From: "Khem Raj" <raj.khem@gmail.com>
To: openembedded-core@lists.openembedded.org
Subject: Re: [OE-core][PATCH] glib-2.0: Backport GMainContext fixes
Date: Fri, 3 Apr 2020 10:16:08 -0700 [thread overview]
Message-ID: <04a9bc5c-fa6f-ea78-36b2-5db00bdcaeb7@gmail.com> (raw)
In-Reply-To: <20200403072018.2952371-1-daniel@qtec.com>
On 4/3/20 12:20 AM, Daniel Gomez wrote:
> Backport fixes introduced in 2.63.6 for memory leaks and memory corruption in
> GMainContext
>
> Upstream merge: https://gitlab.gnome.org/GNOME/glib/-/merge_requests/1353
>
> Fixes SIGSEGV in GStreamer:
>
> Thread 2 "multihandlesink" received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7ffff6bb9700 (LWP 18045)]
> 0x00007ffff7d65992 in g_source_unref_internal (source=0x7ffff00047d0, context=0x55555561c800, have_lock=1) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:2146
> 2146 ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c: No such file or directory.
> (gdb) bt
> #0 0x00007ffff7d65992 in g_source_unref_internal (source=0x7ffff00047d0, context=0x55555561c800, have_lock=1) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:2146
> #1 0x00007ffff7d65bb6 in g_source_iter_next (iter=iter@entry=0x7ffff6bb8db0, source=source@entry=0x7ffff6bb8da8) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:980
> #2 0x00007ffff7d67ef3 in g_main_context_prepare (context=context@entry=0x55555561c800, priority=priority@entry=0x7ffff6bb8e30) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:944
> #3 0x00007ffff7d6896b in g_main_context_iterate (context=context@entry=0x55555561c800, block=block@entry=1, dispatch=dispatch@entry=1, self=<optimized out>) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:3900
> #4 0x00007ffff7d68b4c in g_main_context_iteration (context=0x55555561c800, may_block=may_block@entry=1) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gmain.c:3981
> #5 0x00007ffff6be4482 in gst_multi_socket_sink_thread (mhsink=0x555555679ab0 [GstMultiSocketSink]) at ../../../gst-plugins-base-1.14.4/gst/tcp/gstmultisocketsink.c:1164
> #6 0x00007ffff7d8fb35 in g_thread_proxy (data=0x55555565c770) at ../../../../../../../repo/workspace/sources/glib-2.0/glib/gthread.c:784
> #7 0x00007ffff7841ebd in start_thread (arg=<optimized out>) at pthread_create.c:486
> #8 0x00007ffff7aa12bf in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
> #8 0x00007ffff7aa12bf in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
>
this change is good
> Signed-off-by: Daniel Gomez <daniel@qtec.com>
> ---
> ...-GSource-iterator-if-iteration-can-m.patch | 43 +++++++
> ...-memory-leaks-and-memory-corruption-.patch | 109 ++++++++++++++++++
> ...e-mutex-unlocking-in-destructor-righ.patch | 36 ++++++
> meta/recipes-core/glib-2.0/glib-2.0_2.62.4.bb | 3 +
> 4 files changed, 191 insertions(+)
> create mode 100644 meta/recipes-core/glib-2.0/glib-2.0/0011-GMainContext-Fix-GSource-iterator-if-iteration-can-m.patch
> create mode 100644 meta/recipes-core/glib-2.0/glib-2.0/0012-GMainContext-Fix-memory-leaks-and-memory-corruption-.patch
> create mode 100644 meta/recipes-core/glib-2.0/glib-2.0/0013-GMainContext-Move-mutex-unlocking-in-destructor-righ.patch
>
> diff --git a/meta/recipes-core/glib-2.0/glib-2.0/0011-GMainContext-Fix-GSource-iterator-if-iteration-can-m.patch b/meta/recipes-core/glib-2.0/glib-2.0/0011-GMainContext-Fix-GSource-iterator-if-iteration-can-m.patch
> new file mode 100644
> index 0000000000..3aae6f0c8c
> --- /dev/null
> +++ b/meta/recipes-core/glib-2.0/glib-2.0/0011-GMainContext-Fix-GSource-iterator-if-iteration-can-m.patch
> @@ -0,0 +1,43 @@
> +From ef2be42998e3fc10299055a5a01f7c791538174c Mon Sep 17 00:00:00 2001
> +From: =?UTF-8?q?Sebastian=20Dr=C3=B6ge?= <sebastian@centricular.com>
> +Date: Mon, 3 Feb 2020 15:38:28 +0200
> +Subject: [PATCH] GMainContext - Fix GSource iterator if iteration can modify
> + the list
> +
> +We first have to ref the next source and then unref the previous one.
