| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
I spotted that 'maint selftest' with an executable loaded into GDB,
would (when GDB was compiled for all targets) crash GDB. I fixed this
with a commit to bfd:
commit 427e4066afd13d1bf52c849849475f536e285d66
Date: Thu May 20 09:16:41 2021 +0100
gdb/bfd: avoid crash when architecture is forced to csky or riscv
However, this issue was not spotted as we currently only run 'maint
selftest' without an executable loaded.
This commit extends the testsuite to run 'maint selftest' both with
and without an executable loaded into GDB.
Currently, when no executable is loaded into GDB all of the selftest
pass (i.e. the fail count is 0), however, when running with an
executable loaded, I am seeing 1 failure (on an x86-64 GNU/Linux
host).
This failure is from the ARM disassembler tests, it appears that the
disassembler somehow gets itself into a state where it thinks it is in
thumb mode; when running the same test without an executable loaded
this doesn't happen.
This commit doesn't fix the ARM disassembler issue, but I thought it
was worth adding this anyway, as this will spot if GDB again starts to
crash when 'maint selftest' is run.
gdb/testsuite/ChangeLog:
* gdb.gdb/unittest.c: New file.
* gdb.gdb/unittest.exp: Run with and without a binary file loaded
into GDB.
|
|
This commit adds support for bare metal core dumps on the ARM target,
and is based off of this patch submitted to the mailing list:
https://sourceware.org/pipermail/gdb-patches/2020-October/172845.html
Compared to the version linked above this version is updated to take
account of recent changes to the core dump infrastructure in GDB,
there is now more shared infrastructure for core dumping within GDB,
and also some common bare metal core dumping infrastructure. As a
result this patch is smaller than the original proposed patch.
Further, the original patch included some unrelated changes to the
simulator that have been removed from this version.
I have written a ChangeLog entry as the original patch was missing
one.
I have done absolutely no testing of this patch. It is based on the
original submitted patch, which I assume was tested, but after my
modifications things might have been broken, however, the original
patch author has tested this version and reported it as being good:
https://sourceware.org/pipermail/gdb-patches/2021-May/178900.html
The core dump format is based around generating an ELF containing
sections for the writable regions of memory that a user could be
using. Which regions are dumped rely on GDB's existing common core
dumping code, GDB will attempt to figure out the stack and heap as
well as copying out writable data sections as identified by the
original ELF.
Register information is added to the core dump using notes, just as it
is for Linux of FreeBSD core dumps. The note types used consist of
the 2 basic types you would expect in a OS based core dump,
NT_PRPSINFO, NT_PRSTATUS, along with the architecture specific
NT_ARM_VFP note.
The data layouts for each note type are described below, in all cases,
all padding fields should be set to zero.
Note NT_PRPSINFO is optional. Its data layout is:
struct prpsinfo_t
{
uint8_t padding[28];
char fname[16];
char psargs[80];
}
Field 'fname' - null terminated string consisting of the basename of
(up to the fist 15 characters of) the executable. Any additional
space should be set to zero. If there's no executable name then
this field can be set to all zero.
Field 'psargs' - a null terminated string up to 80 characters in
length. Any additional space should be filled with zero. This
field contains the full executable path and any arguments passed
to the executable. If there's nothing sensible to write in this
field then fill it with zero.
Note NT_PRSTATUS is required, its data layout is:
struct prstatus_t
{
uint8_t padding_1[12];
uint16_t sig;
uint8_t padding_2[10];
uint32_t thread_id;
uint8_t padding_3[44];
uint32_t gregs[18];
}
Field 'sig' - the signal that stopped this thread. It's implementation
defined what this field actually means. Within GDB this will be
the signal number that the remote target reports as the stop
reason for this thread.
Field 'thread_is' - the thread id for this thread. It's implementation
defined what this field actually means. Within GDB this will be
thread thread-id that is assigned to each remote thread.
Field 'gregs' - holds the general purpose registers $a1 through to $pc
at indices 0 to 15. At index 16 the program status register.
Index 17 should be set to zero.
Note NT_ARM_VFP is optional, its data layout is:
armvfp_t
{
uint64_t regs[32];
uint32_t fpscr;
}
Field 'regs' - holds the 32 d-registers 0 to 31 in order.
Field 'fpscr' - holds the fpscr register.
The rules for ordering the notes is the same as for Linux. The
NT_PRSTATUS note must come before any other notes about additional
register sets. And for multi-threaded targets all registers for a
single thread should be grouped together. This is because only
NT_PRSTATUS includes a thread-id, all additional register notes after
a NT_PRSTATUS are assumed to belong to the same thread until a
different NT_PRSTATUS is seen.
gdb/ChangeLog:
PR gdb/14383
* Makefile.in (ALL_TARGET_OBS): Add arm-none-tdep.o.
(ALLDEPFILES): Add arm-none-tdep.c
* arm-none-tdep.c: New file.
* configure.tgt (arm*-*-*): Add arm-none-tdep.o to cpu_obs.
|
|
Consider the following chain of events:
* GDB is performing an inferior call, and
* the inferior calls longjmp, and
* GDB detects that the longjmp has completed, stops, and enters
check_longjmp_breakpoint_for_call_dummy (in breakpoint.c), and
* GDB tries to unwind the stack in order to check that the dummy
frame (setup for the inferior call) is still on the stack, but
* The unwind fails, possibly due to missing debug information, so
* GDB incorrectly concludes that the inferior has longjmp'd past the
dummy frame, and so deletes the dummy frame, including the dummy
frame breakpoint, but then
* The inferior continues, and eventually returns to the dummy frame,
which is usually (always?) on the stack, the inferior starts
trying to execute the random contents of the stack, this results
in undefined behaviour.
This situation is already warned about in the comment on the function
check_longjmp_breakpoint_for_call_dummy where we say:
You should call this function only at places where it is safe to currently
unwind the whole stack. Failed stack unwind would discard live dummy
frames.
