From patchwork Sun May 6 15:38:48 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Pedro Alves X-Patchwork-Id: 27141 Received: (qmail 25188 invoked by alias); 6 May 2018 15:38:56 -0000 Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org Delivered-To: mailing list gdb-patches@sourceware.org Received: (qmail 25163 invoked by uid 89); 6 May 2018 15:38:54 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-24.5 required=5.0 tests=AWL, BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, KAM_LAZY_DOMAIN_SECURITY, SPF_HELO_PASS autolearn=ham version=3.3.2 spammy=longest, architecture X-HELO: mx1.redhat.com Received: from mx3-rdu2.redhat.com (HELO mx1.redhat.com) (66.187.233.73) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Sun, 06 May 2018 15:38:51 +0000 Received: from smtp.corp.redhat.com (int-mx03.intmail.prod.int.rdu2.redhat.com [10.11.54.3]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 584AFEBFF7; Sun, 6 May 2018 15:38:50 +0000 (UTC) Received: from [127.0.0.1] (ovpn04.gateway.prod.ext.ams2.redhat.com [10.39.146.4]) by smtp.corp.redhat.com (Postfix) with ESMTP id 826791121286; Sun, 6 May 2018 15:38:49 +0000 (UTC) Subject: Re: [PATCH 3/3] Heap-allocate core_target instances To: Tom Tromey References: <20180503162234.15371-1-palves@redhat.com> <20180503162234.15371-4-palves@redhat.com> <87in832x9t.fsf@tromey.com> Cc: gdb-patches@sourceware.org From: Pedro Alves Message-ID: <889a13d0-e1c1-02a8-fd99-cdf872e214e7@redhat.com> Date: Sun, 6 May 2018 16:38:48 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.6.0 MIME-Version: 1.0 In-Reply-To: <87in832x9t.fsf@tromey.com> Hi Tom, Thanks for the review. On 05/04/2018 05:36 PM, Tom Tromey wrote: >>>>>> "Pedro" == Pedro Alves writes: > > Pedro> This gets rid of the core_ops global, and replaces it with > Pedro> heap-allocated core_target instances. In practice, there will only be > Pedro> one such instance, though that will change further ahead as more > Pedro> pieces of multi-target support are merged. > > Pedro> + /* FIXME: kettenis/20031023: Eventually this variable should > Pedro> + disappear. */ > Pedro> + struct gdbarch *m_core_gdbarch = NULL; > > This comment is very gdb. The joys of a codebase with a long history. :-) > > > It was a bit hard to read the patch Note the patch is harder to read if you read it locally using the default diff algorithm. The version posted to the list was generated with git's --patience diff algorithm, which understands this patch better than the default algorithm. but I think I excerpted this correctly: > > Pedro> +void > Pedro> +core_target::close () > [...] > Pedro> + delete this; > > I realize this is what was planned from your c++-ification series, but > today I was wondering if this code path: > > void > core_target::detach (inferior *inf, int from_tty) > { > unpush_target (this); > > => > > int > unpush_target (struct target_ops *t) > ... > target_close (t); > > invokes undefined behavior. Hmm, not seeing a problem. T is not used by unpush_target after the target_close call, and nor is "this" used after the unpush_target call in core_target::detach. > > "delete this" seems to be ok if you are careful not to use "this" > afterward, Yeah, it's a common idiom in a decref()/release() method of reference counted types, for example (when the reference count reaches 0). As long as you don't use 'this' afterwards, it's OK. > but is this particular use ok? I don't know the rule here > but I thought I'd mention it just in case. It seems OK to me. The core_target::detach case might be less obvious? I'm adding some comments. > > Pedro> +/* Deleter for std::unique_ptr. Closes the target if an exception is > Pedro> + thrown. */ > Pedro> +struct target_ops_closer > Pedro> +{ > Pedro> + void operator() (target_ops *target) > Pedro> + { > Pedro> + target->close (); > Pedro> + } > Pedro> +}; > > This seems like something to put in target.h. Done. Let me know what you think of this version. From 18891f9cf0d403f93448cdded9d7c1fbe33186e7 Mon Sep 17 00:00:00 2001 From: Pedro Alves Date: Sun, 6 May 2018 16:24:21 +0100 Subject: [PATCH 3/3] Heap-allocate core_target instances This gets rid of the core_ops global, and replaces it with heap-allocated core_target instances. In practice, there will only be one such instance, though that will change further ahead as more pieces of multi-target