When running test-case gdb.base/catch-syscall.exp on a pinebook (64-bit
aarch64 kernel, 32-bit userland) I run into:
...
(gdb) PASS: $exp: execve: syscall(s) execve appears in 'info breakpoints'
continue^M
Continuing.^M
^M
Catchpoint 18 (call to syscall execve), 0xf7726318 in execve () from \
/lib/arm-linux-gnueabihf/libc.so.6^M
(gdb) PASS: gdb.base/catch-syscall.exp: execve: program has called execve
continue^M
Continuing.^M
process 32392 is executing new program: catch-syscall^M
Cannot access memory at address 0xf77c6a7c^M
(gdb) FAIL: $exp: execve: syscall execve has returned
...
The memory error is thrown by arm_linux_get_syscall_number, when doing:
...
/* PC gets incremented before the syscall-stop, so read the
previous instruction. */
unsigned long this_instr =
read_memory_unsigned_integer (pc - 4, 4, byte_order_for_code);
...
The reason for the error is that we're stopped at the syscall exit of syscall
execve, and the pc is at the first insn of the new exec, which also happens to
be the first insn in the code segment, so consequently we cannot read the
previous insn.
Fix this by detecting the situation by looking at the register state, similar
to what is done in aarch64_linux_get_syscall_number.
Furthermore, catch the memory error by using safe_read_memory_unsigned_integer
and return -1 instead, matching the documented behaviour of
arm_linux_get_syscall_number.
Finally, rather than using a hardcoded constant 11, introduce an ad-hoc
arm_sys_execve.
Tested on pinebook.
PR tdep/31071
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31071
PR 31067
* linker.c (_bfd_generic_link_add_one_symbol): When issuing a warning message, also display a message about the warning not being affected by garbage colleciton.
* ld.texi (Special Sections): New entry in the linker manual. Describes how the .gnu.warning and .gnu.warning.SYM sections behave.
In AIX, we are not able to see the message of a signal recieved if a debugee recieves a signal.
This is a patch to fix the signal handling done incorrectly in AIX.
We remove the status that represent program recieving a signal and allow host_status_to_waitstatus to
handle it for us.
Consider a binary with an erroneous size of the .dynamic section:
$ readelf -S a.out
...
[24] .dynamic DYNAMIC 0000000000004c20 00003c20
000000fffffffa40 0000000000000010 WA 7 0 8
...
This binary causes a segfault in GDB. GDB is trying to write the .dynamic
section into memory allocated on the stack with alloca(). However, the
allocation silently fails and the subsequent access to the memory is
causing the segfault. (On my node at least.)
Stack allocation is a bad idea for something of variable size that GDB has
no control over. So I changed the code to heap allocation.
In addition, I changed the type of sect_size to the type that bfd actually
returns.
There should be no user visible change after this.
Approved-By: Tom Tromey <tom@tromey.com>
An earlier patch (commit b05efa39 "readelf..debug-dump=loc displays
bogus base addresses") inadvertently removed support for displaying
.gdb_index v9 sections.
This patch corrects the oversight. I tested this by using readelf on
an appropriate file.
* dwarf.c (display_gdb_index): Restore v9 display code.
Function Create_large returns a large data structure. On PowerPC, register
r3 contains the address of where the data structure to be returned is to
be stored. However, on exit the ABI does not guarantee that r3 has not
been changed. The GDB finish command prints the return value of the
function at the end of the function. GDB needs to use the
DW_TAG_call_site information to determine the value of r3 on entry to
the function to correctly print the return value at the end of the
function. The test must be compiled with -fvar-tracking for the
DW_TAG_call_site information to be included in the executable file.
This patch adds the -fvar-tracking option to the compile line if the
option is supported.
The patch fixes the one regression error for the test on PowerPC.
The patch has been tested on Power 10 and X86-64 with no regressions.
When CTF finds conflicting types, it usually shoves each definition
into a CTF dictionary named after the compilation unit.
The intent of the obscure "cu-mapped link" feature is to allow you to
implement custom linkers that shove the definitions into other, more
coarse-grained units (say, one per kernel module, even if a module consists
of more than one compilation unit): conflicting types within one of these
larger components are hidden from name lookup so you can only look up (an
arbitrary one of) them by name, but can still be found by chasing type graph
links and are still fully deduplicated.