> +This might be the last reference to the previous source, and freeing the
> +previous source might unref and free the next one which would then leave
> +use with a dangling pointer here.
> +
> +Fixes https://gitlab.gnome.org/GNOME/glib/issues/2031
> +
> +Upstream-Status: Backport [https://gitlab.gnome.org/GNOME/glib/-/commit/b06c48de7554607ff3fb58d6c0510cfa5088e909]
> +
> +---
> + glib/gmain.c | 8 ++++++--
> + 1 file changed, 6 insertions(+), 2 deletions(-)
> +
> +diff --git a/glib/gmain.c b/glib/gmain.c
> +index af979c8..a9a287d 100644
> +--- a/glib/gmain.c
> ++++ b/glib/gmain.c
> +@@ -969,13 +969,17 @@ g_source_iter_next (GSourceIter *iter, GSource **source)
> + * GSourceList to be removed from source_lists (if iter->source is
> + * the only source in its list, and it is destroyed), so we have to
> + * keep it reffed until after we advance iter->current_list, above.
> ++ *
> ++ * Also we first have to ref the next source before unreffing the
> ++ * previous one as unreffing the previous source can potentially
> ++ * free the next one.
> + */
> ++ if (next_source && iter->may_modify)
> ++ g_source_ref (next_source);
> +
> + if (iter->source && iter->may_modify)
> + g_source_unref_internal (iter->source, iter->context, TRUE);
> + iter->source = next_source;
> +- if (iter->source && iter->may_modify)
> +- g_source_ref (iter->source);
> +
> + *source = iter->source;
> + return *source != NULL;
> diff --git a/meta/recipes-core/glib-2.0/glib-2.0/0012-GMainContext-Fix-memory-leaks-and-memory-corruption-.patch b/meta/recipes-core/glib-2.0/glib-2.0/0012-GMainContext-Fix-memory-leaks-and-memory-corruption-.patch
> new file mode 100644
> index 0000000000..8c025f7fc6
> --- /dev/null
> +++ b/meta/recipes-core/glib-2.0/glib-2.0/0012-GMainContext-Fix-memory-leaks-and-memory-corruption-.patch
> @@ -0,0 +1,109 @@
> +From 611430a32a46d0dc806a829161e2dccf9c0196a8 Mon Sep 17 00:00:00 2001
> +From: =?UTF-8?q?Sebastian=20Dr=C3=B6ge?= <sebastian@centricular.com>
> +Date: Mon, 3 Feb 2020 15:35:51 +0200
> +Subject: [PATCH] GMainContext - Fix memory leaks and memory corruption when
> + freeing sources while freeing a context
> +
> +Instead of destroying sources directly while freeing the context, and
> +potentially freeing them if this was the last reference to them, collect
> +new references of all sources in a separate list before and at the same
> +time invalidate their context so that they can't access it anymore. Only
> +once all sources have their context invalidated, destroy them while
> +still keeping a reference to them. Once all sources are destroyed we get
> +rid of the additional references and free them if nothing else keeps a
> +reference to them anymore.
> +
> +This fixes a regression introduced by 26056558be in 2012.
> +
> +The previous code that invalidated the context of each source and then
> +destroyed it before going to the next source without keeping an
> +additional reference caused memory leaks or memory corruption depending
> +on the order of the sources in the sources lists.
