binutils-gdb/gdb/gcore-elf.h

48 lines
1.8 KiB
C
Raw Normal View History

gdb: unify parts of the Linux and FreeBSD core dumping code While reviewing the Linux and FreeBSD core dumping code within GDB for another patch series, I noticed that the code that collects the registers for each thread and writes these into ELF note format is basically identical between Linux and FreeBSD. This commit merges this code and moves it into a new file gcore-elf.c. The function find_signalled_thread is moved from linux-tdep.c to gcore.c despite not being shared. A later commit will make use of this function. I did merge, and then revert a previous version of this patch (commit 82a1fd3a4935 for the original patch and 03642b7189bc for the revert). The problem with the original patch is that it introduced a unconditional dependency between GDB and some ELF specific functions in the BFD library, e.g. elfcore_write_prstatus and elfcore_write_register_note. It was pointed out in this mailing list post: https://sourceware.org/pipermail/gdb-patches/2021-February/175750.html that this change was breaking any build of GDB for non-ELF targets. To confirm this breakage, and to test this new version of GDB I configured and built for the target x86_64-apple-darwin20.3.0. Where the previous version of this patch placed all of the common code into gcore.c, which is included in all builds of GDB, this new patch only places non-ELF specific generic code (i.e. find_signalled_thread) into gcore.c, the ELF specific code is put into the new gcore-elf.c file, which is only included in GDB if BFD has ELF support. The contents of gcore-elf.c are referenced unconditionally from linux-tdep.c and fbsd-tdep.c, this is fine, we previously always assumed that these two targets required ELF support, and we continue to make that assumption after this patch; nothing has changed there. With my previous version of this patch the darwin target mentioned above failed to build, but with the new version, the target builds fine. There are a couple of minor changes to the FreeBSD target after this commit, but I believe that these are changes for the better: (1) For FreeBSD we always used to record the thread-id in the core file by using ptid_t.lwp (). In contrast the Linux code did this: /* For remote targets the LWP may not be available, so use the TID. */ long lwp = ptid.lwp (); if (lwp == 0) lwp = ptid.tid (); Both target now do this: /* The LWP is often not available for bare metal target, in which case use the tid instead. */ if (ptid.lwp_p ()) lwp = ptid.lwp (); else lwp = ptid.tid (); Which is equivalent for Linux, but is a change for FreeBSD. I think that all this means is that in some cases where GDB might have previously recorded a thread-id of 0 for each thread, we might now get something more useful. (2) When collecting the registers for Linux we collected into a zero initialised buffer. By contrast on FreeBSD the buffer is left uninitialised. In the new code the buffer is always zero initialised. I suspect once the registers are copied into the buffer there's probably no gaps left so this makes no difference, but if it does then using zeros rather than random bits of GDB's memory is probably a good thing. Otherwise, there should be no other user visible changes after this commit. Tested this on x86-64/GNU-Linux and x86-64/FreeBSD-12.2 with no regressions. gdb/ChangeLog: * Makefile.in (SFILES): Add gcore-elf.c. (HFILES_NO_SRCDIR): Add gcore-elf.h * configure: Regenerate. * configure.ac: Add gcore-elf.o to CONFIG_OBS if we have ELF support. * fbsd-tdep.c: Add 'gcore-elf.h' include. (struct fbsd_collect_regset_section_cb_data): Delete. (fbsd_collect_regset_section_cb): Delete. (fbsd_collect_thread_registers): Delete. (struct fbsd_corefile_thread_data): Delete. (fbsd_corefile_thread): Delete. (fbsd_make_corefile_notes): Call gcore_elf_build_thread_register_notes instead of the now deleted FreeBSD code. * gcore-elf.c: New file, the content was moved here from linux-tdep.c, functions were renamed and given minor cleanup. * gcore-elf.h: New file. * gcore.c (gcore_find_signalled_thread): Moved here from linux-tdep.c and given a new name. Minor cleanups. * gcore.h (gcore_find_signalled_thread): Declare. * linux-tdep.c: Add 'gcore.h' and 'gcore-elf.h' includes. (struct linux_collect_regset_section_cb_data): Delete. (linux_collect_regset_section_cb): Delete. (linux_collect_thread_registers): Delete. (linux_corefile_thread): Call gcore_elf_build_thread_register_notes. (find_signalled_thread): Delete. (linux_make_corefile_notes): Call gcore_find_signalled_thread.
