Avoid some pointer chasing in DWARF reader

I noticed a spot in the DWARF reader using "per_objfile->per_bfd",
where a local per_bfd variable had already been created.  Looking
through the file, I found a number of such spots.  This patch changes
them to use the already-existing local, avoiding a bit of excess
pointer chasing.

gdb/ChangeLog
2021-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_read_debug_names)
(dwarf2_build_psymtabs_hard, create_addrmap_from_aranges)
(dw2_debug_names_iterator::next, create_type_unit_group):
Simplify.

gdb/testsuite: resolve remaining duplicate test names in gdb.cp/*.exp

This commit resolves the remaining duplicate test names in
gdb.cp/*.exp.  These are all the easy duplicates, I'm either giving
tests a new, unique name, extending an existing name to make it
unique, or changing an existing name to better reflect what the test
is actually doing, and thus, making this test name unique.

There should be no change in what is tested after this commit.

gdb/testsuite/ChangeLog:

* gdb.cp/breakpoint.exp: Extend test names to make them unique.
* gdb.cp/casts.exp: Give tests unique names.
* gdb.cp/filename.exp: Likewise.
* gdb.cp/gdb2495.exp: Likewise.
* gdb.cp/mb-ctor.exp: Extend test names to make them unique.
* gdb.cp/misc.exp: Rename test to make it unique.
* gdb.cp/nsnested.exp: Give tests unique names.
* gdb.cp/ovldbreak.exp: Likewise.
* gdb.cp/pr17494.exp: Rename test to reflect what is actually
being tested.  This also removes the duplicate test name.
* gdb.cp/ref-types.exp: Likewise.
* gdb.cp/temargs.exp: Likewise.

gdb/testsuite: resolve duplicate test name in gdb.cp/cplusfuncs.exp

While resolving duplicate test names I spotted that a test in
gdb.cp/cplusfuncs.exp included an unescaped '[]'.  In TCL square
brackets enclose expressions to evaluate, and so in this case, where
there is no enclosed expression, this just evaluates to the empty
string.

This clearly was not what the test intended, so in this commit I have
escaped the square brackets.  This has extended the test coverage.

gdb/testsuite/ChangeLog:

* gdb.cp/cplusfuncs.exp (test_paddr_operator_functions): Escape
square brackets in test.

gdb/testsuite: remove duplicate test from gdb.cp/maint.exp

I wanted to remove the duplicate test name from gdb.cp/maint.exp.  In
this test we run some checks against different operator names.  For
one operator we test with a variable number of spaces.  However, we
were accidentally testing the one space version twice, and the zero
space version not at all, leading to a duplicate test name.

I could have just changed the duplicate one space version into the
missing zero space version, but I thought it would be neater to wrap
multiple tests in a loop, and check all operators with either zero,
one, or two spaces.

These tests are super quick so take almost no extra time, and this
gives marginally more test coverage.

gdb/testsuite/ChangeLog:

* gdb.cp/maint.exp (test_first_component): Run more tests with a
variable number of spaces, this removes the duplicate testing of
'operator ->' which existed before.

gdb/testsuite: remove duplicate test names from gdb.cp/gdb2384.exp

The test gdb.cp/gdb2384.exp contains some duplicate test names, and
also some test names with a string inside parentheses at the end.  In
order to resolve the duplicates the obvious choice would be to add yet
more strings inside parentheses at the end of names, however, this is
discouraged in our test naming scheme.

The string in parentheses originates from a comment in the test source
code, which naturally leads to including this comment in the test
name.

In this commit I have changed the comment in the test source to remove
the string in parentheses, I then rename the tests in the .exp script
to match, making sure that all test names are unique.

There should be no change in test coverage after this commit.

gdb/testsuite/ChangeLog:

* gdb.cp/gdb2384.cc (main): Change comments used for breakpoints.
* gdb.cp/gdb2384.exp: Change and extend test names to avoid
duplicates, and also to avoid having a string inside parentheses
at the end of test names.

gdb/testsuite: remove duplicate test names for gdb.cp/nsusing.exp

In trying to resolve the duplicate test names for the
gdb.cp/nsusing.exp script, I ended up giving the test script a serious
spring clean.

This reverts some of the changes introduced in commit df83a9bf8b0d,
but I don't think that we have lost any testing.

The test program is made of many functions, the test script wants to
stop in different functions and check which symbols are in scope.

Previously the test script would either restart GDB completely in
order to "progress" to the next function, or the script would restart
the test program using 'runto'.

In this commit I have reordered the steps of the test to correspond to
program order, I then progress through the test program once by just
placing a breakpoint and then continuing.  As I said, the test is
checking which symbols are in scope at each location, so the exact
order of the tests doesn't matter, so long as we check the correct
symbols at each location.

I have also given the comments capital letters and full stops, and
re-wrapped them to a more sensible line length.

There was a duplicate test block introduced in the df83a9bf8b0d
commit which I have removed in this commit, this duplicate code was
responsible for one of the duplicate test names.

The other duplicate test name was due to the same command being run at
different locations, in this case I just gave the two tests explicit,
unique, names.

gdb/testsuite/ChangeLog:

* gdb.cp/nsusing.exp: Rewrite test, remove a duplicate test block.
Avoid repeated uses of 'runto', and instread just progress once
through the test stopping at different breakpoints.  Give comments
a capital letter and full stop.  Give duplicate tests unique names.

x86-64: don't accept supposedly disabled MOVQ forms

While all of MMX, SSE, and SSE2 are included in "generic64", they can be
individually disabled. There are two MOVQ forms lacking respective
attributes. While the MMX one would get refused anyway (due to MMX
registers not recognized with .nommx), the assembler did happily accept
the SSE2 form. Add respective CPU settings to both, paralleling what the
MOVD counterparts have.

Automatic date update in version.in

Fix bkpt-other-inferior.exp race

When testing with "maint set target-non-stop on",
gdb.server/bkpt-other-inferior.exp sometimes fails like so:

 (gdb) inferior 2
 [Switching to inferior 2 [process 368191] (<noexec>)]
 [Switching to thread 2.1 (Thread 368191.368191)]
 [remote] Sending packet: $m7ffff7fd0100,1#5b
 [remote] Packet received: 48
 [remote] Sending packet: $m7ffff7fd0100,1#5b
 [remote] Packet received: 48
 [remote] Sending packet: $m7ffff7fd0100,9#63
 [remote] Packet received: 4889e7e8e80c000049
 #0  0x00007ffff7fd0100 in ?? ()
 (gdb) PASS: gdb.server/bkpt-other-inferior.exp: inf 2: switch to inferior
 break -q main
 Breakpoint 2 at 0x1138: file /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.server/server.c, line 21.
 (gdb) PASS: gdb.server/bkpt-other-inferior.exp: inf 2: set breakpoint
 delete breakpoints
 Delete all breakpoints? (y or n) y
 (gdb) [remote] wait: enter
 [remote] wait: exit
 FAIL: gdb.server/bkpt-other-inferior.exp: inf 2: delete all breakpoints in delete_breakpoints (timeout)
 ERROR: breakpoints not deleted
 Remote debugging from host ::1, port 55876
 monitor exit

The problem is here:

 (gdb) [remote] wait: enter

The testcase isn't expecting any output after the prompt.

Why is that "[remote] wait" output?  What happens is that "delete
breakpoints" queries the user, and `query` disables/reenables target
async, which results in the remote target's async event handler ending
up marked:

 (top-gdb) bt
 #0  mark_async_event_handler (async_handler_ptr=0x556bffffffff) at ../../src/gdb/async-event.c:295
 #1  0x0000556bf71b711f in infrun_async (enable=1) at ../../src/gdb/infrun.c:119
 #2  0x0000556bf7471387 in target_async (enable=1) at ../../src/gdb/target.c:3684
 #3  0x0000556bf748a0bd in gdb_readline_wrapper_cleanup::~gdb_readline_wrapper_cleanup (this=0x7ffe3cf30eb0, __in_chrg=<optimized out>) at ../../src/gdb/top.c:1074
 #4  0x0000556bf74874e2 in gdb_readline_wrapper (prompt=0x556bfa17da60 "Delete all breakpoints? (y or n) ") at ../../src/gdb/top.c:1096
 #5  0x0000556bf75111c5 in defaulted_query(const char *, char, typedef __va_list_tag __va_list_tag *) (ctlstr=0x556bf7717f34 "Delete all breakpoints? ", defchar=0 '\000', args=0x7ffe3cf31020) at ../../src/gdb/utils.c:893
 #6  0x0000556bf751166f in query (ctlstr=0x556bf7717f34 "Delete all breakpoints? ") at ../../src/gdb/utils.c:985
 #7  0x0000556bf6f11404 in delete_command (arg=0x0, from_tty=1) at ../../src/gdb/breakpoint.c:13500
 ...

... which then later results in a target_wait call:

 (top-gdb) bt
 #0  remote_target::wait_ns (this=0x7ffe3cf30f80, ptid=..., status=0xde530314f0802800, options=...) at ../../src/gdb/remote.c:7937
 #1  0x0000556bf7369dcb in remote_target::wait (this=0x556bfa0b2180, ptid=..., status=0x7ffe3cf31568, options=...) at ../../src/gdb/remote.c:8173
 #2  0x0000556bf745e527 in target_wait (ptid=..., status=0x7ffe3cf31568, options=...) at ../../src/gdb/target.c:2000
 #3  0x0000556bf71be686 in do_target_wait_1 (inf=0x556bfa1573d0, ptid=..., status=0x7ffe3cf31568, options=...) at ../../src/gdb/infrun.c:3463
 #4  0x0000556bf71be88b in <lambda(inferior*)>::operator()(inferior *) const (__closure=0x7ffe3cf31320, inf=0x556bfa1573d0) at ../../src/gdb/infrun.c:3526
 #5  0x0000556bf71bebcd in do_target_wait (wait_ptid=..., ecs=0x7ffe3cf31540, options=...) at ../../src/gdb/infrun.c:3539
 #6  0x0000556bf71bf97b in fetch_inferior_event () at ../../src/gdb/infrun.c:3879
 #7  0x0000556bf71a27f8 in inferior_event_handler (event_type=INF_REG_EVENT) at ../../src/gdb/inf-loop.c:42
 #8  0x0000556bf71cc8b7 in infrun_async_inferior_event_handler (data=0x0) at ../../src/gdb/infrun.c:9220
 #9  0x0000556bf6ecb80f in check_async_event_handlers () at ../../src/gdb/async-event.c:327
 #10 0x0000556bf76b011a in gdb_do_one_event () at ../../src/gdbsupport/event-loop.cc:216
 ...

... which returns TARGET_WAITKIND_IGNORE.

Fix this by only enabling remote output around setting the breakpoint.

gdb/testsuite/ChangeLog:

* gdb.server/bkpt-other-inferior.exp: Only enable remote output
around setting the breakpoint.

Change-Id: I2fd152fd9c46b1c5e7fa678cc4d4054dac0b2bd4

Fix problem exposed by gdb.server/stop-reply-no-thread-multi.exp

Running gdb.server/stop-reply-no-thread-multi.exp with "maint set
target-non-stop on" occasionally hit an internal error like this:

  ...
  continue
  Continuing.
  warning: multi-threaded target stopped without sending a thread-id, using first non-exited thread
  /home/pedro/gdb/binutils-gdb/src/gdb/inferior.c:291: internal-error: inferior* find_inferior_pid(process_stratum_target*, int): Assertion `pid != 0' failed.
  A problem internal to GDB has been detected,
  further debugging may prove unreliable.