> +
> +If a source was destroyed it might happen that this was the last
> +reference to this source, and it would then be freed. This would cause
> +the finalize function to be called, which might destroy and unref
> +another source and potentially free it. This other source would then
> +either
> +- go through the normal free logic and change the intern linked list
> + between the sources, while other sources that are unreffed as part of
> + the main context freeing would not. As such the list would be in an
> + inconsistent state and we might dereference freed memory.
> +- go through the normal destroy and free logic but because the context
> + pointer was already invalidated it would simply mark the source as
> + destroyed without actually removing it from the context. This would
> + then cause a memory leak because the reference owned by the context is
> + not freed.
> +
> +Fixes https://github.com/gtk-rs/glib/issues/583 while still keeping
> +https://bugzilla.gnome.org/show_bug.cgi?id=661767 fixes.
> +
> +Upstream-Status: Backport [https://gitlab.gnome.org/GNOME/glib/-/commit/aa20167d419c649f34fed06a9463890b41b1eba0]
> +
> +---
> + glib/gmain.c | 35 ++++++++++++++++++++++++++++++++++-
> + 1 file changed, 34 insertions(+), 1 deletion(-)
> +
> +diff --git a/glib/gmain.c b/glib/gmain.c
> +index a9a287d..10ba2f8 100644
> +--- a/glib/gmain.c
> ++++ b/glib/gmain.c
> +@@ -538,6 +538,7 @@ g_main_context_unref (GMainContext *context)
> + GSourceIter iter;
> + GSource *source;
> + GList *sl_iter;
> ++ GSList *s_iter, *remaining_sources = NULL;
> + GSourceList *list;
> + guint i;
> +
> +@@ -557,10 +558,30 @@ g_main_context_unref (GMainContext *context)
> +
> + /* g_source_iter_next() assumes the context is locked. */
> + LOCK_CONTEXT (context);
> +- g_source_iter_init (&iter, context, TRUE);
> ++
> ++ /* First collect all remaining sources from the sources lists and store a
> ++ * new reference in a separate list. Also set the context of the sources
> ++ * to NULL so that they can't access a partially destroyed context anymore.
> ++ *
> ++ * We have to do this first so that we have a strong reference to all
> ++ * sources and destroying them below does not also free them, and so that
> ++ * none of the sources can access the context from their finalize/dispose
> ++ * functions. */
> ++ g_source_iter_init (&iter, context, FALSE);
> + while (g_source_iter_next (&iter, &source))
> + {
> + source->context = NULL;
> ++ remaining_sources = g_slist_prepend (remaining_sources, g_source_ref (source));
> ++ }
> ++ g_source_iter_clear (&iter);
> ++
> ++ /* Next destroy all sources. As we still hold a reference to all of them,
> ++ * this won't cause any of them to be freed yet and especially prevents any
> ++ * source that unrefs another source from its finalize function to be freed.
> ++ */
> ++ for (s_iter = remaining_sources; s_iter; s_iter = s_iter->next)
> ++ {
> ++ source = s_iter->data;
> + g_source_destroy_internal (source, context, TRUE);
> + }
> + UNLOCK_CONTEXT (context);
> +@@ -585,6 +606,18 @@ g_main_context_unref (GMainContext *context)
> + g_cond_clear (&context->cond);
> +
> + g_free (context);
> ++
> ++ /* And now finally get rid of our references to the sources. This will cause
> ++ * them to be freed unless something else still has a reference to them. Due
> ++ * to setting the context pointers in the sources to NULL above, this won't
> ++ * ever access the context or the internal linked list inside the GSource.
> ++ * We already removed the sources completely from the context above. */
> ++ for (s_iter = remaining_sources; s_iter; s_iter = s_iter->next)
> ++ {
> ++ source = s_iter->data;
> ++ g_source_unref_internal (source, NULL, FALSE);
> ++ }
> ++ g_slist_free (remaining_sources);
> + }
> +
> + /* Helper function used by mainloop/overflow test.
