bfd_perform_relocation should not have special case target code. This
patch moves the code that was there for x86_64 PE linking to ELF
output into the x86_64 PE howto special function, correcting that
function for linking to targets other than ELF too. The fixes in
bfd_perform_relocation were over-complicated due to needing to
compensate for things that had already gone wrong in coff_amd64_reloc.
In particular, an adjustment for pc-relative relocs was done in a way
that meant adjustment for things related to symbol offsets was lost.
I think those two things are orthogonal, but who knows with COFF where
addends and symbol values are found randomly in the section contents.
Note that linking natively to an x86_64 PE output relocates by
coff_pe_amd64_relocate_section, which does not use arelent relocs or
bfd_perform_relocation, but be aware of coff_amd64_rtype_to_howto
hacking addends for relocations. The adjustments for a particular
relocation type there and in coff_amd64_reloc ought to match after
taking into consideration CALC_ADDEND. They don't. For example,
the pc-relative adjustment for R_PCRWORD is 2 bytes in
coff_amd64_reloc and 4 bytes in coff_amd64_rtype_to_howto.
* reloc.c (bfd_perform_relocation): Revert 2021-01-12 and
2020-09-16 changes.
* coff-x86_64.c (coff_amd64_reloc): Do more or less the same
adjustments here instead. Separate pc-relative adjustments
from symbol related adjustments. Tidy comments and formatting.
PR 27147 shows that on sparc64, GDB is unable to properly unwind:
Expected result (from GDB 9.2):
#0 0x0000000000108de4 in puts ()
#1 0x0000000000100950 in hello () at gdb-test.c:4
#2 0x0000000000100968 in main () at gdb-test.c:8
Actual result (from GDB latest git):
#0 0x0000000000108de4 in puts ()
#1 0x0000000000100950 in hello () at gdb-test.c:4
Backtrace stopped: previous frame inner to this frame (corrupt stack?)
The first failing commit is 5b6d1e4fa4 ("Multi-target support"). The cause
of the change in behavior is due to (thanks for Andrew Burgess for finding
this):
- inferior_ptid is no longer set on entry of target_ops::wait, whereas
it was set to something valid previously
- deep down in linux_nat_target::wait (see stack trace below), we fetch
the registers of the event thread
- on sparc64, fetching registers involves reading memory (in
sparc_supply_rwindow, see stack trace below)
- reading memory (target_ops::xfer_partial) relies on inferior_ptid
being set to the thread from which we want to read memory
This is where things go wrong:
#0 linux_nat_target::xfer_partial (this=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, annex=0x0, readbuf=0x7feffe3b000 "", writebuf=0x0, offset=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:3697
#1 0x00000100007f5b10 in raw_memory_xfer_partial (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, readbuf=0x7feffe3b000 "", writebuf=0x0, memaddr=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/target.c:912
#2 0x00000100007f60e8 in memory_xfer_partial_1 (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, readbuf=0x7feffe3b000 "", writebuf=0x0, memaddr=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/target.c:1043
#3 0x00000100007f61b4 in memory_xfer_partial (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, readbuf=0x7feffe3b000 "", writebuf=0x0, memaddr=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/target.c:1072
#4 0x00000100007f6538 in target_xfer_partial (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, annex=0x0, readbuf=0x7feffe3b000 "", writebuf=0x0, offset=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/target.c:1129
#5 0x00000100007f7094 in target_read_partial (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, annex=0x0, buf=0x7feffe3b000 "", offset=8791798050744, len=8, xfered_len=0x7feffe3ae88) at /home/simark/src/binutils-gdb/gdb/target.c:1375
#6 0x00000100007f721c in target_read (ops=0x10000fa2c40 <the_sparc64_linux_nat_target>, object=TARGET_OBJECT_MEMORY, annex=0x0, buf=0x7feffe3b000 "", offset=8791798050744, len=8) at /home/simark/src/binutils-gdb/gdb/target.c:1415
#7 0x00000100007f69d4 in target_read_memory (memaddr=8791798050744, myaddr=0x7feffe3b000 "", len=8) at /home/simark/src/binutils-gdb/gdb/target.c:1218
#8 0x0000010000758520 in sparc_supply_rwindow (regcache=0x10000fea4f0, sp=8791798050736, regnum=-1) at /home/simark/src/binutils-gdb/gdb/sparc-tdep.c:1960
#9 0x000001000076208c in sparc64_supply_gregset (gregmap=0x10000be3190 <sparc64_linux_ptrace_gregmap>, regcache=0x10000fea4f0, regnum=-1, gregs=0x7feffe3b230) at /home/simark/src/binutils-gdb/gdb/sparc64-tdep.c:1974
#10 0x0000010000751b64 in sparc_fetch_inferior_registers (regcache=0x10000fea4f0, regnum=80) at /home/simark/src/binutils-gdb/gdb/sparc-nat.c:170
