My earlier attempt to mask the segment registers in gdbserver for
Windows introduced some bugs. That patch is here:
https://sourceware.org/pipermail/gdb-patches/2023-December/205306.html
The problem turned out to be that these fields in the Windows
'CONTEXT' type are just 16 bits, so while masking the values on read
is fine, writing the truncated values back then causes zeros to be
written to some subsequent field.
This patch cleans this up by arranging never to write too much data to
a field.
I also noticed that two register numbers here were never updated for
the 64-bit port. This patch fixes this as well, and renames the "FCS"
register to follow gdb.
Finally, this patch applies the same treatment to windows-nat.c.
Reviewed-by: John Baldwin <jhb@FreeBSD.org>
This patch removes support for the two instructions above from the GNU
simulator, including the corresponding tests. These instructions do
not really exist in BPF and are not recognized as such by the kernel
verifier. This has now been pointed out during the standardization of
the BPF ISA.
Signed-off-by: Jose E. Marchesi <jose.marchesi@oracle.com>
Since commit 6771fc6f1d "Use a .def file for domain_enum", the
sym-domains.def file has been introduced, and requires the user to
define the DOMAIN(x) macro.
On older systems (centos-7 with glibc-2.17 for example), this DOMAIN
macro conflicts with another macro defined in /usr/include/math.h.
Fix this conflict by changing sym-domains.def to use a macro named
SYM_DOMAIN instead of DOMAIN.
Change-Id: I679df30e2bd2f4333343f16bbd2a3511a37550a3
Approved-By: Tom Tromey <tom@tromey.com>
There are no legacy ldind nor ldabs BPF instructions with BPF_SIZE_DW.
For some reason we were (incorrectly) supporting these. This patch
updates the opcodes so the instructions get removed and modifies the
GAS manual and testsuite accordingly.
See discussion at
https://lore.kernel.org/bpf/110aad7a-f8a3-46ed-9fda-2f8ee54dcb89@linux.dev
Tested in bpf-uknonwn-none target, x86-64-linux-gnu host.
include/ChangeLog:
2024-01-29 Jose E. Marchesi <jose.marchesi@oracle.com>
* opcode/bpf.h (enum bpf_insn_id): Remove BPF_INSN_LDINDDW and
BPF_INSN_LDABSDW instructions.
opcodes/ChangeLog:
2024-01-29 Jose E. Marchesi <jose.marchesi@oracle.com>
* bpf-opc.c (bpf_opcodes): Remove BPF_INSN_LDINDDW and
BPF_INSN_LDABSDW instructions.
gas/ChangeLog:
2024-01-29 Jose E. Marchesi <jose.marchesi@oracle.com>
* doc/c-bpf.texi (BPF Instructions): There is no indirect 64-bit
load instruction.
(BPF Instructions): There is no absolute 64-bit load instruction.
* testsuite/gas/bpf/mem.s: Update test accordingly.
* testsuite/gas/bpf/mem-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-be.d: Likewise.
* testsuite/gas/bpf/mem-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-pseudoc.s: Likewise.
* testsuite/gas/bpf/mem.d: Likewise.
* testsuite/gas/bpf/mem.s: Likewise.
If you have set up a backtrace limit, and the backtrace stops
because of this in an inline frame with arguments, you get an
assertion failure:
```
(gdb) bt
(gdb) set backtrace limit 2
(gdb) bt
C:/src/repos/binutils-gdb.git/gdb/frame.c:3346: internal-error: reinflate: Assertion `m_cached_level >= -1' failed.
```
And if this one is fixed, there is another one as well:
```
(gdb) bt
C:/src/repos/binutils-gdb.git/gdb/dwarf2/loc.c:1160: internal-error: dwarf_expr_reg_to_entry_parameter: Assertion `frame != NULL' failed.
```
The reason for both of them is this kind of loop:
```
while (get_frame_type (frame) == INLINE_FRAME)
frame = get_prev_frame (frame);
```
Since get_prev_frame respects the backtrace limit, it will return
NULL, and from there on you can't continue.
This changes these loops to use get_prev_frame_always instead, so
you always get a non-inline frame in the end.
With this backtrace works:
```
(gdb) bt
(gdb)
```
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29865
Approved-By: Andrew Burgess <aburgess@redhat.com>
This changes cp_search_static_and_baseclasses to only search for
types, functions, and modules. The latter two cases were discovered
by regression testing. I found it somewhat surprising the Fortran
name lookup ends up in this code, but did not attempt to change this.
This changes cp_lookup_rtti_type to only search for types -- not
functions or variables. Due to the symbol-matching hack, this could
just use SEARCH_TYPE_DOMAIN, but I think it's better to be clear; also
I hold on to some hope that perhaps the hack can someday be removed.
This simplifies some symbol searches in linespec.c. In particular,
two separate searches here can now be combined into one, due to the
new use of flags.