  This is a bug, please report it.
  FAIL: gdb.server/stop-reply-no-thread-multi.exp: to_disable=Tthread: continue until exit (GDB internal error)

The backtrace looks like this:

 ...
 #5  0x0000560357b0879c in internal_error (file=0x560357be6c18 "/home/pedro/gdb/binutils-gdb/src/gdb/inferior.c", line=291, fmt=0x560357be6b21 "%s: Assertion `%s' failed.") at /home/pedro/gdb/binutils-gdb/src/gdbsupport/errors.cc:55
 #6  0x000056035762061b in find_inferior_pid (targ=0x5603596e9560, pid=0) at /home/pedro/gdb/binutils-gdb/src/gdb/inferior.c:291
 #7  0x00005603576206e6 in find_inferior_ptid (targ=0x5603596e9560, ptid=...) at /home/pedro/gdb/binutils-gdb/src/gdb/inferior.c:305
 #8  0x00005603577d43ed in remote_target::check_pending_events_prevent_wildcard_vcont (this=0x5603596e9560, may_global_wildcard=0x7fff84fb05f0) at /home/pedro/gdb/binutils-gdb/src/gdb/remote.c:7215
 #9  0x00005603577d2a9c in remote_target::commit_resumed (this=0x5603596e9560) at /home/pedro/gdb/binutils-gdb/src/gdb/remote.c:6680
 ...

pid is 0 in this case because the queued event is a process exit event
with no pid associated:

 (top-gdb) p event->ws
 During symbol reading: .debug_line address at offset 0x563c9a is 0 [in module /home/pedro/gdb/binutils-gdb/build/gdb/gdb]
 $1 = {kind = TARGET_WAITKIND_EXITED, value = {integer = 0, sig = GDB_SIGNAL_0, related_pid = {m_pid = 0, m_lwp = 0, m_tid = 0}, execd_pathname = 0x0, syscall_number = 0}}
 (top-gdb)

This fixes it, and adds a "maint set target-non-stop on/off" axis to the testcase.

gdb/ChangeLog:

* remote.c
(remote_target::check_pending_events_prevent_wildcard_vcont):
Check whether the event's ptid is not null_ptid before looking up
the corresponding inferior.

gdb/testsuite/ChangeLog:

* gdb.server/stop-reply-no-thread-multi.exp (run_test): Add
"target_non_stop" parameter and use it.
(top level): Add "maint set target-non-stop on/off" testing axis.

Change-Id: Ia30cf275305ee4dcbbd33f731534cd71d1550eaa

ld: do not rely on the exact size of the CTF symtypetabs in test results

The data object and function info sections (collectively "symtypetabs")
usually (i.e. if non-indexed) have sizes defined by the size of the ELF
dynamic symbol table in the object they are linked to.  This means test
results should not depend on the exact sizes of these sections, because
adding entirely irrelevant symbols to the dynsym can cause spurious test
failures.  (This also means we should not match the offset of sections
that follow them, since those too depend on the exact size of the
symtypetab sections.)

Spotted by turning the sanitizer on, which introduced new dynsym entries
and expanded the symtypetab sizes to match.

ld/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

* testsuite/ld-ctf/array.d: Only check that the data object
section is nonempty: do not check its exact size.
* testsuite/ld-ctf/diag-parlabel.d: Likewise.
* testsuite/ld-ctf/slice.d: Likewise.
* testsuite/ld-ctf/data-func-conflicted.d: Likewise, and for the
func info section too.
* testsuite/ld-ctf/function.d: Likewise, for the func info section.

libctf: fix ELF-in-BFD checks in the presence of ASAN

The address sanitizer contains a redirector that captures dlopen calls,
so checks for dlopen with AC_SEARCH_LIBS will always conclude that
dlopen is present when the sanitizer is on.  This means it won't add
-ldl to LIBS even if needed, and the immediately-following attempt to
actually link with -lbfd will fail because libbfd also needs dlsym,
which ASAN does *not* contain a redirector for.

If we check for dlsym instead of dlopen, the check works whether ASAN is
on or off.  (bfd uses both in close proximity: if it needs one, it will
always need the other.)

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

* configure.ac: Check for dlsym, not dlopen.
* configure: Regenerate.

libctf: fix memory leak in a test

Harmless, but causes noise that makes it harder to spot other leaks.

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

* testsuite/libctf-writable/symtypetab-nonlinker-writeout.c: Don't
leak buf.

libctf: don't dereference out-of-bounds locations in the qualifier hashtab

isqualifier, which is used by ctf_lookup_by_name to figure out if a
given word in a type name is a qualifier, takes the address of a
possibly out-of-bounds location before checking its bounds.

In any reasonable compiler this will just lead to a harmless address
computation that is then discarded if out-of-bounds, but it's still
undefined behaviour and the sanitizer rightly complains.

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

PR libctf/27628
* ctf-lookup.c (isqualifier): Don't dereference out-of-bounds
qhash values.

libctf: make ctf_bfdopen_ctfsect a debugger entry point

This makes it possible to use LIBCTF_DEBUG to debug things that happen
before the ctf_bfdopen_internal call that ctf_bfdopen_ctfsect eventually
thunks down to (symtab/strtab lookup, archive opening, etc).

This is not important for ctf_open callers, since ctf_fdopen already
calls libctf_init_debug, but ctf_bfdopen_ctfsect is a public entry point
that can be called directly (e.g. objdump and readelf both do so).

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

* ctf-open-bfd.c (ctf_bfdopen_ctfsect): Initialize debugging.

libctf, serialize: functions with no args have a NULL dtd_vlen

Every place that accesses a function's dtd_vlen accesses it only if the
number of args is nonzero, except the serializer, which always tries to
memcpy it.  The number of bytes it memcpys in this case is zero, but it
is still undefined behaviour to copy zero bytes from a null pointer.
So check for this case explicitly.

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

PR libctf/27628
* ctf-serialize.c (ctf_emit_type_sect): Allow for a NULL vlen in
CTF_K_FUNCTION types.

include: always do unsigned left-shift in CTF_SET_STID

This turns into a signed left shift by 31 bits, otherwise.  This is an
offset and is always treated as unsigned in any case, so add an
appropriate cast.

include/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

PR libctf/27628
* ctf-api.h: Fix some indentation.
(CTF_SET_STID): Always do an unsigned shift, even if STID is
signed.

libctf, dump: do not emit size or alignment if it would error

When we dump normal types, we emit their size and/or alignment:
but size and alignment dumping can return errors if the type is
part of a chain that terminates in a forward.

Emitting 0xffffffff as a size or alignment is unhelpful, so simply
skip emitting this info for any type for which size or alignment
checks return an error, no matter what the error is.

libctf/ChangeLog
2021-03-25  Nick Alcock  <nick.alcock@oracle.com>

* ctf-dump.c (ctf_dump_format_type): Don't emit size or alignment
on error.

gdb/testsuite: use -wrap with gdb_test_multiple in lib/ada.exp

I ran into a new failure in gdb.base/gdb-caching-proc.exp:

  FAIL: gdb.base/gdb-caching-proc.exp: supports_memtag: initial: memory-tag check

This is a failure from the `supports_memtag` proc added recently (this
new proc is in lib/gdb.exp).

The problem here is that `supports_memtag` is hitting one of the
default error cases in gdb_test_multiple, specifically it is finding a
$gdb_prompt left unmatched from an earlier call to gdb_test_multiple.

Looking back through the test output I found that the problem is the
proc `gnat_runtime_has_debug_info` in lib/ada.exp.  This proc is not
matching the trailing $gdb_prompt.  This leaves the prompt in the
expect buffer, then when we run `supports_memtag` it sees the prompt
and thinks that the test completed with no output.

Fixed by making use of `-wrap` in `gnat_runtime_has_debug_info` to
ensure the trailing prompt gets matched.

gdb/testsuite/ChangeLog:

* lib/ada.exp (gnat_runtime_has_debug_info): Use -wrap with
gdb_test_multiple.

gdb/riscv: fix creating breakpoints at invalid addresses

To allow breakpoints to be created at invalid addresses,
target_read_code is used instead of read_code.  This was fixed in
commit:

  commit c01660c625766e848195285cc20581b9ed7ecfe2
  Date:   Wed Apr 17 00:31:43 2019 +0100

      gdb/riscv: Allow breakpoints to be created at invalid addresses

Unfortunately, the call to read_code was left in by mistake.  The
result is that GDB will fail when trying to create the breakpoint,
rather than when trying to install the breakpoint (as is the case with
other targets).

This commit fixes this mistake and removes the offending call to
read_code.

gdb/ChangeLog:

* riscv-tdep.c (riscv_breakpoint_kind_from_pc): Remove call to
read_code.

[NIOS2] Fix disassembly of br.n instruction.

The code was checking wrong bit for sign extension. It caused it
to zero-extend instead of sign-extend the immediate value.

2021-03-25  Abid Qadeer  <abidh@codesourcery.com>

opcodes/
* nios2-dis.c (nios2_print_insn_arg): Fix sign extension of
immediate in br.n instruction.

gas/
* testsuite/gas/nios2/brn.s: New.
* testsuite/gas/nios2/brn.d: New.

x86: fix CMPXCHG8B special case when disallowing q suffix outside of 64-bit mode

In match_template() i.tm hasn't been filled yet, so it is necessarily t
which needs checking. This is only a latent issue as no other templates
with the same base_opcode have an extension_opcode of 1.

x86: flag bad S/G insn operand combinations

For VEX-encoded ones, all three involved vector registers have to be
distinct. For EVEX-encoded ones an actual mask register has to be in use
and zeroing-masking cannot be used (violation of either will #UD).
Additionally both involved vector registers have to be distinct for
EVEX-encoded gathers.

x86: flag as bad AVX512 insns with EVEX.z set but EVEX.aaa clear

This combination makes no sense and is documented to cause #UD.

x86: fix AMD Zen3 insns

For INVLPGB the operand count was wrong (besides %edx there's also %ecx
which is an input to the insn). In this case I see little sense in
retaining the bogus 2-operand template. Plus swapping of the operands
wasn't properly suppressed for Intel syntax.

For PVALIDATE, RMPADJUST, and RMPUPDATE bogus single operand templates
were specified. These get retained, as the address operand is the only
one really needed to expressed non-default address size, but only for
compatibility reasons. Proper multi-operand insn get introduced and the
testcases get adjusted / extended accordingly.

While at it also drop the redundant definition of __amd64__ - we already
have x86_64 defined (or not) to distinguish 64-bit and non-64-bit cases.

x86-64: limit breakage from gcc movdir64b et al workaround

This is only a partial fix for PR/gas 27419, in that it limits the bad
behavior of accepting mismatched operands to just x32 mode. The full fix
would be to revert commits 27f134698ac5 and b3a3496f83a1, and to address
the issue in gcc instead.