The warning here is fine, the problem is that, even though we call the
function from a location within GDB where we hope to be able to
unwind, sometime the state of the inferior means that the unwind will
not succeed.
This commit tries to improve the situation by adding the following
additional check; when GDB fails to find the dummy frame on the stack,
instead of just assuming that the dummy frame can be garbage
collected, first find the stop_reason for the last frame on the stack.
If this stop_reason indicates that the stack unwinding may have failed
then we assume that the dummy frame is still in use. However, if the
last frame's stop_reason indicates that the stack unwind completed
successfully then we can be confident that the dummy frame is no
longer in use, and we garbage collect it.
Tested on x86-64 GNU/Linux.
gdb/ChangeLog:
* breakpoint.c (check_longjmp_breakpoint_for_call_dummy): Add
check for why the backtrace stopped.
gdb/testsuite/ChangeLog:
* gdb.base/premature-dummy-frame-removal.c: New file.
* gdb.base/premature-dummy-frame-removal.exp: New file.
* gdb.base/premature-dummy-frame-removal.py: New file.
Change-Id: I8f330cfe0f3f33beb3a52a36994094c4abada07e
|
|
All signatured_type constucted (even those used only for lookups in hash
maps) need a signature. Enforce that by passing the signature all the
way to the signatured_type constructor.
gdb/ChangeLog:
* dwarf2/read.h (struct structured_type) <signatured_type>: New.
Update all callers.
(struct dwarf2_per_bfd) <allocate_signatured_type>: Add
signature parameter, update all callers.
* dwar2/read.c (dwarf2_per_bfd::allocate_signatured_type): Add
signature parameter.
Change-Id: I99bc1f88f54127666aa133ddbbabb7f7668fa14a
|
|
Add an alias for std::unique_ptr<signatured_type> and use it where
possible.
gdb/ChangeLog:
* dwarf2/read.h (signatured_type_up): New, use where possible.
Change-Id: I5a41e8345551434c8beeb9f269b03bdcf27989be
|
|
Move them up before dwarf2_per_bfd, this will allow adding and using
signatured_type_up in the next patch.
gdb/ChangeLog:
* dwarf2/read.h (signatured_type, dwarf2_per_cu_data): Move up.
Change-Id: I85acad4476c8236930b6f9e53ddb8bbbad009e5e
|
|
All callers of allocate_signatured_type set the is_debug_types flag on
the result -- in fact, they are required to, because this is the sign
that downcasting the object to signatured_type is safe. This patch
moves this assignment into the allocation function.
2021-05-30 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (dwarf2_per_bfd::allocate_signatured_type): Set
is_debug_types.
(create_signatured_type_table_from_index)
(create_signatured_type_table_from_debug_names, add_type_unit)
(read_comp_units_from_section): Update.
|
|
Now that CUs and TUs are stored together in all_comp_units, the
m_num_psymtabs member is no longer needed -- it is always identical to
the length of the vector. This patch removes it.
2021-05-30 Tom Tromey <tom@tromey.com>
* dwarf2/read.h (struct dwarf2_per_bfd) <num_psymtabs,
m_num_psymtabs>: Remove.
(resize_symtabs): Update.
* dwarf2/read.c (dwarf2_per_bfd::allocate_per_cu)
(dwarf2_per_bfd::allocate_signatured_type): Update.
|
|
Argument fobj was only available in the constructor.
gdb/doc/ChangeLog:
2021-05-29 Hannes Domani <ssbssa@yahoo.de>
* python.texi (Writing a Frame Filter): Fix example.
|
|
I spotted some indentation issues where we had some spaces followed by
tabs at beginning of line, that I wanted to fix. So while at it, I did
a quick grep to find and fix all I could find.
gdb/ChangeLog:
* Fix tab after space indentation issues throughout.
Change-Id: I1acb414dd9c593b474ae2b8667496584df4316fd
|
|
I wrote a small script to spot a pattern of indentation mistakes I saw
happened in breakpoint.c. And while at it I ran it on all files and
fixed what I found. No behavior changes intended, just indentation and
addition / removal of curly braces.
gdb/ChangeLog:
* Fix some indentation mistakes throughout.
gdbserver/ChangeLog:
* Fix some indentation mistakes throughout.
Change-Id: Ia01990c26c38e83a243d8f33da1d494f16315c6e
|
|
Remove it, change users (well, a single one) to use all_bp_locations.
This requires moving all_bp_locations to breakpoint.h to expose it.
gdb/ChangeLog:
* breakpoint.h (iterate_over_bp_locations): Remove. Update
users to use all_bp_locations.
(all_bp_locations): New.
* breakpoint.c (all_bp_locations): Make non-static.
(iterate_over_bp_locations): Remove.
Change-Id: Iaf1f716d6c2c5b2975579b3dc113a86f5d0975be
|
|
Now that we have range functions that let us use ranged for loops, we
can remove iterate_over_breakpoints in favor of those, which are easier
to read and write. This requires exposing the declaration of
all_breakpoints and all_breakpoints_safe in breakpoint.h, as well as the
supporting types.
Change some users of iterate_over_breakpoints to use all_breakpoints,
when they don't need to delete the breakpoint, and all_breakpoints_safe
otherwise.
gdb/ChangeLog:
* breakpoint.h (iterate_over_breakpoints): Remove. Update
callers to use all_breakpoints or all_breakpoints_safe.
(breakpoint_range, all_breakpoints, breakpoint_safe_range,
all_breakpoints_safe): Move here.
* breakpoint.c (all_breakpoints, all_breakpoints_safe): Make
non-static.
(iterate_over_breakpoints): Remove.
* python/py-finishbreakpoint.c (bpfinishpy_detect_out_scope_cb):
Return void.
* python/py-breakpoint.c (build_bp_list): Add comment, reverse
return value logic.
* guile/scm-breakpoint.c (bpscm_build_bp_list): Return void.