> diff --git a/meta/recipes-core/glib-2.0/glib-2.0/0013-GMainContext-Move-mutex-unlocking-in-destructor-righ.patch b/meta/recipes-core/glib-2.0/glib-2.0/0013-GMainContext-Move-mutex-unlocking-in-destructor-righ.patch
> new file mode 100644
> index 0000000000..b10ba0066e
> --- /dev/null
> +++ b/meta/recipes-core/glib-2.0/glib-2.0/0013-GMainContext-Move-mutex-unlocking-in-destructor-righ.patch
> @@ -0,0 +1,36 @@
> +From 3e9d85f1b75e2b1096d9643563d7d17380752fc7 Mon Sep 17 00:00:00 2001
> +From: =?UTF-8?q?Sebastian=20Dr=C3=B6ge?= <sebastian@centricular.com>
> +Date: Tue, 11 Feb 2020 09:34:38 +0200
> +Subject: [PATCH] GMainContext - Move mutex unlocking in destructor right
> + before freeing the mutex
> +
> +This does not have any behaviour changes but is cleaner. The mutex is
> +only unlocked now after all operations on the context are done and right
> +before freeing the mutex and the context itself.
> +
> +Upstream-Status: Backport [https://gitlab.gnome.org/GNOME/glib/-/commit/730a75fc8e8271c38fbd5363d1f77a00876b9ddc]
> +
> +---
> + glib/gmain.c | 2 +-
> + 1 file changed, 1 insertion(+), 1 deletion(-)
> +
> +diff --git a/glib/gmain.c b/glib/gmain.c
> +index 10ba2f8..b1df470 100644
> +--- a/glib/gmain.c
> ++++ b/glib/gmain.c
> +@@ -584,7 +584,6 @@ g_main_context_unref (GMainContext *context)
> + source = s_iter->data;
> + g_source_destroy_internal (source, context, TRUE);
> + }
> +- UNLOCK_CONTEXT (context);
> +
> + for (sl_iter = context->source_lists; sl_iter; sl_iter = sl_iter->next)
> + {
> +@@ -595,6 +594,7 @@ g_main_context_unref (GMainContext *context)
> +
> + g_hash_table_destroy (context->sources);
> +
> ++ UNLOCK_CONTEXT (context);
> + g_mutex_clear (&context->mutex);
> +
> + g_ptr_array_free (context->pending_dispatches, TRUE);
> diff --git a/meta/recipes-core/glib-2.0/glib-2.0_2.62.4.bb b/meta/recipes-core/glib-2.0/glib-2.0_2.62.4.bb
> index 5ed3b4e871..d496235003 100644
> --- a/meta/recipes-core/glib-2.0/glib-2.0_2.62.4.bb
> +++ b/meta/recipes-core/glib-2.0/glib-2.0_2.62.4.bb
> @@ -16,6 +16,9 @@ SRC_URI = "${GNOME_MIRROR}/glib/${SHRT_VER}/glib-${PV}.tar.xz \
> file://0001-Do-not-write-bindir-into-pkg-config-files.patch \
> file://0001-meson-Run-atomics-test-on-clang-as-well.patch \
> file://0001-gio-tests-resources.c-comment-out-a-build-host-only-.patch \
> + file://0011-GMainContext-Fix-GSource-iterator-if-iteration-can-m.patch \
> + file://0012-GMainContext-Fix-memory-leaks-and-memory-corruption-.patch \
> + file://0013-GMainContext-Move-mutex-unlocking-in-destructor-righ.patch \
> "
>
> SRC_URI_append_class-native = " file://relocate-modules.patch"
> --
> 2.25.1
>
>
>
>
prev parent reply other threads:[~2020-04-03 17:16 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-03 7:20 [OE-core][PATCH] glib-2.0: Backport GMainContext fixes Daniel Gomez
2020-04-03 17:16 ` Khem Raj [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=04a9bc5c-fa6f-ea78-36b2-5db00bdcaeb7@gmail.com \
--to=raj.khem@gmail.com \
--cc=openembedded-core@lists.openembedded.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox