Currently, with "maint set target-non-stop on", that is, when gdb
connects with the non-stop/asynchronous variant of the remote
protocol, even with "set non-stop off", GDB always sends one vCont
packet per thread resumed. This patch makes GDB aggregate and
coalesce vCont packets, so we send vCont packets like "vCont;s:p1.1;c"
in non-stop mode too.
Basically, this is done by:
- Adding a new target method target_commit_resume that is called
after calling target_resume one or more times. When resuming a
batch of threads, we'll only call target_commit_resume once after
calling target_resume for all threads.
- Making the remote target defer sending the actual vCont packet to
target_commit_resume.
Special care must be taken to avoid sending a vCont action with a
"wildcard" thread-id (all threads of process / all threads) when that
would resume threads/processes that should not be resumed. See
remote_commit_resume comments for details.
Unlike all-stop's remote_resume implementation, this handles the case
of too many actions resulting in a too-big vCont packet, by flushing
the vCont packet and starting a new one.
E.g., imagining that the "c" action in:
vCont;s:1;c
overflows the packet buffer, we split the actions like:
vCont;s:1
vCont;c
Tested on x86_64 Fedora 20, with and without "maint set
target-non-stop on".
Also tested with a hack that makes remote_commit_resume flush the vCont
packet after every action appended (which caught a few bugs).
gdb/ChangeLog:
2016-10-26 Pedro Alves <palves@redhat.com>
* inferior.h (ALL_NON_EXITED_INFERIORS): New macro.
* infrun.c (do_target_resume): Call target_commit_resume.
(proceed): Defer target_commit_resume while looping over threads,
resuming them. Call target_commit_resume at the end.
* record-btrace.c (record_btrace_commit_resume): New function.
(init_record_btrace_ops): Install it as to_commit_resume method.
* record-full.c (record_full_commit_resume): New function.
(record_full_wait_1): Call the beneath target's to_commit_resume
method.
(init_record_full_ops): Install record_full_commit_resume as
to_commit_resume method.
* remote.c (struct private_thread_info) <last_resume_step,
last_resume_sig, vcont_resumed>: New fields.
(remote_add_thread): Set the new thread's vcont_resumed flag.
(demand_private_info): Delete.
(get_private_info_thread, get_private_info_ptid): New functions.
(remote_update_thread_list): Adjust.
(process_initial_stop_replies): Clear the thread's vcont_resumed
flag.
(remote_resume): If connected in non-stop mode, record the resume
request and return early.
(struct private_inferior): New.
(struct vcont_builder): New.
(vcont_builder_restart, vcont_builder_flush)
(vcont_builder_push_action): New functions.
(MAX_ACTION_SIZE): New macro.
(remote_commit_resume): New function.
(thread_pending_fork_status, is_pending_fork_parent_thread): New
functions.
(check_pending_event_prevents_wildcard_vcont_callback)
(check_pending_events_prevent_wildcard_vcont): New functions.
(process_stop_reply): Adjust. Clear the thread's vcont_resumed
flag.
(init_remote_ops): Install remote_commit_resume.
* target-delegates.c: Regenerate.
* target.c (defer_target_commit_resume): New global.
(target_commit_resume, make_cleanup_defer_target_commit_resume):
New functions.
* target.h (struct target_ops) <to_commit_resume>: New field.
(target_resume): Update comments.
(target_commit_resume): New declaration.
This simple commit consolidates the API of
target_supports_multi_process. Since both GDB and gdbserver use the
same function prototype, all that was needed was to move create this
prototype on gdb/target/target.h and turn the macros declared on
gdb/{,gdbserver/}target.h into actual functions.
Regtested (clean pass) on the BuildBot.
gdb/ChangeLog:
2016-10-06 Sergio Durigan Junior <sergiodj@redhat.com>
* target.c (target_supports_multi_process): New function, moved
from...
* target.h (target_supports_multi_process): ... here. Remove
macro.
* target/target.h (target_supports_multi_process): New prototype.
gdb/gdbserver/ChangeLog:
2016-10-06 Sergio Durigan Junior <sergiodj@redhat.com>
* target.c (target_supports_multi_process): New function, moved
from...
* target.h (target_supports_multi_process): ... here. Remove
macro.
This patch consolidates the API of target_mourn_inferior between GDB
and gdbserver, in my continuing efforts to make sharing the
fork_inferior function possible between both.
GDB's version of the function did not care about the inferior's ptid
being mourned, but gdbserver's needed to know this information. Since
it actually makes sense to pass the ptid as an argument, instead of
depending on a global value directly (which GDB's version did), I
decided to make the generic API to accept it. I then went on and
extended all calls being made on GDB to include a ptid argument (which
ended up being inferior_ptid most of the times, anyway), and now we
have a more sane interface.
On GDB's side, after talking to Pedro a bit about it, we decided that
just an assertion to make sure that the ptid being passed is equal to
inferior_ptid would be enough for now, on the GDB side. We can remove
the assertion and perform more operations later if we ever pass
anything different than inferior_ptid.
Regression tested on our BuildBot, everything OK.
I'd appreciate a special look at gdb/windows-nat.c's modification
because I wasn't really sure what to do there. It seemed to me that
maybe I should build a ptid out of the process information there, but
then I am almost sure the assertion on GDB's side would trigger.
gdb/ChangeLog:
2016-09-19 Sergio Durigan Junior <sergiodj@redhat.com>
* darwin-nat.c (darwin_kill_inferior): Adjusting call to
target_mourn_inferior to include ptid_t argument.
* fork-child.c (startup_inferior): Likewise.
* gnu-nat.c (gnu_kill_inferior): Likewise.
* inf-ptrace.c (inf_ptrace_kill): Likewise.
* infrun.c (handle_inferior_event_1): Likewise.
* linux-nat.c (linux_nat_attach): Likewise.
(linux_nat_kill): Likewise.
* nto-procfs.c (interrupt_query): Likewise.
(procfs_interrupt): Likewise.
(procfs_kill_inferior): Likewise.
* procfs.c (procfs_kill_inferior): Likewise.
* record.c (record_mourn_inferior): Likewise.
* remote-sim.c (gdbsim_kill): Likewise.
* remote.c (remote_detach_1): Likewise.
(remote_kill): Likewise.
* target.c (target_mourn_inferior): Change declaration to accept
new ptid_t argument; use gdb_assert on it.
* target.h (target_mourn_inferior): Move function prototype from
here...
* target/target.h (target_mourn_inferior): ... to here. Adjust it
to accept new ptid_t argument.
* windows-nat.c (get_windows_debug_event): Adjusting call to
target_mourn_inferior to include ptid_t argument.
gdb/gdbserver/ChangeLog:
2016-09-19 Sergio Durigan Junior <sergiodj@redhat.com>
* server.c (start_inferior): Call target_mourn_inferior instead of
mourn_inferior; pass ptid_t argument to it.
(resume): Likewise.
(handle_target_event): Likewise.
* target.c (target_mourn_inferior): New function.
* target.h (mourn_inferior): Delete macro.
This commit moves the target_wait prototype from the GDB-specific
target.h header to the common target/target.h header. Then, it
creates a compatible implementation of target_wait on gdbserver using
the_target->wait, and adjusts the (only) caller (mywait function).
Pretty straightforward, no regressions introduced.
gdb/gdbserver/ChangeLog:
2016-09-01 Sergio Durigan Junior <sergiodj@redhat.com>
* target.c (mywait): Call target_wait instead of
the_target->wait.
(target_wait): New function.
gdb/ChangeLog:
2016-09-01 Sergio Durigan Junior <sergiodj@redhat.com>
* target.c (target_wait): Mention that the function's prototype
can be found at target/target.h.
* target.h (target_wait): Move prototype from here...
* target/target.h (target_wait): ... to here.
When executing commands on a secondary UI running the MI interpreter,
some commands that should be synchronous are not. MI incorrectly
continues processing input right after the synchronous command is
sent, before the target stops.
The problem happens when we emit MI async events (=library-loaded,
etc.), and we go about restoring the previous terminal state, we end
up calling target_terminal_ours, which incorrectly always installs the
current UI's input_fd in the event loop... That is, code like this:
old_chain = make_cleanup_restore_target_terminal ();
target_terminal_ours_for_output ();
fprintf_unfiltered (mi->event_channel, "library-loaded");
...
do_cleanups (old_chain);
The fix is to move the add_file_handler/delete_file_handler calls out
of target_terminal_$foo, making these completely no-ops unless called
with the main UI as current UI.
gdb/ChangeLog:
2016-08-09 Pedro Alves <palves@redhat.com>
PR gdb/20418
* event-top.c (ui_register_input_event_handler)
(ui_unregister_input_event_handler): New functions.
(async_enable_stdin): Register input in the event loop.
(async_disable_stdin): Unregister input from the event loop.
(gdb_setup_readline): Register input in the event loop.
* infrun.c (check_curr_ui_sync_execution_done): Register input in
the event loop.
* target.c (target_terminal_inferior): Don't unregister input from
the event loop.
(target_terminal_ours): Don't register input in the event loop.
* target.h (target_terminal_inferior)
(target_terminal_ours_for_output, target_terminal_ours): Update
comments.
* top.h (ui_register_input_event_handler)
(ui_unregister_input_event_handler): New declarations.
* utils.c (ui_unregister_input_event_handler_cleanup)
(prepare_to_handle_input): New functions.