You do this by calling
ctf_link_add_cu_mapping (fp, "CU name", "bigger lump name"), repeatedly,
with different "CU name"s: the ctf_link() following that will put all
conflicting types found in "CU name"s sharing a "bigger lump name" into a
child dict in an archive member named "bigger lump name".
So it's clear enough what happens if you call it repeatedly with the same
"bigger lump name" more than once, because that's the whole point of it: but
what if you call it with the same "CU name" repeatedly?
ctf_link_add_cu_mapping (fp, "CU name", "bigger lump name");
ctf_link_add_cu_mapping (fp, "CU name", "other name");
This is meant to be the same as just doing the second of these, as if the
first was never called. Alas, this isn't what happens, and what you get is
instead a bit of an inconsistent mess: more or less, the first takes
precedence, which is the exact opposite of what we wanted.
Fix this to work the right way round.
(I plan to add support for CU-mapped links to GNU ld, mainly so that we can
properly *test* this machinery.)
libctf/ChangeLog:
* ctf-link.c (ctf_create_per_cu): Note the behaviour of
repeatedly adding FROMs.
(ctf_link_add_cu_mapping): Implement that behavour.
Following on from this commit:
commit f2c4f78c81
Date: Thu Sep 21 16:35:30 2023 +0100
gdb: fix reread_symbols when an objfile has target: prefix
In this commit I update reopen_exec_file to correctly handle
executables with a target: prefix. Before this commit we used the
system 'stat' call, which obviously isn't going to work for files with
a target: prefix (files located on a possibly remote target machine).
By switching to bfd_stat we will use remote fileio to stat the remote
files, which means we should now correctly detect changes in a remote
executable.
The program_space::ebfd_mtime variable, with which we compare the
result of bfd_stat is set with a call to bfd_get_mtime, which in turn
calls bfd_stat, so comparing to the result of calling bfd_stat makes
sense (I think).
As I discussed in the commit f2c4f78c81, if a BFD is an in-memory
BFD, then calling bfd_stat will always return 0, while bfd_get_mtime
will always return the time at which the BFD was created. As a result
comparing the results will always show the file having changed.
I don't believe that GDB can set the main executable to an in-memory
BFD object, so, in this commit, I simply assert that the executable is
not in-memory. If this ever changes then we would need to decide how
to handle this case -- always reload, or never reload. The assert
doesn't appear to trigger for our current test suite.
Approved-By: Tom Tromey <tom@tromey.com>
In the following commit I ran into a problem. The next commit aims to
improve GDB's handling of the main executable being a file on a remote
target (i.e. one with a 'target:' prefix).
To do this I have replaced a system 'stat' call with a bfd_stat call.
However, doing this caused a regression in gdb.base/attach.exp.
The problem is that the bfd library caches open FILE* handles for bfd
objects that it has accessed, which is great for short-lived, non
interactive programs (e.g. the assembler, or objcopy, etc), however,
for GDB this caching causes us a problem.
If we open the main executable as a bfd then the bfd library will
cache the open FILE*. If some time passes, maybe just sat at the GDB
prompt, or with the inferior running, and then later we use bfd_stat
to check if the underlying, on-disk file has changed, then the bfd
library will actually use fstat on the underlying file descriptor.
This is of course slightly different than using system stat on with
the on-disk file name.
If the on-disk file has changed then system stat will give results for
the current on-disk file. But, if the bfd cache is still holding open
the file descriptor for the original on-disk file (from before the
change) then fstat will return a result based on the original file,
and so show no change as having happened.
This is a known problem in GDB, and so far this has been solved by
scattering bfd_cache_close_all() calls throughout GDB. But, as I
said, in the next commit I've made a change and run into a
problem (gdb.base/attach.exp) where we are apparently missing a
bfd_cache_close_all() call.
Now I could solve this problem by adding a bfd_cache_close_all() call
before the bfd_stat call that I plan to add in the next commit, that
would for sure solve the problem, but feels a little crude.
