* [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
@ 2022-02-11 23:19 H.J. Lu
2022-02-14 0:16 ` H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2022-02-11 23:19 UTC (permalink / raw)
To: binutils
Don't change indirect symbol defined in IR to undefined if it is
referenced from shared object.
bfd/
PR ld/28879
* elflink.c (_bfd_elf_merge_symbol): Don't change indirect
symbol defined in IR to undefined if it is referenced from
shared object.
ld/
PR ld/28879
* testsuite/ld-plugin/lto.exp: Run PR ld/28879 tests.
* testsuite/ld-plugin/pr28879a.cc: New file.
* testsuite/ld-plugin/pr28879b.cc: Likewise.
---
bfd/elflink.c | 5 ++---
ld/testsuite/ld-plugin/lto.exp | 26 ++++++++++++++++++++++++++
ld/testsuite/ld-plugin/pr28879a.cc | 7 +++++++
ld/testsuite/ld-plugin/pr28879b.cc | 8 ++++++++
4 files changed, 43 insertions(+), 3 deletions(-)
create mode 100644 ld/testsuite/ld-plugin/pr28879a.cc
create mode 100644 ld/testsuite/ld-plugin/pr28879b.cc
diff --git a/bfd/elflink.c b/bfd/elflink.c
index 6fa18d92007..f8521426cad 100644
--- a/bfd/elflink.c
+++ b/bfd/elflink.c
@@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
h->root.non_ir_ref_dynamic = true;
hi->root.non_ir_ref_dynamic = true;
}
-
- if ((oldbfd->flags & BFD_PLUGIN) != 0
- && hi->root.type == bfd_link_hash_indirect)
+ else if ((oldbfd->flags & BFD_PLUGIN) != 0
+ && hi->root.type == bfd_link_hash_indirect)
{
/* Change indirect symbol from IR to undefined. */
hi->root.type = bfd_link_hash_undefined;
diff --git a/ld/testsuite/ld-plugin/lto.exp b/ld/testsuite/ld-plugin/lto.exp
index a70a84562b8..64b880265ee 100644
--- a/ld/testsuite/ld-plugin/lto.exp
+++ b/ld/testsuite/ld-plugin/lto.exp
@@ -471,6 +471,32 @@ set lto_link_elf_tests [list \
[list {pr27441c.so} \
{-shared -fPIC -Wl,--as-needed tmpdir/pr27441c.o tmpdir/pr27441b.so tmpdir/pr27441a.so} {-fPIC} \
{dummy.c} {{readelf {-dW} pr27441c.d}} {pr27441c.so}] \
+ [list \
+ "Build libpr28879a.so" \
+ "-shared" \
+ "-O0 -fpic" \
+ {pr28879a.cc} \
+ {} \
+ "libpr28879a.so" \
+ "c++" \
+ ] \
+ [list \
+ "Build libpr28879b.so" \
+ "-shared -Wl,--no-as-needed tmpdir/libpr28879a.so" \
+ "-O2 -fpic" \
+ {dummy.c} \
+ {} \
+ "libpr28879b.so" \
+ ] \
+ [list \
+ "Build pr28879" \
+ "-Wl,--no-as-needed tmpdir/libpr28879b.so -Wl,-rpath-link,." \
+ "-O0 -flto -D_GLIBCXX_ASSERTIONS" \
+ {pr28879b.cc} \
+ {} \
+ "pr28879" \
+ "c++" \
+ ] \
]
# PR 14918 checks that libgcc is not spuriously included in a shared link of
diff --git a/ld/testsuite/ld-plugin/pr28879a.cc b/ld/testsuite/ld-plugin/pr28879a.cc
new file mode 100644
index 00000000000..8307a42e2fb
--- /dev/null
+++ b/ld/testsuite/ld-plugin/pr28879a.cc
@@ -0,0 +1,7 @@
+#include <string>
+
+void
+func (std::string *s)
+{
+ delete s;
+}
diff --git a/ld/testsuite/ld-plugin/pr28879b.cc b/ld/testsuite/ld-plugin/pr28879b.cc
new file mode 100644
index 00000000000..02fc351366c
--- /dev/null
+++ b/ld/testsuite/ld-plugin/pr28879b.cc
@@ -0,0 +1,8 @@
+#include <string>
+
+int
+main (void)
+{
+ std::string header;
+ return 0;
+}
--
2.34.1
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
2022-02-11 23:19 [PATCH] ld: Keep indirect symbol from IR if referenced from shared object H.J. Lu
@ 2022-02-14 0:16 ` H.J. Lu
2022-02-14 2:47 ` Alan Modra
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2022-02-14 0:16 UTC (permalink / raw)
To: Binutils, Nick Clifton, Alan Modra
On Fri, Feb 11, 2022 at 3:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> Don't change indirect symbol defined in IR to undefined if it is
> referenced from shared object.