(defaulted_query, prompt_for_continue): Use
prepare_to_handle_input.
gdb/testsuite/ChangeLog:
2016-08-09 Pedro Alves <palves@redhat.com>
Simon Marchi <simon.marchi@ericsson.com>
PR gdb/20418
* gdb.mi/new-ui-mi-sync.c, gdb.mi/new-ui-mi-sync.exp: New files.
* lib/mi-support.exp (mi_expect_interrupt): Remove anchors.
Some analysis we did here showed that increasing the cap on the
transfer size in target.c:memory_xfer_partial could give 20% or more
improvement in remote load across JTAG. Transfer sizes were capped
to 4K bytes because of performance problems encountered with the
restore command, documented here:
https://sourceware.org/ml/gdb-patches/2013-07/msg00611.html
and in commit 67c059c29e ("Improve performance of large restore
commands").
The 4K cap was introduced because in a case where the restore command
requested a 100MB transfer, memory_xfer_partial would repeatedy
allocate and copy an entire 100MB buffer in order to properly handle
breakpoint shadow instructions, even though memory_xfer_partial would
actually only write a small portion of the buffer contents.
A couple of alternative solutions were suggested:
* change the algorithm for handling the breakpoint shadow instructions
* throttle the transfer size up or down based on the previous actual
transfer size
I tried implementing the throttling approach, and my implementation
reduced the performance in some cases.
This patch implements a new target function that returns that target's
limit on memory transfer size. It defaults to ULONGEST_MAX bytes,
because for native targets there is no marshaling and thus no limit is
needed. For remote targets it uses get_memory_write_packet_size.
gdb/ChangeLog:
* remote.c (remote_get_memory_xfer_limit): New function.
* target-delegates.c: Regenerate.
* target.c (memory_xfer_partial): Call
target_ops.to_get_memory_xfer_limit.
* target.h (struct target_ops)
<to_get_memory_xfer_limit>: New member.
Several targets have a copy of the same code that prints
"Detaching from program ..."
in their target_detach implementation. Factor that out to a common
function.
(For now, I left the couple targets that print this a bit differently
alone. Maybe this could be further pulled out into infcmd.c. If we
did that, and those targets want to continue printing differently,
this new function could be converted to a target method.)
gdb/ChangeLog:
2016-07-01 Pedro Alves <palves@redhat.com>
* darwin-nat.c (darwin_detach): Use target_announce_detach.
* inf-ptrace.c (inf_ptrace_detach): Likewise.
* nto-procfs.c (procfs_detach): Likewise.
* remote.c (remote_detach_1): Likewise.
* target.c (target_announce_detach): New function.
* target.h (target_announce_detach): New declaration.
If the user presses Ctrl-C immediately before target_terminal_inferior
is called and the target is resumed, instead of after, the Ctrl-C ends
up pending in the quit flag until the target next stops.
remote.c has this bit to handle this:
if (!target_is_async_p ())
{
ofunc = signal (SIGINT, sync_remote_interrupt);
/* If the user hit C-c before this packet, or between packets,
pretend that it was hit right here. */
if (check_quit_flag ())
sync_remote_interrupt (SIGINT);
}
But that's only reachable if async is off, while async is on by
default nowadays. It's also obviously not reacheable on native
targets.
This patch generalizes that to all targets.
We can't remove that remote.c bit yet, until we get rid of the sync
SIGINT handler though. That'll be done later in the series.
gdb/ChangeLog:
2016-04-12 Pedro Alves <palves@redhat.com>
* remote.c (remote_pass_ctrlc): New function.
(init_remote_ops): Install it.
* target.c (target_terminal_inferior): Pass pending Ctrl-C to the
target.
(target_pass_ctrlc, default_target_pass_ctrlc): New functions.
* target.h (struct target_ops) <to_pass_ctrlc>: New method.
(target_pass_ctrlc, default_target_pass_ctrlc): New declarations.
* target-delegates.c: Regenerate.
I see a timeout in gdb.base/random-signal.exp,
Continuing.^M
PASS: gdb.base/random-signal.exp: continue
^CPython Exception <type 'exceptions.KeyboardInterrupt'> <type
exceptions.KeyboardInterrupt'>: ^M
FAIL: gdb.base/random-signal.exp: stop with control-c (timeout)
it can be reproduced by running random-signal.exp with native-gdbserver
in a loop, like this, and the fail will be shown in about 20 runs,
$ (set -e; while true; do make check RUNTESTFLAGS="--target_board=native-gdbserver random-signal.exp"; done)
In the test, the program is being single-stepped for software watchpoint,
and in each internal stop, python unwinder sniffer is used,
#0 pyuw_sniffer (self=<optimised out>, this_frame=<optimised out>, cache_ptr=0xd554f8) at /home/yao/SourceCode/gnu/gdb/git/gdb/python/py-unwind.c:608
#1 0x00000000006a10ae in frame_unwind_try_unwinder (this_frame=this_frame@entry=0xd554e0, this_cache=this_cache@entry=0xd554f8, unwinder=0xecd540)
at /home/yao/SourceCode/gnu/gdb/git/gdb/frame-unwind.c:107
#2 0x00000000006a143f in frame_unwind_find_by_frame (this_frame=this_frame@entry=0xd554e0, this_cache=this_cache@entry=0xd554f8)
at /home/yao/SourceCode/gnu/gdb/git/gdb/frame-unwind.c:163
#3 0x000000000069dc6b in compute_frame_id (fi=0xd554e0) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:454
#4 get_prev_frame_if_no_cycle (this_frame=this_frame@entry=0xd55410) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1781
#5 0x000000000069fdb9 in get_prev_frame_always_1 (this_frame=0xd55410) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1955
#6 get_prev_frame_always (this_frame=this_frame@entry=0xd55410) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1971
#7 0x00000000006a04b1 in get_prev_frame (this_frame=this_frame@entry=0xd55410) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:2213
when GDB goes to python extension, or other language extension, the
SIGINT handler is changed, and is restored when GDB leaves extension
language. GDB only stays in extension language for a very short period
in this case, but if ctrl-c is pressed at that moment, python extension
will handle the SIGINT, and exceptions.KeyboardInterrupt is shown.
Language extension is used in GDB side rather than inferior side,
so GDB should only change SIGINT handler for extension language when
the terminal is ours (not inferior's). This is what this patch does.
With this patch applied, I run random-signal.exp in a loop for 18
hours, and no fail is shown.
gdb:
2016-01-08 Yao Qi <yao.qi@linaro.org>
* extension.c: Include target.h.
(set_active_ext_lang): Only call install_gdb_sigint_handler,
check_quit_flag, and set_quit_flag if target_terminal_is_ours
returns false.
(restore_active_ext_lang): Likewise.
* target.c (target_terminal_is_ours): New function.
* target.h (target_terminal_is_ours): Declare.
When testing with "maint set target-non-stop on", a few
threading-related tests expose an issue that requires new RSP packets.
Say there are 3 threads running, 1-3. If GDB tries to stop thread 1,
2 and 3, and then waits for their stops, but meanwhile say, thread 2
exits, GDB hangs forever waiting for a stop for thread 2 that won't
ever happen.
This patch fixes the issue by adding support for thread exit events to
the protocol. However, we don't want these always enabled, as they're
useless most of the time, and would slow down remote debugging. So I
made it so that GDB can enable/disable them, and then made gdb do that
around the cases that need it, which currently is only
infrun.c:stop_all_threads.
In turn, if we have thread exit events, then the extra "thread x
exited" traffic slows down attach-many-short-lived-threads.exp enough
that gdb has trouble keeping up with new threads that are spawned
while gdb tries to stop existing ones. To fix that I added support
for the counterpart thread created events too. Enabling those when we
try to stop threads ensures that new threads never get a chance to
themselves start new threads, killing the race.
gdb/doc/ChangeLog:
2015-11-30 Pedro Alves <palves@redhat.com>
* gdb.texinfo (Remote Configuration): List "set/show remote
thread-events" command in configuration table.
(Stop Reply Packets): Document "T05 create" stop
reason and 'w' stop reply.
(General Query Packets): Document QThreadEvents packet. Document
QThreadEvents qSupported feature.
gdb/gdbserver/ChangeLog:
2015-11-30 Pedro Alves <palves@redhat.com>
* linux-low.c (handle_extended_wait): Assert that the LWP's
waitstatus is TARGET_WAITKIND_IGNORE. If GDB wants to hear about
thread create events, leave the new child's status pending.
(linux_low_filter_event): If GDB wants to hear about thread exit
events, leave the LWP marked dead and don't delete it.
(linux_wait_for_event_filtered): Don't check for thread exit.
(filter_exit_event): New function.
(linux_wait_1): Use it, when returning an exit event.
(linux_resume_one_lwp_throw): Assert that the LWP's
waitstatus is TARGET_WAITKIND_IGNORE.
* remote-utils.c (prepare_resume_reply): Handle
TARGET_WAITKIND_THREAD_CREATED and TARGET_WAITKIND_THREAD_EXITED.
* server.c (report_thread_events): New global.
(handle_general_set): Handle QThreadEvents.
(handle_query) <qSupported>: Handle and report QThreadEvents+;
(handle_target_event): Handle TARGET_WAITKIND_THREAD_CREATED and
TARGET_WAITKIND_THREAD_EXITED.