support are merged. Notice that this replaces one heap-allocated object for another, the number of allocations is the same. Specifically, currently we heap-allocate the 'core_data' object, which holds the core's section table. With this patch, that object is made a field of the core_target class, and no longer allocated separately. Note that this bit: - /* Looks semi-reasonable. Toss the old core file and work on the - new. */ - - unpush_target (&core_ops); does not need a replacement, because by the time we get here, the target_preopen call at the top of core_target_open has already unpushed any previous target. gdb/ChangeLog: yyyy-mm-dd Pedro Alves * corelow.c (core_target) : No longer inline. Initialize m_core_gdbarch, m_core_vec and build the section table here. <~core_target>: New. : New methods. : New fields, factored out from ... : ... these deleted globals. (core_ops): Delete. (sniff_core_bfd): Add gdbarch parameter. (core_close): Delete, merged into ... (core_target::close): ... here. Delete self. (core_close_cleanup): Delete. (core_target_open): Allocate a core_target on the heap. Use a unique_ptr instead of a cleanup. Bits moved into the core_target ctor. Adjust to use core_target methods instead of globals. (get_core_register_section): Rename to ... (core_target::get_core_register_section): ... this and adjust. (struct get_core_registers_cb_data): New. (get_core_registers_cb): Use it. Use bool. (core_target::fetch_registers, core_target::files_info) (core_target::xfer_partial, core_target::read_description) (core_target::pid_to, core_target::thread_name): Adjust to reference class fields instead of globals. * target.h (struct target_ops_deleter, target_ops_up): New. --- gdb/corelow.c | 265 +++++++++++++++++++++++++++++++--------------------------- gdb/target.h | 14 ++++ 2 files changed, 157 insertions(+), 122 deletions(-) diff --git a/gdb/corelow.c b/gdb/corelow.c index 3b71a1b856..439fe1338a 100644 --- a/gdb/corelow.c +++ b/gdb/corelow.c @@ -50,6 +50,9 @@ #define O_LARGEFILE 0 #endif +static core_fns *sniff_core_bfd (gdbarch *core_gdbarch, + bfd *abfd); + /* The core file target. */ static const target_info core_target_info = { @@ -61,8 +64,8 @@ static const target_info core_target_info = { class core_target final : public target_ops { public: - core_target () - { to_stratum = process_stratum; } + core_target (); + ~core_target () override; const target_info &info () const override { return core_target_info; } @@ -90,42 +93,75 @@ public: bool has_stack () override; bool has_registers () override; bool info_proc (const char *, enum info_proc_what) override; + + /* A few helpers. */ + + /* Getter, see variable definition. */ + struct gdbarch *core_gdbarch () + { + return m_core_gdbarch; + } + + /* See definition. */ + void get_core_register_section (struct regcache *regcache, + const struct regset *regset, + const char *name, + int min_size, + int which, + const char *human_name, + bool required); + +private: /* per-core data */ + + /* The core's section table. Note that these target sections are + *not* mapped in the current address spaces' set of target + sections --- those should come only from pure executable or + shared library bfds. The core bfd sections are an implementation + detail of the core target, just like ptrace is for unix child + targets. */ + target_section_table m_core_section_table {}; + + /* The core_fns for a core file handler that is prepared to read the + core file currently open on core_bfd. */ + core_fns *m_core_vec = NULL; + + /* FIXME: kettenis/20031023: Eventually this field should + disappear. */ + struct gdbarch *m_core_gdbarch = NULL; }; +core_target::core_target () +{ + to_stratum = process_stratum; + + m_core_gdbarch = gdbarch_from_bfd (core_bfd); + + /* Find a suitable core file handler to munch on core_bfd */ + m_core_vec = sniff_core_bfd (m_core_gdbarch, core_bfd); + + /* Find the data section */ + if (build_section_table (core_bfd, + &m_core_section_table.sections, + &m_core_section_table.sections_end)) + error (_("\"%s\": Can't find sections: %s"), + bfd_get_filename (core_bfd), bfd_errmsg (bfd_get_error ())); +} + +core_target::~core_target () +{ + xfree (m_core_section_table.sections); +} + /* List of all available core_fns. On gdb startup, each core file register reader calls deprecated_add_core_fns() to