Arguably the STRUCT_DOMAIN searches should perhaps not even be done.
Only C really has these, and C doesn't have member functions.
However, it seems relatively harmless -- and clearly compatible -- to
leave this in.
This changes some of the Ada code to simplify symbol searches. For
example, if a function is being looked for, the search is narrowed to
use SEARCH_FUNCTION_DOMAIN rather than SEARCH_VFT. In one spot, a
search of the "struct" domain is removed, because Ada does not have a
tag domain.
This patch changes the DWARF reader to use the new symbol domains. It
also adjusts many bits of associated code to adapt to this change.
The non-DWARF readers are updated on a best-effort basis. This is
somewhat simpler since most of them only support C and C++. I have no
way to test a few of these.
I went back and forth a few times on how to handle the "tag"
situation. The basic problem is that C has a special namespace for
tags, which is separate from the type namespace. Other languages
don't do this. So, the question is, should a DW_TAG_structure_type
end up in the tag domain, or the type domain, or should it be
language-dependent?
I settled on making it language-dependent using a thought experiment.
Suppose there was a Rust compiler that only emitted nameless
DW_TAG_structure_type objects, and specified all structure type names
using DW_TAG_typedef. This DWARF would be correct, in that it
faithfully represents the source language -- but would not work with a
purely struct-domain implementation in gdb. Therefore gdb would be
wrong.
Now, this approach is a little tricky for C++, which uses tags but
also enters a typedef for them. I notice that some other readers --
like stabsread -- actually emit a typedef symbol as well. And, I
think this is a reasonable approach. It uses more memory, but it
makes the internals simpler. However, DWARF never did this for
whatever reason, and so in the interest of keeping the series slightly
shorter, I've left some C++-specific hacks in place here.
Note that this patch includes language_minimal as a language that uses
tags. I did this to avoid regressing gdb.dwarf2/debug-names-tu.exp,
which doesn't specify the language for a type unit. Arguably this
test case is wrong.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30164
Nothing calls the variant of symbol_matches_domain that accepts a
domain_enum for searching, so this patch removes it and the
corresponding symbol::matches.
This changes lookup_symbol and associated APIs to accept
domain_search_flags rather than a domain_enum.
Note that this introduces some new constants to Python and Guile. I
chose to break out the documentation patch for this, because the
internals here do not change until a later patch, and it seemed
simpler to patch the docs just once, rather than twice.
This changes quick_symbol_functions::lookup_global_symbol_language to
accept domain_search_flags rather than just a domain_enum, and fixes
up the fallout.
To avoid introducing any regressions, any code passing VAR_DOMAIN now
uses SEARCH_VFT.
That is, no visible changes should result from this patch. However,
it sets the stage to refine some searches later on.
The Python and Guile code exposed the internal domain constants both
as attributes of symbols and as values to pass to lookup functions.
Now, perfect backward compatibility here can't be achieved: some
symbols are going to have domain changes by the end of this series.
However, it seemed to me that we can preserve lookups using the basic
domain values.
This patch implements this by exporting the "or"-able search constants
with an extra bit set. Then it introduces some functions to convert
such constants to domain_search_flags. This will be used by the
Python and Guile code, so that both old- and new-style lookups will
work properly; and while preserving the idea that the domain constants
can be compared to a symbol's domain.
completion_list_add_symbol checks that the returned symbol has
VAR_DOMAIN, but also checks that its address class is LOC_BLOCK. The
domain check is redundant -- only functions can possibly be LOC_BLOCK
-- and leaving this in place will cause a regression when combined
with a later patch in this series. This patch preemptively removes
the redundant check.
This adds a new flag enum type, domain_search_flags, which is the flag
version of domain_enum. Nothing uses this yet, but the goal here is
to have all symbol searches and lookups use these flags. The new
names are chosen to exactly parallel domain_enum.
This adds two new symbol domain constants, TYPE_DOMAIN and
FUNCTION_DOMAIN.
Historically, gdb was a C debugger, and the symbol tables continue to
reflect this. In particular, symbol domains match the C language,
with VAR_DOMAIN including variables, functions, and types.
However, other languages have other approaches to namespacing. And,
in any case, it is often useful for other parts of gdb to be able to
distinguish between some domains at lookup time, without resorting to
examining a symbol's location -- in some situations, this sort of
filtering happens too late.
Nothing uses these new domains yet, but the idea behind the patch is
to separate symbols into more domains and then let the
language-specific parts of gdb implement their semantics in terms of
these categories.
Future patches will change and reuse the names from domain_enum. This
patch makes this less error-prone by having a single point to define
these names, using the typical gdb ".def" file.
global_symbol_searcher::add_matching_symbols in symtab.c has a
gigantic 'if' statement -- 33 lines of conditional expression. This
patch splits it up into a series of separate 'if's.