* server.h (report_thread_events): Declare.
gdb/ChangeLog:
2015-11-30 Pedro Alves <palves@redhat.com>
* NEWS (New commands): Mention "set/show remote thread-events"
commands.
(New remote packets): Mention thread created/exited stop reasons
and QThreadEvents packet.
* infrun.c (disable_thread_events): New function.
(stop_all_threads): Disable/enable thread create/exit events.
Handle TARGET_WAITKIND_THREAD_EXITED.
(handle_inferior_event_1): Handle TARGET_WAITKIND_THREAD_CREATED
and TARGET_WAITKIND_THREAD_EXITED.
* remote.c (remove_child_of_pending_fork): Also remove threads of
threads that have TARGET_WAITKIND_THREAD_EXITED events.
(remote_parse_stop_reply): Handle "create" magic register. Handle
'w' stop reply.
(initialize_remote): Install remote_thread_events as
to_thread_events target hook.
(remote_thread_events): New function.
* target-delegates.c: Regenerate.
* target.c (target_thread_events): New function.
* target.h (struct target_ops) <to_thread_events>: New field.
(target_thread_events): Declare.
* target/waitstatus.c (target_waitstatus_to_string): Handle
TARGET_WAITKIND_THREAD_CREATED and TARGET_WAITKIND_THREAD_EXITED.
* target/waitstatus.h (enum target_waitkind)
<TARGET_WAITKIND_THREAD_CREATED, TARGET_WAITKIND_THREAD_EXITED):
New values.
This patch adds support for thread names in the remote protocol, and
updates gdb/gdbserver to use it. The information is added to the XML
description sent in response to the qXfer:threads:read packet.
gdb/ChangeLog:
* linux-nat.c (linux_nat_thread_name): Replace implementation by call
to linux_proc_tid_get_name.
* nat/linux-procfs.c (linux_proc_tid_get_name): New function,
implementation inspired by linux_nat_thread_name.
* nat/linux-procfs.h (linux_proc_tid_get_name): New declaration.
* remote.c (struct private_thread_info) <name>: New field.
(free_private_thread_info): Free name field.
(remote_thread_name): New function.
(thread_item_t) <name>: New field.
(clear_threads_listing_context): Free name field.
(start_thread): Get name xml attribute.
(thread_attributes): Add "name" attribute.
(remote_update_thread_list): Copy name field.
(init_remote_ops): Assign remote_thread_name callback.
* target.h (target_thread_name): Update comment.
* NEWS: Mention remote thread name support.
gdb/gdbserver/ChangeLog:
* linux-low.c (linux_target_ops): Use linux_proc_tid_get_name.
* server.c (handle_qxfer_threads_worker): Refactor to include thread
name in reply.
* target.h (struct target_ops) <thread_name>: New field.
(target_thread_name): New macro.
gdb/doc/ChangeLog:
* gdb.texinfo (Thread List Format): Mention thread names.
Since this code path returns a string owned by the target (we don't know how
it's allocated, could be a static read-only string), it's safer if we return
a constant string. If, for some reasons, the caller wishes to modify the
string, it should make itself a copy.
gdb/ChangeLog:
* linux-nat.c (linux_nat_thread_name): Constify return value.
* target.h (struct target_ops) <to_thread_name>: Likewise.
(target_thread_name): Likewise.
* target.c (target_thread_name): Likewise.
* target-delegates.c (debug_thread_name): Regenerate.
* python/py-infthread.c (thpy_get_name): Constify local variables.
* thread.c (print_thread_info): Likewise.
(thread_find_command): Likewise.
Fixes:
src/gdb/remote.c: In function ‘void remote_unpush_target()’:
src/gdb/remote.c:4610:45: error: invalid conversion from ‘int’ to ‘strata’ [-fpermissive]
pop_all_targets_above (process_stratum - 1);
^
In file included from src/gdb/inferior.h:38:0,
from src/gdb/remote.c:25:
src/gdb/target.h:2299:13: error: initializing argument 1 of ‘void pop_all_targets_above(strata)’ [-fpermissive]
extern void pop_all_targets_above (enum strata above_stratum);
^
I used to carry a patch in the C++ branch that just did:
- pop_all_targets_above (process_stratum - 1);
+ pop_all_targets_above ((enum strata) (process_stratum - 1));
But then thought that maybe adding a routine that does exactly what we
need results in clearer code. This is the result.
gdb/ChangeLog:
2015-11-19 Pedro Alves <palves@redhat.com>
* remote.c (remote_unpush_target): Use
pop_all_targets_at_and_above instead of pop_all_targets_above.
* target.c (unpush_target_and_assert): New function, factored out
from ...
(pop_all_targets_above): ... here.
(pop_all_targets_at_and_above): New function.
* target.h (pop_all_targets_at_and_above): Declare.
Add a new target method to_record_will_replay to query if there is a record
target that will replay at least one thread matching the argument PTID if it
were executed in the argument execution direction.
gdb/
* record-btrace.c ((record_btrace_will_replay): New.
(init_record_btrace_ops): Initialize to_record_will_replay.
* record-full.c ((record_full_will_replay): New.
(init_record_full_ops): Initialize to_record_will_replay.
* target-delegates.c: Regenerated.
* target.c (target_record_will_replay): New.
* target.h (struct target_ops) <to_record_will_replay>: New.
(target_record_will_replay): New.
Signed-off-by: Markus Metzger <markus.t.metzger@intel.com>
The to_record_is_replaying target method is used to query record targets if
they are replaying. This is currently interpreted as "is any thread being
replayed".
Add a PTID argument and change the interpretation to "is any thread matching
PTID being replayed".
Change all users to pass minus_one_ptid to preserve the old meaning.
The record full target does not really support multi-threading and ignores
the PTID argument.
gdb/
* record-btrace.c (record_btrace_is_replaying): Add ptid argument.
Update users to pass minus_one_ptid.
* record-full.c (record_full_is_replaying): Add ptid argument (ignored).
* record.c (cmd_record_delete): Pass inferior_ptid to
target_record_is_replaying.
* target-delegates.c: Regenerated.
* target.c (target_record_is_replaying): Add ptid argument.
* target.h (struct target_ops) <to_record_is_replaying>: Add ptid
argument.
(target_record_is_replaying): Add ptid argument.
Nowadays, GDB only knows whether architecture supports hardware single
step or software single step (through gdbarch hook software_single_step),
and for a given instruction or instruction sequence, GDB knows how to
do single step (hardware or software). However, GDB doesn't know whether
the target supports hardware single step. It is possible that the
architecture doesn't support hardware single step, such as arm, but
the target supports, such as simulator. This was discussed in this
thread https://www.sourceware.org/ml/gdb/2009-12/msg00033.html before.
I encounter this problem for aarch64 multi-arch support. When aarch64
debugs arm program, gdbarch is arm, so software single step is still
used. However, the underneath linux kernel does support hardware
single step, so IWBN to use it.
This patch is to add a new target_ops hook to_can_do_single_step, and
only use it in arm_linux_software_single_step to decide whether or not
to use hardware single step. On the native aarch64 linux target, 1 is
returned. On other targets, -1 is returned. On the remote target, if
the target supports s and S actions in the vCont? reply, then target
can do single step. However, old GDBserver will send s and S in the
reply to vCont?, which will confuse new GDB. For example, old GDBserver
on arm-linux will send s and S in the reply to vCont?, but it doesn't
support hardware single step. On the other hand, new GDBserver, on
arm-linux for example, will not send s and S in the reply to vCont?,
but old GDB thinks it doesn't support vCont packet at all. In order
to address this problem, I add a new qSupported feature vContSupported,
which indicates GDB wants to know the supported actions in the reply
to vCont?, and qSupported response contains vContSupported if the
stub is able tell supported vCont actions in the reply of vCont?.
If the patched GDB talks with patched GDBserver on x86, the RSP traffic
is like this:
-> $qSupported:...+;vContSupported+
<- ...+;vContSupported+
...
-> $vCont?
<- vCont;c;C;t;s;S;r
then, GDB knows the stub can do single step, and may stop using software
single step even the architecture doesn't support hardware single step.
If the patched GDB talks with patched GDBserver on arm, the last vCont?
reply will become:
<- vCont;c;C;t
GDB thinks the target doesn't support single step, so it will use software
single step.
If the patched GDB talks with unpatched GDBserver, the RSP traffic is like
this:
-> $qSupported:...+;vContSupported+
<- ...+
...
-> $vCont?
<- vCont;c;C;t;s;S;r
although GDBserver returns s and S, GDB still thinks GDBserver may not
support single step because it doesn't support vContSupported.
If the unpatched GDB talks with patched GDBserver on x86, the RSP traffic
is like:
-> $qSupported:...+;
<- ...+;vContSupported+
...
-> $vCont?
<- vCont;c;C;t;s;S;r
Since GDB doesn't sent vContSupported in the qSupported feature, GDBserver
sends s and S regardless of the support of hardware single step.
gdb:
2015-09-15 Yao Qi <yao.qi@linaro.org>
* aarch64-linux-nat.c (aarch64_linux_can_do_single_step): New
function.
(_initialize_aarch64_linux_nat): Install it to to_can_do_single_step.
* arm-linux-tdep.c (arm_linux_software_single_step): Return 0
if target_can_do_single_step returns 1.
* remote.c (struct vCont_action_support) <s, S>: New fields.
(PACKET_vContSupported): New enum.
(remote_protocol_features): New element for vContSupported.