register information on each core format it is prepared to read. */ static struct core_fns *core_file_fns = NULL; -/* The core_fns for a core file handler that is prepared to read the - core file currently open on core_bfd. */ - -static struct core_fns *core_vec = NULL; - -/* FIXME: kettenis/20031023: Eventually this variable should - disappear. */ - -static struct gdbarch *core_gdbarch = NULL; - -/* Per-core data. Currently, only the section table. Note that these - target sections are *not* mapped in the current address spaces' set - of target sections --- those should come only from pure executable - or shared library bfds. The core bfd sections are an - implementation detail of the core target, just like ptrace is for - unix child targets. */ -static struct target_section_table *core_data; - -static struct core_fns *sniff_core_bfd (bfd *); - static int gdb_check_format (bfd *); -static void core_close_cleanup (void *ignore); - static void add_to_thread_list (bfd *, asection *, void *); -static core_target core_ops; - /* An arbitrary identifier for the core inferior. */ #define CORELOW_PID 1 @@ -159,7 +195,7 @@ default_core_sniffer (struct core_fns *our_fns, bfd *abfd) selected. */ static struct core_fns * -sniff_core_bfd (bfd *abfd) +sniff_core_bfd (struct gdbarch *core_gdbarch, bfd *abfd) { struct core_fns *cf; struct core_fns *yummy = NULL; @@ -217,11 +253,10 @@ gdb_check_format (bfd *abfd) return (0); } -/* Discard all vestiges of any previous core file and mark data and - stack spaces as empty. */ +/* Close the core target. */ -static void -core_close () +void +core_target::close () { if (core_bfd) { @@ -235,30 +270,13 @@ core_close () comments in clear_solib in solib.c. */ clear_solib (); - if (core_data) - { - xfree (core_data->sections); - xfree (core_data); - core_data = NULL; - } - gdb_bfd_unref (core_bfd); core_bfd = NULL; } - core_vec = NULL; - core_gdbarch = NULL; -} -static void -core_close_cleanup (void *ignore) -{ - core_close (); -} - -void -core_target::close () -{ - core_close (); + /* Core targets are heap-allocated (see core_target_open), so here + we delete ourselves. */ + delete this; } /* Look for sections whose names start with `.reg/' so that we can @@ -388,29 +406,15 @@ core_target_open (const char *arg, int from_tty) filename.get (), bfd_errmsg (bfd_get_error ())); } - /* Looks semi-reasonable. Toss the old core file and work on the - new. */ - - unpush_target (&core_ops); core_bfd = temp_bfd.release (); - old_chain = make_cleanup (core_close_cleanup, 0 /*ignore*/); - core_gdbarch = gdbarch_from_bfd (core_bfd); + core_target *target = new core_target (); - /* Find a suitable core file handler to munch on core_bfd */ - core_vec = sniff_core_bfd (core_bfd); + /* Own the target until it is successfully pushed. */ + target_ops_up target_holder (target); validate_files (); - core_data = XCNEW (struct target_section_table); - - /* Find the data section */ - if (build_section_table (core_bfd, - &core_data->sections, - &core_data->sections_end)) - error (_("\"%s\": Can't find sections: %s"), - bfd_get_filename (core_bfd), bfd_errmsg (bfd_get_error ())); - /* If we have no exec file, try to set the architecture from the core file. We don't do this unconditionally since an exec file typically contains more information that helps us determine the @@ -418,8 +422,8 @@ core_target_open (const char *arg, int from_tty) if (!exec_bfd) set_gdbarch_from_file (core_bfd); - push_target (&core_ops); - discard_cleanups (old_chain); + push_target (target); + target_holder.release (); /* Do this before acknowledging the inferior, so if post_create_inferior throws (can happen easilly if you're loading @@ -463,7 +467,7 @@ core_target_open (const char *arg, int from_tty) switch_to_thread (thread->ptid); } - post_create_inferior (&core_ops, from_tty); + post_create_inferior (target, from_tty); /* Now go through the target stack looking for threads since there may be a thread_stratum target loaded on top of target core by @@ -490,6 +494,8 @@ core_target_open (const char *arg, int from_tty) siggy = bfd_core_file_failing_signal (core_bfd); if (siggy > 0) { + gdbarch *core_gdbarch = target->core_gdbarch (); + /* If we don't have a CORE_GDBARCH to work with, assume a native core (map gdb_signal from host signals). If we do have CORE_GDBARCH to work with, but no