PR27647 PowerPC extended conditional branch mnemonics

opcodes/
PR 27647
* ppc-opc.c (XLOCB_MASK): Delete.
(XLBOBB_MASK, XLBOBIBB_MASK, XLBOCBBB_MASK): Define using
XLBH_MASK.
(powerpc_opcodes): Accept a BH field on all extended forms of
bclr, bclrl, bcctr, bcctrl, bctar, bctarl.
gas/
PR 27647
* testsuite/gas/ppc/a2.d: Update expected output.
* testsuite/gas/ppc/power8.d: Likewise.

Automatic date update in version.in

gnulib: import gitlog-to-changelog

This is going to be used to generate ChangeLog files.

gdb: remove current_top_target function

The current_top_target function is a hidden dependency on the current
inferior.  Since I'd like to slowly move towards reducing our dependency
on the global current state, remove this function and make callers use

  current_inferior ()->top_target ()

There is no expected change in behavior, but this one step towards
making those callers use the inferior from their context, rather than
refer to the global current inferior.

gdb/ChangeLog:

* target.h (current_top_target): Remove, make callers use the
current inferior instead.
* target.c (current_top_target): Remove.

Change-Id: Iccd457036f84466cdaa3865aa3f9339a24ea001d

gdb: move all "current target" wrapper implementations to target.c

The following patch removes the current_top_target function, replacing
uses with `current_inferior ()->top_target ()`.  This is a problem for
uses in target.h, because they don't have access to the current_inferior
function and the inferior structure: target.h can't include inferior.h,
otherwise that would make a cyclic inclusion.

Avoid this by moving all implementations of the wrappers that call
target methods with the current target to target.c.  Many of them are
changed from a macro to a function, which is an improvement for
readability and debuggability, IMO.

target_shortname and target_longname were not function-like macros, so a
few adjustments are needed.

gdb/ChangeLog:

* target.h (target_shortname): Change to function declaration.
(target_longname): Likewise.
(target_attach_no_wait): Likewise.
(target_post_attach): Likewise.
(target_prepare_to_store): Likewise.
(target_supports_enable_disable_tracepoint): Likewise.
(target_supports_string_tracing): Likewise.
(target_supports_evaluation_of_breakpoint_conditions): Likewise.
(target_supports_dumpcore): Likewise.
(target_dumpcore): Likewise.
(target_can_run_breakpoint_commands): Likewise.
(target_files_info): Likewise.
(target_post_startup_inferior): Likewise.
(target_insert_fork_catchpoint): Likewise.
(target_remove_fork_catchpoint): Likewise.
(target_insert_vfork_catchpoint): Likewise.
(target_remove_vfork_catchpoint): Likewise.
(target_insert_exec_catchpoint): Likewise.
(target_remove_exec_catchpoint): Likewise.
(target_set_syscall_catchpoint): Likewise.
(target_rcmd): Likewise.
(target_can_lock_scheduler): Likewise.
(target_can_async_p): Likewise.
(target_is_async_p): Likewise.
(target_execution_direction): Likewise.
(target_extra_thread_info): Likewise.
(target_pid_to_exec_file): Likewise.
(target_thread_architecture): Likewise.
(target_find_memory_regions): Likewise.
(target_make_corefile_notes): Likewise.
(target_get_bookmark): Likewise.
(target_goto_bookmark): Likewise.
(target_stopped_by_watchpoint): Likewise.
(target_stopped_by_sw_breakpoint): Likewise.
(target_supports_stopped_by_sw_breakpoint): Likewise.
(target_stopped_by_hw_breakpoint): Likewise.
(target_supports_stopped_by_hw_breakpoint): Likewise.
(target_have_steppable_watchpoint): Likewise.
(target_can_use_hardware_watchpoint): Likewise.
(target_region_ok_for_hw_watchpoint): Likewise.
(target_can_do_single_step): Likewise.
(target_insert_watchpoint): Likewise.
(target_remove_watchpoint): Likewise.
(target_insert_hw_breakpoint): Likewise.
(target_remove_hw_breakpoint): Likewise.
(target_can_accel_watchpoint_condition): Likewise.
(target_can_execute_reverse): Likewise.
(target_get_ada_task_ptid): Likewise.
(target_filesystem_is_local): Likewise.
(target_trace_init): Likewise.
(target_download_tracepoint): Likewise.
(target_can_download_tracepoint): Likewise.
(target_download_trace_state_variable): Likewise.
(target_enable_tracepoint): Likewise.
(target_disable_tracepoint): Likewise.
(target_trace_start): Likewise.
(target_trace_set_readonly_regions): Likewise.
(target_get_trace_status): Likewise.
(target_get_tracepoint_status): Likewise.
(target_trace_stop): Likewise.
(target_trace_find): Likewise.
(target_get_trace_state_variable_value): Likewise.
(target_save_trace_data): Likewise.
(target_upload_tracepoints): Likewise.
(target_upload_trace_state_variables): Likewise.
(target_get_raw_trace_data): Likewise.
(target_get_min_fast_tracepoint_insn_len): Likewise.
(target_set_disconnected_tracing): Likewise.
(target_set_circular_trace_buffer): Likewise.
(target_set_trace_buffer_size): Likewise.
(target_set_trace_notes): Likewise.
(target_get_tib_address): Likewise.
(target_set_permissions): Likewise.
(target_static_tracepoint_marker_at): Likewise.
(target_static_tracepoint_markers_by_strid): Likewise.
(target_traceframe_info): Likewise.
(target_use_agent): Likewise.
(target_can_use_agent): Likewise.
(target_augmented_libraries_svr4_read): Likewise.
(target_log_command): Likewise.
* target.c (target_shortname): New.
(target_longname): New.
(target_attach_no_wait): New.
(target_post_attach): New.
(target_prepare_to_store): New.
(target_supports_enable_disable_tracepoint): New.
(target_supports_string_tracing): New.
(target_supports_evaluation_of_breakpoint_conditions): New.
(target_supports_dumpcore): New.
(target_dumpcore): New.
(target_can_run_breakpoint_commands): New.
(target_files_info): New.
(target_post_startup_inferior): New.
(target_insert_fork_catchpoint): New.
(target_remove_fork_catchpoint): New.
(target_insert_vfork_catchpoint): New.
(target_remove_vfork_catchpoint): New.
(target_insert_exec_catchpoint): New.
(target_remove_exec_catchpoint): New.
(target_set_syscall_catchpoint): New.
(target_rcmd): New.
(target_can_lock_scheduler): New.
(target_can_async_p): New.
(target_is_async_p): New.
(target_execution_direction): New.
(target_extra_thread_info): New.
(target_pid_to_exec_file): New.
(target_thread_architecture): New.
(target_find_memory_regions): New.
(target_make_corefile_notes): New.
(target_get_bookmark): New.
(target_goto_bookmark): New.
(target_stopped_by_watchpoint): New.
(target_stopped_by_sw_breakpoint): New.
(target_supports_stopped_by_sw_breakpoint): New.
(target_stopped_by_hw_breakpoint): New.
(target_supports_stopped_by_hw_breakpoint): New.
(target_have_steppable_watchpoint): New.
(target_can_use_hardware_watchpoint): New.
(target_region_ok_for_hw_watchpoint): New.
(target_can_do_single_step): New.
(target_insert_watchpoint): New.
(target_remove_watchpoint): New.
(target_insert_hw_breakpoint): New.
(target_remove_hw_breakpoint): New.
(target_can_accel_watchpoint_condition): New.
(target_can_execute_reverse): New.
(target_get_ada_task_ptid): New.
(target_filesystem_is_local): New.
(target_trace_init): New.
(target_download_tracepoint): New.
(target_can_download_tracepoint): New.
(target_download_trace_state_variable): New.
(target_enable_tracepoint): New.
(target_disable_tracepoint): New.
(target_trace_start): New.
(target_trace_set_readonly_regions): New.
(target_get_trace_status): New.
(target_get_tracepoint_status): New.
(target_trace_stop): New.
(target_trace_find): New.
(target_get_trace_state_variable_value): New.
(target_save_trace_data): New.
(target_upload_tracepoints): New.
(target_upload_trace_state_variables): New.
(target_get_raw_trace_data): New.
(target_get_min_fast_tracepoint_insn_len): New.
(target_set_disconnected_tracing): New.
(target_set_circular_trace_buffer): New.
(target_set_trace_buffer_size): New.
(target_set_trace_notes): New.
(target_get_tib_address): New.
(target_set_permissions): New.
(target_static_tracepoint_marker_at): New.
(target_static_tracepoint_markers_by_strid): New.
(target_traceframe_info): New.
(target_use_agent): New.
(target_can_use_agent): New.
(target_augmented_libraries_svr4_read): New.
(target_log_command): New.
* bfin-tdep.c (bfin_sw_breakpoint_from_kind): Adjust.
* infrun.c (set_schedlock_func): Adjust.
* mi/mi-main.c (exec_reverse_continue): Adjust.
* reverse.c (exec_reverse_once): Adjust.
* sh-tdep.c (sh_sw_breakpoint_from_kind): Adjust.
* tui/tui-stack.c (tui_locator_window::make_status_line): Adjust.
* remote-sim.c (gdbsim_target::detach): Adjust.
(gdbsim_target::files_info): Adjust.

Change-Id: I72ef56e9a25adeb0b91f1ad05e34c89f77ebeaa8

Remove 'kind' parameter from dw2_map_matching_symbols

I noticed that dw2_map_matching_symbols does not use its 'kind'
parameter.  This patch removes it.  Tested by rebuilding.

2021-03-24  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dw2_map_matching_symbols): Update.
(dw2_expand_symtabs_matching_symbol): Remove 'kind' parameter.
(check_match, dw2_expand_symtabs_matching)
(dwarf2_debug_names_index::map_matching_symbols)
(dwarf2_debug_names_index::expand_symtabs_matching): Update.

Fix TYPE_DECLARED_CLASS thinko

Simon pointed out an error that I made in
compile_cplus_conver_struct_or_union in my original C++ compile submission:

  if (type->code () == TYPE_CODE_STRUCT)
    {
      const char *what = TYPE_DECLARED_CLASS (type) ? "struct" : "class";

      resuld = instance->plugin ().build_decl
        (what, name.get (), (GCC_CP_SYMBOL_CLASS | nested_access
                             | (TYPE_DECLARED_CLASS (type)
                                ? GCC_CP_FLAG_CLASS_NOFLAG
                                : GCC_CP_FLAG_CLASS_IS_STRUCT)),
         0, nullptr, 0, filename, line);
    }

Notice that WHAT will contain "struct" for TYPE_DECLARED_CLASS. Whoops.

Fortunately this first parameter of build_decl is only used for
debugging.

gdb/ChangeLog
2021-03-24  Keith Seitz  <keiths@redhat.com>

* compile/compile-cplus-types.c
(compile_cplus_convert_struct_or_union): Fix TYPE_DECLARED_CLASS
thinko.

gdb: make gdbarch_data_registry static

This variable was made static in:

  6bd434d6caa4 ("gdb: make some variables static")

But I modified gdbarch.c instead of gdbarch.sh, so the change was
later reverted when gdbarch.c was re-generated.

Do it right this time.

gdb/ChangeLog:

* gdbarch.sh (gdbarch_data_registry): Make static.
* gdbarch.c: Re-generate.