(remote_query_supported): Append "vContSupported+".
(remote_vcont_probe): Remove support_s and support_S, use
rs->supports_vCont.s and rs->supports_vCont.S instead. Disable
vCont packet if c and C actions are not supported.
(remote_can_do_single_step): New function.
(init_remote_ops): Install it to to_can_do_single_step.
(_initialize_remote): Call add_packet_config_cmd.
* target.h (struct target_ops) <to_can_do_single_step>: New field.
(target_can_do_single_step): New macro.
* target-delegates.c: Re-generated.
gdb/gdbserver:
2015-09-15 Yao Qi <yao.qi@linaro.org>
* server.c (vCont_supported): New global variable.
(handle_query): Set vCont_supported to 1 if "vContSupported+"
matches. Append ";vContSupported+" to own_buf.
(handle_v_requests): Append ";s;S" to own_buf if target supports
hardware single step or vCont_supported is false.
(capture_main): Set vCont_supported to zero.
gdb/doc:
2015-09-15 Yao Qi <yao.qi@linaro.org>
* gdb.texinfo (General Query Packets): Add vContSupported to
tables of 'gdbfeatures' and 'stub features' supported in the
qSupported packet, as well as to the list containing stub
feature details.
This patch implements support for exec events on extended-remote Linux
targets. Follow-exec-mode and rerun behave as expected. Catchpoints and
test updates are implemented in subsequent patches.
This patch was derived from a patch posted last October:
https://sourceware.org/ml/gdb-patches/2014-10/msg00877.html.
It was originally based on some work done by Luis Machado in 2013.
IMPLEMENTATION
----------------
Exec events are enabled via ptrace options.
When an exec event is detected by gdbserver, the existing process
data, along with all its associated lwp and thread data, is deleted
and replaced by data for a new single-threaded process. The new
process data is initialized with the appropriate parts of the state
of the execing process. This approach takes care of several potential
pitfalls, including:
* deleting the data for an execing non-leader thread before any
wait/sigsuspend occurs
* correctly initializing the architecture of the execed process
We then report the exec event using a new RSP stop reason, "exec".
When GDB receives an "exec" event, it saves the status in the event
structure's target_waitstatus field, like what is done for remote fork
events. Because the original and execed programs may have different
architectures, we skip parsing the section of the stop reply packet
that contains register data. The register data will be retrieved
later after the inferior's architecture has been set up by
infrun.c:follow_exec.
At that point the exec event is handled by the existing event handling
in GDB. However, a few changes were necessary so that
infrun.c:follow_exec could accommodate the remote target.
* Where follow-exec-mode "new" is handled, we now call
add_inferior_with_spaces instead of add_inferior with separate calls
to set up the program and address spaces. The motivation for this
is that add_inferior_with_spaces also sets up the initial architecture
for the inferior, which is needed later by target_find_description
when it calls target_gdbarch.
* We call a new target function, target_follow_exec. This function
allows us to store the execd_pathname in the inferior, instead of
using the static string remote_exec_file from remote.c. The static
string didn't work for follow-exec-mode "new", since once you switched
to the execed program, the original remote exec-file was lost. The
execd_pathname is now stored in the inferior's program space as a
REGISTRY field. All of the requisite mechanisms for this are
defined in remote.c.
gdb/gdbserver/ChangeLog:
* linux-low.c (linux_mourn): Static declaration.
(linux_arch_setup): Move in front of
handle_extended_wait.
(linux_arch_setup_thread): New function.
(handle_extended_wait): Handle exec events. Call
linux_arch_setup_thread. Make event_lwp argument a
pointer-to-a-pointer.
(check_zombie_leaders): Do not check stopped threads.
(linux_low_ptrace_options): Add PTRACE_O_TRACEEXEC.
(linux_low_filter_event): Add lwp and thread for exec'ing
non-leader thread if leader thread has been deleted.
Refactor code into linux_arch_setup_thread and call it.
Pass child lwp pointer by reference to handle_extended_wait.
(linux_wait_for_event_filtered): Update comment.
(linux_wait_1): Prevent clobbering exec event status.
(linux_supports_exec_events): New function.
(linux_target_ops) <supports_exec_events>: Initialize new member.
* lynx-low.c (lynx_target_ops) <supports_exec_events>: Initialize
new member.
* remote-utils.c (prepare_resume_reply): New stop reason 'exec'.
* server.c (report_exec_events): New global variable.
(handle_query): Handle qSupported query for exec-events feature.
(captured_main): Initialize report_exec_events.
* server.h (report_exec_events): Declare new global variable.
* target.h (struct target_ops) <supports_exec_events>: New
member.
(target_supports_exec_events): New macro.
* win32-low.c (win32_target_ops) <supports_exec_events>:
Initialize new member.
gdb/ChangeLog:
* infrun.c (follow_exec): Use process-style ptid for
exec message. Call add_inferior_with_spaces and
target_follow_exec.
* nat/linux-ptrace.c (linux_supports_traceexec): New function.
* nat/linux-ptrace.h (linux_supports_traceexec): Declare.
* remote.c (remote_pspace_data): New static variable.
(remote_pspace_data_cleanup): New function.
(get_remote_exec_file): New function.
(set_remote_exec_file_1): New function.
(set_remote_exec_file): New function.
(show_remote_exec_file): New function.
(remote_exec_file): Delete static variable.
(anonymous enum) <PACKET_exec_event_feature> New
enumeration constant.
(remote_protocol_features): Add entry for exec-events feature.
(remote_query_supported): Add client side of qSupported query
for exec-events feature.
(remote_follow_exec): New function.
(remote_parse_stop_reply): Handle 'exec' stop reason.
(extended_remote_run, extended_remote_create_inferior): Call
get_remote_exec_file and set_remote_exec_file_1.
(init_extended_remote_ops) <to_follow_exec>: Initialize new
member.
(_initialize_remote): Call
register_program_space_data_with_cleanup. Call
add_packet_config_cmd for remote exec-events feature.
Modify call to add_setshow_string_noescape_cmd for exec-file
to use new functions set_remote_exec_file and
show_remote_exec_file.
* target-debug.h, target-delegates.c: Regenerated.
* target.c (target_follow_exec): New function.
* target.h (struct target_ops) <to_follow_exec>: New member.
(target_follow_exec): Declare new function.
This patch makes the execution control code use largely the same
mechanisms in both sync- and async-capable targets. This means using
continuations and use the event loop to react to target events on sync
targets as well. The trick is to immediately mark infrun's event loop
source after resume instead of calling wait_for_inferior. Then
fetch_inferior_event is adjusted to do a blocking wait on sync
targets.
Tested on x86_64 Fedora 20, native and gdbserver, with and without
"maint set target-async off".
gdb/ChangeLog:
2015-09-09 Pedro Alves <palves@redhat.com>
* breakpoint.c (bpstat_do_actions_1, until_break_command): Don't
check whether the target can async.
* inf-loop.c (inferior_event_handler): Only call target_async if
the target can async.
* infcall.c: Include top.h and interps.h.
(run_inferior_call): For the interpreter to sync mode while
running the infcall. Call wait_sync_command_done instead of
wait_for_inferior plus normal_stop.
* infcmd.c (prepare_execution_command): Don't check whether the
target can async when running in the foreground.
(step_1): Delete synchronous case handling.
(step_once): Always install a continuation, even in sync mode.
(until_next_command, finish_forward): Don't check whether the
target can async.
(attach_command_post_wait, notice_new_inferior): Always install a
continuation, even in sync mode.
* infrun.c (mark_infrun_async_event_handler): New function.
(proceed): In sync mode, mark infrun's event source instead of
waiting for events here.
(fetch_inferior_event): If the target can't async, do a blocking
wait.
(prepare_to_wait): In sync mode, mark infrun's event source.
(infrun_async_inferior_event_handler): No longer bail out if the
target can't async.
* infrun.h (mark_infrun_async_event_handler): New declaration.
* linux-nat.c (linux_nat_wait_1): Remove calls to
set_sigint_trap/clear_sigint_trap.
(linux_nat_terminal_inferior): No longer check whether the target
can async.
* mi/mi-interp.c (mi_on_sync_execution_done): Update and simplify
comment.
(mi_execute_command_input_handler): No longer check whether the
target is async. Update and simplify comment.
* target.c (default_target_wait): New function.
* target.h (struct target_ops) <to_wait>: Now defaults to
default_target_wait.
(default_target_wait): Declare.
* top.c (wait_sync_command_done): New function, factored out from
...
(maybe_wait_sync_command_done): ... this.
* top.h (wait_sync_command_done): Declare.
* target-delegates.c: Regenerate.
Currently, when remote debugging, if you type Ctrl-C just while the
target stopped for an internal event, and GDB is busy doing something
that takes a while (e.g., fetching chunks of a shared library off of
the target, with vFile, to process ELF headers and debug info), the
Ctrl-C is lost.
The patch hooks up the QUIT macro to a new target method that lets the
target react to the double-Ctrl-C before the event loop is reached,
which allows reacting to a double-Ctrl-C even when GDB is busy doing
some long operation and not waiting for a stop reply. That end result
is:
(gdb) c
Continuing.
^C
^C
Interrupted while waiting for the program.
Give up waiting? (y or n) y
Quit
(gdb) info threads
Id Target Id Frame
* 1 Thread 11673 0x00007ffff7deb240 in _dl_debug_state () from target:/lib64/ld-linux-x86-64.so.2
(gdb)
If, however, GDB is waiting for a stop reply (because the target has
been resumed, with e.g., vCont;c), but the target isn't responding, we
now get:
(gdb) c
Continuing.
