote_target::get_tib_address): Likewise.
RISC-V supports instructions of varying lengths. Standard existing
instructions in the base ISA are 4 bytes in length, but the 'C'
extension adds support for compressed, 2 byte instructions. RISC-V
supports two different breakpoint instructions: EBREAK is a 4 byte
instruction in the base ISA, and C.EBREAK is a 2 byte instruction only
available on processors implementing the 'C' extension. Using EBREAK
to set breakpoints on compressed instructions causes problems as the
second half of EBREAK will overwrite the first 2 bytes of the
following instruction breaking other threads in the process if their
PC is the following instruction. Thus, breakpoints on compressed
instructions need to use C.EBREAK instead of EBREAK.
Previously, the riscv architecture checked the MISA register to
determine if the 'C' extension was available. If so, it used C.EBREAK
for all breakpoints. However, the MISA register is not necessarily
available to supervisor mode operating systems. While native targets
could provide a fake MISA register value, this patch instead examines
the existing instruction at a breakpoint target to determine which
breakpoint instruction to use. If the existing instruction is a
compressed instruction, C.EBREAK is used, otherwise EBREAK is used.
gdb/ChangeLog:
* disasm-selftests.c (print_one_insn_test): Add bfd_arch_riscv to
case with explicit breakpoint kind.
* riscv-tdep.c (show_use_compressed_breakpoints): Remove
'additional_info' and related logic.
(riscv_debug_breakpoints): New variable.
(riscv_breakpoint_kind_from_pc): Use the length of the existing
instruction to determine the breakpoint kind.
(_initialize_riscv_tdep): Add 'set/show debug riscv breakpoints'
flag. Update description of 'set/show riscv
use-compressed-breakpoints' flag.
This commit improves the prologue scanning stack unwinder, to better
support AUIPC, LUI, and more variants of ADD and ADDI.
This allows unwinding over frames containing large local variables,
where the frame size does not fit into a single instruction immediate,
and is first loaded into a temporary register, before being added to
the stack pointer.
A new test is added that tests this behaviour. As there's nothing
truely RiscV specific about this test I've added it into gdb.base, but
as this depends on target specific code to perform the unwind it is
possible that some targets might fail this new test.
gdb/ChangeLog:
* riscv-tdep.c (riscv_insn::decode): Decode c.lui.
(riscv_scan_prologue): Split handling of AUIPC, LUI, ADD, ADDI,
and NOP.
gdb/testsuite/ChangeLog:
* gdb.base/large-frame-1.c: New file.
* gdb.base/large-frame-2.c: New file.
* gdb.base/large-frame.exp: New file.
* gdb.base/large-frame.h: New file.
This patch fixes an ARI violation in riscv-tdep.c (line ends with
'+').
gdb/ChangeLog:
* riscv-tdep.c (riscv_frame_cache): Fix ARI warning, don't end a
line with '+'.
Collects information during the prologue scan and uses this to unwind
registers when no DWARF information is available.
This patch has been tested by disabling the DWARF stack unwinders, and
running the complete GDB testsuite against a range of RISC-V targets.
The results are comparable to running with the DWARF unwinders in
place.
gdb/ChangeLog:
* riscv-tdep.c: Add 'prologue-value.h' include.
(struct riscv_unwind_cache): New struct.
(riscv_debug_unwinder): New global.
(riscv_scan_prologue): Update arguments, capture register details
from prologue scan.
(riscv_skip_prologue): Reformat arguments line, move end of
prologue calculation into riscv_scan_prologue.
(riscv_frame_cache): Update return type, create
riscv_unwind_cache, scan the prologue, and fill in remaining cache
details.
(riscv_frame_this_id): Use frame id computed in riscv_frame_cache.
(riscv_frame_prev_register): Use the trad_frame within the
riscv_unwind_cache.
(_initialize_riscv_tdep): Add 'set/show debug riscv unwinder'
flag.
Extends the instruction decoder used during prologue scan and software
single step to cover more instructions. These instructions are
encountered when running the current GDB testsuite with the DWARF
stack unwinders turned off.
gdb/ChangeLog:
* riscv-tdep.c (riscv_insn::decode): Decode c.addi4spn, c.sd,
c.sw, c.swsp, and c.sdsp.
The RISC-V had a mechanism in place to cache the contents of the misa
register per-inferior, the original intention behind this was to
reduce the number of times the misa register had to be read (as the
contents should be constant), but it was pointed out on the mailing
list[1] that the register cache will mean the register is only
accessed once each time GDB stops, and any additional caching is
probably just unneeded extra complexity.
As such, until it can be shown that there's a real need for additional
caching, this commit removes all of the additional caching of the misa
register, and just accesses the misa register like a normal register.
[1] https://sourceware.org/ml/gdb-patches/2018-03/msg00136.html
gdb/ChangeLog:
* riscv-tdep.c (struct riscv_inferior_data): Delete.
(riscv_read_misa_reg): Don't cache value read into inferior data.
(riscv_new_inferior_data): Delete.
(riscv_inferior_data_cleanup): Delete.
(riscv_inferior_data): Delete.
(riscv_invalidate_inferior_data): Delete.
(_initialize_riscv_tdep): Remove initialisation of inferior data.
This adds software single step support that is needed by the linux native port.
This is modeled after equivalent code in the MIPS port.
This also fixes a few bugs in the compressed instruction decode support. Some
instructions are RV32/RV64 specific, and this wasn't being checked. Also, a
few instructions were accidentally using the non-compressed is_* function.
This has been tested on a HiFive Unleashed running Fedora, by putting a
breakpoint on start, typing stepi, and then holding down the return key until
it finishes, and observing that I see everything I expect to see along the way.
There is a problem in _dl_addr where I get into an infinite loop, but it seems
to be some synchronization code that doesn't agree with single step, so I have
to find the end of the loop, put a breakpoint there, continue, and then single
step again until the end.
gdb/
* riscv-tdep.c (enum opcode): Add jump, branch, lr, and sc opcodes.
(decode_register_index_short): New.
(decode_j_type_insn, decode_cj_type_insn): New.
(decode_b_type_insn, decode_cb_type_insn): New.
(riscv_insn::decode): Add support for jumps, branches, lr, and sc. New
local xlen. Check xlen when decoding ambiguous compressed insns. In
compressed decode, use is_c_lui_insn instead of is_lui_insn, and
is_c_sw_insn instead of is_sw_insn.
(riscv_next_pc, riscv_next_pc_atomic_sequence): New.
(riscv_software_single_step): New.
* riscv-tdep.h (riscv_software_single_step): Declare.
This allows the function to be used from riscv OS files, which also need to
depend on XLEN size.
gdb/
* riscv-tdep.c (riscv_isa_xlen): Drop static.
* riscv-tdep.h (riscv_isa_xlen): Add extern declaration.
gdb/
* riscv-tdep.c (riscv_has_feature): Delete comment that refers to
set_gdbarch_decr_pc_after_break. Call riscv_read_misa_reg always.
(riscv_gdbarch_init): Delete local has_compressed_isa. Delete now
unecessary braces after EF_RISCV_RVC test. Delete call to
set_gdbarch_decr_pc_after_break.
Simon Cook
Andrew Burgess
Tom Tromey
Keith Seitz
KONRAD Frederic
Jim Wilson
Joel Brobecker
Alan Hayward
John Baldwin