NR_DOMAINS is only used for a static assert, but we no longer need it
now. If we add too many constants to this enum, GCC will warn about
the bitfield overflow:
error: ‘symbol::m_domain’ is too small to hold all values of ‘enum domain_enum’
A patch later in this series will change check_typedef to also look in
the type domain. This change by itself caused a regression, but one
that revealed some peculiar behavior.
The regression is in nullptr_t.exp, where examining a std::nullptr_t
will change from the correct:
typedef decltype(nullptr) std::nullptr_t;
to
typedef void std::nullptr_t;
Right now, the DWARF reader marks all unspecified types as stub types.
However, this interacts weirdly with check_typedef, which currently
does not try to resolve types -- only struct-domain objects.
My first attempt here was to fix this by changing void types not to be
stub types, as I didn't see what value that provided. However, this
caused another regression, because call_function_by_hand_dummy checks
for stub-ness:
if (values_type == NULL || values_type->is_stub ())
values_type = default_return_type;
I'm not really sure why it does this rather than check for
TYPE_CODE_VOID.
While looking into this, I found another oddity: the DWARF reader
correctly creates a type named 'decltype(nullptr)' when it seems a
DW_TAG_unspecified_type -- but it creates a symbol named "void"
instead.
This patch changes the DWARF reader to give the symbol the correct
name. This avoids the regression.
nsalias.exp tries to detect a complaint that is issued when expanding
a CU. However, the test is a bit funny in that, while gdb does
currently expand the CU and issue the complaint, it also emits this
error:
No symbol "N100" in current context.
This series will change gdb such that this CU is not expanded -- which
makes sense, the symbol in question doesn't actually match the lookups
that are done.
So, to make the test more robust, a direct request to expand symtabs
is done instead.
A DW_TAG_entry_point symbol inherits its extern/static property from
the enclosing subroutine. This is encoded in new_symbol -- but the
cooked indexer does not agree.
I noticed a couple of spots in dwarf/read.c:new_symbol that call
add_symbol_to_list. However, this function is generally written to
set list_to_add, and then have a single call to add_symbol_to_list at
the end. This patch cleans up this discrepancy.
Note that new_symbol is overlong and should probably be split up.
Testing this entire series pointed out that the cooked index scanner
disagrees with new_symbol about certain symbols. In particular,
new_symbol has this comment:
Ada and Fortran subprograms, whether marked external or
not, are always stored as a global symbol, because we want
This patch updates the scanner to match.
I don't know why the current code does not cause failures.
It's maybe worth noting that incremental CU expansion -- creating
symtabs directly from the index -- would eliminate this sort of bug.
PR gas/31284
Currently, if an indirect jump is seen, GCFG (a CFG of ginsns) cannot be
created, and the SCFI machinery bails out with a warning:
"Warning: Untraceable control flow for func 'foo'; Skipping SCFI"
It is, however, better suited if this is a hard error. Change it to a
hard error. Also change the message to skip mentioning "SCFI", because
the error itself may also useful when ginsns are used for other passes
(distinct from SCFI) involving GCFG, like a pass to detect if there is
unreachable code. Hence, simply say:
"Error: untraceable control flow for func 'foo'"
gas/
PR gas/31284
* ginsn.c (ginsn_data_end): Use as_bad instead of as_warn.
gas/testsuite/
PR gas/31284
* gas/scfi/x86_64/ginsn-cofi-1.l: Adjust to the expected output
in case of errors.
* gas/scfi/x86_64/scfi-unsupported-cfg-1.l: Error not Warning.
The testcase for change of flow instructions in its current shape is not
doing much: it checks that SCFI issues an appropriate warning. The same
warning is covered by another testcase (scfi-unsupported-cfg-1); It is
better to test the ginsn translation instead, for these 'change of flow
instructions'.
gas/testsuite/
* gas/scfi/x86_64/scfi-cofi-1.s: Moved to...
* gas/scfi/x86_64/ginsn-cofi-1.s: ...here.
* gas/scfi/x86_64/scfi-x86-64.exp: Adjust tests.
* gas/scfi/x86_64/scfi-cofi-1.d: Removed.
* gas/scfi/x86_64/scfi-cofi-1.l: Removed.
* gas/scfi/x86_64/ginsn-cofi-1.l: New test.
SHA512 instructions were added to the architecture at the same time as SHA3
instructions, but later than the SHA1 and SHA256 instructions. Furthermore,
implementations must support either both or neither of the SHA512 and SHA3
instruction sets. However, SHA512 instructions were originally (and
incorrectly) added to Binutils under the +sha2 flag.
This patch moves SHA512 instructions under the +sha3 flag, which matches the
architecture constraints and existing GCC and LLVM behaviour.
Tom Tromey
GDB Administrator
Alan Modra
Jose E. Marchesi
Lancelot SIX
Nick Clifton
Hannes Domani
Tom Tromey
Indu Bhagat
H.J. Lu
Andrew Carlotti