^C
^C
The target is not responding to interrupt requests.
Stop debugging it? (y or n) y
Disconnected from target.
(gdb) info threads
No threads.
This offers to disconnect, because when we're waiting for a stop
reply, there's nothing else we can send the target other than an
interrupt request. And if that doesn't work, there's nothing else we
can do.
The Ctrl-C is presently lost because until we get to a user-visible
stop, the SIGINT handler that is installed is the one that forwards
the interrupt to the remote side, with the \003 "packet" [1]. But,
gdbserver ignores an interrupt request if the program is stopped.
Still, even if it didn't, the server can only report back a
stop-because-of-SIGINT when the program is next resumed. And it may
take a while to actually re-resume the target.
[1] - In the old sync days, the remote target would react to a
double-Ctrl-C by asking users whether they wanted to give up waiting
and disconnect. The code is still there, but it it isn't reacheable
on most hosts, which support serial connections in async mode
(probably only DJGPP doesn't). Even then, in sync mode, remote.c's
SIGINT handler is only installed while the target is resumed, and is
removed as soon as the target sends back a stop reply. That means
that a Ctrl-C just while GDB is processing an internal event can end
up with an odd "Quit" at the prompt instead of "Program stopped by
SIGINT". In contrast, in async mode, remote.c's SIGINT handler is set
up as long as target_terminal_inferior or
target_terminal_ours_for_output are in effect (IOW, until we get a
user-visible stop and call target_terminal_ours), so the user
shouldn't get back a spurious Quit. However, it's still desirable to
be able to interrupt a long-running GDB operation, if GDB takes a
while to re-resume the target or get back to the event loop.
Tested on x86_64 Fedora 20.
gdb/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* defs.h (maybe_quit): Declare.
(QUIT): Now calls maybe_quit.
* event-loop.c (clear_async_signal_handler)
(async_signal_handler_is_marked): New functions.
* event-loop.h (async_signal_handler_is_marked)
(clear_async_signal_handler): New declarations.
* remote.c (remote_check_pending_interrupt): New function.
(interrupt_query): Use make_cleanup_restore_target_terminal. No
longer check whether the target is async. If waiting for a stop
reply, and a Ctrl-C as been sent to the target, offer to
disconnect, and throw TARGET_CLOSE_ERROR instead of a quit.
Otherwise do not disconnect and throw a quit.
(_initialize_remote): Install remote_check_pending_interrupt as
to_check_pending_interrupt.
* target.c (target_check_pending_interrupt): New function.
* target.h (struct target_ops) <to_check_pending_interrupt>: New
field.
(target_check_pending_interrupt): New declaration.
* utils.c (maybe_quit): New function.
* target-delegates.c: Regenerate.
GDB provides no indicator of progress during file operations, and can
appear to have locked up during slow remote transfers. This commit
updates GDB to print a warning each time a file is accessed over RSP.
An additional message detailing how to avoid remote transfers is
printed for the first transfer only.
gdb/ChangeLog:
* target.h (struct target_ops) <to_fileio_open>: New argument
warn_if_slow. Update comment. All implementations updated.
(target_fileio_open_warn_if_slow): New declaration.
* target.c (target_fileio_open): Renamed as...
(target_fileio_open_1): ...this. New argument warn_if_slow.
Pass warn_if_slow to implementation. Update debug printing.
(target_fileio_open): New function.
(target_fileio_open_warn_if_slow): Likewise.
* gdb_bfd.c (gdb_bfd_iovec_fileio_open): Use new function
target_fileio_open_warn_if_slow.
gdb/testsuite/ChangeLog:
* gdb.trace/pending.exp: Cope with remote transfer warnings.
With "maint set target-non-stop on" we get:
@@ -66,13 +66,16 @@ Continuing.
interrupt
(gdb) PASS: gdb.base/interrupt-noterm.exp: interrupt
-Program received signal SIGINT, Interrupt.
-PASS: gdb.base/interrupt-noterm.exp: inferior received SIGINT
-testcase src/gdb/testsuite/gdb.base/interrupt-noterm.exp completed in 0 seconds
+[process 12119] #1 stopped.
+0x0000003615ebc6d0 in __nanosleep_nocancel () at ../sysdeps/unix/syscall-template.S:81
+81 T_PSEUDO (SYSCALL_SYMBOL, SYSCALL_NAME, SYSCALL_NARGS)
+FAIL: gdb.base/interrupt-noterm.exp: inferior received SIGINT (timeout)
+testcase src/gdb/testsuite/gdb.base/interrupt-noterm.exp completed in 10 seconds
That is, we get "[$thread] #1 stopped" instead of SIGINT.
The issue is that we don't currently distinguish send
"interrupt/ctrl-c" to target terminal vs "stop/pause" thread well;
both cases go through "target_stop".
And then, the native Linux backend (linux-nat.c) implements
target_stop with SIGSTOP in non-stop mode, and SIGINT in all-stop
mode. Since "maint set target-non-stop on" forces the backend to be
always running in non-stop mode, even though the user-visible behavior
is "set non-stop" is "off", "interrupt" causes a SIGSTOP instead of
the SIGINT the test expects.
Fix this by introducing a target_interrupt method to use in the
"interrupt/ctrl-c" case, so "set non-stop off" can always work the
same irrespective of "maint set target-non-stop on/off". I'm
explictly considering changing the "set non-stop on" behavior as out
of scope here.
Most of the patch is an across-the-board rename of to_stop hook
implementations to to_interrupt. The only targets where something
more than a rename is being done are linux-nat.c and remote.c, which
are the only targets that support async, and thus are the only ones
the core side calls target_stop on.
gdb/ChangeLog:
2015-08-07 Pedro Alves <palves@redhat.com>
* darwin-nat.c (darwin_stop): Rename to ...
(darwin_interrupt): ... this.
(_initialize_darwin_inferior): Adjust.
* gnu-nat.c (gnu_stop): Delete.
(gnu_target): Don't install gnu_stop.
* inf-ptrace.c (inf_ptrace_stop): Rename to ...
(inf_ptrace_interrupt): ... this.
(inf_ptrace_target): Adjust.
* infcmd.c (interrupt_target_1): Use target_interrupt instead of
target_stop.
* linux-nat (linux_nat_stop): Rename to ...
(linux_nat_interrupt): ... this.
(linux_nat_stop): Reimplement.
(linux_nat_add_target): Install linux_nat_interrupt.
* nto-procfs.c (nto_interrupt_twice): Rename to ...
(nto_handle_sigint_twice): ... this.
(nto_interrupt): Rename to ...
(nto_handle_sigint): ... this. Call target_interrupt instead of
target_stop.
(procfs_wait): Adjust.
(procfs_stop): Rename to ...
(procfs_interrupt): ... this.
(init_procfs_targets): Adjust.
* procfs.c (procfs_stop): Rename to ...
(procfs_interrupt): ... this.
(procfs_target): Adjust.
* remote-m32r-sdi.c (m32r_stop): Rename to ...
(m32r_interrupt): ... this.
(init_m32r_ops): Adjust.
* remote-sim.c (gdbsim_stop_inferior): Rename to ...
(gdbsim_interrupt_inferior): ... this.
(gdbsim_stop): Rename to ...
(gdbsim_interrupt): ... this.
(gdbsim_cntrl_c): Adjust.
(init_gdbsim_ops): Adjust.
* remote.c (sync_remote_interrupt): Adjust comments.
(remote_stop_as): Rename to ...
(remote_interrupt_as): ... this.
(remote_stop): Adjust comment.
(remote_interrupt): New function.
(init_remote_ops): Install remote_interrupt.
* target.c (target_interrupt): New function.
* target.h (struct target_ops) <to_interrupt>: New field.
(target_interrupt): New declaration.
* windows-nat.c (windows_stop): Rename to ...
(windows_interrupt): ... this.
* target-delegates.c: Regenerate.
This finally implements user-visible all-stop mode running with the
target_ops backend always in non-stop mode. This is a stepping stone
towards finer-grained control of threads, being able to do interesting
things like thread groups, associating groups with breakpoints, etc.
From the user's perspective, all-stop mode is really just a special
case of being able to stop and resume specific sets of threads, so it
makes sense to do this step first.
With this, even in all-stop, the target is no longer in charge of
stopping all threads before reporting an event to the core -- the core
takes care of it when it sees fit. For example, when "next"- or
"step"-ing, we can avoid stopping and resuming all threads at each
internal single-step, and instead only stop all threads when we're
about to present the stop to the user.
The implementation is almost straight forward, as the heavy lifting
has been done already in previous patches. Basically, we replace
checks for "set non-stop on/off" (the non_stop global), with calls to
a new target_is_non_stop_p function. In a few places, if "set
non-stop off", we stop all threads explicitly, and in a few other
places we resume all threads explicitly, making use of existing
methods that were added for teaching non-stop to step over breakpoints
without displaced stepping.
This adds a new "maint set target-non-stop on/off/auto" knob that
allows both disabling the feature if we find problems, and
force-enable it for development (useful when teaching a target about
this. The default is "auto", which means the feature is enabled if a
new target method says it should be enabled. The patch implements the
method in linux-nat.c, just for illustration, because it still returns
false. We'll need a few follow up fixes before turning it on by
default. This is a separate target method from indicating regular
non-stop support, because e.g., while e.g., native linux-nat.c is
close to regression free with all-stop-non-stop (with following
patches will fixing the remaining regressions), remote.c+gdbserver
will still need more fixing, even though it supports "set non-stop
on".