Change-Id: Idde764a1f577de0423e4f2444a7d5cdb01ba5e48
|
|
Add the all_bp_locations_at_addr function, which returns a range of all
breakpoint locations at exactly the given address. This lets us
replace:
bp_location *loc, **loc2p, *locp;
ALL_BP_LOCATIONS_AT_ADDR (loc2p, locp, address)
{
loc = *loc2p;
// use loc
}
with
for (bp_location *loc : all_bp_locations_at_addr (address))
{
// use loc
}
The all_bp_locations_at_addr returns a bp_locations_at_addr_range
object, which is really just a wrapper around two std::vector iterators
representing the beginning and end of the interesting range. These
iterators are found when constructing the bp_locations_at_addr_range
object using std::equal_range, which seems a perfect fit for this use
case.
One thing I noticed about the current ALL_BP_LOCATIONS_AT_ADDR is that
if you call it with a NULL start variable, that variable gets filled in
and can be re-used for subsequent iterations. This avoids the cost of
finding the start of the interesting range again for the subsequent
iterations. This happens in build_target_command_list, for example.
The same effect can be achieved by storing the range in a local
variable, it can be iterated on multiple times.
Note that the original comment over ALL_BP_LOCATIONS_AT_ADDR says:
Iterates through locations with address ADDRESS for the currently
selected program space.
I don't see anything restricting the iteration to a given program space,
as we iterate over all bp_locations, which as far as I know contains all
breakpoint locations, regardless of the program space. So I just
dropped that part of the comment.
gdb/ChangeLog:
* breakpoint.c (get_first_locp_gte_addr): Remove.
(ALL_BP_LOCATIONS_AT_ADDR): Remove. Replace all uses with
all_bp_locations_at_addr.
(struct bp_locations_at_addr_range): New.
(all_bp_locations_at_addr): New.
(bp_locations_compare_addrs): New.
Change-Id: Icc8c92302045c47a48f507b7f1872bdd31d4ba59
|
|
Add the all_bp_locations function to replace the ALL_BP_LOCATIONS macro.
For simplicity, all_bp_locations simply returns a const reference to the
bp_locations vector. But the callers just treat it as a range to
iterate on, so if we ever change the breakpoint location storage, we can
change the all_bp_locations function to return some other range type,
and the callers won't need to be changed.
gdb/ChangeLog:
* breakpoint.c (ALL_BP_LOCATIONS): Remove, update users to use
all_bp_locations.
(all_bp_locations): New.
Change-Id: Iae71a1ba135c1a5bcdb4658bf3cf9793f0e9f81c
|
|
Change the type of the global location list, bp_locations, to be an
std::vector.
Adjust the users to deal with that, mostly in an obvious way by using
.data() and .size(). The user where it's slightly less obvious is
update_global_location_list. There, we std::move the old location list
out of the global vector into a local variable. The code to fill the
new location list gets simpler, as it's now simply using .push_back(),
no need to count the locations beforehand.
In the rest of update_global_location_list, the code is adjusted to work
with indices instead of `bp_location **`, to iterate on the location
list. I believe it's a bit easier to understand this way. But more
importantly, when we build with _GLIBCXX_DEBUG, the operator[] of the
vector does bound checking, so we will know if we ever access past a
vector size (which we won't if we access by raw pointer). I think that
work can further be done to make that function easier to understand,
notably find better names than "loc" and "loc2" for variables, but
that's work for later.
gdb/ChangeLog:
* breakpoint.c (bp_locations): Change to std::vector, update all
users.
(bp_locations_count): Remove.
(update_global_location_list): Change to work with indices
rather than bp_location**.
Change-Id: I193ce40f84d5dc930fbab8867cf946e78ff0df0b
|
|
Add the breakpoint::locations method, which returns a range that can be
used to iterate over a breakpoint's locations. This shortens
for (bp_location *loc = b->loc; loc != nullptr; loc = loc->next)
into
for (bp_location *loc : b->locations ())
Change all the places that I found that could use it.
gdb/ChangeLog:
* breakpoint.h (bp_locations_range): New.
(struct breakpoint) <locations>: New. Use where possible.
Change-Id: I1ba2f7d93d57e544e1f8609124587dcf2e1da037
|
|
Same idea as the previous patches, but to replace the ALL_TRACEPOINTS
macro. Define a new filtered_iterator that only keeps the breakpoints
for which is_tracepoint returns true (just like the macro did).
I would have like to make it so tracepoint_range yields some
`tracepoint *` instead of some `breakpoint *`, that would help simplify
the callers, who wouldn't have to do the cast themselves. But I didn't
find an obvious way to do it. It can always be added later.
It turns out there is already an all_tracepoints function, which returns
a vector containing all the breakpoints that are tracepoint. Remove it,
most users will just work seamlessly with the new function. The
exception is start_tracing, which iterated multiple times on the vector.
Adapt this one so it iterates multiple times on the returned range.
Since the existing users of all_tracepoints are outside of breakpoint.c,
this requires defining all_tracepoints and a few supporting types in
breakpoint.h. So, move breakpoint_iterator from breakpoint.c to
breakpoint.h.
gdb/ChangeLog:
* breakpoint.h (all_tracepoints): Remove.
(breakpoint_iterator): Move here.
(struct tracepoint_filter): New.
(tracepoint_iterator): New.
(tracepoint_range): New.
(all_tracepoints): New.
* breakpoint.c (ALL_TRACEPOINTS): Remove, replace all users with
all_tracepoints.
(breakpoint_iterator): Move to header.
(all_tracepoints): New.
* tracepoint.c (start_tracing): Adjust.
Change-Id: I76b1bba4215dbec7a03846c568368aeef7f1e05a
|
|
Same as the previous patch, but intended to replace the
ALL_BREAKPOINTS_SAFE macro, which allows deleting the current breakpoint
while iterating. The new range type simply wraps the range added by the
previous patch with basic_safe_range.