gdb_signal_from_target @@ -538,14 +544,18 @@ core_target_open (const char *arg, int from_tty) void core_target::detach (inferior *inf, int from_tty) { + /* Note that 'this' is dangling after this call. unpush_target + closes the target, and our close implementation deletes + 'this'. */ unpush_target (this); + reinit_frame_cache (); maybe_say_no_core_file_now (from_tty); } /* Try to retrieve registers from a section in core_bfd, and supply - them to core_vec->core_read_registers, as the register set numbered - WHICH. + them to m_core_vec->core_read_registers, as the register set + numbered WHICH. If ptid's lwp member is zero, do the single-threaded thing: look for a section named NAME. If ptid's lwp @@ -556,18 +566,17 @@ core_target::detach (inferior *inf, int from_tty) HUMAN_NAME is a human-readable name for the kind of registers the NAME section contains, for use in error messages. - If REQUIRED is non-zero, print an error if the core file doesn't - have a section by the appropriate name. Otherwise, just do - nothing. */ + If REQUIRED is true, print an error if the core file doesn't have a + section by the appropriate name. Otherwise, just do nothing. */ -static void -get_core_register_section (struct regcache *regcache, - const struct regset *regset, - const char *name, - int min_size, - int which, - const char *human_name, - int required) +void +core_target::get_core_register_section (struct regcache *regcache, + const struct regset *regset, + const char *name, + int min_size, + int which, + const char *human_name, + bool required) { struct bfd_section *section; bfd_size_type size; @@ -614,12 +623,19 @@ get_core_register_section (struct regcache *regcache, return; } - gdb_assert (core_vec); - core_vec->core_read_registers (regcache, contents, size, which, - ((CORE_ADDR) - bfd_section_vma (core_bfd, section))); + gdb_assert (m_core_vec != nullptr); + m_core_vec->core_read_registers (regcache, contents, size, which, + ((CORE_ADDR) + bfd_section_vma (core_bfd, section))); } +/* Data passed to gdbarch_iterate_over_regset_sections's callback. */ +struct get_core_registers_cb_data +{ + core_target *target; + struct regcache *regcache; +}; + /* Callback for get_core_registers that handles a single core file register note section. */ @@ -628,12 +644,12 @@ get_core_registers_cb (const char *sect_name, int size, const struct regset *regset, const char *human_name, void *cb_data) { - struct regcache *regcache = (struct regcache *) cb_data; - int required = 0; + auto *data = (get_core_registers_cb_data *) cb_data; + bool required = false; if (strcmp (sect_name, ".reg") == 0) { - required = 1; + required = true; if (human_name == NULL) human_name = "general-purpose"; } @@ -645,8 +661,8 @@ get_core_registers_cb (const char *sect_name, int size, /* The 'which' parameter is only used when no regset is provided. Thus we just set it to -1. */ - get_core_register_section (regcache, regset, sect_name, - size, -1, human_name, required); + data->target->get_core_register_section (data->regcache, regset, sect_name, + size, -1, human_name, required); } /* Get the registers out of a core file. This is the machine- @@ -662,8 +678,9 @@ core_target::fetch_registers (struct regcache *regcache, int regno) int i; struct gdbarch *gdbarch; - if (!(core_gdbarch && gdbarch_iterate_over_regset_sections_p (core_gdbarch)) - && (core_vec == NULL || core_vec->core_read_registers == NULL)) + if (!(m_core_gdbarch != nullptr + && gdbarch_iterate_over_regset_sections_p (m_core_gdbarch)) + && (m_core_vec == NULL || m_core_vec->core_read_registers == NULL)) { fprintf_filtered (gdb_stderr, "Can't fetch registers from this type of core file\n"); @@ -672,9 +689,12 @@ core_target::fetch_registers (struct regcache *regcache, int regno) gdbarch = regcache->arch (); if (gdbarch_iterate_over_regset_sections_p (gdbarch)) - gdbarch_iterate_over_regset_sections (gdbarch, - get_core_registers_cb, - (void *) regcache, NULL); + { + get_core_registers_cb_data data = { this, regcache }; + gdbarch_iterate_over_regset_sections (gdbarch, + get_core_registers_cb, + (void *) &data, NULL); + } else { get_core_register_section (regcache, NULL, @@ -692,7 +712,7 @@ core_target::fetch_registers (struct regcache *regcache, int regno) void core_target::files_info () { - print_section_info (core_data, core_bfd); + print_section_info (&m_core_section_table, core_bfd); } struct spuid_list @@ -733,11 +753,12 @@ core_target::xfer_partial (enum target_object object, const char *annex, switch (object) { case TARGET_OBJECT_MEMORY: - return section_table_xfer_memory_partial (readbuf, writebuf, - offset, len, xfered_len, - core_data->sections, - core_data->sections_end, - NULL); + return (section_table_xfer_memory_partial + (readbuf, writebuf, + offset, len, xfered_len, + m_core_section_table.sections, + m_core_section_table.sections_end, + NULL)); case TARGET_OBJECT_AUXV: if (readbuf) @@ -810,14 +831,14 @@ core_target::xfer_partial (enum target_object object, const char *annex, return TARGET_XFER_E_IO; case TARGET_OBJECT_LIBRARIES: - if (core_gdbarch - && gdbarch_core_xfer_shared_libraries_p (core_gdbarch)) + if (m_core_gdbarch != nullptr + && gdbarch_core_xfer_shared_libraries_p (m_core_gdbarch)) { if (writebuf) return TARGET_XFER_E_IO; else { - *xfered_len = gdbarch_core_xfer_shared_libraries (core_gdbarch, + *xfered_len = gdbarch_core_xfer_shared_libraries (m_core_gdbarch, readbuf, offset, len); @@ -830,15 +851,15 @@ core_target::xfer_partial (enum target_object object, const char *annex, /* FALL THROUGH */ case TARGET_OBJECT_LIBRARIES_AIX: - if (core_gdbarch - && gdbarch_core_xfer_shared_libraries_aix_p (core_gdbarch)) + if (m_core_gdbarch != nullptr + && gdbarch_core_xfer_shared_libraries_aix_p (m_core_gdbarch)) { if (writebuf) return TARGET_XFER_E_IO; else { *xfered_len - = gdbarch_core_xfer_shared_libraries_aix (core_gdbarch, + = gdbarch_core_xfer_shared_libraries_aix (m_core_gdbarch, readbuf, offset, len); @@ -911,10 +932,10 @@ core_target::xfer_partial (enum target_object object, const char *annex, case TARGET_OBJECT_SIGNAL_INFO: if (readbuf) { - if (core_gdbarch - && gdbarch_core_xfer_siginfo_p (core_gdbarch)) + if (m_core_gdbarch != nullptr + && gdbarch_core_xfer_siginfo_p (m_core_gdbarch)) { - LONGEST l = gdbarch_core_xfer_siginfo (core_gdbarch, readbuf, + LONGEST l = gdbarch_core_xfer_siginfo (m_core_gdbarch, readbuf, offset, len); if (l >= 0) @@ -953,16 +974,16 @@ core_target::thread_alive (ptid_t ptid) /* Ask the current architecture what it knows about this core file. That will be used, in turn, to pick a better architecture. This wrapper could be avoided if targets got a chance to specialize - core_ops. */ + core_target. */ const struct target_desc * core_target::read_description () { - if (core_gdbarch && gdbarch_core_read_description_p (core_gdbarch)) + if (m_core_gdbarch && gdbarch_core_read_description_p (m_core_gdbarch)) { const struct target_desc *result; - result = gdbarch_core_read_description (core_gdbarch, this, core_bfd); + result = gdbarch_core_read_description (m_core_gdbarch, this, core_bfd); if (result != NULL) return result; } @@ -979,9 +1000,9 @@ core_target::pid_to_str (ptid_t ptid) /* The preferred way is to have a gdbarch/OS specific implementation. */ - if (core_gdbarch - && gdbarch_core_pid_to_str_p (core_gdbarch)) - return gdbarch_core_pid_to_str (core_gdbarch, ptid); + if (m_core_gdbarch != nullptr + && gdbarch_core_pid_to_str_p (m_core_gdbarch)) + return gdbarch_core_pid_to_str (m_core_gdbarch, ptid); /* Otherwise, if we don't have one, we'll just fallback to "process", with normal_pid_to_str. */ @@ -1005,9 +1026,9 @@ core_target::pid_to_str (ptid_t ptid) const char * core_target::thread_name (struct thread_info *thr) { - if (core_gdbarch - && gdbarch_core_thread_name_p (core_gdbarch)) - return gdbarch_core_thread_name (core_gdbarch, thr); + if (m_core_gdbarch != nullptr + && gdbarch_core_thread_name_p (m_core_gdbarch)) + return gdbarch_core_thread_name (m_core_gdbarch, thr); return NULL; } diff --git a/gdb/target.h b/gdb/target.h index b37702b501..e2d1e61cdc 100644 --- a/gdb/target.h +++ b/gdb/target.h @@ -1243,6 +1243,20 @@ struct target_ops TARGET_DEFAULT_IGNORE (); }; +/* Deleter for std::unique_ptr. See comments in + target_ops::~target_ops and target_ops::close about heap-allocated + targets. */ +struct target_ops_deleter +{ + void operator() (target_ops *target) + { + target->close (); + } +}; + +/* A unique pointer for target_ops. */ +typedef std::unique_ptr target_ops_up; + /* Native target backends call this once at initialization time to inform the core about which is the target that can respond to "run" or "attach". Note: native targets are always singletons. */