Tested on x86_64 Fedora 20, native, with and without "set displaced
off", and with and without "maint set target-non-stop on"; and also
against gdbserver.
gdb/ChangeLog:
2015-08-07 Pedro Alves <palves@redhat.com>
* NEWS: Mention "maint set/show target-non-stop".
* breakpoint.c (update_global_location_list): Check
target_is_non_stop_p instead of non_stop.
* infcmd.c (attach_command_post_wait, attach_command): Likewise.
* infrun.c (show_can_use_displaced_stepping)
(can_use_displaced_stepping_p, start_step_over_inferior):
Likewise.
(internal_resume_ptid): New function.
(resume): Use it.
(proceed): Check target_is_non_stop_p instead of non_stop. If in
all-stop mode but the target is always in non-stop mode, start all
the other threads that are implicitly resumed too.
(for_each_just_stopped_thread, fetch_inferior_event)
(adjust_pc_after_break, stop_all_threads): Check
target_is_non_stop_p instead of non_stop.
(handle_inferior_event): Likewise. Handle detach-fork in all-stop
with the target always in non-stop mode.
(handle_signal_stop) <random signal>: Check target_is_non_stop_p
instead of non_stop.
(switch_back_to_stepped_thread): Check target_is_non_stop_p
instead of non_stop.
(keep_going_stepped_thread): Use internal_resume_ptid.
(stop_waiting): If in all-stop mode, and the target is in non-stop
mode, stop all threads.
(keep_going_pass): Likewise, when starting a new in-line step-over
sequence.
* linux-nat.c (get_pending_status, select_event_lwp)
(linux_nat_filter_event, linux_nat_wait_1, linux_nat_wait): Check
target_is_non_stop_p instead of non_stop.
(linux_nat_always_non_stop_p): New function.
(linux_nat_stop): Check target_is_non_stop_p instead of non_stop.
(linux_nat_add_target): Install linux_nat_always_non_stop_p.
* target-delegates.c: Regenerate.
* target.c (target_is_non_stop_p): New function.
(target_non_stop_enabled, target_non_stop_enabled_1): New globals.
(maint_set_target_non_stop_command)
(maint_show_target_non_stop_command): New functions.
(_initilize_target): Install "maint set/show target-non-stop"
commands.
* target.h (struct target_ops) <to_always_non_stop_p>: New field.
(target_non_stop_enabled): New declaration.
(target_is_non_stop_p): New declaration.
gdb/doc/ChangeLog:
2015-08-07 Pedro Alves <palves@redhat.com>
* gdb.texinfo (Maintenance Commands): Document "maint set/show
target-non-stop".
That is, step past breakpoints by:
- pausing all threads
- removing breakpoint at PC
- single-step
- reinsert breakpoint
- restart threads
similarly to all-stop (with displaced stepping disabled). This allows
non-stop to work on targets/architectures without displaced stepping
support. That is, it makes displaced stepping an optimization instead
of a requirement. For example, in principle, all GNU/Linux ports
support non-stop mode at the target_ops level, but not all
corresponding gdbarch's implement displaced stepping. This should
make non-stop work for all (albeit, not as efficiently). And then
there are scenarios where even if the architecture supports displaced
stepping, we can't use it, because we e.g., don't find a usable
address to use as displaced step scratch pad. It should also fix
stepping past watchpoints on targets that have non-continuable
watchpoints in non-stop mode (e.g., PPC, untested). Running the
instruction out of line in the displaced stepping scratch pad doesn't
help that case, as the copied instruction reads/writes the same
watched memory... We can fix that too by teaching GDB to only remove
the watchpoint from the thread that we want to move past the
watchpoint (currently, removing a watchpoint always removes it from
all threads), but again, that can be considered an optimization; not
all targets would support it.
For those familiar with the gdb and gdbserver Linux target_ops
backends, the implementation should look similar, except it is done on
the core side. When we pause threads, we may find they stop with an
interesting event that should be handled later when the thread is
re-resumed, thus we store such events in the thread object, and mark
the event as pending. We should only consume pending events if the
thread is indeed resumed, thus we add a new "resumed" flag to the
thread object. At a later stage, we might add new target methods to
accelerate some of this, like "pause all threads", with corresponding
RSP packets, but we'd still need a fallback method for remote targets
that don't support such packets, so, again, that can be deferred as
optimization.
My _real_ motivation here is making it possible to reimplement
all-stop mode on top of the target always working on non-stop mode, so
that e.g., we can send RSP packets to a remote target even while the
target is running -- can't do that in the all-stop RSP variant, by
design).
Tested on x86_64 Fedora 20, with and without "set displaced off"
forced. The latter forces the new code paths whenever GDB needs to
step past a breakpoint.
gdb/ChangeLog:
2015-08-07 Pedro Alves <pedro@codesourcery.com>
* breakpoint.c (breakpoints_should_be_inserted_now): If any thread
has a pending status, return true.
* gdbthread.h: Include target/waitstatus.h.
(struct thread_suspend_state) <stop_reason, waitstatus_pending_p,
stop_pc>: New fields.
(struct thread_info) <resumed>: New field.
(set_resumed): Declare.
* infrun.c: Include "event-loop.h".
(infrun_async_inferior_event_token, infrun_is_async): New globals.
(infrun_async): New function.
(clear_step_over_info): Add debug output.
(displaced_step_in_progress_any_inferior): New function.
(displaced_step_fixup): New returns int.
(start_step_over): Handle in-line step-overs too. Assert the
thread is marked resumed.
(resume_cleanups): Clear the thread's resumed flag.
(resume): Set the thread's resumed flag. Return early if the
thread has a pending status. Allow stepping a breakpoint with no
signal.
(proceed): Adjust to check 'resumed' instead of 'executing'.
(clear_proceed_status_thread): If the thread has a pending status,
and that status is a finished step, discard the pending status.
(clear_proceed_status): Don't clear step_over_info here.
(random_pending_event_thread, do_target_wait): New functions.
(prepare_for_detach, wait_for_inferior, fetch_inferior_event): Use
do_target_wait.
(wait_one): New function.
(THREAD_STOPPED_BY): New macro.
(thread_stopped_by_watchpoint, thread_stopped_by_sw_breakpoint)
(thread_stopped_by_hw_breakpoint): New functions.
(switch_to_thread_cleanup, save_waitstatus, stop_all_threads): New
functions.
(handle_inferior_event): Also call set_resumed(false) on all
threads implicitly stopped by the event.
(restart_threads, resumed_thread_with_pending_status): New
functions.
(finish_step_over): If we were doing an in-line step-over before,
and no longer are after trying to start a new step-over, restart
all threads. If we have multiple threads with pending events,
save the current event and go through the event loop again.
(handle_signal_stop): Return early if finish_step_over returns
false.
<random signal>: If we get a signal while stepping over a
breakpoint in-line in non-stop mode, restart all threads. Clear
step_over_info before delivering the signal.
(keep_going_stepped_thread): Use internal_error instead of
gdb_assert. Mark the thread as resumed.
(keep_going_pass_signal): Assert the thread isn't already resumed.
If some other thread is doing an in-line step-over, defer the
resume. If we just started a new in-line step-over, stop all
threads. Don't clear step_over_info.
(infrun_async_inferior_event_handler): New function.
(_initialize_infrun): Create async event handler with
infrun_async_inferior_event_handler as callback.
(infrun_async): New declaration.
* target.c (target_async): New function.
* target.h (target_async): Declare macro and readd as function
declaration.
* target/waitstatus.h (enum target_stop_reason)
<TARGET_STOPPED_BY_SINGLE_STEP>: New value.
* thread.c (new_thread): Clear the new waitstatus field.
(set_resumed): New function.
Clarify that currently_stepping works at a higher level than
target_resume.
gdb/ChangeLog:
2015-08-07 Pedro Alves <palves@redhat.com>
* infrun.c (currently_stepping): Extend intro comment.
* target.h (target_resume): Extend intro comment.
We see the following regressions in testing on x86_64-linux,
reverse-step^M
Cannot access memory at address 0x2aaaaaed26c0^M
(gdb) FAIL: gdb.reverse/solib-precsave.exp: reverse-step into solib function one
when GDB reverse step into a function, GDB wants to skip prologue so
it requests TARGET_OBJECT_CODE_MEMORY to read some code memory in
memory_xfer_partial_1. However in dcache_read_memory_partial, the object
becomes TARGET_OBJECT_MEMORY
return ops->to_xfer_partial (ops, TARGET_OBJECT_MEMORY, NULL,
myaddr, NULL, memaddr, len,
xfered_len);
in reverse debugging, ops->to_xfer_partial is record_full_core_xfer_partial
and it will return TARGET_XFER_E_IO because it can't find any records.
The test fails.
At this moment, the delegate relationship is like
dcache -> record-core -> core -> exec
and we want to GDB read memory across targets, which means if the
requested memory isn't found in record-core, GDB can read memory from
core, and exec even further if needed. I find raw_memory_xfer_partial
is exactly what I want.
gdb:
2015-07-29 Yao Qi <yao.qi@linaro.org>
PR record/18691
* dcache.c (dcache_read_memory_partial): Call
raw_memory_xfer_partial.
* target.c (raw_memory_xfer_partial): Make it non-static.
* target.h (raw_memory_xfer_partial): Declare.
This should be just a move with no changes.
gdb/ChangeLog
2015-07-15 Aleksandar Ristovski <aristovski@qnx.com
Jan Kratochvil <jan.kratochvil@redhat.com>
Move linux_find_memory_regions_full & co.
* linux-tdep.c (nat/linux-maps.h): Include.
(gdb_regex.h): Remove the include.
(enum filterflags, struct smaps_vmflags, read_mapping, decode_vmflags)
(mapping_is_anonymous_p, dump_mapping_p): Moved to nat/linux-maps.c.
(linux_find_memory_region_ftype): Moved typedef to nat/linux-maps.h.
(linux_find_memory_regions_full): Moved definition to nat/linux-maps.c.
* nat/linux-maps.c: Include ctype.h, target/target-utils.h, gdb_regex.h
and target/target.h.
(struct smaps_vmflags, read_mapping, decode_vmflags)
(mapping_is_anonymous_p, dump_mapping_p): Move from linux-tdep.c.
(linux_find_memory_regions_full): Move from linux-tdep.c.
* nat/linux-maps.h (read_mapping): New declaration.
(linux_find_memory_region_ftype, enum filterflags): Moved from
linux-tdep.c.
(linux_find_memory_regions_full): New declaration.
* target.c (target/target-utils.h): Include.
(read_alloc_pread_ftype): Moved typedef to target/target-utils.h.
(read_alloc, read_stralloc_func_ftype, read_stralloc): Moved
definitions to target/target-utils.c.
* target.h (target_fileio_read_stralloc): Move it to target/target.h.
* target/target-utils.c (read_alloc, read_stralloc): Move definitions
from target.c.
* target/target-utils.h (read_alloc_pread_ftype): New typedef.
(read_alloc): New declaration.
(read_stralloc_func_ftype): New typedef.
(read_stralloc): New declaration.
* target/target.h (target_fileio_read_stralloc): Move it from target.h.
gdb/gdbserver/ChangeLog
2015-07-15 Aleksandar Ristovski <aristovski@qnx.com
Jan Kratochvil <jan.kratochvil@redhat.com>
* target.c: Include target/target-utils.h and fcntl.h.
(target_fileio_read_stralloc_1_pread, target_fileio_read_stralloc_1)
(target_fileio_read_stralloc): New functions.
If we are reading/writing from a memory object, the length represents
the number of "addresses" to read/write, so the addressable unit size
needs to be taken into account when allocating memory on gdb's side.
gdb/ChangeLog:
* target.c (target_read): Consider addressable unit size when
reading from a memory object.
(read_memory_robust): Same.
(read_whatever_is_readable): Same.
(target_write_with_progress): Consider addressable unit size
when writing to a memory object.
* target.h (target_read): Update documentation.
(target_write): Add documentation.
This contains various cleanups in the target memory read and write code.
They are not directly related to the non-8-bits changes, but they
clarify things a bit down the line.
gdb/ChangeLog:
* target.c (target_read): Rename variables and use
TARGET_XFER_E_IO.
(target_read_with_progress): Same.
(read_memory_robust): Constify parameters and rename
variables.
(read_whatever_is_readable): Constify parameters,
rename variables, adjust formatting.
* target.h (read_memory_robust): Constify parameters.
This commit adds a new argument to all target_fileio functions with
filename arguments to allow the desired inferior to be specified.
This allows GDB to support systems where processes do not necessarily
share a common filesystem.
gdb/ChangeLog:
* target.h (struct inferior): New forward declaration.
(struct target_ops) <to_filesystem_is_local>: Update comment.
(struct target_ops) <to_fileio_open>: New argument inf.
Update comment. All implementations updated.
(struct target_ops) <to_fileio_unlink>: Likewise.
(struct target_ops) <to_fileio_readlink>: Likewise.
(target_filesystem_is_local): Update comment.
(target_fileio_open): New argument inf. Update comment.
(target_fileio_unlink): Likewise.
(target_fileio_readlink): Likewise.
(target_fileio_read_alloc): Likewise.
(target_fileio_read_stralloc): Likewise.
* target.c (target_fileio_open): New argument inf.
Pass inf to implementation. Update debug printing.
(target_fileio_unlink): Likewise.
(target_fileio_readlink): Likewise.
(target_fileio_read_alloc_1): New argument inf. Pass inf
to target_fileio_open.
(target_fileio_read_alloc): New argument inf. Pass inf to
target_fileio_read_alloc_1.
(target_fileio_read_stralloc): Likewise.
* gdb_bfd.c (inferior.h): New include.
(gdb_bfd_iovec_fileio_open): Replace unused "open_closure"
argument with new argument "inferior". Pass inferior to
target_fileio_open.
(gdb_bfd_open): Supply inferior argument to
gdb_bfd_iovec_fileio_open.
* linux-tdep.c (linux_info_proc): Supply inf argument to
relevant target_fileio calls.
(linux_find_memory_regions_full): Likewise.
(linux_fill_prpsinfo): Likewise.
* remote.c (remote_filesystem_is_local): Supply inf
argument to remote_hostio_open.
(remote_file_put): Likewise.
(remote_file_get): Likewise.
(remote_file_delete): Supply inf argument to
remote_hostio_unlink.
The return value of target_can_use_hardware_watchpoint isn't well
documented, so this patch is to update the comments to reflect the
fact. This patch also removes a trailing ";" which is picked up
from Pedro's patch https://sourceware.org/ml/gdb-patches/2015-04/msg00527.html
gdb:
2015-04-17 Yao Qi <yao.qi@linaro.org>
Pedro Alves <palves@redhat.com>
* target.h (target_can_use_hardware_watchpoint): Update comments.
Remove trailing ";".
This commit adds a new packet "qXfer:exec-file:read" to the remote
protocol that can be used to obtain the pathname of the file that
was executed to create a process on the remote system. Support for
this packet is added to GDB and remote_ops.to_pid_to_exec_file is
implemented using it.
gdb/ChangeLog:
* target.h (TARGET_OBJECT_EXEC_FILE): New enum value.
* remote.c (PACKET_qXfer_exec_file): Likewise.
(remote_protocol_features): Register the
"qXfer:exec-file:read" feature.
(remote_xfer_partial): Handle TARGET_OBJECT_EXEC_FILE.
(remote_pid_to_exec_file): New function.
(init_remote_ops): Initialize to_pid_to_exec_file.
(_initialize_remote): Register new "set/show remote
pid-to-exec-file-packet" command.
* NEWS: Announce new qXfer:exec-file:read packet.
gdb/doc/ChangeLog:
* gdb.texinfo (Remote Configuration): Document the "set/show
remote pid-to-exec-file-packet" command.
(General Query Packets): Document the qXfer:exec-file:read
qSupported features. Document the qXfer:exec-file:read packet.
This commit introduces a new target method target_filesystem_is_local
which can be used to determine whether or not the filesystem accessed
by the target_fileio_* methods is the local filesystem.
gdb/ChangeLog:
* target.h (struct target_ops) <to_filesystem_is_local>:
New field.
(target_filesystem_is_local): New macro.
* target-delegates.c: Regenerate.
* remote.c (remote_filesystem_is_local): New function.
(init_remote_ops): Initialize to_filesystem_is_local.
This commit introduces a new target method target_fileio_fstat
which can be used to retrieve information about files opened with
target_fileio_open.
gdb/ChangeLog:
* target.h (struct target_ops) <to_fileio_fstat>: New field.
(target_fileio_fstat): New declaration.
* target.c (target_fileio_fstat): New function.
* inf-child.c (inf_child_fileio_fstat): Likewise.
(inf_child_target): Initialize to_fileio_fstat.
* remote.c (init_remote_ops): Likewise.
All callers of target_async pass it the same callback
(inferior_event_handler). Since both common code and target backends
need to be able to put the target in and out of target async mode at
any given time, there's really no way that a different callback could
be passed. This commit simplifies things, and removes the indirection
altogether. Bonus: with this, gdb's target_async method ends up with
the same signature as gdbserver's.
Tested on x86_64 Fedora 20, native and gdbserver.
gdb/ChangeLog:
2015-03-25 Pedro Alves <palves@redhat.com>
* target.h <to_async>: Replace 'callback' and 'context' parameters
with boolean 'enable' parameter.
(target_async): Replace CALLBACK and CONTEXT parameters with
boolean ENABLE parameter.
* inf-loop.c (inferior_event_handler): Adjust.
* linux-nat.c (linux_nat_attach, linux_nat_resume)
(linux_nat_resume): Adjust.
(async_client_callback, async_client_context): Delete.
(handle_target_event): Call inferior_event_handler directly.
(linux_nat_async): Replace 'callback' and 'context' parameters
with boolean 'enable' parameter. Adjust. Remove references to
async_client_callback and async_client_context.
(linux_nat_close): Adjust.
* record-btrace.c (record_btrace_async): Replace 'callback' and
'context' parameters with boolean 'enable' parameter. Adjust.
(record_btrace_resume): Adjust.
* record-full.c (record_full_async): Replace 'callback' and
'context' parameters with boolean 'enable' parameter. Adjust.
(record_full_resume, record_full_core_resume): Adjust.
* remote.c (struct remote_state) <async_client_callback,
async_client_context>: Delete fields.
(remote_start_remote, extended_remote_attach_1, remote_resume)
(extended_remote_create_inferior): Adjust.
(remote_async_serial_handler): Call inferior_event_handler
directly.
(remote_async): Replace 'callback' and 'context' parameters with
boolean 'enable' parameter. Adjust.
* top.c (gdb_readline_wrapper_cleanup, gdb_readline_wrapper):
Adjust.
* target-delegates.c: Regenerate.
The "set serial parity" command allows the user to control which
parity to use when communicating over a serial connection, rather
than having the parity hardcoded to none.
gdb/ChangeLog:
* NEWS: Mention set/show serial parity command.
* monitor.c (monitor_open): Call serial_setparity.
* remote.c (remote_open_1): Likewise.
* ser-base.c (ser_base_serparity): New function.
* ser-base.h (ser_base_setparity): Add declaration.
* ser-go32.c (dos_ops): Set "setparity" field.
* ser-mingw.c (ser_windows_raw): Do not set state.fParity and
state.Parity.
(ser_windows_setparity): New function.
(hardwire_ops): Add ser_windows_setparity.
(tty_ops): Add NULL for setparity field.
(pipe_ops): Add ser_base_setparity.
(tcp_ops): Likewise.
* ser-pipe.c (pipe_ops): Likewise.
* ser-tcp.c (tcp_ops): Likewise.
* ser-unix.c (hardwire_setparity): Add declaration.
(hardwire_raw): Don't reset PARENB flag.
(hardwire_setparity): New function.
(hardwire_ops): Add hardwire_setparity.
* serial.c (serial_setparity): New function.
(serial_parity): New global.
(parity_none, parity_odd, parity_even, parity_enums, parity):
New static globals.
(set_parity): New function.
(_initialize_serial): Add set/show serial parity commands.
* serial.h (GDBPARITY_NONE): Define.
(GDBPARITY_ODD): Define.
(GDBPARITY_EVEN): Define.
(serial_setparity) Add declaration.
(struct serial_ops): Add setparity field.
* target.h (serial_parity): Add declaration.
gdb/doc/ChangeLog:
* gdb.texinfo (Remote configuration): Document "set/show
serial parity" command.
record-btrace was the only target making use of this, and it no longer
uses it.
gdb/ChangeLog:
2015-03-04 Pedro Alves <palves@redhat.com>
* target.h (struct target_ops) <to_decr_pc_after_break>: Delete.
(target_decr_pc_after_break): Delete declaration.
* target.c (default_target_decr_pc_after_break)
(target_decr_pc_after_break): Delete.
* linux-nat.c (check_stopped_by_breakpoint, linux_nat_wait_1): Use
gdbarch_decr_pc_after_break instead of target_decr_pc_after_break.
* linux-thread-db.c (check_event): Likewise.
* infrun.c (adjust_pc_after_break): Likewise.
* darwin-nat.c (cancel_breakpoint): Likewise.
* aix-thread.c (aix_thread_wait): Likewise.
* target-delegates.c: Regenerate.
The moribund locations heuristics are problematic. This patch teaches
GDB about targets that can reliably tell whether a trap was caused by
a software or hardware breakpoint, and thus don't need moribund
locations, thus bypassing all the problems that mechanism has.
The non-stop-fair-events.exp test is frequently failing currently.
E.g., see https://sourceware.org/ml/gdb-testers/2015-q1/msg03148.html.
The root cause is a fundamental problem with moribund locations. For
example, the stepped_breakpoint logic added by af48d08f breaks in this
case (which is what happens with that test):
- Step thread A, no breakpoint is set at PC.
- The kernel doesn't schedule thread A yet.
- Insert breakpoint at A's PC, for some reason (e.g., a step-resume
breakpoint for thread B).
- Kernel finally schedules thread A.
- thread A's stepped_breakpoint flag is not set, even though it now
stepped a breakpoint instruction.
- adjust_pc_after_break gets the PC wrong, because PC == PREV_PC, but
stepped_breakpoint is not set.
We needed the stepped_breakpoint logic to workaround moribund
locations, because otherwise adjust_pc_after_break could apply an
adjustment when it shouldn't just because there _used_ to be a
breakpoint at PC (a moribund breakpoint location). For example, on
x86, that's wrong if the thread really hasn't executed an int3, but
instead executed some other 1-byte long instruction. Getting the PC
adjustment wrong of course leads to the inferior executing the wrong
instruction.
Other problems with moribund locations are:
- if a true SIGTRAP happens to be raised when the program is
executing the PC that used to have a breakpoint, GDB will assume
that is a trap for a breakpoint that has recently been removed, and
thus we miss reporting the random signal to the user.
- to minimize that, we get rid of moribund location after a while.
That while is defined as just a certain number of events being
processed. That number of events sometimes passes by before a
delayed breakpoint is processed, and GDB confuses the trap for a
random signal, thus reporting the random trap. Once the user
resumes the thread, the program crashes because the PC was not
adjusted...
The fix for all this is to bite the bullet and get rid of heuristics
and instead rely on the target knowing accurately what caused the
SIGTRAP. The target/kernel/stub is in the best position to know what
that, because it can e.g. consult priviledged CPU flags GDB has no
access to, or by knowing which exception vector entry was called when
the instruction trapped, etc. Most debug APIs I've seen to date
report breakpoint hits as a distinct event in some fashion. For
example, on the Linux kernel, whether a breakpoint was executed is
exposed to userspace in the si_code field of the SIGTRAP's siginfo.
On Windows, the debug API reports a EXCEPTION_BREAKPOINT exception
code.
We needed to keep around deleted breakpoints in an on-the-side list
(the moribund locations) for two main reasons:
- Know that a SIGTRAP actually is a delayed event for a hit of a
breakpoint that was removed before the event was processed, and
thus should not be reported as a random signal.
- So we still do the decr_pc_after_break adjustment in that case, so
that the thread is resumed at the correct address.
In the new model, if GDB processes an event the target tells is a
breakpoint trap, and GDB doesn't find the corresponding breakpoint in
its breakpoint tables, it means that event is a delayed event for a
breakpoint that has since been removed, and thus the event should be
ignored.
For the decr_pc_after_after issue, it ends up being much simpler that
on targets that can reliably tell whether a breakpoint trapped, for
the breakpoint trap to present the PC already adjusted. Proper
multi-threading support already implies that targets needs to be doing
decr_pc_after_break adjustment themselves, otherwise for example, in
all-stop if two threads hit a breakpoint simultaneously, and the user
does "info threads", he'll see the non-event thread that hit the
breakpoint stopped at the wrong PC.
This way (target adjusts) also ends up eliminating the need for some
awkward re-incrementing of the PC in the record-full and Linux targets
that we do today, and the need for the target_decr_pc_after_break
hook.
If the target always adjusts, then there's a case where GDB needs to
re-increment the PC. Say, on x86, an "int3" instruction that was
explicitly written in the program traps. In this case, GDB should
report a random SIGTRAP signal to the user, with the PC pointing at
the instruction past the int3, just like if GDB was not debugging the
program. The user may well decide to pass the SIGTRAP to the program
because the program being debugged has a SIGTRAP handler that handles
its own breakpoints, and expects the PC to be unadjusted.
Tested on x86-64 Fedora 20.
gdb/ChangeLog:
2015-03-04 Pedro Alves <palves@redhat.com>
* breakpoint.c (need_moribund_for_location_type): New function.
(bpstat_stop_status): Don't skipping checking moribund locations
of breakpoint types which the target tell caused a stop.
(program_breakpoint_here_p): New function, factored out from ...
(bp_loc_is_permanent): ... this.
(update_global_location_list): Don't create a moribund location if
the target supports reporting stops of the type of the removed
breakpoint.
* breakpoint.h (program_breakpoint_here_p): New declaration.
* infrun.c (adjust_pc_after_break): Return early if the target has
already adjusted the PC. Add comments.
(handle_signal_stop): If nothing explains a signal, and the target
tells us the stop was caused by a software breakpoint, check if
there's a breakpoint instruction in the memory. If so, adjust the
PC before presenting the stop to the user. Otherwise, ignore the
trap. If nothing explains a signal, and the target tells us the
stop was caused by a hardware breakpoint, ignore the trap.
* target.h (struct target_ops) <to_stopped_by_sw_breakpoint,
to_supports_stopped_by_sw_breakpoint, to_stopped_by_hw_breakpoint,
to_supports_stopped_by_hw_breakpoint>: New fields.
(target_stopped_by_sw_breakpoint)
(target_supports_stopped_by_sw_breakpoint)
(target_stopped_by_hw_breakpoint)
(target_supports_stopped_by_hw_breakpoint): Define.
* target-delegates.c: Regenerate.
find_default_create_inferior and find_default_attach were removed in b3ccfe11.
gdb/ChangeLog:
* target.h (find_default_create_inferior): Remove declaration.
(find_default_attach): Likewise.
Fixes:
src/gdb/target.h:753:10: error: use of enum ‘exec_direction_kind’ without previous declaration
in C++ mode. We can't forward declare enums.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* target.h: Include "infrun.h".
The definition was removed a year ago, but the declaration managed to
stay behind.
gdb/ChangeLog
2015-02-20 Pedro Alves <palves@redhat.com>
* target.h (forward_target_decr_pc_after_break): Delete
declaration.
Pedro Alves