I didn't remove the ALL_BREAKPOINTS_SAFE macro, because there is one
spot where it's more tricky to remove, in the
check_longjmp_breakpoint_for_call_dummy function. More thought it
needed for this one.
gdb/ChangeLog:
* breakpoint.c (breakpoint_safe_range): New.
(all_breakpoints_safe): New. Use instead of
ALL_BREAKPOINTS_SAFE where possible.
Change-Id: Ifccab29f135e1f85700e3697ed60f0b643c7682f
|
|
Introduce the all_breakpoints function, which returns a range that can
be used to iterate on breakpoints. Replace all uses of the
ALL_BREAKPOINTS macro with this.
In one instance, I could replace the breakpoint iteration with a call to
get_breakpoint.
gdb/ChangeLog:
* breakpoint.c (ALL_BREAKPOINTS): Remove, replace all uses with
all_breakpoints.
(breakpoint_iterator): New.
(breakpoint_range): New.
(all_breakpoints): New.
Change-Id: I229595bddad7c9100b179a9dd56b04b8c206e86c
|
|
To prevent flickering when first calling erase, then write, this new
argument indicates that the passed string contains the full contents of
the window. This fills every unused cell of the window with a space, so
it's not necessary to call erase beforehand.
gdb/ChangeLog:
2021-05-27 Hannes Domani <ssbssa@yahoo.de>
* python/py-tui.c (tui_py_window::output): Add full_window
argument.
(gdbpy_tui_write): Parse "full_window" argument.
gdb/doc/ChangeLog:
2021-05-27 Hannes Domani <ssbssa@yahoo.de>
* python.texi (TUI Windows In Python): Document "full_window"
argument.
|
|
An earlier patch in this series fixed a dependency problem between two
_initialize functions. That problem was uncovered by reversing the
order of the initialize function calls.
In short, symtab.c tried to add the alias "maintenance
flush-symbol-cache" for the command "maintenance flush symbol-cache".
Because the "maintenance flush" prefix command was not yet created (it
happens in maint.c, initialized later in this reversed order), the
add_alias_cmd function returned NULL. That result was passed to
deprecate_cmd, which didn't expected that value, and that caused a
segfault. This was fixed by changing alias creation functions to take
the target command as a cmd_list_element, instead of by name.
This patch adds a runtime option to reverse the order of the initialize
calls at will. I chose to use an environment variable for this, over a
parameter (even a "maintenance" one), because:
- The init functions are called before the early init commands are
executed, so we could use -iex to turn this mode on early enough.
This is obvious when you remember that commands / parameters are
created by initialize funcitions :).
- This is not something anybody would want to tweak after startup
anyway.
gdb/ChangeLog:
* make-init-c: Add option to reverse function calls.
gdb/testsuite/ChangeLog:
* gdb.base/reverse-init-functions.exp: New.
Change-Id: I543e609cf526e7cb145a006a794d0e6851b63f45
|
|
I would like to modify how the init.c file is generated (its content).
But as it is, a shell script with multiple sed invocations in a Makefile
target, it's not very maintainable. Replace that with a shell script
that does the same, but in a more readable way.
The Makefile rule uses the "-" prefix in front of the for loop, I
presume to ignore any error coming from the fact that xml-builtin.c and
cp-name-parser.c are not found in the srcdir (they are generated source
files). I prefer not to blindly ignore errors, so filter these files
out of INIT_FILES instead (we already filter out other files).
There are no expected meaningful changes to the generated init.c file.
Just the _initialize_all_file declaration that is moved down and "void"
in parenthesis that is removed.
The new regular expression is a bit tighter than the existing one, it
requires the init function to be followed by exactly ` ()`. Update
bpf-tdep.c accordingly.
gdb/ChangeLog:
* Makefile.in (INIT_FILES_FILTER_OUT): New.
(INIT_FILES): Use INIT_FILES_FILTER_OUT.
(stamp-init): Use make-init-c.
* bpf-tdep.c (_initialize_bpf_tdep): Remove "void".
* silent-rules.mk (ECHO_INIT_C): Change.
* make-init-c: New file.
Change-Id: I6d6b12cbccf24ab79d1219bff05df01624c684f9
|
|
Same idea as previous patch, but for add_alias_cmd. Remove the overload
that accepts the target command as a string (the target command name),
leaving only the one that takes the cmd_list_element.
gdb/ChangeLog:
* command.h (add_alias_cmd): Accept target as
cmd_list_element. Update callers.
Change-Id: I546311f411e9e7da9302322d6ffad4e6c56df266
|
|
Same idea as previous patch, but for add_info_alias.
gdb/ChangeLog:
* command.h (add_info_alias): Accept target as
cmd_list_element. Update callers.
Change-Id: If830d423364bf42d7bea5ac4dd3a81adcfce6f7a
|
|
The alias creation functions currently accept a name to specify the
target command. They pass this to add_alias_cmd, which needs to lookup
the target command by name.
Given that:
- We don't support creating an alias for a command before that command
exists.
- We always use add_info_alias just after creating that target command,
and therefore have access to the target command's cmd_list_element.
... change add_com_alias to accept the target command as a
cmd_list_element (other functions are done in subsequent patches). This
ensures we don't create the alias before the target command, because you
need to get the cmd_list_element from somewhere when you call the alias
creation function. And it avoids an unecessary command lookup. So it
seems better to me in every aspect.
gdb/ChangeLog:
* command.h (add_com_alias): Accept target as
cmd_list_element. Update callers.
Change-Id: I24bed7da57221cc77606034de3023fedac015150
|
|
In add_setshow_generic, we create set/show commands using add_setshow_*
functions, then look up the commands by name to set the context pointer.
It would be simpler and more efficient to use the return values of the
add_setshow_* functions, do that.
gdb/ChangeLog:
* python/py-param.c (add_setshow_generic): Use return values of
add_setshow functions.
Change-Id: I04d50736e1001ddb732d81e088468876df9c88ff
|
|
Remove a few instances where we look up a command by name, but could
just use the return value of a previous "add command" function call
instead.
gdb/ChangeLog:
* mi/mi-main.c (_initialize_mi_main):
* python/py-auto-load.c (gdbpy_initialize_auto_load):
* remote.c (_initialize_remote):
Change-Id: I6d06f9ca636e340c88c1064ae870483ad392607d
|
|
Some add_set_show commands return a single cmd_list_element, the one for
the "set" command. A subsequent patch will need to access the show
command's cmd_list_element as well. Change these functions to return a
new structure type that holds both pointers.
I initially only modified add_setshow_boolean_cmd (the one I needed),
but I think it's better to change the whole chain to keep everything in
sync.
gdb/ChangeLog:
* command.h (set_show_commands): New.
(add_setshow_enum_cmd, add_setshow_auto_boolean_cmd,
add_setshow_boolean_cmd, add_setshow_filename_cmd,
add_setshow_string_cmd, add_setshow_string_noescape_cmd,
add_setshow_optional_filename_cmd, add_setshow_integer_cmd,
add_setshow_uinteger_cmd, add_setshow_zinteger_cmd,
add_setshow_zuinteger_cmd, add_setshow_zuinteger_unlimited_cmd):
Return set_show_commands. Adjust callers.
* cli/cli-decode.c (add_setshow_cmd_full): Return
set_show_commands, remove result parameters, adjust callers.
Change-Id: I17492b01b76002d09effc84830f9c6db26f1db7a
|
|
Looks like it was missing from the beginning.
gdb/doc/ChangeLog:
2021-05-27 Hannes Domani <ssbssa@yahoo.de>
* python.texi (Symbols In Python): Document gdb.SYMBOL_LOC_LABEL.
|
|
When running test-case gdb.dwarf2/dw2-dummy-cu.exp without -readnow, we run
into:
...
(gdb) file outputs/gdb.dwarf2/dw2-dummy-cu/dw2-dummy-cu^M
Reading symbols from outputs/gdb.dwarf2/dw2-dummy-cu/dw2-dummy-cu...^M
ERROR: Couldn't load dw2-dummy-cu into GDB (eof).
...
The problem is that we're running into a segfault:
...
Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
process_psymtab_comp_unit (this_cu=0x2141090, per_objfile=0x1aa4140,
want_partial_unit=false, pretend_language=language_minimal)
at /home/vries/gdb_versions/devel/src/gdb/dwarf2/read.c:7023
7023 switch (reader.comp_unit_die->tag)
...
due to reader.comp_unit_die == nullptr:
...
(gdb) p reader.comp_unit_die
$1 = (die_info *) 0x0
...
Indeed, there's no CU DIE in the test-case:
...
$ readelf -wi outputs/gdb.dwarf2/dw2-dummy-cu/dw2-dummy-cu
Contents of the .debug_info section:
Compilation Unit @ offset 0x0:
Length: 0x7 (32-bit)
Version: 2
Abbrev Offset: 0x0
Pointer Size: 4
$
...
Fix this by handling reader.comp_unit_die == nullptr in
process_psymtab_comp_unit.
Update the test-case to trigger this PR, as per PR27920 - "[gdb/testsuite]
hardcoding -readnow skips testing of partial symbols".
Tested on x86_64-linux.
gdb/ChangeLog:
2021-05-27 Tom de Vries <tdevries@suse.de>
PR symtab/27919
* dwarf2/read.c (process_psymtab_comp_unit):
gdb/testsuite/ChangeLog:
2021-05-27 Tom de Vries <tdevries@suse.de>
PR symtab/27919
PR testsuite/27920
* gdb.dwarf2/dw2-dummy-cu.exp: Use maint expand-symtabs instead of
-readnow.
|
|
When running these two test-cases in this specific order we get:
...
$ make check 'RUNTESTFLAGS=gdb.dwarf2/gdb-index.exp \
gdb.dwarf2/gdb-add-index-symlink.exp'
...
Running gdb.dwarf2/gdb-index.exp ...
Running gdb.dwarf2/gdb-add-index-symlink.exp ...
FAIL: gdb.dwarf2/gdb-add-index-symlink.exp: gdb-index file created
FAIL: gdb.dwarf2/gdb-add-index-symlink.exp: Unable to call \
gdb-add-index with a symlink to a symfile
...
The problem is that gdb-index.exp introduces a proc add_gdb_index which
overrides the one in lib/gdb.exp and stays active after the test is done.
Consequently it's used in gdb-add-index-symlink.exp, which should use the one
from lib/gdb.exp.
Fix this by renaming proc add_gdb_index in gdb-index.exp.
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-05-27 Tom de Vries <tdevries@suse.de>
PR testsuite/27921
* gdb.dwarf2/gdb-index.exp (add_gdb_index): Rename to ...
(local_add_gdb_index): ... this.
|
|
Fix "Cannot not" typo in dwarf error message.
Tested on x86_64-linux.
gdb/ChangeLog:
2021-05-27 Tom de Vries <tdevries@suse.de>
* dwarf2/read.c (find_partial_die): Fix "Cannot not" typo in dwarf
error.
|
|
When loading the debug info package
libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug from openSUSE Leap 15.2, we
run into a dwarf error:
...
$ gdb -q -batch libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug
Dwarf Error: Cannot not find DIE at 0x18a936e7 \
[from module libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug]
...
The DIE @ 0x18a936e7 does in fact exist, and is part of a CU @ 0x18a23e52.
No error message is printed when using -readnow.
What happens is the following:
- a dwarf2_per_cu_data P is created for the CU.
- a dwarf2_cu A is created for the same CU.
- another dwarf2_cu B is created for the same CU.
- the dwarf2_cu B is set in per_objfile->m_dwarf2_cus, such that
per_objfile->get_cu (P) returns B.
- P->load_all_dies is set to 1.
- all dies are read into the A->partial_dies htab
- dwarf2_cu A is destroyed.
- we try to find the partial_die for the DIE @ 0x18a936e7 in B->partial_dies.
We can't find it, but do not try to load all dies, because P->load_all_dies
is already set to 1.
- an error message is generated.
The question is why we're creating dwarf2_cu A and B for the same CU.
The dwarf2_cu A is created here:
...
(gdb) bt
#0 dwarf2_cu::dwarf2_cu (this=0x79a9660, per_cu=0x23c0b30,
per_objfile=0x1ad01b0) at dwarf2/cu.c:38
#1 0x0000000000675799 in cutu_reader::cutu_reader (this=0x7fffffffd040,
this_cu=0x23c0b30, per_objfile=0x1ad01b0, abbrev_table=0x0,
existing_cu=0x0, skip_partial=false) at dwarf2/read.c:6487
#2 0x0000000000676eb3 in process_psymtab_comp_unit (this_cu=0x23c0b30,
per_objfile=0x1ad01b0, want_partial_unit=false,
pretend_language=language_minimal) at dwarf2/read.c:7028
...
And the dwarf2_cu B is created here:
...
(gdb) bt
#0 dwarf2_cu::dwarf2_cu (this=0x885e8c0, per_cu=0x23c0b30,
per_objfile=0x1ad01b0) at dwarf2/cu.c:38
#1 0x0000000000675799 in cutu_reader::cutu_reader (this=0x7fffffffcc50,
this_cu=0x23c0b30, per_objfile=0x1ad01b0, abbrev_table=0x0,
existing_cu=0x0, skip_partial=false) at dwarf2/read.c:6487
#2 0x0000000000678118 in load_partial_comp_unit (this_cu=0x23c0b30,
per_objfile=0x1ad01b0, existing_cu=0x0) at dwarf2/read.c:7436
#3 0x000000000069721d in find_partial_die (sect_off=(unknown: 0x18a55054),
offset_in_dwz=0, cu=0x0) at dwarf2/read.c:19391
#4 0x000000000069755b in partial_die_info::fixup (this=0x9096900,
cu=0xa6a85f0) at dwarf2/read.c:19512
#5 0x0000000000697586 in partial_die_info::fixup (this=0x8629bb0,
cu=0xa6a85f0) at dwarf2/read.c:19516
#6 0x00000000006787b1 in scan_partial_symbols (first_die=0x8629b40,
lowpc=0x7fffffffcf58, highpc=0x7fffffffcf50, set_addrmap=0, cu=0x79a9660)
at dwarf2/read.c:7563
#7 0x0000000000678878 in scan_partial_symbols (first_die=0x796ebf0,
lowpc=0x7fffffffcf58, highpc=0x7fffffffcf50, set_addrmap=0, cu=0x79a9660)
at dwarf2/read.c:7580
#8 0x0000000000676b82 in process_psymtab_comp_unit_reader
(reader=0x7fffffffd040, info_ptr=0x7fffc1b3f29b, comp_unit_die=0x6ea90f0,
pretend_language=language_minimal) at dwarf2/read.c:6954
#9 0x0000000000676ffd in process_psymtab_comp_unit (this_cu=0x23c0b30,
per_objfile=0x1ad01b0, want_partial_unit=false,
pretend_language=language_minimal) at dwarf2/read.c:7057
...
So in frame #9, a cutu_reader is created with dwarf2_cu A. Then a fixup takes
us to the following CU @ 0x18aa33d6, in frame #5. And a similar fixup in
frame #4 takes us back to CU @ 0x18a23e52. At that point, there's no
information available that we're already trying to read that CU, and we end up
creating another cutu_reader with dwarf2_cu B.
It seems that there are two related problems:
- creating two dwarf2_cu's is not optimal
- the unoptimal case is not handled correctly
This patch addresses the last problem, by moving the load_all_dies flag from
dwarf2_per_cu_data to dwarf2_cu, such that it is paired with the partial_dies
field, which ensures that the two can be kept in sync.
Tested on x86_64-linux.
gdb/ChangeLog:
2021-05-27 Tom de Vries <tdevries@suse.de>
PR symtab/27898
* dwarf2/cu.c (dwarf2_cu::dwarf2_cu): Add load_all_dies init.
* dwarf2/cu.h (dwarf2_cu): Add load_all_dies field.
* dwarf2/read.c (load_partial_dies, find_partial_die): Update.
* dwarf2/read.h (dwarf2_per_cu_data::dwarf2_per_cu_data): Remove
load_all_dies init.
(dwarf2_per_cu_data): Remove load_all_dies field.
|
|
This was wrong: dwarf_die_debug is used as an integer, for example where
it is passed to dump_die. It is documented in the command's help, which
I missed the first time.
This reverts commit 749369c430d88c4abc9acde5cfc7b5218651de10.
Change-Id: I1d09c3da57f8885f4f9fe9f4eae0cf86006e617a
|
|
gdb/ChangeLog:
* dwarf2/read.c (dwarf_die_debug): Change type to bool.
(_initialize_dwarf2_read): Update.
Change-Id: I6d9e2fe4b662409a540acb2c0b82c7d5314d541b
|
|
The reg_buffer constructor zero-initializes (value-initializes, in C++
speak) the gdb_bytes of the m_registers array. This is not necessary,
as these bytes are only meaningful if the corresponding register_status
is REG_VALID. If the corresponding register_status is REG_VALID, then
they will have been overwritten with the actual register data when
reading the registers from the system into the reg_buffer.
Fix that by removing the empty parenthesis following the new expression,
meaning that the bytes will now be default-initialized, meaning they'll
be left uninitialized. For reference, this is explained here:
https://en.cppreference.com/w/cpp/language/new#Construction
These new expressions were added in 835dcf92618e ("Use std::unique_ptr
in reg_buffer"). As mentioned in that commit message, the use of
value-initialisation was done on purpose to keep existing behavior, but
now there is some data that suggest it would be beneficial not to do it,
which is why I suggest changing it.
This doesn't make a big difference on typical architectures where the
register buffer is not that big. However, on ROCm (AMD GPU), the
register buffer is about 65000 bytes big, so the reg_buffer constructor
shows up in profiling. If you want to make some tests and profile it on
a standard system, it's always possible to change:
- m_registers.reset (new gdb_byte[m_descr->sizeof_raw_registers] ());
+ m_registers.reset (new gdb_byte[65000] ());
and run a program that constantly hits a breakpoint with a false
condition. For example, by doing this change and running the following
program:
static void break_here () {}
int main ()
{
for (int i = 0; i < 100000; i++)
break_here ();
}
with the following GDB incantation:
/usr/bin/time ./gdb -nx --data-directory=data-directory -q test -ex "b break_here if 0" -ex r -batch
I get, for value-intializing:
11.75user 7.68system 0:18.54elapsed 104%CPU (0avgtext+0avgdata 56644maxresident)k
And for default-initializing:
6.83user 8.42system 0:14.12elapsed 108%CPU (0avgtext+0avgdata 56512maxresident)k
gdb/ChangeLog:
* regcache.c (reg_buffer::reg_buffer): Default-initialize
m_registers array.
Change-Id: I5071a4444dee0530ce1bc58ebe712024ddd2b158
|
|
In a bigger series I'm working on, it is convenient to have a
libiberty hash table that manages objects allocated with 'new'. To
make this simpler, I wrote a small template function to serve as a
concise wrapper. Then I realized that this could be reused in a few
other places.
gdb/ChangeLog
2021-05-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (allocate_type_unit_groups_table)
(handle_DW_AT_stmt_list, allocate_dwo_file_hash_table): Use
htab_delete_entry.
(free_line_header_voidp): Remove.
* completer.c
(completion_tracker::completion_hash_entry::deleter): Remove.
(completion_tracker::discard_completions): Use htab_delete_entry.
* utils.h (htab_delete_entry): New template function.
|
|
gdb/doc/ChangeLog:
2021-05-25 Hannes Domani <ssbssa@yahoo.de>
* python.texi (Symbols In Python): Fix gdb.SYMBOL_LOC_COMMON_BLOCK.
|
|
tui_win_info::refresh_window first redraws the background window, then
tui_wrefresh draws the python text on top of it, which flickers.
By using wnoutrefresh for the background window, the actual drawing on
the screen is only done once, without flickering.
gdb/ChangeLog:
2021-05-24 Hannes Domani <ssbssa@yahoo.de>
* python/py-tui.c (tui_py_window::refresh_window):
Avoid flickering.
|
|
Add '@:' after 'e.g.' to let texinfo know that this is not the end of
a sentence. This is only needed when there's a space immediately
after the last '.'.
gdb/doc/ChangeLog:
* gdb.texi (Initialization Files): Add '@:' after 'e.g.'.
(Source Path): Likewise.
(GDB/MI Development and Front Ends): Likewise.
(ARM Features): Likewise.
(gdb man): Likewise.
|
|
In a linux kernel mailing list discussion, it was mentioned that "gdb has
this odd thing where it takes the 64-bit vs 32-bit data for the whole process
from one thread, and picks the worst possible thread to do it (ie explicitly
not even the main thread, ...)" [1].
The picking of the thread is done here in
x86_linux_nat_target::read_description:
...
/* GNU/Linux LWP ID's are process ID's. */
tid = inferior_ptid.lwp ();
if (tid == 0)
tid = inferior_ptid.pid (); /* Not a threaded program. */
...
To understand what this code does, let's investigate a scenario in which
inferior_ptid.lwp () != inferior_ptid.pid ().
Say we start exec jit-attach-pie, identified with pid x. The main thread
starts another thread that sleeps, and then the main thread waits for the
sleeping thread. So we have two threads, identified with LWP IDs x and x+1:
...
PID LWP CMD
x x ./jit-attach-pie
x x+1 ./jit-attach-pie
...
[ The thread with LWP x is known as the thread group leader. ]
When attaching to this exec using the pid, gdb does a stop_all_threads which
iterates over all the threads, first LWP x, and then LWP x+1.
So the state we arrive with at x86_linux_nat_target::read_description is:
...
(gdb) p inferior_ptid
$1 = {m_pid = x, m_lwp = x+1, m_tid = 0}
...
and consequently we probe 64/32-bitness from thread LWP x+1.
[ Note that this is different from when gdb doesn't attach but instead
launches the exec itself, in which case there's just one thread to begin with,
and consequently the probed thread is LWP x. ]
According to aforementioned remark, a better choice would have been the main
thread, that is, LWP x.
This patch implement that choice, by simply doing:
...
tid = inferior_ptid.pid ();
...
The fact that gdb makes a per-process permanent choice for 64/32-bitness is a
problem in itself: each thread can be in either 64 or 32 bit mode, and change
forth and back. That is a problem that this patch doesn't fix.
Now finally: why does this matter in the context of the linux kernel
discussion? The discussion was related to a patch that exposed io_uring
threads to user-space. This made it possible that one of those threads would
be picked out to select 64/32-bitness. Given that such threads are atypical
user-space threads in the sense that they don't return to user-space and don't
have a userspace register state, reading their registers returns garbage, and
so it could f.i. occur that in a 64-bit process with all normal user-space
threads in 64-bit mode, the probing would return 32-bit.
It may be that this is worked-around on the kernel side by providing userspace
register state in those threads such that current gdb is happy. Nevertheless,
it seems prudent to fix this on the gdb size as well.
Tested on x86_64-linux.
[1] https://lore.kernel.org/io-uring/CAHk-=wh0KoEZXPYMGkfkeVEerSCEF1AiCZSvz9TRrx=Kj74D+Q@mail.gmail.com/
gdb/ChangeLog:
2021-05-23 Tom de Vries <tdevries@suse.de>
PR tdep/27822
* target.h (struct target_ops): Mention target_thread_architecture in
read_description comment.
* x86-linux-nat.c (x86_linux_nat_target::read_description): Use
pid to determine if process is 64-bit or 32-bit.
* aarch64-linux-nat.c (aarch64_linux_nat_target::read_description):
Same.
* ppc-linux-nat.c (ppc_linux_nat_target::read_description): Same.
* riscv-linux-nat.c (riscv_linux_nat_target::read_description): Same.
* s390-linux-nat.c (s390_linux_nat_target::read_description): Same.
* arm-linux-nat.c (arm_linux_nat_target::read_description): Same.
Likewise, use pid to determine if kernel supports reading VFP
registers.
|
|
The comments in the enum cmdarg_kind were using -sx and -sex instead
of -eix and -eiex.
(Note that gdb --help does not speak about these options).
(pushed as obvious)
|
|
Add target board cc-with-gnu-debuglink.exp that splits off debuginfo into a
seperate .debug file and links to it using .gnu_debuglink.
Tested on x86_64-linux.
gdb/ChangeLog:
2021-05-21 Tom de Vries <tdevries@suse.de>
PR testsuite/25047
* contrib/cc-with-tweaks.sh: Handle -l.
gdb/testsuite/ChangeLog:
2021-05-21 Tom de Vries <tdevries@suse.de>
PR testsuite/25047
* boards/cc-with-gnu-debuglink.exp: New file.
|
|
The test in gdb.dwarf2/dw2-inline-with-lexical-scope.c fails with icc.
The reason is, icc did not emit code for a dead statement, which in
turn caused some labels to be collapsed. Fix this by replacing the
dead code with assignment to a global value. The statement itself
does not change the test scenario.
Also fix a whitespacing problem around an assignment operator.
gdb/testsuite/ChangeLog:
2021-05-21 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* gdb.dwarf2/dw2-inline-with-lexical-scope.c (func): Replace
a dead code with an assignment to a global var. Fix a
whitespacing problem around an assignment operator.
|
|
Consider a minimal test-case test.c:
...
int main (void) { return 0; }
...
which we can compile into llvm byte code using clang:
...
$ clang -g -S -emit-llvm --target=x86_64-unknown-unknown-elf test.c
...
and then run using lli, which uses the llvm jit:
...
$ lli test.ll
...
If we run this under gdb, we run into an assert:
...
$ gdb -q -batch -ex run --args /usr/bin/lli test.ll
Dwarf Error: Cannot not find DIE at 0x18a936e7 \
[from module libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
src/gdb/jit.c:1178: internal-error: \
void jit_event_handler(gdbarch*, objfile*): \
Assertion `jiter->jiter_data != nullptr' failed.
...
This is caused by the following.
When running jit_breakpoint_re_set_internal, we first handle
libLLVM.so.10.debug, and set a jit breakpoint.
Next we handle libLLVM.so.10:
...
(gdb) p the_objfile.original_name
$42 = 0x2494170 "libLLVM.so.10"
...
but the minimal symbols we find are from libLLVM.so.10.debug:
...
(gdb) p reg_symbol.objfile.original_name
$43 = 0x38e7c50 "libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug"
(gdb) p desc_symbol.objfile.original_name
$44 = 0x38e7c50 "libLLVM.so.10-10.0.1-lp152.30.4.x86_64.debug"
...
and consequently, the objf_data is the one from libLLVM.so.10.debug:
...
jiter_objfile_data *objf_data
= get_jiter_objfile_data (reg_symbol.objfile);
...
and so we hit this:
...
if (objf_data->cached_code_address == addr)
continue;
...
and no second jit breakpoint is inserted.
Subsequently, the jit breakpoint is triggered and handled, but when finding
the symbol for the breakpoint address we get:
...
(gdb) p jit_bp_sym.objfile.original_name
$52 = 0x2494170 "libLLVM.so.10"
...
The assert 'jiter->jiter_data != nullptr' triggers because it checks
libLLVM.so.10 while the one with jiter_data setup is libLLVM.so.10.debug.
This fixes the assert:
...
jiter_objfile_data *objf_data
- = get_jiter_objfile_data (reg_symbol.objfile);
- = get_jiter_objfile_data (the_objfile);
...
but consequently we'll have two jit breakpoints, so we also make sure we don't
set a jit breakpoint on separate debug objects like libLLVM.so.10.debug.
Tested on x86_64-linux.
gdb/ChangeLog:
2021-05-21 Tom de Vries <tdevries@suse.de>
PR breakpoint/27889
* jit.c (jit_breakpoint_re_set_internal): Skip separate debug
objects. Call get_jiter_objfile_data with the_objfile.
|
|
I guess this was used with the old VEC implementation, but there is no
reason to have this typedef anymore.
gdb/ChangeLog:
* linespec.c (linespec_p): Remove. Replace all uses with
"linespec *".
Change-Id: I4cea59ae1cd46985da9c08d3a69686846b1ad028
|
|
In cli/cli-script.c, process_next_line() allocates memory
which will eventually end up being assigned to the 'next'
field in struct command_line. However, in a case
recurse_read_control_structure returns 'invalid_control'
this memory is leaked. This commit uses std::unique_ptr
as appropriate to prevent this leakage.
This issue was found by coverity scanning.
gdb/ChangeLog:
* cli/cli-script.h (command_line_up): New unique_ptr typedef.
* cli/cli-script.c (multi_line_command_p): Use unique_ptr
command_line_up instead of struct command_line.
(build_command_line): Likewise.
(get_command_line): Update the cmd function call parameter.
(process_next_line): Use unique_ptr command_line_up instead
of struct command_line.
(recurse_read_control_structure): Change the the type of
next to command_line_up.
(read_command_lines_1): Change type of `next' to be
command_line_up and update all references of `next'
accordingly.
|
|
gdb/ChangeLog:
* MAINTAINERS (Write After Approval): Add myself.
|
|
|