Change-Id: I4048ba99a0cf47acd9da050934965db222fbd159

Add memory tagging testcases

Add an AArch64-specific test and a more generic memory tagging test that
other architectures can run.

Even though architectures not supporting memory tagging can run the memory
tagging tests, the runtime check will make the tests bail out early, as it
would make no sense to proceed without proper support.

It is also tricky to do any further runtime tests for memory tagging, given
we'd need to deal with tags, and those are arch-specific.  Therefore the
test in gdb.base is more of a smoke test.

If an architecture wants to implement memory tagging, then it makes sense to
have tests within gdb.arch instead.

gdb/testsuite/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* gdb.arch/aarch64-mte.c: New file.
* gdb.arch/aarch64-mte.exp: New test.
* gdb.base/memtag.c: New file.
* gdb.base/memtag.exp: New test.
* lib/gdb.exp (supports_memtag): New function.

Add NEWS entry.

Mention the new packets and memory tagging features.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* NEWS: Mention memory tagging changes.

Document new "x" and "print" memory tagging extensions

Document the changes to the "print" and "x" commands to support memory
tagging.

gdb/doc/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* gdb.texinfo (Data): Document memory tagging changes to the "print"
command.
(Examining Memory): Document memory tagging changes to the "x"
command.
(Memory Tagging): Update with more information on changes to the "x"
and "print" commands.

Extend "x" and "print" commands to support memory tagging

Extend the "x" and "print" commands to make use of memory tagging
functionality, if supported by the architecture.

The "print" command will point out any possible tag mismatches it finds
when dealing with pointers, in case such a pointer is tagged.  No additional
modifiers are needed.

Suppose we have a pointer "p" with value 0x1234 (logical tag 0x0) and that we
have an allocation tag of 0x1 for that particular area of memory. This is the
expected output:

(gdb) p/x p
Logical tag (0x0) does not match the allocation tag (0x1).
$1 = 0x1234

The "x" command has a new 'm' modifier that will enable displaying of
allocation tags alongside the data dump.  It will display one allocation
tag per line.

AArch64 has a tag granule of 16 bytes, which means we can have one tag for
every 16 bytes of memory. In this case, this is what the "x" command will
display with the new 'm' modifier:

(gdb) x/32bxm p
<Allocation Tag 0x1 for range [0x1230,0x1240)>
0x1234: 0x01 0x02 0x00 0x00 0x00 0x00 0x00 0x00
0x123c: 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00
<Allocation Tag 0x1 for range [0x1240,0x1250)>
0x1244: 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00
0x124c: 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00

(gdb) x/4gxm a
<Allocation Tag 0x1 for range [0x1230,0x1240)>
0x1234: 0x0000000000000201 0x0000000000000000
<Allocation Tag 0x1 for range [0x1240,0x1250)>
0x1244: 0x0000000000000000 0x0000000000000000

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* printcmd.c (decode_format): Handle the 'm' modifier.
(do_examine): Display allocation tags when required/supported.
(should_validate_memtags): New function.
(print_command_1): Display memory tag mismatches.
* valprint.c (show_memory_tag_violations): New function.
(value_print_option_defs): Add new option "memory-tag-violations".
(user_print_options) <memory_tag_violations>: Initialize to 1.
* valprint.h (struct format_data) <print_tags>: New field.
(value_print_options) <memory_tag_violations>: New field.

gdb/testsuite/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* gdb.base/options.exp: Adjust for new print options.
* gdb.base/with.exp: Likewise.

Documentation for the new mtag commands

Document the new "memory-tag" command prefix and all of its subcommands.

gdb/doc/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* gdb.texinfo (Memory Tagging): New subsection and node.
(AArch64 Memory Tagging Extension): New subsection.

New memory-tag commands

Add new commands under the "memory-tag" prefix to allow users to inspect,
modify and check memory tags in different ways.

The available subcommands are the following:

- memory-tag print-logical-tag <expression>: Prints the logical tag for a
  particular address.

- memory-tag withltag <expression> <tag>: Prints the address tagged with the
  logical tag <tag>.

- memory-tag print-allocation-tag <expression>: Prints the allocation tag for
  a particular address.

- memory-tag setatag <expression> <length> <tags>: Sets one or more allocation
  tags to the specified tags.

- memory-tag check <expression>: Checks if the logical tag in <address>
  matches its allocation tag.

These commands make use of the memory tagging gdbarch methods, and are still
available, but disabled, when memory tagging is not supported by the
architecture.

I've pondered about a way to make these commands invisible when memory tagging
is not available, but given the check is at runtime (and support may come and go
based on a process' configuration), that is a bit too late in the process to
either not include the commands or get rid of them.

Ideas are welcome.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* printcmd.c: Include gdbsupport/rsp-low.h.
(memory_tag_list): New static global.
(process_print_command_args): Factored out of
print_command_1.
(print_command_1): Use process_print_command_args.
(show_addr_not_tagged, show_memory_tagging_unsupported)
(memory_tag_command, memory_tag_print_tag_command)
(memory_tag_print_logical_tag_command)
(memory_tag_print_allocation_tag_command, parse_with_logical_tag_input)
(memory_tag_with_logical_tag_command, parse_set_allocation_tag_input)
(memory_tag_set_allocation_tag_command, memory_tag_check_command): New
functions.
(_initialize_printcmd): Add "memory-tag" prefix and subcommands.

gdbsupport/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* rsp-low.cc (fromhex, hex2bin): Move to ...
* common-utils.cc: ... here.
(fromhex) Change error message text to not be RSP-specific.
* rsp-low.h (fromhex, hex2bin): Move to ...
* common-utils.h: ... here.

AArch64: Add MTE register set support for core files

This patch handles the tagged_addr_ctrl register that is exported when
generating a core file.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-tdep.c
(aarch64_linux_iterate_over_regset_sections): Handle MTE register set.
* aarch64-linux-tdep.h (AARCH64_LINUX_SIZEOF_MTE_REGSET): Define.

AArch64: Add gdbserver MTE support

Adds the AArch64-specific memory tagging support (MTE) by implementing the
required hooks and checks for GDBserver.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* Makefile.in (SFILES): Add /../gdb/nat/aarch64-mte-linux-ptrace.c.
* configure.srv (aarch64*-*-linux*): Add arch/aarch64-mte-linux.o and
nat/aarch64-mte-linux-ptrace.o.
* linux-aarch64-low.cc: Include nat/aarch64-mte-linux-ptrace.h.
(class aarch64_target) <supports_memory_tagging>
<fetch_memtags, store_memtags>: New method overrides.
(aarch64_target::supports_memory_tagging)
(aarch64_target::fetch_memtags)
(aarch64_target::store_memtags): New methods.

AArch64: Report tag violation error information

Whenever a memory tag violation occurs, we get a SIGSEGV. Additional
information can be obtained through the siginfo data structure.

For AArch64 the Linux kernel may expose the fault address and tag
information, if we have a synchronous event. Otherwise there is
no fault address available.

The synchronous event looks like this:

--
(gdb) continue
Continuing.

Program received signal SIGSEGV, Segmentation fault
Memory tag violation while accessing address 0x0500fffff7ff8000
Allocation tag 0x1.
Logical tag 0x5
--

The asynchronous event looks like this:

--
(gdb) continue
Continuing.

Program received signal SIGSEGV, Segmentation fault
Memory tag violation
Fault address unavailable.
--

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-tdep.c
(aarch64_linux_report_signal_info): New function.
(aarch64_linux_init_abi): Register
aarch64_linux_report_signal_info as the report_signal_info hook.
* arch/aarch64-linux.h (SEGV_MTEAERR): Define.
(SEGV_MTESERR): Define.

AArch64: Add unit testing for logical tag set/get operations

Add some unit testing to exercise setting/getting logical tags in the
AArch64 implementation.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-tdep.c: Include gdbsupport/selftest.h.
(aarch64_linux_ltag_tests): New function.
(_initialize_aarch64_linux_tdep): Register aarch64_linux_ltag_tests.

AArch64: Implement the memory tagging gdbarch hooks

This patch implements the memory tagging gdbarch hooks for AArch64, for
the MTE feature.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-tdep.c: Include target.h, arch-utils.h, value.h.
(aarch64_mte_get_atag, aarch64_linux_tagged_address_p)
(aarch64_linux_memtag_mismatch_p, aarch64_linux_set_memtags)
(aarch64_linux_get_memtag, aarch64_linux_memtag_to_string): New
functions.
(aarch64_linux_init_abi): Initialize MTE-related gdbarch hooks.
* arch/aarch64-mte-linux.c (aarch64_mte_make_ltag_bits)
(aarch64_mte_make_ltag, aarch64_linux_set_ltag)
(aarch64_linux_get_ltag): New functions.
* arch/aarch64-mte-linux.h (AARCH64_MTE_LOGICAL_TAG_START_BIT)
(AARCH64_MTE_LOGICAL_MAX_VALUE): Define.
(aarch64_mte_make_ltag_bits, aarch64_mte_make_ltag)
(aarch64_mte_set_ltag, aarch64_mte_get_ltag): New prototypes.

Refactor parsing of /proc/<pid>/smaps

The Linux kernel exposes the information about MTE-protected pages via the
proc filesystem, more specifically through the smaps file.

What we're looking for is a mapping with the 'mt' flag, which tells us that
mapping was created with a PROT_MTE flag and, thus, is capable of using memory
tagging.

We already parse that file for other purposes (core file
generation/filtering), so this patch refactors the code to make the parsing
of the smaps file reusable for memory tagging.

The function linux_address_in_memtag_page uses the refactored code to allow
querying for memory tag support in a particular address, and it gets used in the
next patch.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* linux-tdep.c (struct smaps_vmflags) <memory_tagging>: New flag
bit.
(struct smaps_data): New struct.
(decode_vmflags): Handle the 'mt' flag.
(parse_smaps_data): New function, refactored from
linux_find_memory_regions_full.
(linux_address_in_memtag_page): New function.
(linux_find_memory_regions_full): Refactor into parse_smaps_data.
* linux-tdep.h (linux_address_in_memtag_page): New prototype.

Convert char array to std::string in linux_find_memory_regions_full

This is a quick cleanup that removes the use of fixed-length char arrays and
uses std::string instead.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* linux-tdep.c (linux_find_memory_regions_full): Use std::string
instead of char arrays.

AArch64: Implement memory tagging target methods for AArch64

The patch implements the memory tagging target hooks for AArch64, so we
can handle MTE.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* Makefile.in (ALL_64_TARGET_OBS): Add arch/aarch64-mte-linux.o.
(HFILES_NO_SRCDIR): Add arch/aarch64-mte-linux.h and
nat/aarch64-mte-linux-ptrace.h.
* aarch64-linux-nat.c: Include nat/aarch64-mte-linux-ptrace.h.
(aarch64_linux_nat_target) <supports_memory_tagging>: New method
override.
<fetch_memtags>: New method override.
<store_memtags>: New method override.
(aarch64_linux_nat_target::supports_memory_tagging): New method.
(aarch64_linux_nat_target::fetch_memtags): New method.
(aarch64_linux_nat_target::store_memtags): New method.
* arch/aarch64-mte-linux.c: New file.
* arch/aarch64-mte-linux.h: Include gdbsupport/common-defs.h.
(AARCH64_MTE_GRANULE_SIZE): Define.
(aarch64_memtag_type): New enum.
(aarch64_mte_get_tag_granules): New prototype.
* configure.nat (NATDEPFILES): Add nat/aarch64-mte-linux-ptrace.o.
* configure.tgt (aarch64*-*-linux*): Add arch/aarch64-mte-linux.o.
* nat/aarch64-mte-linux-ptrace.c: New file.
* nat/aarch64-mte-linux-ptrace.h: New file.