#11 0x0000010000759d68 in sparc64_linux_nat_target::fetch_registers (this=0x10000fa2c40 <the_sparc64_linux_nat_target>, regcache=0x10000fea4f0, regnum=80) at /home/simark/src/binutils-gdb/gdb/sparc64-linux-nat.c:38
#12 0x00000100008146ec in target_fetch_registers (regcache=0x10000fea4f0, regno=80) at /home/simark/src/binutils-gdb/gdb/target.c:3287
#13 0x00000100006a8c5c in regcache::raw_update (this=0x10000fea4f0, regnum=80) at /home/simark/src/binutils-gdb/gdb/regcache.c:584
#14 0x00000100006a8d94 in readable_regcache::raw_read (this=0x10000fea4f0, regnum=80, buf=0x7feffe3b7c0 "") at /home/simark/src/binutils-gdb/gdb/regcache.c:598
#15 0x00000100006a93b8 in readable_regcache::cooked_read (this=0x10000fea4f0, regnum=80, buf=0x7feffe3b7c0 "") at /home/simark/src/binutils-gdb/gdb/regcache.c:690
#16 0x00000100006b288c in readable_regcache::cooked_read<unsigned long, void> (this=0x10000fea4f0, regnum=80, val=0x7feffe3b948) at /home/simark/src/binutils-gdb/gdb/regcache.c:777
#17 0x00000100006a9b44 in regcache_cooked_read_unsigned (regcache=0x10000fea4f0, regnum=80, val=0x7feffe3b948) at /home/simark/src/binutils-gdb/gdb/regcache.c:791
#18 0x00000100006abf3c in regcache_read_pc (regcache=0x10000fea4f0) at /home/simark/src/binutils-gdb/gdb/regcache.c:1295
#19 0x0000010000507920 in save_stop_reason (lp=0x10000fc5b10) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:2612
#20 0x00000100005095a4 in linux_nat_filter_event (lwpid=520983, status=1407) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:3050
#21 0x0000010000509f9c in linux_nat_wait_1 (ptid=..., ourstatus=0x7feffe3c8f0, target_options=...) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:3194
#22 0x000001000050b1d0 in linux_nat_target::wait (this=0x10000fa2c40 <the_sparc64_linux_nat_target>, ptid=..., ourstatus=0x7feffe3c8f0, target_options=...) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:3432
#23 0x00000100007f8ac0 in target_wait (ptid=..., status=0x7feffe3c8f0, options=...) at /home/simark/src/binutils-gdb/gdb/target.c:2000
#24 0x00000100004ac17c in do_target_wait_1 (inf=0x1000116d280, ptid=..., status=0x7feffe3c8f0, options=...) at /home/simark/src/binutils-gdb/gdb/infrun.c:3464
#25 0x00000100004ac3b8 in operator() (__closure=0x7feffe3c678, inf=0x1000116d280) at /home/simark/src/binutils-gdb/gdb/infrun.c:3527
#26 0x00000100004ac7cc in do_target_wait (wait_ptid=..., ecs=0x7feffe3c8c8, options=...) at /home/simark/src/binutils-gdb/gdb/infrun.c:3540
#27 0x00000100004ad8c4 in fetch_inferior_event () at /home/simark/src/binutils-gdb/gdb/infrun.c:3880
#28 0x0000010000485568 in inferior_event_handler (event_type=INF_REG_EVENT) at /home/simark/src/binutils-gdb/gdb/inf-loop.c:42
#29 0x000001000050d394 in handle_target_event (error=0, client_data=0x0) at /home/simark/src/binutils-gdb/gdb/linux-nat.c:4060
#30 0x0000010000ab5c8c in handle_file_event (file_ptr=0x10001207270, ready_mask=1) at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:575
#31 0x0000010000ab6334 in gdb_wait_for_event (block=0) at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:701
#32 0x0000010000ab487c in gdb_do_one_event () at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:212
#33 0x0000010000542668 in start_event_loop () at /home/simark/src/binutils-gdb/gdb/main.c:348
#34 0x000001000054287c in captured_command_loop () at /home/simark/src/binutils-gdb/gdb/main.c:408
#35 0x0000010000544e84 in captured_main (data=0x7feffe3d188) at /home/simark/src/binutils-gdb/gdb/main.c:1242
#36 0x0000010000544f2c in gdb_main (args=0x7feffe3d188) at /home/simark/src/binutils-gdb/gdb/main.c:1257
#37 0x00000100000c1f14 in main (argc=4, argv=0x7feffe3d548) at /home/simark/src/binutils-gdb/gdb/gdb.c:32
There is a target_read_memory call in sparc_supply_rwindow, whose return
value is not checked. That call fails, because inferior_ptid does not
contain a valid ptid, and uninitialized buffer contents is used.
Ultimately it results in a corrupt stop_pc.
target_ops::fetch_registers can be (and should remain, in my opinion)
independent of inferior_ptid, because the ptid of the thread from which
to fetch registers can be obtained from the regcache. In other words,
implementations of target_ops::fetch_registers should not rely on
inferior_ptid having a sensible value on entry.
The sparc64_linux_nat_target::fetch_registers case is special, because it calls
a target method that is dependent on the inferior_ptid value
(target_read_inferior, and ultimately target_ops::xfer_partial). So I would
say it's the responsibility of sparc64_linux_nat_target::fetch_registers to set
up inferior_ptid correctly prior to calling target_read_inferior.
This patch makes sparc64_linux_nat_target::fetch_registers (and
store_registers, since it works the same) temporarily set inferior_ptid. If we
ever make target_ops::xfer_partial independent of inferior_ptid, setting
inferior_ptid won't be necessary, we'll simply pass down the ptid as a
parameter in some way.
I chose to set/restore inferior_ptid in sparc_fetch_inferior_registers, because
I am not convinced that doing so in an inner location (in sparc_supply_rwindow
for instance) would always be correct. We have access to the ptid in
sparc_supply_rwindow (from the regcache), so we _could_ set inferior_ptid
there. However, I don't want to just set inferior_ptid, as that would make it
not desync'ed with `current_thread ()` and `current_inferior ()`. It's
preferable to use switch_to_thread instead, as that switches all the global
"current" stuff in a coherent way. But doing so requires a `thread_info *`,
and getting a `thread_info *` from a ptid requires a `process_stratum_target
*`. We could use `current_inferior()->process_target()` in
sparc_supply_rwindow for this (using target_read_memory uses the current
inferior's target stack anyway). However, sparc_supply_rwindow is also used in
the context of BSD uthreads, where a thread stratum target defines threads. I
presume the ptid in the regcache would be the ptid of the uthread, defined by
the thread stratum target (bsd_uthread_target). Using
`current_inferior()->process_target()` would look up a ptid defined by the
thread stratum target using the process stratum target. I don't think it would
give good results. So I prefer playing it safe and looking up the thread
earlier, in sparc_fetch_inferior_registers.
I added some assertions (in sparc_supply_rwindow and others) to verify
that the regcache's ptid matches inferior_ptid. That verifies that the
caller has properly set the correct global context. This would have
caught (though a failed assertion) the current problem.
gdb/ChangeLog:
PR gdb/27147
* sparc-nat.h (sparc_fetch_inferior_registers): Add
process_stratum_target parameter,
sparc_store_inferior_registers): update callers.
* sparc-nat.c (sparc_fetch_inferior_registers,
sparc_store_inferior_registers): Add process_stratum_target
parameter. Switch current thread before calling
sparc_supply_gregset / sparc_collect_rwindow.
(sparc_store_inferior_registers): Likewise.
* sparc-obsd-tdep.c (sparc32obsd_supply_uthread): Add assertion.
(sparc32obsd_collect_uthread): Likewise.
* sparc-tdep.c (sparc_supply_rwindow, sparc_collect_rwindow):
Add assertion.
* sparc64-obsd-tdep.c (sparc64obsd_collect_uthread,
sparc64obsd_supply_uthread): Add assertion.
Change-Id: I16c658cd70896cea604516714f7e2428fbaf4301
Without setting an image base address and without naming at least .text,
this test produces entirely bogus PE output. To be honest, even the ELF
output looks odd: .text gets placed at 0x10204, and both foo and bar get
associated with .text despite living below its start address.
Since neither image base nor .text placement are the subject of this
test, specify .text placement explicitly and in the PE case force the
image base to zero.
It is my understanding that IMAGE_SCN_LNK_* are supposed to communicate
information to the (static) linker, and become at best meaningless in PE
images. I wouldn't call loaders wrong which would refuse to process
sections with any of these bits set. While there's no replacement for
IMAGE_SCN_LNK_COMDAT, use IMAGE_SCN_MEM_DISCARDABLE in place of
IMAGE_SCN_LNK_REMOVE in this case.
It is the very nature of absolute symbols that they don't change even
if the loader decides to put the image at other than its link-time base
address. Of the linker-defined (and PE-specific) symbols __image_base__
(and its alias) needs special casing, as it'll still appear to be
absolute at this point.
A new inquiry function in ldexp.c is needed because PE base relocations
get generated before ldexp_finalize_syms() runs, yet whether a
relocation is needed depends on the ultimate property of a symbol.
Christian suggested switching an "int" in ada-lang.c to "bool"
instead. This patch makes this change. Tested on x86-64 Fedora 32.
gdb/ChangeLog
2021-03-04 Tom Tromey <tromey@adacore.com>
* ada-lang.c (struct match_data) <found_sym>: Now bool.
(aux_add_nonlocal_symbols): Update.
(ada_add_block_symbols): Change "found_sym" to bool.
PR 27478
* objdump.c (process_links): New variable.
(usage): Add --process-links.
(long_options): Likewise.
(dump_bfd): Stop processing once the bfd has been loaded unless
this is the main file or process_links has been enabled.
(main): Handle the process-links option.
* readelf.c (process_links): New variable.
(struct filedata): Add is_separate field.
(options): Add --process-links.
(usage): Likewise.
(parse_args): Likewise.
(process_file_header): Include the filename when dumping
information for separate debuginfo files.
(process_program_headers): Likewise.
(process_section_headers): Likewise.
(process_section_groups): Likewise.
(process_relocs): Likewise.
(process_dynamic_section): Likewise.
(process_version_sections): Likewise.
(display_lto_symtab): Likewise.
(process_symbol_table): Likewise.
(process_syminfo): Likewise.
(initialise_dumps_by_name): Likewise.
(process_section_contents): Likewise.
(process_notes_at): Likewise.
(process_notes): Likewise.
(open_file): Add is_separate parameter. Use to initialise the
is_separate field in the filedata structure.
(open_deug): Update call to open_file.
(process_object): Add processing of the contents of separate
debuginfo files, gated by the process_links variable.
(process_archive): Update call to open_file.
(process_file): Initialise the is_separate field in the filedata
structure.
* dwarf.c (load_separate_debug_info_file): Only report the
loading of a separate file if debug links are being dumped.
* objcopy.c (keep_section_symbols): New variable.
(enum command_line_switch): Add OPTION_KEEP_SYMBOLS.
(strip_options): Add keep-section-symbols.
(copy_options): Likewise.
(copy_usage): Likewise.
(strip_usage): Likewise.
(copy_object): Keep section symbols if requested by command line
option.
(strip_main): Handle --keep-section-symbols.
(copy_main): Likewise.
* doc/binutils.texi: Document the new options.
* NEWS: Mention the new features.
* testsuite/binutils-all/compress.exp (test_gnu_debuglink):
Update options passed to objdump. Use diff rather than cmp to
compare the dumped data.
* testsuite/binutils-all/objdump.WK2: Update regexp.
* testsuite/binutils-all/objdump.WK3: Update regexp.
* testsuite/binutils-all/objdump.exp: Use --process-links
instead of --dwarf=follow-links.
* testsuite/binutils-all/readelf.exp (readelf_test): Include
readelf's output in the log when the test fails.
Add the -P option to the -wKis test.
* testsuite/binutils-all/readelf.wKis: Update expected output.
The patch adds a missing elf64_sparc_copy_solaris_special_section_fields
function that enables to fill sh_link and sh_info fields in .SUNW_* sections.
Note that elf64_sparc_copy_solaris_special_section_fields is empty since
the default handling is currently sufficient for GNU strip command.
This is a followup patch of the following upstream commits:
commit 5522f910cb
Author: Nick Clifton <nickc@redhat.com>
Date: Fri Apr 29 09:24:42 2016 +0100
Enhance support for copying and stripping Solaris and ARM binaries.
commit 8486501545
Author: Nick Clifton <nickc@redhat.com>
Date: Thu Apr 14 12:04:09 2016 +0100
Fix copying Solaris binaries with objcopy.
gdb/ChangeLog:
2021-03-01 Libor Bukata <libor.bukata@oracle.com>
* bfd/elf64-sparc.c: Fix GNU strip on Solaris SPARC64.
This patch addresses some review comments that I forgot to deal with
in an earlier patch. See the comments here:
https://sourceware.org/pipermail/gdb-patches/2021-February/176278.html
For the most part this is fixing up comments, but it also includes
adding a constructor and initializers to "match_data".
Regression tested on x86-64 Fedora 32.
gdb/ChangeLog
2021-03-03 Tom Tromey <tromey@adacore.com>
* ada-lang.c (ada_resolve_function): Update comment.
(is_nonfunction, add_symbols_from_enclosing_procs)
(remove_extra_symbols): Likewise.
(struct match_data): Add constructor, initializers.
(add_nonlocal_symbols): Remove memset.
(aux_add_nonlocal_symbols): Update comment.
(ada_add_block_renamings, add_nonlocal_symbols)
(ada_add_all_symbols): Likewise.
* ada-exp.y (write_var_or_type): Clean up trailing whitespace.
Having this count explicitly in the table is redundant and (even if just
slightly) disturbs clarity. Infer the count from the number of operands
actually found.
Also convert the "no operands" indicator from "{ 0 }" to just "{}", as
that (now) ends up being easier to parse.
In gdb.btrace/rn-dl-bind.exp we test that we can reverse-step over
recorded dynamic linking. The test covers specific behaviour to support
_dl_runtime_resolve calling the resolved function by returning to it.
This would normally mess up stepping as we'd end up with backtraces that
contain the same functions but different frame ids.
Since GDB needs to recognize a return from _dl_runtime_resolve, the test
only passes when debug information for _dl_runtime_resolve is available.
The test requires that symbols are bound lazily. Otherwise, we won't
record dynamic linking and the test will be fairly pointless.
Recent GCC pass -z now by default to bind symbols eagerly. Add -z lazy to
the test's ldflags to enforce lazy binding.
In gdb.btrace/non-stop.exp, we hard-code expected source lines assuming we
know how they would match to the recorded trace. Despite the fact that we
should really have been using an assembly source, the assumptions work
pretty well.
With clang-6 -m32, we found a case where the assumptions do not hold.
Adjust the expected source lines a little bit to cover that case, as well.
Should we run into more cases like this, we will have to switch to an
assembly source file.
We use pre-compiled assembly including debug information for stepi, yet we
compiled with -g, which was implicitly set by prepare_for_testing.
Add {} options to avoid the implicit {debug}.
Clang generates slightly different debug information. Adjust the expected
output of gdb.btrace/function_call_history.exp to work with both gcc and
clang.
Also modify gdb.btrace/exception.cc to reliably trace into main and update
the corresponding patterns in gdb.btrace/exception.exp.
In gdb.btrace/unknown_functions.exp we need the linker to discard local
symbols so GDB wouldn't know about them from the symbol table.
When building with clang, it complains about the option not being used at
compile-time. Move the option to ldflags to only pass it at link-time.
Clang generates slightly different debug information causing
gdb.btrace/rn-dl-bind.exp to fail on its way to the actual test.
Modify the test to remove that dependency.
In gdb.btrace/delta.exp, we test that we do not extend the trace
unintentionally. This can be tested by checking the number of
instructions.
If we wanted to check the instruction history, as well, we'd need to work
on an assembly file to have deterministic behaviour. This isn't really
necessary for this test, however, and covered elsewhere. Also remove the
function call history check for the same reason.
Some older GCC, e.g. 7.5.0 on Ubuntu 18.04 need -fno-pie to be passed to
the compiler in addition to -no-pie to be passed to the linker for non-pie
code generation.
The gdb,nopie_flag is already documented as getting passed to the
compiler, not the linker. Use that for the new -fno-pie compiler flag and
add a new gdb,nopie_ldflag for the existing -no-pie linker flag.
CAUTION: this might break existing board files that specify
gdb,nopie_flag. Affected board files need to rename
gdb,nopie_flag into gdb,nopie_ldflag.
The testcases added here show situations where synthesized start/stop
symbols don't cause their associated input sections to be marked.
Fixed with the elflink.c and ldlang.c changes.
bfd/
PR 27500
* elflink.c (_bfd_elf_gc_mark_rsec): Do special start/stop
processing not when start/stop symbol section is unmarked but
on first time a start/stop symbol is processed.
ld/
* ldlang.c (insert_undefined): Don't mark symbols here.
(lang_mark_undefineds): Do so here instead, new function.
(lang_process): Call lang_mark_undefineds.
* testsuite/ld-gc/start3.d,
* testsuite/ld-gc/start3.s: New test.
* testsuite/ld-gc/start4.d,
* testsuite/ld-gc/start4.s: New test.
* testsuite/ld-gc/gc.exp: Run them.
Adjust tests to reference __start and __stop syms with an extra
leading underscore when appropriate, and run tests on more targets.
* testsuite/ld-gc/gc.exp: Define UNDERSCORE in ASFLAGS.
Move tests with ELF section directives to is_elf_format block.
* testsuite/ld-gc/abi-note.d: Run on more targets.
* testsuite/ld-gc/pr19167.d: Likewise and adjust xfails.
* testsuite/ld-gc/start.d: Likewise.
* testsuite/ld-gc/start2.d: Likewise.
* testsuite/ld-gc/stop.d: Likewise.
* testsuite/ld-gc/pr19167a.s: Add support for underscore targets.
* testsuite/ld-gc/start.s: Likewise.
* testsuite/ld-gc/start2.s: Likewise.
The mingw build fails with Readline 8.1, because sigprocmask is called
unconditionally. This patch adds the missing check for
HAVE_POSIX_SIGNALS.
I reported this upstream here:
https://lists.gnu.org/archive/html/bug-readline/2021-01/msg00011.html
readline/ChangeLog
2021-03-02 Tom Tromey <tom@tromey.com>
* readline/signals.c (_rl_handle_signal): Add missing check for
HAVE_POSIX_SIGNALS.
This imports readline 8.1. I did this via various hackery in a
readline git repository to make a version of readline identical to
gdb's, then did a git merge.
readline/ChangeLog
2021-03-02 Tom Tromey <tom@tromey.com>
* Import readline 8.1.
gdb currently supports two different styles of fixed-point. The
original style, where fixed point types are "GNAT encoded", is handled
primarily in the Ada code. The newer style, encoded using DWARF, is
handled by the core of gdb.
This patch changes gdb to read the GNAT encodings in the DWARF reader
as well. This removes some code and unifies the two paths. As a
result, GNAT-encoded fixed-point now works a bit better.
One possible drawback of this change is that, if someone uses stabs,
then fixed-point might now stop working. I consider stabs to be fully
obsolete, though, so I don't intend to address this.
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.c (cast_from_gnat_encoded_fixed_point_type)
(cast_to_gnat_encoded_fixed_point_type): Remove.
(ada_value_cast, ada_evaluate_subexp): Update.
(gnat_encoded_fixed_point_type_info)
(ada_is_gnat_encoded_fixed_point_type)
(gnat_encoded_fixed_point_delta)
(gnat_encoded_fixed_point_scaling_factor): Remove.
* ada-lang.h (ada_is_gnat_encoded_fixed_point_type)
(gnat_encoded_fixed_point_delta)
(gnat_encoded_fixed_point_scaling_factor): Don't declare.
* ada-typeprint.c (print_gnat_encoded_fixed_point_type): Remove.
(ada_print_type): Update.
* ada-valprint.c (ada_value_print_num): Update.
* dwarf2/read.c (ada_get_gnat_encoded_number)
(ada_get_gnat_encoded_ratio): New functions.
(finish_fixed_point_type): Use them. Add parameters.
(GNAT_FIXED_POINT_SUFFIX): New define.
(gnat_encoded_fixed_point_type_info): New function.
(read_base_type): Handle gnat encodings.
gdb/testsuite/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* gdb.ada/fixed_points.exp: Remove most special cases for minimal
encodings.
This removes the "GROW_VECT" macro and helper function in favor of
simply using std::string in a few spots.
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.c (ada_fold_name, ada_variant_discrim_name)
(ada_enum_name, scan_discrim_bound, to_fixed_range_type): Use
std::string.
(GROW_VECT): Remove.
(grow_vect): Remove.
This changes ada_lookup_symbol_list to return a std::vector, and
changes various other helper functions to follow. This simplifies the
code, and makes it more type-safe (by using a vector where an obstack
had been used).
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.h (ada_lookup_symbol_list): Return a vector.
* ada-lang.c (resolve_subexp): Update.
(ada_resolve_function): Accept a vector.
(is_nonfunction, add_defn_to_vec)
(add_symbols_from_enclosing_procs): Likewise.
(num_defns_collected, defns_collected): Remove.
(remove_extra_symbols): Return a vector.
(remove_irrelevant_renamings): Return void.
(ada_add_local_symbols): Accept a vector.
(struct match_data) <obstackp>: Remove.
<resultp>: New member.
(aux_add_nonlocal_symbols): Update.
(ada_add_block_renamings, add_nonlocal_symbols)
(ada_add_all_symbols): Accept a vector.
(ada_lookup_symbol_list_worker, ada_lookup_symbol_list): Return a
vector.
(ada_lookup_symbol): Update.
(ada_add_block_symbols): Accept a vector.
(get_var_value, iterate_over_symbols): Update.
* ada-exp.y (block_lookup, write_var_or_type, write_name_assoc):
Update.
This changes resolve_subexp to use any_of and the erase-remove idiom
to simplify the code somewhat. This simplifies the next patch a bit.
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.c (resolve_subexp): Use any_of and erase-remove idiom.
This changes the ada_symbol_cache to be allocated with 'new' and
managed via unique_ptr. This simplifies the code somewhat. Also,
ada_clear_symbol_cache is changed so that it does not allocate a
symbol cache just to clear it.
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.c (struct ada_symbol_cache) <cache_space>: Now an
auto_obstack.
<root>: Initialize.
(ada_pspace_data): Remove destructor.
<sym_cache>: Now a unique_ptr.
(ada_init_symbol_cache, ada_free_symbol_cache): Remove.
(ada_get_symbol_cache): Use 'new'.
(ada_clear_symbol_cache): Rewrite.
Most places in gdb that reference objfile->sf also check that it is
not null. It is valid for it to be null, because find_sym_fns can
return null for some kinds of object file. However, it's rare to
encounter this scenario with Ada code. I only encountered it when
looking at a fork of gdb that, I believe, makes its own objfiles
without setting 'sf'.
This patch changes ada-lang.c to check this field before using it.
This avoids any potential crash here. There's no test case because
I'm not even sure this is possible to trip over with an unmodified
gdb.
There are some other unchecked uses in gdb, but at a quick glance they
all seem to be involved with symbol reading, which of course won't
happen when sf==null.
gdb/ChangeLog
2021-03-02 Tom Tromey <tromey@adacore.com>
* ada-lang.c (add_nonlocal_symbols): Handle case where objfile->sf
is null.
This is a tricky one. BFD, on the linker's behalf, reports symbols to
libctf via the ctf_new_symbol and ctf_new_dynsym callbacks, which
ultimately call ctf_link_add_linker_symbol. But while this happens
after strtab offsets are finalized, it happens before the .dynstr is
actually laid out, so we can't iterate over it at this stage and
it is not clear what the reported symbols are actually called. So
a second callback, examine_strtab, is called after the .dynstr is
finalized, which calls ctf_link_add_strtab and ultimately leads
to ldelf_ctf_strtab_iter_cb being called back repeatedly until the
offsets of every string in the .dynstr is passed to libctf.
libctf can then use this to get symbol names out of the input (which
usually stores symbol types in the form of a name -> type mapping at
this stage) and extract the types of those symbols, feeding them back
into their final form as a 1:1 association with the real symtab's
STT_OBJ and STT_FUNC symbols (with a few skipped, see
ctf_symtab_skippable).
This representation is compact, but has one problem: if libctf somehow
gets confused about the st_type of a symbol, it'll stick an entry into
the function symtypetab when it should put it into the object
symtypetab, or vice versa, and *every symbol from that one on* will have
the wrong CTF type because it's actually looking up the type for a
different symbol.
And we have just such a bug. ctf_link_add_strtab was not taking the
refcounts of strings into consideration, so even strings that had been
eliminated from the strtab by virtue of being in objects eliminated via
--as-needed etc were being reported. This is harmful because it can
lead to multiple strings with the same apparent offset, and if the last
duplicate to be reported relates to an eliminated symbol, we look up the
wrong symbol from the input and gets its type wrong: if it's unlucky and
the eliminated symbol is also of the wrong st_type, we will end up with
a corrupted symtypetab.
Thankfully the wrong-st_type case is already diagnosed by a
this-can-never-happen paranoid warning:
CTF warning: Symbol 61a added to CTF as a function but is of type 1
or the converse
* CTF warning: Symbol a3 added to CTF as a data object but is of type 2
so at least we can tell when the corruption has spread to more than one
symbol's type.
Skipping zero-refcounted strings is easy: teach _bfd_elf_strtab_str to
skip them, and ldelf_ctf_strtab_iter_cb to loop over skipped strings
until it falls off the end or finds one that isn't skipped.
bfd/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* elf-strtab.c (_bfd_elf_strtab_str): Skip strings with zero refcount.
ld/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ldelfgen.c (ldelf_ctf_strtab_iter_cb): Skip zero-refcount strings.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-create.c (symtypetab_density): Report the symbol name as
well as index in the name != object error; note the likely
consequences.
* ctf-link.c (ctf_link_shuffle_syms): Report the symbol index
as well as name.
In the "no symbols" case (commonplace for executables), we were freeing
the ctf_dynsyms using free(), instead of ctf_dynhash_destroy(), leaking
a little memory.
(This is harmless in the common case of ld usage, but libctf might be
used by persistent processes too.)
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-link.c (ctf_link_shuffle_syms): Free ctf_dynsyms properly.
Comparing an encoding's cte_bits to a ctf_type_size needs a cast:
one is a uint32_t and the other is an ssize_t.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-dump.c (ctf_dump_format_type): Fix signed/unsigned confusion.
A transient bug in the preceding change (fixed before commit) exposed a
new failure, of ld/testsuite/ld-ctf/diag-parname.d. This attempts to
ensure that if we link a dict with child type IDs but no attached
parent, we get a suitable ECTF_NOPARENT error. This was happening
before this commit, but only by chance, because ctf_variable_iter and
ctf_variable_next check to see if the dict they're passed is a child
dict without an associated parent. We forgot error-checking on the
ctf_variable_next call, and as a result this was concealed -- and
looking for the problem exposed a new bug.
If any of the lookups beneath ctf_dedup_hash_type fail, the CTF link
does *not* fail, but acts quite bizarrely, skipping the type but
emitting an error to the CTF error/warning log -- so the linker will
report an error, emit a partial CTF dict missing some types, and exit
with exitcode 0 as if nothing went wrong. Since ctf_dedup_hash_type is
never expected to fail in normal operation, this is surely wrong:
failures at emission time do not emit partial CTF dicts, so failures
at hashing time should not either.
So propagate the error back up.
Also fix a couple of smaller bugs where we fail to properly free things
and/or propagate error codes on various rare link-time errors and
out-of-memory conditions.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-dedup.c (ctf_dedup): Pass on errors from ctf_dedup_hash_type.
Call ctf_dedup_fini properly on other errors.
(ctf_dedup_emit_type): Set the errno on dynhash insertion failure.
* ctf-link.c (ctf_link_deduplicating_per_cu): Close outputs beyond
output 0 when asserting because >1 output is found.
(ctf_link_deduplicating): Likewise, when asserting because the
shared output is not the same as the passed-in fp.
When CTF linking is done, the linker has to track the association
between types in the inputs and types in the outputs. The deduplicator
does this via the cd_output_emission_hashes, which maps from hashes of
types (valid in both the input and output) to the IDs of types in the
specific dict in which the cd_emission_hashes is held. However, the
nondeduplicating linker and ctf_add_type used a different mechanism, a
dedicated hashtab stored in the ctf_link_type_mapping, populated via
ctf_add_type_mapping and queried via the ctf_type_mapping function. To
allow the same functions to be used for variable and symbol population
in both the deduplicating and nondeduplicating linker, the deduplicator
carefully transferred all its input->output mappings into this hashtab
before returning.
This is *expensive*. The number of entries in this hashtab scales as the
number of input types, and unlike the hashing machinery the type mapping
machinery (the only other thing which scales that way) has not been much
optimized.
Now the nondeduplicating linker is gone, we can throw this out, move
the existing type mapping machinery to ctf-create.c and dedicate it to
ctf_add_type alone, and add a new function ctf_dedup_type_mapping which
uses the deduplicator's built-in knowledge of type mappings directly,
without requiring an expensive repopulation phase.
This speeds up a test link of nouveau.ko (a good worst-case candidate
with a lot of types in each of a lot of input files) from 9.11s to 7.15s
in my testing, a speedup of over 20%.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-impl.h (ctf_dict_t) <ctf_link_type_mapping>: No longer used
by the nondeduplicating linker.
(ctf_add_type_mapping): Removed, now static.
(ctf_type_mapping): Likewise.
(ctf_dedup_type_mapping): New.
(ctf_dedup_t) <cd_input_nums>: New.
* ctf-dedup.c (ctf_dedup_init): Populate it.
(ctf_dedup_fini): Free it again. Emphasise that this has to be
the last thing called.
(ctf_dedup): Populate it.
(ctf_dedup_populate_type_mapping): Removed.
(ctf_dedup_populate_type_mappings): Likewise.
(ctf_dedup_emit): No longer call it. No longer call
ctf_dedup_fini either.
(ctf_dedup_type_mapping): New.
* ctf-link.c (ctf_unnamed_cuname): New.
(ctf_create_per_cu): Arguments must be non-null now.
(ctf_in_member_cb_arg): Removed.
(ctf_link): No longer populate it. No longer discard the
mapping table.
(ctf_link_deduplicating_one_symtypetab): Use
ctf_dedup_type_mapping, not ctf_type_mapping. Use
ctf_unnamed_cuname.
(ctf_link_one_variable): Likewise. Pass in args individually: no
longer a ctf_variable_iter callback.
(empty_link_type_mapping): Removed.
(ctf_link_deduplicating_variables): Use ctf_variable_next, not
ctf_variable_iter. No longer pack arguments to
ctf_link_one_variable into a struct.
(ctf_link_deduplicating_per_cu): Call ctf_dedup_fini once
all link phases are done.
(ctf_link_deduplicating): Likewise.
(ctf_link_intern_extern_string): Improve comment.
(ctf_add_type_mapping): Migrate...
(ctf_type_mapping): ... these functions...
* ctf-create.c (ctf_add_type_mapping): ... here...
(ctf_type_mapping): ... and make static, for the sole use of
ctf_add_type.
There is no such thing, and the comment makes no sense, and doesn't
match what the code is doing. We always want to put variables in the
same dicts as the types they relate to if at all possible.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-link.c (ctf_link_one_variable): Remove reference to
"unconflicted link mode".
The nondeduplicating CTF linker was kept around when the deduplicating
one was added so that people had something to fall back to in case the
deduplicating linker turned out to be buggy. It's now much more stable
than the nondeduplicating linker, in addition to much faster, using much
less memory and producing much better output. In addition, while
libctf has a linker flag to invoke the nondeduplicating linker, ld does
not expose it: the only way to turn it on within ld is an intentionally-
undocumented environment variable. So we can remove it without any ABI
or user-visibility concerns (the only thing we leave around is the
CTF_LINK_NONDEDUP flag, which can easily be interpreted as "deduplicate
less", though right now it does nothing).
This lets us remove a lot of complexity associated with tracking
filenames and CU names separately (something the deduplcating linker
never bothered with, since the cunames are always reliable and ld never
hands us useful filenames anyway)
The biggest lacuna left behind is the ctf_type_mapping machinery, which
slows down deduplicating links quite a lot. We can't just ditch it
because ctf_add_type uses it: removing the slowdown from the
deduplicating linker is a job for another commit.
include/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-api.h (CTF_LINK_SHARE_DUPLICATED): Note that this might
merely change how much deduplication is done.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-link.c (ctf_create_per_cu): Drop FILENAME now that it is
always identical to CUNAME.
(ctf_link_deduplicating_one_symtypetab): Adjust.
(ctf_link_one_type): Remove.
(ctf_link_one_input_archive_member): Likewise.
(ctf_link_close_one_input_archive): Likewise.
(ctf_link_one_input_archive): Likewise.
(ctf_link): No longer call it. Drop CTF_LINK_NONDEDUP path.
Improve header comment a bit (dicts, not files). Adjust
ctf_create_per_cu call.
(ctf_link_deduplicating_variables): Simplify.
(ctf_link_in_member_cb_arg_t) <cu_name>: Remove.
<in_input_cu_file>: Likewise.
<in_fp_parent>: Likewise.
<done_parent>: Likewise.
(ctf_link_one_variable): Turn uses of in_file_name to in_cuname.
Ever since the generator-style _next iterators were introduced, there
have been separate implementations of the functional-style _iter
iterators that do the same thing as _next.
This is annoying and adds more dependencies on the internal guts of the
file format. Rip them all out and replace them with the corresponding
_next iterators. Only ctf_archive_raw_iter and ctf_label_iter survive,
the former because there is no access to the raw binary data of archives
via any _next iterator, and the latter because ctf_label_next hasn't
been implemented (because labels are currently not used for anything).
Tested by reverting the change (already applied) that reimplemented
ctf_member_iter in terms of ctf_member_next, then verifying that the
_iter and _next iterators produced the same results for every iterable
entity within a large type archive.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-types.c (ctf_member_iter): Move 'rc' to an inner scope.
(ctf_enum_iter): Reimplement in terms of ctf_enum_next.
(ctf_type_iter): Reimplement in terms of ctf_type_next.
(ctf_type_iter_all): Likewise.
(ctf_variable_iter): Reimplement in terms of ctf_variable_next.
* ctf-archive.c (ctf_archive_iter_internal): Remove.
(ctf_archive_iter): Reimplement in terms of ctf_archive_next.
The top level of CTF containers is a "CTF archive", which contains a
collection of named members (each a CTF dictionary). In the serialized
file format, this is optional and skipped if the archive would have only
one member, as when no ambiguous types are present: so it is commonplace
to have a simple ctf_dict_t written out, with no archive container
wrapped around it.
But, unlike ctf_archive_iter, ctf_archive_next didn't quite handle this
case right. It should set the name of this fake "member" to
_CTF_SECTION, i.e. ".ctf", but it was failing to do so, so callers got
an unintialized variable back instead and were understandably confused.
So set the name properly.
libctf/ChangeLog
2021-03-02 Nick Alcock <nick.alcock@oracle.com>
* ctf-archive.c (ctf_archive_next): Set the name of parents in
single-member archives.
PowerPC64 has its own gc_mark_dynamic_ref.
bfd/
PR 27451
* elf64-ppc.c (ppc64_elf_gc_mark_dynamic_ref): Ignore synthesized
linker defined start/stop symbols when start_stop_gc.
ld/
* testsuite/ld-powerpc/startstop.d,
* testsuite/ld-powerpc/startstop.r,
* testsuite/ld-powerpc/startstop.s: New test.
* testsuite/ld-powerpc/powerpc.exp: Run it.
Undefined weak symbols with non-default visibility are seen as local
by SYMBOL_REFERENCES_LOCAL. This stops a got indirect to relative
optimisation for them, so that pies and dlls don't get non-zero values
when loading somewhere other than the address they are linked at
(which always happens). The optimisation could be allowed for pdes,
but I thought it best not to allow it there too.
bfd/
* elf64-ppc.c (ppc64_elf_relocate_section): Don't optimise got
indirect to pc-relative or toc-relative for undefined symbols.
ld/
* testsuite/ld-powerpc/weak1.d,
* testsuite/ld-powerpc/weak1.r,
* testsuite/ld-powerpc/weak1.s,
* testsuite/ld-powerpc/weak1so.d,
* testsuite/ld-powerpc/weak1so.r: New tests.
* testsuite/ld-powerpc/powerpc.exp: Run them.