>
> bfd/
>
> PR ld/28879
> * elflink.c (_bfd_elf_merge_symbol): Don't change indirect
> symbol defined in IR to undefined if it is referenced from
> shared object.
>
> ld/
>
> PR ld/28879
> * testsuite/ld-plugin/lto.exp: Run PR ld/28879 tests.
> * testsuite/ld-plugin/pr28879a.cc: New file.
> * testsuite/ld-plugin/pr28879b.cc: Likewise.
> ---
> bfd/elflink.c | 5 ++---
> ld/testsuite/ld-plugin/lto.exp | 26 ++++++++++++++++++++++++++
> ld/testsuite/ld-plugin/pr28879a.cc | 7 +++++++
> ld/testsuite/ld-plugin/pr28879b.cc | 8 ++++++++
> 4 files changed, 43 insertions(+), 3 deletions(-)
> create mode 100644 ld/testsuite/ld-plugin/pr28879a.cc
> create mode 100644 ld/testsuite/ld-plugin/pr28879b.cc
>
> diff --git a/bfd/elflink.c b/bfd/elflink.c
> index 6fa18d92007..f8521426cad 100644
> --- a/bfd/elflink.c
> +++ b/bfd/elflink.c
> @@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
> h->root.non_ir_ref_dynamic = true;
> hi->root.non_ir_ref_dynamic = true;
> }
> -
> - if ((oldbfd->flags & BFD_PLUGIN) != 0
> - && hi->root.type == bfd_link_hash_indirect)
> + else if ((oldbfd->flags & BFD_PLUGIN) != 0
> + && hi->root.type == bfd_link_hash_indirect)
> {
> /* Change indirect symbol from IR to undefined. */
> hi->root.type = bfd_link_hash_undefined;
> diff --git a/ld/testsuite/ld-plugin/lto.exp b/ld/testsuite/ld-plugin/lto.exp
> index a70a84562b8..64b880265ee 100644
> --- a/ld/testsuite/ld-plugin/lto.exp
> +++ b/ld/testsuite/ld-plugin/lto.exp
> @@ -471,6 +471,32 @@ set lto_link_elf_tests [list \
> [list {pr27441c.so} \
> {-shared -fPIC -Wl,--as-needed tmpdir/pr27441c.o tmpdir/pr27441b.so tmpdir/pr27441a.so} {-fPIC} \
> {dummy.c} {{readelf {-dW} pr27441c.d}} {pr27441c.so}] \
> + [list \
> + "Build libpr28879a.so" \
> + "-shared" \
> + "-O0 -fpic" \
> + {pr28879a.cc} \
> + {} \
> + "libpr28879a.so" \
> + "c++" \
> + ] \
> + [list \
> + "Build libpr28879b.so" \
> + "-shared -Wl,--no-as-needed tmpdir/libpr28879a.so" \
> + "-O2 -fpic" \
> + {dummy.c} \
> + {} \
> + "libpr28879b.so" \
> + ] \
> + [list \
> + "Build pr28879" \
> + "-Wl,--no-as-needed tmpdir/libpr28879b.so -Wl,-rpath-link,." \
> + "-O0 -flto -D_GLIBCXX_ASSERTIONS" \
> + {pr28879b.cc} \
> + {} \
> + "pr28879" \
> + "c++" \
> + ] \
> ]
>
> # PR 14918 checks that libgcc is not spuriously included in a shared link of
> diff --git a/ld/testsuite/ld-plugin/pr28879a.cc b/ld/testsuite/ld-plugin/pr28879a.cc
> new file mode 100644
> index 00000000000..8307a42e2fb
> --- /dev/null
> +++ b/ld/testsuite/ld-plugin/pr28879a.cc
> @@ -0,0 +1,7 @@
> +#include <string>
> +
> +void
> +func (std::string *s)
> +{
> + delete s;
> +}
> diff --git a/ld/testsuite/ld-plugin/pr28879b.cc b/ld/testsuite/ld-plugin/pr28879b.cc
> new file mode 100644
> index 00000000000..02fc351366c
> --- /dev/null
> +++ b/ld/testsuite/ld-plugin/pr28879b.cc
> @@ -0,0 +1,8 @@
> +#include <string>
> +
> +int
> +main (void)
> +{
> + std::string header;
> + return 0;
> +}
> --
> 2.34.1
>
Hi Nick, Alan,
Is this OK for master and 2.38 branch?
Thanks.
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
2022-02-14 0:16 ` H.J. Lu
@ 2022-02-14 2:47 ` Alan Modra
2022-02-14 2:58 ` H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: Alan Modra @ 2022-02-14 2:47 UTC (permalink / raw)
To: H.J. Lu; +Cc: Binutils, Nick Clifton
On Sun, Feb 13, 2022 at 04:16:35PM -0800, H.J. Lu wrote:
> On Fri, Feb 11, 2022 at 3:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >
> > Don't change indirect symbol defined in IR to undefined if it is
> > referenced from shared object.