Better I think would be to track down where the bfd is being opened
and add a corresponding bfd_cache_close_all() call elsewhere in GDB
once we've finished doing whatever it is that caused us to open the
bfd in the first place.
This second solution felt like the better choice, so I tracked the
problem down to elf_locate_base and fixed that. But that just exposed
another problem in gdb_bfd_map_section which was also re-opening the
bfd, so I fixed this (with another bfd_cache_close_all() call), and
that exposed another issue in gdbarch_lookup_osabi... and at this
point I wondered if I was approaching this problem the wrong way...
.... And so, I wonder, is there a _better_ way to handle these
bfd_cache_close_all() calls?
I see two problems with the current approach:
1. It's fragile. Folk aren't always aware that they need to clear
the bfd cache, and this feels like something that is easy to
overlook in review. So adding new code to GDB can innocently touch
a bfd, which populates the cache, which will then be a bug that can
lie hidden until an on-disk file just happens to change at the wrong
time ... and GDB fails to spot the change. Additionally,
2. It's in efficient. The caching is intended to stop the bfd
library from continually having to re-open the on-disk file. If we
have a function that touches a bfd then often that function is the
obvious place to call bfd_cache_close_all. But if a single GDB
command calls multiple functions, each of which touch the bfd, then
we will end up opening and closing the same on-disk file multiple
times. It feels like we would be better postponing the
bfd_cache_close_all call until some later point, then we can benefit
from the bfd cache.
So, in this commit I propose a new approach. We now clear the bfd
cache in two places:
(a) Just before we display a GDB prompt. We display a prompt after
completing a command, and GDB is about to enter an idle state
waiting for further input from the user (or in async mode, for an
inferior event). If while we are in this idle state the user
changes the on-disk file(s) then we would like GDB to notice this
the next time it leaves its idle state, e.g. the next time the user
executes a command, or when an inferior event arrives,
(b) When we resume the inferior. In synchronous mode, resuming the
inferior is another time when GDB is blocked and sitting idle, but
in this case we don't display a prompt. As with (a) above, when an
inferior event arrives we want GDB to notice any changes to on-disk
files.
It turns out that there are existing observers for both of these
cases (before_prompt and target_resumed respectively), so my initial
thought was that I should attach to these observers in gdb_bfd.c, and
in both cases call bfd_cache_close_all().
And this does indeed solve the gdb.base/attach.exp problem that I see
with the following commit.
However, I see a problem with this solution.
Both of the observers I'm using are exposed through the Python API as
events that a user can hook into. The user can potentially run any
GDB command (using gdb.execute), so Python code might end up causing
some bfds to be reopened, and inserted into the cache.
To solve this one solution would be to add a bfd_cache_close_all()
call into gdbpy_enter::~gdbpy_enter(). Unfortunately, there's no
similar enter/exit object for Guile, though right now Guile doesn't
offer the same event API, so maybe we could just ignore that
problem... but this doesn't feel great.
So instead, I think a better solution might be to not use observers
for the bfd_cache_close_all() calls. Instead, I'll call
bfd_cache_close_all() directly from core GDB after we've notified the
before_prompt and target_resumed observers, this was we can be sure
that the cache is cleared after the observers have run, and before GDB
enters an idle state.
This commit also removes all of the other bfd_cache_close_all() calls
from GDB. My claim is that these are no longer needed.
Approved-By: Tom Tromey <tom@tromey.com>
The previous commit introduced some local variables to make some if
statements simpler. This commit uses them more liberally throughout the
process_event_stop_test in order to simplify the function a little more.
No functional changes are expected.
Approved-By: Tom Tromey <tom@tromey.com>
Currently, when GDB is reverse stepping out of a function into the same
function due to a recursive call, it doesn't print frame information, as
reported by PR record/29178. This happens because when the inferior
leaves the current frame, GDB decides to refresh the step information,
clobbering the original step_frame_id, making it impossible to figure
out later on that the frame has been changed.
This commit changes GDB so that, if we notice we're in this exact
situation, we won't refresh the step information.
Because of implementation details, this change can cause some debug
information to be read when it normally wouldn't before, which showed up
as a regression on gdb.dwarf2/dw2-out-of-range-end-of-seq. Since that
isn't a problem, the test was changed to allow for the new output.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29178
Approved-By: Tom Tromey <tom@tromey.com>
2023-11-18 Jose E. Marchesi <jemarch@gnu.org>
* config/tc-bpf.c (parse_bpf_register): Move before
bpf_parse_name.
(bpf_parse_name): Do not allow using symbols that are also
register names as operands in pseudo-c syntax.
* testsuite/gas/bpf/regs-for-symbols-pseudoc.d: New file.
* testsuite/gas/bpf/regs-for-symbols-pseudoc.s: Likewise.
* testsuite/gas/bpf/regs-for-symbols-pseudoc.l: Likewise.
* doc/c-bpf.texi (BPF Registers): Document that it is not possible
to refer to register names as symbols in instruction operands.
To support the "pseudo-C" asm dialect in BPF, the BPF parser must often
attempt multiple different templates for a single instruction. In some
cases this can cause the parser to incorrectly parse part of the
instruction opcode as an expression, which leads to the creation of a
new undefined symbol.
Once the parser recognizes the error, the expression is discarded and it
tries again with a new instruction template. However, symbols created
during the process are added to the symbol table and are not removed
even if the expression is discarded.
This is a problem for BPF: generally the assembled object will be loaded
directly to the Linux kernel, without being linked. These erroneous
parser-created symbols are rejected by the kernel BPF loader, and the
entire object is refused.
This patch remedies the issue by tentatively creating symbols while
parsing instruction operands, and storing them in a temporary list
rather than immediately inserting them into the symbol table. Later,
after the parser is sure that it has correctly parsed the instruction,
those symbols are committed to the real symbol table.
This approach is modeled directly after Jan Beulich's patch for RISC-V:
commit 7a29ee2903
RISC-V: adjust logic to avoid register name symbols
Many thanks to Jan for recognizing the problem as similar, and pointing
me to that patch.
gas/
* config/tc-bpf.c (parsing_insn_operands): New.
(parse_expression): Set it here.
(deferred_sym_rootP, deferred_sym_lastP): New.
(orphan_sym_rootP, orphan_sym_lastP): New.
(bpf_parse_name): New.
(parse_error): Clear deferred symbol list on error.
(md_assemble): Clear parsing_insn_operands. Commit deferred
symbols to symbol table on successful parse.
* config/tc-bpf.h (md_parse_name): Define to...
(bpf_parse_name): ...this. New prototype.
* testsuite/gas/bpf/asm-extra-sym-1.s: New test source.
* testsuite/gas/bpf/asm-extra-sym-1.d: New test.
* testsuite/gas/bpf/bpf.exp: Run new test.
A simple refactor to make the reference to current_program_space bubble
up one level. No behavior changes expected.
Change-Id: I237cf2f45ae73c35bcb433ce40e3c03cef6b87e2
Remove get_current_regcache, inlining the call to get_thread_regcache in
callers. When possible, pass the right thread_info object known from
the local context. Otherwise, fall back to passing `inferior_thread ()`.
This makes the reference to global context bubble up one level, a small
step towards the long term goal of reducing the number of references to
global context (or rather, moving those references as close as possible
to the top of the call tree).
No behavior change expected.
Change-Id: Ifa6980c88825d803ea586546b6b4c633c33be8d6
While looking at the regcache code, I noticed that the address space
(passed to regcache when constructing it, and available through
regcache::aspace) wasn't relevant for the regcache itself. Callers of
regcache::aspace use that method because it appears to be a convenient
way of getting the address space for a thread, if you already have the
regcache. But there is always another way to get the address space, as
the callers pretty much always know which thread they are dealing with.
The regcache code itself doesn't use the address space.
This patch removes anything related to address_space from the regcache
code, and updates callers to get it from the thread in context. This
removes a bit of unnecessary complexity from the regcache code.
The current get_thread_arch_regcache function gets an address_space for
the given thread using the target_thread_address_space function (which
calls the target_ops::thread_address_space method). This suggest that
there might have been the intention of supporting per-thread address
spaces. But digging through the history, I did not find any such case.
Maybe this method was just added because we needed a way to get an
address space from a ptid (because constructing a regcache required an
address space), and this seemed like the right way to do it, I don't
know.
The only implementations of thread_address_space and
process_stratum_target::thread_address_space and
linux_nat_target::thread_address_space, which essentially just return
the inferior's address space. And thread_address_space is only used in
the current get_thread_arch_regcache, which gets removed. So, I think
that the thread_address_space target method can be removed, and we can
assume that it's fine to use the inferior's address space everywhere.
Callers of regcache::aspace are updated to get the address space from
the relevant inferior, either using some context they already know
about, or in last resort using the current global context.
So, to summarize:
- remove everything in regcache related to address spaces
- in particular, remove get_thread_arch_regcache, and rename
get_thread_arch_aspace_regcache to get_thread_arch_regcache
- remove target_ops::thread_address_space, and
target_thread_address_space
- adjust all users of regcache::aspace to get the address space another
way
Change-Id: I04fd41b22c83fe486522af7851c75bcfb31c88c7
dwarf2/read.c:new_symbol has some extra braces in a couple of 'case's.
These read weirdly to me, and since they aren't necessary, this patch
removes the braces and reindents the bodies. Tested by rebuilding.
bfd/dwarf2.c:read_ranges compares bfd_vma values against -1UL, which
doesn't work correctly when long is 32-bit and bfd_vma is 64-bit
(observed as "nm -l" being very slow for mingw64 host; probably causes
issues on 32-bit hosts as well as IL32LLP64 cases such as mingw64).
Fix by using (bfd_vma) -1 in place of -1UL, as done elsewhere.
DAP specifies that a request can fail with the "notStopped" message if
the inferior is running but the request requires that it first be
stopped.
This patch implements this for gdb. Most requests are assumed to
require a stopped inferior, and the exceptions are noted by a new
'request' parameter.
You may notice that the implementation is a bit racy. I think this is
inherent -- unless the client waits for a stop event before sending a
request, the request may be processed at any time relative to a stop.
https://sourceware.org/bugzilla/show_bug.cgi?id=31037
Reviewed-by: Kévin Le Gouguec <legouguec@adacore.com>
ExecutionInvoker is no longer really needed, due to the previous DAP
refactoring. This patch removes it in favor of an ordinary function.
One spot (the 'continue' request) could still have used it, but is
more succinctly expressed as a lambda.
Reviewed-by: Kévin Le Gouguec <legouguec@adacore.com>
Nearly every DAP request implementation forwards its work to the gdb
thread, using send_gdb_with_response. This patch refactors the
'request' decorator to make this automatic, and to provide some
parameters so that the unusual requests can express their needs as
well.
In a few spots this simplifies the code by removing an unnecessary
helper function. This could be done in more places as well if we
wanted.
The main motivation for this patch is that I thought it would be
helpful for cancellation. I am still working on that, but meanwhile
the parameterization of 'request' makes it easy to handle the
'notStopped' response as well.
Reviewed-by: Kévin Le Gouguec <legouguec@adacore.com>
DAP specifies a StackFrameFormat object that can be used to change how
the "name" part of a stack frame is constructed. While this output
can already be done in a nicer way (and also letting the client choose
the formatting), nevertheless it is in the spec, so I figured I'd
implement it.
While implementing this, I discovered that the current code does not
correctly preserve frame IDs across requests. I rewrote frame
iteration to preserve this, and it turned out to be simpler to combine
these patches.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30475
The AMD_DBGAPI_SCOPED_DEBUG_START_END macro in gdb/amd-dbgapi-target.c
is incorrectly controlled by "set debug infrun", while it should be
controlled by "set debug amd-dbgapi" instead. This commit fixes it.
Change-Id: I8ec2b1a4b9980c2d565a8aafd060ed070eeb3b29
PR gas/31043
"unsupported instruction ..." can mean about anything, and can also be
mistaken to mean something that isn't meant. Replace most of its uses by
more specific diagnostics,
While there also take the opportunity and purge the no longer used
invalid_register_operand enumerator.
{disp16} is invalid to use in 64-bit mode, while {disp32} is invalid to
use on pre-386 CPUs. The latter, also affecting other (real) prefixes,
further requires that like for insns we fully check the CPU flags; till
now only Cpu64/CpuNo64 were taken into consideration.
ELF-only functions don't need to be built when dealing with a non-ELF
target. md_section_align() also doesn't need to be a function when
dealing with non-AOUT targets. Similarly tc_fix_adjustable() can be a
simple macro when building non-ELF targets.
Furthermore x86_elf_abi is already used in ELF-only code sections, with
one exception. By adjusting that, the otherwise bogusly named variable
can also be confined to just ELF builds.
Simplify the code follow ing what check_{,q}word_reg() already do. This
the also eliminates a stale comment talking about a warning when an
error is raised. While there, correct a similarly stale comment in
check_qword_reg() while there.
Due to a missing check "crc32q %al, %rax" was wrongly translated to the
encoding of "crc32q %rax, %rax", rather than being rejected as invalid.
(The mnemonic suffix describes the source operand, not the destination
one.)
Note that check_{word,long}_reg() do not (currently) appear to require
similar amending, as there are no insn templates permitting an L or W
suffix and having an operand which allows for Reg8 and Reg64, but
neither Reg16 nor Reg32.
Previously the condition prevented shared objects from being relaxed.
To remove the limitation, we need to update program header size and
.eh_frame_hdr size before relaxation.
Remove uses of obstack in the code generating the libraries XML for
Windows and AIX. Use std::string instead. I'm not able to test this
change, unfortunately.
Change-Id: I28480913337e3fe8d6c31e551626931e6b1367ef
Approved-By: Tom Tromey <tom@tromey.com>
compile.exp generally does not work for me on Fedora 38. However, I
sent a GCC patch to fix the plugin crash. With that patch, I get this
error from one test in compile.exp:
gdb command line:1:22: warning: initialization of 'int (*)(int)' from incompatible pointer type 'int (*)()' [-Wincompatible-pointer-types]
This patch adds a cast to compile.exp. This makes the test pass.
Reviewed-by: Keith Seitz <keiths@redhat.com>
This commit:
commit 8f6c452b5a
Date: Sun Oct 15 22:48:42 2023 +0100
gdb: implement missing debug handler hook for Python
introduced a use of str.isascii(), which was only added in Python 3.7.
This commit switches to use curses.ascii.isascii(), as this was
available in 3.6.
The same is true for str.isalnum(), which is replaced with
curses.ascii.isalnum().
There should be no user visible changes after this commit.
This patch adds the permission model enhancement and memory
attribute index enhancement features and their corresponding
system registers in AArch64 assembler.
Permission Indirection Extension (FEAT_S1PIE, FEAT_S2PIE)
Permission Overlay Extension (FEAT_S1POE, FEAT_S2POE)
Memory Attribute Index Enhancement (FEAT_AIE)
Extension to Translation Control Registers (FEAT_TCR2)
These features are available by default from Armv9.4-A architecture.
This patch also adds support for:
1. FEAT_RASv2 feature and "ERXGSR_EL1" system register.
RASv2 feature is enabled by passing +rasv2 to -march
(eg: -march=armv8-a+rasv2).
2. FEAT_SCTLR2 and following system registers.
SCTLR2_EL1, SCTLR2_EL12, SCTLR2_EL2 and SCTLR2_EL3.
3. FEAT_FGT2 and following system registers.
HDFGRTR2_EL2, HDFGWTR2_EL2, HFGRTR2_EL2, HFGWTR2_EL2
4. FEAT_PFAR and following system registers.
PFAR_EL1, PFAR_EL2 and PFAR_EL12.
FEAT_RASv2, FEAT_SCTLR2, FEAT_FGT2 and FEAT_PFAR features are by default
enabled from Armv9.4-A architecture.
This patch also adds support for two read only system registers
id_aa64mmfr3_el1 and id_aa64mmfr4_el1, which are available from
Armv8-A Architecture.
This patch adds features to the Statistical Profiling Extension,
identified as FEAT_SPEv1p4, FEAT_SPE_FDS, and FEAT_SPE_CRR, which
are enabled by default from Armv9.4-A.
Also adds support for system register "pmsdsfr_el1".