AArch64: Add MTE ptrace requests

This patch adds the required ptrace request definitions into a new include
file that will be used by the next patches.

They are PTRACE_PEEKMTETAGS and PTRACE_POKEMTETAGS.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* Makefile.in (HFILES_NO_SRCDIR): Add nat/aarch64-mte-linux-ptrace.h.
* nat/aarch64-mte-linux-ptrace.h: New file.

AArch64: Add MTE register set support for GDB and gdbserver

AArch64 MTE support in the Linux kernel exposes a new register
through ptrace.  This patch adds the required code to support it.

include/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* elf/common.h (NT_ARM_TAGGED_ADDR_CTRL): Define.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-nat.c (fetch_mteregs_from_thread): New function.
(store_mteregs_to_thread): New function.
(aarch64_linux_nat_target::fetch_registers): Update to call
fetch_mteregs_from_thread.
(aarch64_linux_nat_target::store_registers): Update to call
store_mteregs_to_thread.
* aarch64-tdep.c (aarch64_mte_register_names): New struct.
(aarch64_cannot_store_register): Handle MTE registers.
(aarch64_gdbarch_init): Initialize and setup MTE registers.
* aarch64-tdep.h (gdbarch_tdep) <mte_reg_base>: New field.
<has_mte>: New method.
* arch/aarch64-linux.h (AARCH64_LINUX_SIZEOF_MTE): Define.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* linux-aarch64-low.cc (aarch64_fill_mteregset): New function.
(aarch64_store_mteregset): New function.
(aarch64_regsets): Add MTE register set entry.
(aarch64_sve_regsets): Add MTE register set entry.

AArch64: Add target description/feature for MTE registers

This patch adds a target description and feature "mte" for aarch64.

It includes one new register, tag_ctl, that can be used to configure the
tag generation rules and sync/async modes.  It is 64-bit in size.

The patch also adjusts the code that creates the target descriptions at
runtime based on CPU feature checks.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* aarch64-linux-nat.c
(aarch64_linux_nat_target::read_description): Take MTE flag into
account.
Slight refactor to hwcap flag checking.
* aarch64-linux-tdep.c
(aarch64_linux_core_read_description): Likewise.
* aarch64-tdep.c (tdesc_aarch64_list): Add one more dimension for
MTE.
(aarch64_read_description): Add mte_p parameter and update to use it.
Update the documentation.
(aarch64_gdbarch_init): Update call to aarch64_read_description.
* aarch64-tdep.h (aarch64_read_description): Add mte_p parameter.
* arch/aarch64.c: Include ../features/aarch64-mte.c.
(aarch64_create_target_description): Add mte_p parameter and update
the code to use it.
* arch/aarch64.h (aarch64_create_target_description): Add mte_p
parameter.
* features/Makefile (FEATURE_XMLFILES): Add aarch64-mte.xml.
* features/aarch64-mte.c: New file, generated.
* features/aarch64-mte.xml: New file.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* linux-aarch64-ipa.cc (get_ipa_tdesc): Update call to
aarch64_linux_read_description.
(initialize_low_tracepoint): Likewise.
* linux-aarch64-low.cc (aarch64_target::low_arch_setup): Take MTE flag
into account.
* linux-aarch64-tdesc.cc (tdesc_aarch64_list): Add one more dimension
for MTE.
(aarch64_linux_read_description): Add mte_p parameter and update to
use it.
* linux-aarch64-tdesc.h (aarch64_linux_read_description): Add mte_p
parameter.

AArch64: Add MTE CPU feature check support

This patch is a preparation for the next patches implementing MTE. It just adds
a HWCAP2 constant for MTE, creates a new generic arch/aarch64-mte-linux.h file
and includes that file in the source files that will use it.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* Makefile.in (HFILES_NO_SRCDIR): Add arch/aarch64-mte-linux.h.
* aarch64-linux-nat.c: Include arch/aarch64-mte-linux.h.
* aarch64-linux-tdep.c: Likewise
* arch/aarch64-mte-linux.h: New file.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* linux-aarch64-low.cc: Include arch/aarch64-mte-linux.h.

Documentation for memory tagging remote packets

Document the remote packet changes to support memory tagging.

gdb/doc/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* gdb.texinfo (General Query Packets): Document qMemTags and
QMemTags.  Document the "memory-tagging" feature.
(ARM-Specific Protocol Details): Document memory tag types.

Unit tests for gdbserver memory tagging remote packets

Add some unit testing to exercise the functions handling the qMemTags and
QMemTags packets as well as feature support.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* server.cc (test_memory_tagging_functions): New function.
(captured_main): Register test_memory_tagging_functions.

GDBserver remote packet support for memory tagging

This patch adds the generic remote bits to gdbserver so it can check for memory
tagging support and handle fetch tags and store tags requests.

gdbserver/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* remote-utils.cc (decode_m_packet_params): Renamed from ...
(decode_m_packet): ... this, which now calls decode_m_packet_params.
Make char * param/return const char *.
(decode_M_packet): Use decode_m_packet_params and make char * param
const char *.
* remote-utils.h (decode_m_packet_params): New prototype.
(decode_m_packet): Constify char pointers.
(decode_M_packet): Likewise.
* server.cc (create_fetch_memtags_reply)
(parse_store_memtags_request): New
functions.
(handle_general_set): Handle the QMemTags packet.
(parse_fetch_memtags_request): New function.
(handle_query): Handle the qMemTags packet and advertise memory
tagging support.
(captured_main): Initialize memory tagging flag.
* server.h (struct client_state): Initialize memory tagging flag.
* target.cc (process_stratum_target::supports_memory_tagging)
(process_stratum_target::fetch_memtags)
(process_stratum_target::store_memtags): New methods.
* target.h: Include gdbsupport/byte-vector.h.
(class process_stratum_target) <supports_memory_tagging>
<fetch_memtags, store_memtags>: New class virtual methods.
(target_supports_memory_tagging): Define.

Unit testing for GDB-side remote memory tagging handling

Include some unit testing for the functions handling the new qMemTags and
QMemTags packets.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* remote: Include gdbsupport/selftest.h.
(test_memory_tagging_functions): New function.
(_initialize_remote): Register test_memory_tagging_functions.

Add GDB-side remote target support for memory tagging

This patch adds memory tagging support to GDB's remote side, with
packet string checks, new packet support and an implementation of
the two new tags methods fetch_memtags and store_memtags.

GDBserver needs to know how to read/write allocation tags, since that is
done via ptrace.  It doesn't need to know about logical tags.

The new packets are:

qMemTags:<address>,<length>:<type>
--

Reads tags of the specified type from the address range
[<address>, <address + length>)

QMemTags:<address>,<length>:<type>:<uninterpreted tag bytes>
--
Writes the tags of specified type represented by the uninterpreted bytes to
the address range [<address>, <address + length>).

The interpretation of what to do with the tag bytes is up to the arch-specific
code.

Note that these new packets consider the case of packet size overflow as an
error, given the common use case is to read/write only a few memory tags at
a time.  Having to use a couple new packets for multi-part transfers wouldn't
make sense for the little use it would have.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* remote.c (PACKET_memory_tagging_feature): New enum.
(remote_memory_tagging_p): New function.
(remote_protocol_features): New "memory-tagging" entry.
(remote_target::remote_query_supported): Handle memory tagging
support.
(remote_target::supports_memory_tagging): Implement.
(create_fetch_memtags_request, parse_fetch_memtags_reply)
(create_store_memtags_request): New functions.
(remote_target::fetch_memtags): Implement.
(remote_target::store_memtags): Implement.
(_initialize_remote): Add new "memory-tagging-feature"
config command.

New gdbarch memory tagging hooks

We need some new gdbarch hooks to help us manipulate memory tags without having
to have GDB call the target methods directly.

This patch adds the following hooks:

gdbarch_memtag_to_string
--
Returns a printable string corresponding to the tag.

gdbarch_tagged_address_p
--
Checks if a particular address is protected with memory tagging.

gdbarch_memtag_matches_p
--
Checks if the logical tag of a pointer and the allocation tag from the address
the pointer points to matches.

gdbarch_set_memtags:
--
Sets either the allocation tag or the logical tag for a particular value.

gdbarch_get_memtag:
--
Gets either the allocation tag or the logical tag for a particular value.

gdbarch_memtag_granule_size
--
Sets the memory tag granule size, which represents the number of bytes a
particular allocation tag covers. For example, this is 16 bytes for
AArch64's MTE.

I've used struct value as opposed to straight CORE_ADDR so other architectures
can use the infrastructure without having to rely on a particular type for
addresses/pointers.  Some architecture may use pointers of 16 bytes that don't
fit in a CORE_ADDR, for example.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* arch-utils.c (default_memtag_to_string, default_tagged_address_p)
(default_memtag_matches_p, default_set_memtags)
(default_get_memtag): New functions.
* arch-utils.h (default_memtag_to_string, default_tagged_address_p)
(default_memtag_matches_p, default_set_memtags)
(default_get_memtag): New prototypes.
* gdbarch.c: Regenerate.
* gdbarch.h: Regenerate.
* gdbarch.sh (memtag_to_string, tagged_address_p, memtag_matches_p)
(set_memtags, get_memtag, memtag_granule_size): New gdbarch hooks.
(enum memtag_type): New enum.

New target methods for memory tagging support

This patch starts adding some of the generic pieces to accomodate memory
tagging.

We have three new target methods:

- supports_memory_tagging: Checks if the target supports memory tagging. This
  defaults to false for targets that don't support memory tagging.

- fetch_memtags: Fetches the allocation tags associated with a particular
  memory range [address, address + length).

  The default is to return 0 without returning any tags. This should only
  be called if memory tagging is supported.

- store_memtags: Stores a set of allocation tags for a particular memory
  range [address, address + length).

  The default is to return 0. This should only
  be called if memory tagging is supported.

gdb/ChangeLog:

2021-03-24  Luis Machado  <luis.machado@linaro.org>

* remote.c (remote_target) <supports_memory_tagging>: New method
override.
<fetch_memtags>: New method override.
<store_memtags>: New method override.
(remote_target::supports_memory_tagging): New method.
(remote_target::fetch_memtags): New method.
(remote_target::store_memtags): New method.
* target-delegates.c: Regenerate.
* target.h (struct target_ops) <supports_memory_tagging>: New virtual
method.
<fetch_memtags>: New virtual method.
<store_memtags>: New virtual method.
(target_supports_memory_tagging): Define.
(target_fetch_memtags): Define.
(target_store_memtags): Define.
* target-debug.h (target_debug_print_size_t)
(target_debug_print_const_gdb_byte_vector_r)
(target_debug_print_gdb_byte_vector_r): New functions.

x86: derive opcode length from opcode value

In the majority of cases we can easily determine the length from the
encoding, irrespective of whether a prefix is specified there as well.
We further don't even need to record the value in the table entries, as
it's easy enough to determine it (without any guesswork, unless an insn
with major opcode 00 appeared that requires a 2nd opcode byte to be
specified explicitly) when installing the chosen template for further
processing.

Should an encoding appear which
- has a major opcode byte of 66, F3, or F2,
- requires a 2nd opcode byte to be specified explicitly,
- doesn't have a mandatory prefix
we'd need to convert all templates presently encoding a mandatory prefix
this way to the Prefix_0X<nn> model to eliminate the respective guessing
i386-gen does.

x86: derive mandatory prefix attribute from base opcode

Just like is already done for legacy encoded insns, record the mandatory
prefix information in the respective opcode modifier field. Do this
without changing the source table, but rather by deriving the values from
their existing source representation.

x86: don't use opcode_length to identify pseudo prefixes

This is in preparation of opcode_length going away as a field in the
templates. Identify pseudo prefixes by a base opcode of zero instead:
No real prefix has an opcode of zero. This at the same time allows
dropping a curious special case from i386-gen.

Since most attributes are identical for all pseudo prefixes, take the
opportunity and also template them.

Automatic date update in version.in

x86: re-number PREFIX_0X<nn>

In preparation to use PREFIX_0X<nn> attributes also in VEX/XOP/EVEX
encoding templates, renumber the pseudo-enumerators such that their
values can then also be used directly in the respective prefix bit
fields.

x86: re-order two fields of struct insn_template

To facilitate a subsequent table parser change, re-order CPU flags and
opcode modifier fields. No functional change intended.

x86: split opcode prefix and opcode space representation

Commit 8b65b8953af2 ("x86: Remove the prefix byte from non-VEX/EVEX
base_opcode") used the opcodeprefix field for two distinct purposes. In
preparation of having VEX/XOP/EVEX and non-VEX templates become similar
in the representatioon of both encoding space and opcode prefixes, split
the field to have a separate one holding an insn's opcode space.

gdb: remote target_longname

I noticed it was unused.

gdb/ChangeLog:

* target.h (target_longname): Remove.

Change-Id: Id4f514ea038a6d8d40e11179db587b11793cbbd8

gdb: remove target_is_pushed free function

Same principle as the previous patches.

gdb/ChangeLog:

* target.h (target_is_pushed): Remove, update callers to use
inferior::target_is_pushed instead.
* target.c (target_is_pushed): Remove.

Change-Id: I9862e6205acc65672da807cbe4b46cde009e7b9d

gdb: remove push_target free functions

Same as the previous patch, but for the push_target functions.

The implementation of the move variant is moved to a new overload of
inferior::push_target.

gdb/ChangeLog:

* target.h (push_target): Remove, update callers to use
inferior::push_target.
* target.c (push_target): Remove.
* inferior.h (class inferior) <push_target>: New overload.

Change-Id: I5a95496666278b8f3965e5e8aecb76f54a97c185

gdb: remove unpush_target free function

unpush_target unpushes the passed-in target from the current inferior's
target stack.  Calling it is therefore an implicit dependency on the
current global inferior.  Remove that function and make the callers use
the inferior::unpush_target method directly.  This sometimes allows
using the inferior from the context rather than the global current
inferior.

target_unpusher::operator() now needs to be implemented in target.c,
otherwise target.h and inferior.h both need to include each other, and
that wouldn't work.

gdb/ChangeLog:

* target.h (unpush_target): Remove, update all callers
to use `inferior::unpush_target` instead.
(struct target_unpusher) <operator()>: Just declare.
* target.c (unpush_target): Remove.
(target_unpusher::operator()): New.

Change-Id: Ia5172dfb3f373e0a75b991885b50322ca2142a8c

x86: don't open-code PREFIX_NONE

Use the constant rather than literal zero. While at it fold two
conditionals (using the same base opcode and prefix) in load_insn_p().

x86: unbreak certain MPX insn operand forms

Commit 8b65b8953af2 ("x86: Remove the prefix byte from non-VEX/EVEX
base_opcode") dropped the mandatory prefix bytes from legacy encoded
insn templates, but failed to also adjust affected MPX-specific checks
in two places.

For the expressions to remain halfway readable, introduce local
variables to hold current_templates->start.

bfd: avoid "shadowing" of glibc function name

Old enough glibc has an (unguarded) declaration of index() in string.h,
which triggers a "shadows a global declaration" warning.

Automatic date update in version.in

gdb: handle invalid DWARF  when compilation unit is missing

Replace an abort call in process_psymtab_comp_unit with a real error,
and add a test to cover this case.  The case is question is when badly
formed DWARF is missing a DW_TAG_compile_unit, DW_TAG_partial_unit, or
DW_TAG_type_unit as its top level tag.

I then tested with --target_board=readnow and added additional code to
also validate the top-level tag in this case.

I added an assert that would trigger for the readnow case before I
added the fix.  I suspect there's lots of places where badly formed
DWARF could result in the builder being nullptr when it shouldn't be,
but I only added this one assert, as this is the one that would have
helped me in this case.

gdb/ChangeLog:

* dwarf2/read.c (process_psymtab_comp_unit): Replace abort with an
error.
(process_full_comp_unit): Validate the top-level tag before
processing the first DIE.
(read_func_scope): Ensure we have a valid builder.

gdb/testsuite/ChangeLog:

* gdb.dwarf2/dw2-missing-cu-tag.c: New file.
* gdb.dwarf2/dw2-missing-cu-tag.exp: New file.

gdb/objc: make objc_demangle a member function of objc_language

Makes the objc_demangle helper function a member function of
objc_language (by renaming it to be the demangle_symbol member
function).

I also fixed some of the obvious coding standard violations in
obj_demangle, so the '&&' operators are now at the start of the line,
not the end.  Comparison to nullptr are now made explicit, as are
comparisons to the null character.

There should be no user visible changes after this commit.

gdb/ChangeLog:

* objc-lang.c (objc_demangle): Renamed to
objc_language::demangle_symbol, and moved later in the file.
(objc_language::sniff_from_mangled_name): Call demangle_symbol
member function.
(objc_language::demangle_symbol): Defined outside of class
declaration.  The definition is the old objc_demangle with NULL
changed to nullptr, and if conditions relating to nullptr pointers
or null character checks made explicit.
* objc-lang.h (objc_demangle): Delete declaration.

Add startswith function and use it instead of CONST_STRNEQ.

bfd/ChangeLog:

* bfd-in.h (startswith): Add startswith function.
(CONST_STRNEQ): Remove.
* bfd-in2.h (startswith): Regenerate with make headers.
* archive.c (bfd_slurp_armap): Replace usage of CONST_STRNEQ with startswith.
(_bfd_slurp_extended_name_table): Likewise.
* archive64.c (_bfd_archive_64_bit_slurp_armap): Likewise.
* bfd.c (bfd_get_sign_extend_vma): Likewise.
(bfd_convert_section_size): Likewise.
(bfd_convert_section_contents): Likewise.
* coff-stgo32.c (go32exe_create_stub): Likewise.
(go32exe_check_format): Likewise.
* coffcode.h (styp_to_sec_flags): Likewise.
(GNU_DEBUGALTLINK): Likewise.
* coffgen.c (_bfd_coff_section_already_linked): Likewise.
(coff_gc_sweep): Likewise.
(bfd_coff_gc_sections): Likewise.
* cofflink.c (coff_link_add_symbols): Likewise.
(process_embedded_commands): Likewise.
* compress.c (bfd_is_section_compressed_with_header): Likewise.
(bfd_init_section_decompress_status): Likewise.
* dwarf2.c (find_debug_info): Likewise.
(place_sections): Likewise.
* ecoff.c (_bfd_ecoff_slurp_armap): Likewise.
* elf-m10300.c (_bfd_mn10300_elf_size_dynamic_sections): Likewise.
* elf.c (_bfd_elf_make_section_from_shdr): Likewise.
(assign_section_numbers): Likewise.
(elfcore_grok_win32pstatus): Likewise.
* elf32-arm.c (cmse_scan): Likewise.
(elf32_arm_gc_mark_extra_sections): Likewise.
(elf32_arm_size_dynamic_sections): Likewise.
(is_arm_elf_unwind_section_name): Likewise.
* elf32-bfin.c (bfin_size_dynamic_sections): Likewise.
* elf32-cr16.c (_bfd_cr16_elf_size_dynamic_sections): Likewise.
* elf32-cris.c (elf_cris_size_dynamic_sections): Likewise.
* elf32-csky.c (csky_elf_size_dynamic_sections): Likewise.
* elf32-hppa.c (elf32_hppa_size_dynamic_sections): Likewise.
* elf32-iq2000.c (iq2000_elf_check_relocs): Likewise.
* elf32-lm32.c (lm32_elf_size_dynamic_sections): Likewise.
* elf32-m32r.c (m32r_elf_size_dynamic_sections): Likewise.
* elf32-m68k.c (elf_m68k_size_dynamic_sections): Likewise.
* elf32-metag.c (elf_metag_size_dynamic_sections): Likewise.
* elf32-msp430.c (msp430_elf_relax_delete_bytes): Likewise.
* elf32-nios2.c (nios2_elf32_size_dynamic_sections): Likewise.
* elf32-or1k.c (or1k_elf_size_dynamic_sections): Likewise.
* elf32-ppc.c (ppc_elf_size_dynamic_sections): Likewise.
* elf32-s390.c (elf_s390_size_dynamic_sections): Likewise.
* elf32-score.c (s3_bfd_score_elf_size_dynamic_sections): Likewise.
* elf32-score7.c (s7_bfd_score_elf_size_dynamic_sections): Likewise.
* elf32-sh.c (sh_elf_size_dynamic_sections): Likewise.
* elf32-tic6x.c (is_tic6x_elf_unwind_section_name): Likewise.
(elf32_tic6x_size_dynamic_sections): Likewise.
* elf32-vax.c (elf_vax_size_dynamic_sections): Likewise.
* elf32-xtensa.c (elf_xtensa_size_dynamic_sections): Likewise.
(xtensa_is_insntable_section): Likewise.
(xtensa_is_littable_section): Likewise.
(xtensa_is_proptable_section): Likewise.
(xtensa_property_section_name): Likewise.
(xtensa_callback_required_dependence): Likewise.
* elf64-alpha.c (elf64_alpha_size_dynamic_sections): Likewise.
* elf64-hppa.c (elf64_hppa_size_dynamic_sections): Likewise.
* elf64-ia64-vms.c (is_unwind_section_name): Likewise.
(get_reloc_section): Likewise.
(elf64_ia64_size_dynamic_sections): Likewise.
(elf64_ia64_object_p): Likewise.
* elf64-mmix.c (mmix_elf_add_symbol_hook): Likewise.
* elf64-ppc.c (ppc64_elf_size_dynamic_sections): Likewise.
* elf64-s390.c (elf_s390_size_dynamic_sections): Likewise.
* elflink.c (elf_link_add_object_symbols): Likewise.
(_bfd_elf_gc_mark_extra_sections): Likewise.
(bfd_elf_parse_eh_frame_entries): Likewise.
(_bfd_elf_section_already_linked): Likewise.
* elfnn-aarch64.c (elfNN_aarch64_size_dynamic_sections): Likewise.
* elfnn-ia64.c (is_unwind_section_name): Likewise.
(elfNN_ia64_size_dynamic_sections): Likewise.
(elfNN_ia64_object_p): Likewise.
* elfxx-mips.c (FN_STUB_P): Likewise.
(CALL_STUB_P): Likewise.
(CALL_FP_STUB_P): Likewise.
(_bfd_mips_elf_section_from_shdr): Likewise.
(_bfd_mips_elf_fake_sections): Likewise.
(_bfd_mips_elf_size_dynamic_sections): Likewise.
(_bfd_mips_final_write_processing): Likewise.
(_bfd_mips_elf_final_link): Likewise.
* elfxx-sparc.c (_bfd_sparc_elf_size_dynamic_sections): Likewise.
* elfxx-x86.c (elf_i386_is_reloc_section): Likewise.
(elf_x86_64_is_reloc_section): Likewise.
* hpux-core.c (thread_section_p): Likewise.
* libcoff.h (bfd_pei_p): Likewise.
* linker.c (REAL): Likewise.
(unwrap_hash_lookup): Likewise.
(_bfd_generic_link_add_one_symbol): Likewise.
* mmo.c (mmo_internal_write_section): Likewise.
* osf-core.c (osf_core_core_file_p): Likewise.
* pef.c (bfd_pef_print_symbol): Likewise.
* pei-x86_64.c (pex64_print_all_pdata_sections): Likewise.
* som.c (som_slurp_symbol_table): Likewise.
(som_slurp_armap): Likewise.
* wasm-module.c (wasm_compute_custom_section_file_position): Likewise.

binutils/ChangeLog:

* dlltool.c (scan_drectve_symbols): Replace usage of CONST_STRNEQ with startswith.
* emul_aix.c (ar_emul_aix_parse_arg): Likewise.
* objcopy.c (is_mergeable_note_section): Likewise.
* objdump.c (dump_dwarf_section): Likewise.
* prdbg.c (pr_method_type): Likewise.
(pr_class_baseclass): Likewise.
(tg_class_baseclass): Likewise.
* readelf.c (process_lto_symbol_tables): Likewise.
* stabs.c (ULLHIGH): Likewise.
(parse_stab_argtypes): Likewise.
(stab_demangle_function_name): Likewise.

gas/ChangeLog:

* config/tc-i386.c (md_parse_option): Replace usage of CONST_STRNEQ with startswith.
(x86_64_section_word): Likewise.
* config/tc-sparc.c (md_parse_option): Likewise.

gdb/ChangeLog:

* arm-tdep.c (show_disassembly_style_sfunc): Replace usage of CONST_STRNEQ with startswith.
(_initialize_arm_tdep): Likewise.

ld/ChangeLog:

* emultempl/aix.em: Replace usage of CONST_STRNEQ with startswith.
* emultempl/beos.em: Likewise.
* emultempl/elf.em: Likewise.
* emultempl/pe.em: Likewise.
* emultempl/pep.em: Likewise.
* emultempl/xtensaelf.em: Likewise.
* ldctor.c (ctor_prio): Likewise.
* ldelf.c (ldelf_try_needed): Likewise.
(ldelf_parse_ld_so_conf): Likewise.
(ldelf_after_open): Likewise.
(output_rel_find): Likewise.
(ldelf_place_orphan): Likewise.
* ldfile.c (ldfile_add_library_path): Likewise.
* ldlang.c (lang_add_input_file): Likewise.
* ldmain.c (get_sysroot): Likewise.
(get_emulation): Likewise.
(add_archive_element): Likewise.
* ldwrite.c (unsplittable_name): Likewise.
(clone_section): Likewise.
* lexsup.c (parse_args): Likewise.
* pe-dll.c (is_import): Likewise.
(pe_implied_import_dll): Likewise.

opcodes/ChangeLog:

* aarch64-dis.c (parse_aarch64_dis_option): Replace usage of CONST_STRNEQ with startswith.
* arc-dis.c (parse_option): Likewise.
* arm-dis.c (parse_arm_disassembler_options): Likewise.
* cris-dis.c (print_with_operands): Likewise.
* h8300-dis.c (bfd_h8_disassemble): Likewise.
* i386-dis.c (print_insn): Likewise.
* ia64-gen.c (fetch_insn_class): Likewise.
(parse_resource_users): Likewise.
(in_iclass): Likewise.
(lookup_specifier): Likewise.
(insert_opcode_dependencies): Likewise.
* mips-dis.c (parse_mips_ase_option): Likewise.
(parse_mips_dis_option): Likewise.
* s390-dis.c (disassemble_init_s390): Likewise.
* wasm32-dis.c (parse_wasm32_disassembler_options): Likewise.

gdb/testsuite: use the correct .debug_str section name for DW_FORM_strp

When handling DWARF attributes of the form DW_FORM_strp the strings
should be placed in the .debug_str section, not .debug_string as they
currently are by the DWARF assembler (in lib/dwarf.exp).

I've added a test.  This is as much to test the DWARF generator as it
is to test GDB as GCC makes frequent use of DW_FORM_strp so we can be
pretty sure this part of GDB is already well tested.

gdb/testsuite/ChangeLog:

* gdb.dwarf2/dw2-using-debug-str.c: New file.
* gdb.dwarf2/dw2-using-debug-str.exp: New file.
* lib/dwarf.exp (Dwarf::DW_FORM_strp): Create .debug_str section,
not .debug_string.

gdbserver: convert the global dll list into a process_info field

The 'all_dlls' list is global.  This would cause the complete dll list
to be reported for individual processes.  Move the list into the
process_info struct.

Currently the dll list is used only by the win32-low target, which
does not support the multi-process feature.  Therefore, it practically
does not matter whether the list is global or per-process.  However,
there may be targets that are outside the binutils-gdb repo (e.g. we,
at Intel, have such a target) that have multi-process and use the dll
list.  So, it makes sense to do the right thing.

gdbserver/ChangeLog:
2021-03-22  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* inferiors.h (struct process_info) <all_dlls, dlls_changed>: New
fields.
* dll.h (loaded_dll)
(unloaded_dll): Declare an overloaded version that takes a proc
parameter.
* dll.cc (loaded_dll)
(unloaded_dll): Implement the overloaded versions.
(clear_dlls): Clear all process' dll lists.
(all_dlls, dlls_changed): Remove the global variables.
* remote-utils.cc (prepare_resume_reply): Update to consider a dll
list per proc.
* server.cc (handle_qxfer_libraries): Ditto.
(handle_v_attach): Ditto.
(captured_main): Ditto.

bfd: add translation wrappers to PE image section RVA diagnostics

In commit 87fa7d568ddd ("bfd: don't silently wrap or truncate PE image
section RVAs") I forgot about the need to wrap string literals for
translation.

Automatic date update in version.in

Provide an inline startswith function in bfd.h

bfd/
* bfd-in.h (startswith): New inline.
(CONST_STRNEQ): Use startswith.
* bfd-in2.h: Regenerate.
gdbsupport/
* common-utils.h (startswith): Delete version now supplied by bfd.h.
libctf/
* ctf-impl.h: Include string.h.

Automatic date update in version.in

Allow multiple partial symbol readers per objfile

This patch finally changes gdb so that an objfile can have multiple
sources of partial symbols (or mixed partial symbols and other kinds
of indices).

This is done by having each symbol reader create its own
psymbol_functions object and add it to the 'qf' list in the objfile.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* xcoffread.c (xcoff_initial_scan): Create partial symtabs.
* symfile.c (syms_from_objfile_1, reread_symbols): Update.
* psymtab.h (make_psymbol_functions): Don't declare.
* psymtab.c (make_psymbol_functions): Remove.
(maintenance_print_psymbols): Update.
* psympriv.h (struct psymbol_functions): Add no-argument
constructor.
* objfiles.h (struct objfile) <reset_psymtabs>: Remove.
<partial_symtabs>: Remove.
* mdebugread.c (mdebug_build_psymtabs): Create partial symtabs.
* elfread.c (read_partial_symbols): Update.
(elf_symfile_read): Remove check for existing partial symbols.
Don't clear "qf".
* dwarf2/read.c (dwarf2_has_info): Remove check for existing
partial symbols.
(dwarf2_build_psymtabs): Add psymbol_functions parameter.  Create
partial symtabs.
* dwarf2/public.h (dwarf2_build_psymtabs): Add psymbol_functions
parameter.
* dbxread.c (dbx_symfile_read): Create partial symtabs.
* ctfread.c (elfctf_build_psymtabs): Create partial symtabs.

Switch objfile to hold a list of psymbol readers

This changes objfile::qf to be a forward_list, and then updates all
the uses to iterate over the list.  Note that there is still only ever
a single element in the list; this is handled by clearing the list
whenever an object is added.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_build_psymtabs): Update.
* symfile.c (syms_from_objfile_1, reread_symbols): Update.
* symfile-debug.c (objfile::has_partial_symbols)
(objfile::find_last_source_symtab)
(objfile::forget_cached_source_info)
(objfile::map_symtabs_matching_filename, objfile::lookup_symbol)
(objfile::print_stats, objfile::dump)
(objfile::expand_symtabs_for_function)
(objfile::expand_all_symtabs)
(objfile::expand_symtabs_with_fullname)
(objfile::map_matching_symbols)
(objfile::expand_symtabs_matching)
(objfile::find_pc_sect_compunit_symtab)
(objfile::map_symbol_filenames)
(objfile::find_compunit_symtab_by_address)
(objfile::lookup_global_symbol_language)
(objfile::require_partial_symbols): Update.
* psymtab.c (maintenance_print_psymbols)
(maintenance_info_psymtabs, maintenance_check_psymtabs): Update.
* objfiles.h (struct objfile) <qf>: Now a forward_list.
* objfiles.c (objfile_relocate1): Update.
* elfread.c (elf_symfile_read): Update.

Remove objfile::psymtabs

The method objfile::psymtabs is no longer used and can be removed.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* objfiles.h (struct objfile) <psymtabs>: Remove method.

Change count_psyms to be a method on psymbol_functions

This removes a use of objfile->psymtabs by changing count_psyms to be
a method on psymbol_functions.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* psymtab.c (psymbol_functions::count_psyms): Rename.
(psymbol_functions::print_stats): Update.
* psympriv.h (struct psymbol_functions) <count_psyms>: Declare
method.

Remove last objfile partial_symtab references from psymtab.c

This removes the last references to the partial_symtab via the objfile
from psymtab.c.  require_partial_symbols is now a method on
psymbol_functions, and maintenance_print_psymbols is changed to use
dynamic_cast to verify that it is examining partial symbols.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* psymtab.c (psymbol_functions::require_partial_symbols): Rename.
(psymbol_functions::find_pc_sect_psymtab): Rename.
(psymbol_functions::find_pc_sect_compunit_symtab)
(maintenance_print_psymbols, maintenance_check_psymtabs): Update.
* psympriv.h (struct psymbol_functions) <require_partial_symbols>:
Declare new method.
<get_partial_symtabs, find_pc_sect_psymtab>: Likewise.

Add partial_symtabs parameter to psymtab construction functions

This adds a partial_symtabs parameter to the psymtab constructors and
to add_psymbol.  This helps with the overall project of removing the
partial symtabs from the objfile.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* xcoffread.c (xcoff_start_psymtab): Add partial_symtabs parameter.
(xcoff_end_psymtab, scan_xcoff_symtab): Update.
* psymtab.c (partial_symtab::partial_symtab): Add partial_symtabs
parameter.
(add_psymbol_to_bcache): Remove.
(partial_symtab::add_psymbol): Add partial_symtabs parameter.
(partial_symtab::add_psymbol, partial_symtab::partial_symtab):
Likewise.
* psympriv.h (partial_symtab): Add partial_symtabs parameter.
<add_psymbol>: Likewise.
(standard_psymtab, legacy_psymtab): Likewise.
* mdebugread.c (parse_partial_symbols): Update.
(handle_psymbol_enumerators): Add partial_symtabs parameter.
(handle_psymbol_enumerators): Update.
(new_psymtab): Add partial_symtabs parameter.
* dwarf2/read.h (dwarf2_psymtab): Add partial_symtabs parameter.
* dwarf2/read.c (dwarf2_include_psymtab): Add partial_symtabs
parameter.
(dwarf2_create_include_psymtab): Add partial_symtabs parameter.
(create_partial_symtab, add_partial_symbol, dwarf_decode_lines):
Update.
* dbxread.c (read_dbx_symtab): Update.
(start_psymtab): Add partial_symtabs parameter.
(dbx_end_psymtab): Update.
* ctfread.c (struct ctf_context) <partial_symtabs>: New member.
(ctf_psymtab): Add partial_symtabs parameter.
(create_partial_symtab, ctf_psymtab_type_cb, ctf_psymtab_var_cb):
Update.
(scan_partial_symbols): Add partial_symtabs parameter.
(scan_partial_symbols, elfctf_build_psymtabs)
(ctf_psymtab_add_enums): Update.

Introduce objfile::require_partial_symbols

This adds a new method, objfile::require_partial_symbols.  This reuses
most of the code from the old function in psymtab.c.  That function is
now made static, and simplified.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* symfile.c (read_symbols): Use objfile method.
* symfile-debug.c (objfile::require_partial_symbols): New method.
* psymtab.h (require_partial_symbols): Don't declare.
* psymtab.c (require_partial_symbols): Use objfile method.  Now
static.
(psymbol_functions::map_symtabs_matching_filename, OBJFILE)
(psymbol_functions::lookup_symbol)
(psymbol_functions::lookup_global_symbol_language)
(psymbol_functions::find_last_source_symtab)
(psymbol_functions::forget_cached_source_info)
(psymbol_functions::print_stats)
(psymbol_functions::expand_symtabs_for_function)
(psymbol_functions::expand_all_symtabs)
(psymbol_functions::expand_symtabs_with_fullname)
(psymbol_functions::map_symbol_filenames)
(psymbol_functions::map_matching_symbols)
(psymbol_functions::expand_symtabs_matching)
(psymbol_functions::find_compunit_symtab_by_address)
(maintenance_print_psymbols, maintenance_info_psymtabs)
(maintenance_check_psymtabs): Update.
* objfiles.h (struct objfile) <require_partial_symbols>: Declare
new method.

Remove sym_fns::sym_read_psymbols

Partial symbols are read via the sym_fns::sym_read_psymbols function
pointer.  In order to separate the partial symbols from the objfile,
this must instead be done via a virtual method on
quick_symbol_functions.  This patch implements this change.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* xcoffread.c (xcoff_sym_fns): Update.
* symfile.h (struct sym_fns) <sym_read_psymbols>: Remove.
* symfile-debug.c (objfile::has_partial_symbols): Use
can_lazily_read_symbols.
(debug_sym_read_psymbols): Remove.
(debug_sym_fns, install_symfile_debug_logging): Update.
* quick-symbol.h (struct quick_symbol_functions)
<can_lazily_read_symbols, read_partial_symbols>: New methods.
* psymtab.c (require_partial_symbols): Use new 'qf' methods.
* mipsread.c (ecoff_sym_fns): Update.
* machoread.c (macho_sym_fns): Update.
* elfread.c (struct lazy_dwarf_reader): New.
(elf_symfile_read): Update.
(read_psyms): Now a method of lazy_dwarf_reader.
(elf_sym_fns): Update.
(elf_sym_fns_lazy_psyms): Remove.
* dbxread.c (aout_sym_fns): Update.
* coffread.c (coff_sym_fns): Update.

Rearrange psymtab_storage construction

This changes objfile so that it doesn't construct a psymtab_storage
object until the psymtab functions are installed.  It also applies a
similar treatment to reread_symbols.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* symfile.c (syms_from_objfile_1): Call reset_psymtabs.
(reread_symbols): Move reset_psymtabs call later.
* objfiles.c (objfile::objfile): Don't initialize
partial_symtabs.

Attach partial symtab storage to psymbol_functions

Currently, the storage for partial symtabs is attached to the objfile.
Ultimately, though, this direct assocation will be removed, and the
storage will be owned by the psymbol_functions object.

This patch is a step toward this goal.  The storage is already managed
as a shared_ptr, to enable cross-objfile sharing, so this adds a
reference from the psymbol_functions, and changes some code in
psymtab.c to use this reference instead.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_build_psymtabs): Call
set_partial_symtabs.
* symfile.c (syms_from_objfile_1, reread_symbols): Update.
* psymtab.h (make_psymbol_functions): Add partial_symtabs
parameter.
* psymtab.c (find_pc_sect_psymtab): Add partial_symtabs
parameter.
(psymbol_functions::find_pc_sect_compunit_symtab)
(psymbol_functions::print_stats, psymbol_functions::dump)
(psymbol_functions::has_symbols): Update.
(make_psymbol_functions, dump_psymtab_addrmap): Add
partial_symtabs parameter.
(maintenance_print_psymbols): Update.
(psymbol_functions::expand_symtabs_matching): Update.
* psympriv.h (struct psymbol_functions): Add constructor.
<m_partial_symtabs>: New member.
<set_partial_symtabs>: New method.

Reference psymtabs via per_bfd in DWARF reader

This changes the DWARF reader to reference the psymtabs via the
per_bfd object, rather than via the objfile.  This helps isolate the
partial symtabs from the objfile.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_create_include_psymtab): Add per_bfd
parameter.
(process_psymtab_comp_unit_reader)
(build_type_psymtab_dependencies, dwarf2_build_psymtabs_hard)
(add_partial_subprogram, dwarf2_ranges_read, dwarf_decode_lines):
Reference psymtabs via per_bfd.

Change how DWARF index writer finds address map

This changes the DWARF index writer to find the psymtab address map in
the per_bfd object, rather than go via the objfile.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/index-write.c (struct addrmap_index_data) <objfile>:
Remove.
(add_address_entry): Remove objfile parameter.
(add_address_entry_worker): Update.
(write_address_map): Replace objfile parameter with per_bfd.
(write_gdbindex, write_psymtabs_to_index): Update.

Move psymtab statistics printing to psymtab.c

This moves all the psymtab statistics printing code form symmisc.c to
psymtab.c.  This changes the formatting of the output a little, but
considering that it is a maint command (and, I assume, a rarely used
one), this seems fine to me.

This change helps further dissociate the psymtab from the objfile.  In
the end there will be no direct connect -- only via the
quick_symbol_functions interface.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_base_index_functions::print_stats): Add
print_bcache parameter.
* symfile-debug.c (objfile::print_stats): Add print_bcache
parameter.
* quick-symbol.h (struct quick_symbol_functions)
<print_stats>: Add print_bcache parameter.
* symmisc.c (print_symbol_bcache_statistics, count_psyms): Move
code to psymtab.c.
(print_objfile_statistics): Move psymtab code to psymtab.c.
* psymtab.c (count_psyms): Move from symmisc.c.
(psymbol_functions::print_stats): Print partial symbol and bcache
statistics.  Add print_bcache parameter.
* objfiles.h (print_symbol_bcache_statistics): Don't declare.
(struct objfile) <print_stats>: Add print_bcache parameter.
* maint.c (maintenance_print_statistics): Update.

gdb/testsuite/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* gdb.base/maint.exp: Update "maint print statistics" output.

Change how DWARF indices use addrmap

Currently the DWARF index readers reuse the objfile's partial symbol
table in order to store an addrmap.  We're going to be remove the
partial symbol object, so this patch changes the DWARF reader to store
this addrmap in the per_bfd object.  This object is chosen, rather
than the quick_symbol_functions subclass, because the addrmap can be
shared across objfiles.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.h (struct dwarf2_per_bfd) <psymtabs_addrmap>: New
member.
* dwarf2/read.c (create_addrmap_from_index)
(create_addrmap_from_aranges): Set per_bfd addrmap.
(dwarf2_read_gdb_index): Don't set partial_symtabs.
(dwarf2_base_index_functions::find_pc_sect_compunit_symtab): Use
per_bfd addrmap.
(dwarf2_read_debug_names): Don't set partial_symtabs.
(dwarf2_initialize_objfile): Likewise.

Set per_bfd->partial_symtabs earlier

dwarf2_build_psymtabs sets per_bfd->partial_symtabs at its end, in
order to facilitate sharing.

However, in order to diassociate partial symtabs from the objfile, we
want to change the DWARF reader to reference partial symtabs via the
per_bfd.  So, we want to set this reference before reading psymtabs.
This is safe to do because the assignment is done unconditionally.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_build_psymtabs): Set partial_symtabs
earlier.

Do not pass objfile to psymtab_discarder

This changes the psymtab_discarder to not assume that partial symtabs
are attached to the objfile.  Instead, a psymtab_storage object is
passed directly to it.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* psympriv.h (psymtab_discarder): Take psymtab_storage parameter.
(~psymtab_discarder, keep): Update.
<m_objfile>: Remove.
<m_partial_symtabs>: New member.
* dwarf2/read.c (dwarf2_build_psymtabs): Update.

Change how some psymbol readers access the psymtab storage

Currently, all psymbol readers access the psymtab storage via the
objfile.  This is done directly at any spot requiring the storage.

In order to move psymbols out of the objfile, the psymtab_storage must
be passed in explicitly.  This patch consolidates the access of the
storage in a single place in these readers, updating various functions
to pass the storage object around.  "Hidden" uses, like
"objfile->psymtabs ()", are also updated.

The DWARF reader is not yet touched.  That requires more complicated
changes.

gdb/ChangeLog
2021-03-20  Tom Tromey  <tom@tromey.com>

* xcoffread.c (xcoff_end_psymtab): Add partial_symtabs parameter.
(xcoff_end_psymtab): Update.
(scan_xcoff_symtab): Add partial_symtabs parameter.
(xcoff_initial_scan): Update.
* stabsread.h (dbx_end_psymtab): Add partial_symtabs parameter.
* mdebugread.c (mdebug_build_psymtabs): Update.
(parse_partial_symbols): Add partial_symtabs parameter.
* dbxread.c (dbx_symfile_read): Update.
(read_dbx_symtab): Add partial_symtabs parameter.
(read_dbx_symtab): Update.
(dbx_end_psymtab): Add partial_symtabs parameter.