The description above
> > diff --git a/bfd/elflink.c b/bfd/elflink.c
> > index 6fa18d92007..f8521426cad 100644
> > --- a/bfd/elflink.c
> > +++ b/bfd/elflink.c
> > @@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
> > h->root.non_ir_ref_dynamic = true;
> > hi->root.non_ir_ref_dynamic = true;
> > }
> > -
> > - if ((oldbfd->flags & BFD_PLUGIN) != 0
> > - && hi->root.type == bfd_link_hash_indirect)
> > + else if ((oldbfd->flags & BFD_PLUGIN) != 0
> > + && hi->root.type == bfd_link_hash_indirect)
> > {
> > /* Change indirect symbol from IR to undefined. */
> > hi->root.type = bfd_link_hash_undefined;
and the patch do not exactly match. The code logic is old sym
indirect in IR, new sym not IR, both in shared library *or both not*.
Is that correct?
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
2022-02-14 2:47 ` Alan Modra
@ 2022-02-14 2:58 ` H.J. Lu
2022-02-14 4:13 ` Alan Modra
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2022-02-14 2:58 UTC (permalink / raw)
To: Alan Modra; +Cc: Binutils, Nick Clifton
On Sun, Feb 13, 2022 at 6:47 PM Alan Modra <amodra@gmail.com> wrote:
>
> On Sun, Feb 13, 2022 at 04:16:35PM -0800, H.J. Lu wrote:
> > On Fri, Feb 11, 2022 at 3:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> > >
> > > Don't change indirect symbol defined in IR to undefined if it is
> > > referenced from shared object.
>
> The description above
>
> > > diff --git a/bfd/elflink.c b/bfd/elflink.c
> > > index 6fa18d92007..f8521426cad 100644
> > > --- a/bfd/elflink.c
> > > +++ b/bfd/elflink.c
> > > @@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
> > > h->root.non_ir_ref_dynamic = true;
> > > hi->root.non_ir_ref_dynamic = true;
> > > }
> > > -
> > > - if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > - && hi->root.type == bfd_link_hash_indirect)
> > > + else if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > + && hi->root.type == bfd_link_hash_indirect)
> > > {
> > > /* Change indirect symbol from IR to undefined. */
> > > hi->root.type = bfd_link_hash_undefined;
>
> and the patch do not exactly match. The code logic is old sym
> indirect in IR, new sym not IR, both in shared library *or both not*.
> Is that correct?
>
The existing code is
if (newdyn != olddyn)
{
/* Handle a case where plugin_notice won't be called and thus
won't set the non_ir_ref flags on the first pass over
symbols. */
h->root.non_ir_ref_dynamic = true;
hi->root.non_ir_ref_dynamic = true;
}
If an indirect symbol defined in IR is marked with
non_ir_ref_dynamic, we can't change it to undefined.
That is why I added "else" so that we don't get the
error of undefined reference from shared object.
if ((oldbfd->flags & BFD_PLUGIN) != 0
&& hi->root.type == bfd_link_hash_indirect)
{
/* Change indirect symbol from IR to undefined. */
hi->root.type = bfd_link_hash_undefined;
hi->root.u.undef.abfd = oldbfd;
}
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
2022-02-14 2:58 ` H.J. Lu
@ 2022-02-14 4:13 ` Alan Modra
2022-02-14 15:28 ` H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: Alan Modra @ 2022-02-14 4:13 UTC (permalink / raw)
To: H.J. Lu; +Cc: Binutils, Nick Clifton
On Sun, Feb 13, 2022 at 06:58:48PM -0800, H.J. Lu wrote:
> On Sun, Feb 13, 2022 at 6:47 PM Alan Modra <amodra@gmail.com> wrote:
> >
> > On Sun, Feb 13, 2022 at 04:16:35PM -0800, H.J. Lu wrote:
> > > On Fri, Feb 11, 2022 at 3:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > >
> > > > Don't change indirect symbol defined in IR to undefined if it is
> > > > referenced from shared object.
> >
> > The description above
> >
> > > > diff --git a/bfd/elflink.c b/bfd/elflink.c
> > > > index 6fa18d92007..f8521426cad 100644
> > > > --- a/bfd/elflink.c
> > > > +++ b/bfd/elflink.c
> > > > @@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
> > > > h->root.non_ir_ref_dynamic = true;
> > > > hi->root.non_ir_ref_dynamic = true;
> > > > }
> > > > -
> > > > - if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > > - && hi->root.type == bfd_link_hash_indirect)
> > > > + else if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > > + && hi->root.type == bfd_link_hash_indirect)
> > > > {
> > > > /* Change indirect symbol from IR to undefined. */
> > > > hi->root.type = bfd_link_hash_undefined;
> >
> > and the patch do not exactly match. The code logic is old sym
> > indirect in IR, new sym not IR, both in shared library *or both not*.