2021-01-19 00:00:38 +08:00
/* Copyright (C) 2021 Free Software Foundation, Inc.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
/* This file contains generic functions for writing ELF based core files. */
#if !defined (GCORE_ELF_H)
#define GCORE_ELF_H 1
#include "gdb_bfd.h"
#include "gdbsupport/gdb_signals.h"
#include "gcore.h"
struct gdbarch;
struct thread_info;
/* Add content to *NOTE_DATA (and update *NOTE_SIZE) to describe the
registers of thread INFO. Report the thread as having stopped with
STOP_SIGNAL. The core file is being written to OBFD, and GDBARCH is the
architecture for which the core file is being generated. */
extern void gcore_elf_build_thread_register_notes
(struct gdbarch *gdbarch, struct thread_info *info, gdb_signal stop_signal,
bfd *obfd, gdb::unique_xmalloc_ptr<char> *note_data, int *note_size);
gdb: write target description into core file When a core file is created from within GDB add the target description into a note within the core file. When loading a core file, if the target description note is present then load the target description from the core file. The benefit of this is that we can be sure that, when analysing the core file within GDB, that we are using the exact same target description as was in use at the time the core file was created. GDB already supports a mechanism for figuring out the target description from a given corefile; gdbarch_core_read_description. This new mechanism (GDB adding the target description) is not going to replace the old mechanism. Core files generated outside of GDB will not include a target description, and so GDB still needs to be able to figure out a target description for these files. My primary motivation for adding this feature is that, in a future commit, I will be adding support for bare metal core dumps on some targets. For RISC-V specifically, I want to be able to dump all the available control status registers. As different targets will present different sets of register in their target description, including registers that are possibly not otherwise known to GDB I wanted a way to capture these registers in the core dump. I therefore need a mechanism to write out an arbitrary set of registers, and to then derive a target description from this arbitrary set when later loading the core file. The obvious approach (I think) is to just reuse the target description. Once I'd decided to add support for writing out the target description I could either choose to make this RISC-V only, or make it generic. I figure that having the target description in the core file doesn't hurt, and _might_ be helpful. So that's how I got here, general support for including the target description in GDB generated core files. In previous versions of this patch I added the target description from generic code (in gcore.c). However, doing this creates a dependency between GDB's common code and bfd ELF support. As ELF support in gdb is optional (for example the target x86_64-apple-darwin20.3.0 does not include ELF support) then having gcore.c require ELF support would break the GDB build in some cases. Instead, in this version of the patch, writing the target description note is done from each specific targets make notes function. Each of these now calls a common function in gcore-elf.c (which is only linked in when bfd has ELF support). And so only targets that are ELF based will call the new function and we can therefore avoid an unconditional dependency on ELF support. gdb/ChangeLog: * corelow.c: Add 'xml-tdesc.h' include. (core_target::read_description): Load the target description from the core file when possible. * fbsd-tdep.c (fbsd_make_corefile_notes): Add target description note. * gcore-elf.c: Add 'gdbsupport/tdesc.h' include. (gcore_elf_make_tdesc_note): New function. * gcore-elf.h (gcore_elf_make_tdesc_note): Declare. * linux-tdep.c (linux_make_corefile_notes): Add target description note.
2020-11-27 23:41:52 +08:00
/* Add content to *NOTE_DATA (and update *NOTE_SIZE) to include a note
containing the current targtet's target description. The core file is
being written to OBFD. If something goes wrong then *NOTE_DATA can be
set to nullptr. */
extern void gcore_elf_make_tdesc_note
(bfd *obfd, gdb::unique_xmalloc_ptr<char> *note_data, int *note_size);
gdb: unify parts of the Linux and FreeBSD core dumping code While reviewing the Linux and FreeBSD core dumping code within GDB for another patch series, I noticed that the code that collects the registers for each thread and writes these into ELF note format is basically identical between Linux and FreeBSD. This commit merges this code and moves it into a new file gcore-elf.c. The function find_signalled_thread is moved from linux-tdep.c to gcore.c despite not being shared. A later commit will make use of this function. I did merge, and then revert a previous version of this patch (commit 82a1fd3a4935 for the original patch and 03642b7189bc for the revert). The problem with the original patch is that it introduced a unconditional dependency between GDB and some ELF specific functions in the BFD library, e.g. elfcore_write_prstatus and elfcore_write_register_note. It was pointed out in this mailing list post: https://sourceware.org/pipermail/gdb-patches/2021-February/175750.html that this change was breaking any build of GDB for non-ELF targets. To confirm this breakage, and to test this new version of GDB I configured and built for the target x86_64-apple-darwin20.3.0. Where the previous version of this patch placed all of the common code into gcore.c, which is included in all builds of GDB, this new patch only places non-ELF specific generic code (i.e. find_signalled_thread) into gcore.c, the ELF specific code is put into the new gcore-elf.c file, which is only included in GDB if BFD has ELF support. The contents of gcore-elf.c are referenced unconditionally from linux-tdep.c and fbsd-tdep.c, this is fine, we previously always assumed that these two targets required ELF support, and we continue to make that assumption after this patch; nothing has changed there. With my previous version of this patch the darwin target mentioned above failed to build, but with the new version, the target builds fine. There are a couple of minor changes to the FreeBSD target after this commit, but I believe that these are changes for the better: (1) For FreeBSD we always used to record the thread-id in the core file by using ptid_t.lwp (). In contrast the Linux code did this: /* For remote targets the LWP may not be available, so use the TID. */ long lwp = ptid.lwp (); if (lwp == 0) lwp = ptid.tid (); Both target now do this: /* The LWP is often not available for bare metal target, in which case use the tid instead. */ if (ptid.lwp_p ()) lwp = ptid.lwp (); else lwp = ptid.tid (); Which is equivalent for Linux, but is a change for FreeBSD. I think that all this means is that in some cases where GDB might have previously recorded a thread-id of 0 for each thread, we might now get something more useful. (2) When collecting the registers for Linux we collected into a zero initialised buffer. By contrast on FreeBSD the buffer is left uninitialised. In the new code the buffer is always zero initialised. I suspect once the registers are copied into the buffer there's probably no gaps left so this makes no difference, but if it does then using zeros rather than random bits of GDB's memory is probably a good thing. Otherwise, there should be no other user visible changes after this commit. Tested this on x86-64/GNU-Linux and x86-64/FreeBSD-12.2 with no regressions. gdb/ChangeLog: * Makefile.in (SFILES): Add gcore-elf.c. (HFILES_NO_SRCDIR): Add gcore-elf.h * configure: Regenerate. * configure.ac: Add gcore-elf.o to CONFIG_OBS if we have ELF support. * fbsd-tdep.c: Add 'gcore-elf.h' include. (struct fbsd_collect_regset_section_cb_data): Delete. (fbsd_collect_regset_section_cb): Delete. (fbsd_collect_thread_registers): Delete. (struct fbsd_corefile_thread_data): Delete. (fbsd_corefile_thread): Delete. (fbsd_make_corefile_notes): Call gcore_elf_build_thread_register_notes instead of the now deleted FreeBSD code. * gcore-elf.c: New file, the content was moved here from linux-tdep.c, functions were renamed and given minor cleanup. * gcore-elf.h: New file. * gcore.c (gcore_find_signalled_thread): Moved here from linux-tdep.c and given a new name. Minor cleanups. * gcore.h (gcore_find_signalled_thread): Declare. * linux-tdep.c: Add 'gcore.h' and 'gcore-elf.h' includes. (struct linux_collect_regset_section_cb_data): Delete. (linux_collect_regset_section_cb): Delete. (linux_collect_thread_registers): Delete. (linux_corefile_thread): Call gcore_elf_build_thread_register_notes. (find_signalled_thread): Delete. (linux_make_corefile_notes): Call gcore_find_signalled_thread.
2021-01-19 00:00:38 +08:00
#endif /* GCORE_ELF_H */