> > Is that correct?
> >
>
> The existing code is
>
> if (newdyn != olddyn)
> {
> /* Handle a case where plugin_notice won't be called and thus
> won't set the non_ir_ref flags on the first pass over
> symbols. */
> h->root.non_ir_ref_dynamic = true;
> hi->root.non_ir_ref_dynamic = true;
> }
>
> If an indirect symbol defined in IR is marked with
> non_ir_ref_dynamic, we can't change it to undefined.
> That is why I added "else" so that we don't get the
> error of undefined reference from shared object.
The patch is OK, I wasn't thinking clearly.
> if ((oldbfd->flags & BFD_PLUGIN) != 0
> && hi->root.type == bfd_link_hash_indirect)
> {
> /* Change indirect symbol from IR to undefined. */
> hi->root.type = bfd_link_hash_undefined;
> hi->root.u.undef.abfd = oldbfd;
> }
>
>
> --
> H.J.
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] ld: Keep indirect symbol from IR if referenced from shared object
2022-02-14 4:13 ` Alan Modra
@ 2022-02-14 15:28 ` H.J. Lu
0 siblings, 0 replies; 6+ messages in thread
From: H.J. Lu @ 2022-02-14 15:28 UTC (permalink / raw)
To: Alan Modra; +Cc: Binutils, Nick Clifton
On Sun, Feb 13, 2022 at 8:13 PM Alan Modra <amodra@gmail.com> wrote:
>
> On Sun, Feb 13, 2022 at 06:58:48PM -0800, H.J. Lu wrote:
> > On Sun, Feb 13, 2022 at 6:47 PM Alan Modra <amodra@gmail.com> wrote:
> > >
> > > On Sun, Feb 13, 2022 at 04:16:35PM -0800, H.J. Lu wrote:
> > > > On Fri, Feb 11, 2022 at 3:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > > >
> > > > > Don't change indirect symbol defined in IR to undefined if it is
> > > > > referenced from shared object.
> > >
> > > The description above
> > >
> > > > > diff --git a/bfd/elflink.c b/bfd/elflink.c
> > > > > index 6fa18d92007..f8521426cad 100644
> > > > > --- a/bfd/elflink.c
> > > > > +++ b/bfd/elflink.c
> > > > > @@ -1294,9 +1294,8 @@ _bfd_elf_merge_symbol (bfd *abfd,
> > > > > h->root.non_ir_ref_dynamic = true;
> > > > > hi->root.non_ir_ref_dynamic = true;
> > > > > }
> > > > > -
> > > > > - if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > > > - && hi->root.type == bfd_link_hash_indirect)
> > > > > + else if ((oldbfd->flags & BFD_PLUGIN) != 0
> > > > > + && hi->root.type == bfd_link_hash_indirect)
> > > > > {
> > > > > /* Change indirect symbol from IR to undefined. */
> > > > > hi->root.type = bfd_link_hash_undefined;
> > >
> > > and the patch do not exactly match. The code logic is old sym
> > > indirect in IR, new sym not IR, both in shared library *or both not*.
> > > Is that correct?
> > >
> >
> > The existing code is
> >
> > if (newdyn != olddyn)
> > {
> > /* Handle a case where plugin_notice won't be called and thus
> > won't set the non_ir_ref flags on the first pass over
> > symbols. */
> > h->root.non_ir_ref_dynamic = true;
> > hi->root.non_ir_ref_dynamic = true;
> > }
> >
> > If an indirect symbol defined in IR is marked with
> > non_ir_ref_dynamic, we can't change it to undefined.
> > That is why I added "else" so that we don't get the
> > error of undefined reference from shared object.
>
> The patch is OK, I wasn't thinking clearly.
I checked it into master branch and will backport it to
2.38 branch later this week.
> > if ((oldbfd->flags & BFD_PLUGIN) != 0
> > && hi->root.type == bfd_link_hash_indirect)
> > {
> > /* Change indirect symbol from IR to undefined. */
> > hi->root.type = bfd_link_hash_undefined;
> > hi->root.u.undef.abfd = oldbfd;
> > }
> >
> >
> > --
> > H.J.
>
> --
> Alan Modra
> Australia Development Lab, IBM
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2022-02-14 15:29 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-02-11 23:19 [PATCH] ld: Keep indirect symbol from IR if referenced from shared object H.J. Lu
2022-02-14 0:16 ` H.J. Lu
2022-02-14 2:47 ` Alan Modra
2022-02-14 2:58 ` H.J. Lu
2022-02-14 4:13 ` Alan Modra
2022-02-14 15:28 ` H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).