public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Szabolcs Nagy <szabolcs.nagy@arm.com>
To: "H.J. Lu" <hjl.tools@gmail.com>,
	GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH v3 2/5] Make libc symbols hidden in static PIE
Date: Thu, 14 Jan 2021 11:17:57 +0000	[thread overview]
Message-ID: <20210114111757.GA11466@arm.com> (raw)
In-Reply-To: <20210113095053.GA2379@arm.com>

The 01/13/2021 09:50, Szabolcs Nagy via Libc-alpha wrote:
> The 01/12/2021 17:19, H.J. Lu wrote:
> > On Tue, Jan 12, 2021 at 4:33 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > On Tue, Jan 12, 2021 at 4:02 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > > See:
> > > >
> > > > https://sourceware.org/bugzilla/show_bug.cgi?id=14961
> > > >
> > >  /* Mark all symbols hidden in static PIE libc to avoid GOT indirections.  */
> > > -#if BUILD_PIE_DEFAULT && IS_IN (libc) && !defined LIBC_NONSHARED
> > > +#if BUILD_PIE_DEFAULT && !defined NO_HIDDEN_EXTERN_FUNC_IN_PIE \
> > > +    && IS_IN (libc) && !defined LIBC_NONSHARED
> > >  # pragma GCC visibility push(hidden)
> > >  #endif
> > >
> > 
> > This works on i686.

The series i plan to commit today is in nsz/bug27072 now,

This is the v4 of this patch:

Hidden matters with static PIE: extern symbol access in position
independent code usually involves GOT indirections which needs
RELATIVE relocs in a static linked PIE. Hidden visibility avoids
indirections and RELATIVE relocs on targets that can access symbols
pc-relative.

The check should use IS_IN_LIB instead of IS_IN(libc) since all
static libraries can use hidden visibility to avoid indirections,
however the test system links objects from libcrypt.a into dynamic
linked test binaries so hidden does not work there.  I think mixing
static and shared libc components in the same binary should not be
supported usage, but to be safe only use hidden in libc.a.

There are targets (i686) where hidden visibility functions are
problematic in PIE code so hidden cannot be applied to all symbols.
Then static PIE requires extern object access without relocations
(e.g. by relying on copy relocations in shared libraries instead of
GOT access in PIE code). See bug 14961.

From -static-pie linked 'int main(){}' this shaves off 73 relative
relocs on aarch64 and reduces code size too.
---
 include/libc-symbols.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/include/libc-symbols.h b/include/libc-symbols.h
index ea126ae70c..f4dd735555 100644
--- a/include/libc-symbols.h
+++ b/include/libc-symbols.h
@@ -434,13 +434,18 @@ for linking")
   strong_alias(real, name)
 #endif
 
-#if defined SHARED || defined LIBC_NONSHARED \
-  || (BUILD_PIE_DEFAULT && IS_IN (libc))
+#if defined SHARED || defined LIBC_NONSHARED
 # define attribute_hidden __attribute__ ((visibility ("hidden")))
 #else
 # define attribute_hidden
 #endif
 
+/* Mark all symbols hidden in static PIE libc to avoid GOT indirections.  */
+#if BUILD_PIE_DEFAULT && !defined NO_HIDDEN_EXTERN_FUNC_IN_PIE \
+    && IS_IN (libc) && !defined LIBC_NONSHARED
+# pragma GCC visibility push(hidden)
+#endif
+
 #define attribute_tls_model_ie __attribute__ ((tls_model ("initial-exec")))
 
 #define attribute_relro __attribute__ ((section (".data.rel.ro")))
-- 
2.17.1


  reply	other threads:[~2021-01-14 11:18 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-12 17:21 [PATCH v3 0/5] fix ifunc with static pie [BZ #27072] Szabolcs Nagy
2021-01-12 17:21 ` [PATCH v3 1/5] configure: Require PI_STATIC_AND_HIDDEN for static pie Szabolcs Nagy
2021-01-12 18:38   ` Adhemerval Zanella
2021-01-12 17:22 ` [PATCH v3 2/5] Make libc symbols hidden in static PIE Szabolcs Nagy
2021-01-12 23:09   ` H.J. Lu
2021-01-13  0:02     ` H.J. Lu
2021-01-13  0:33       ` H.J. Lu
2021-01-13  1:19         ` H.J. Lu
2021-01-13  9:50           ` Szabolcs Nagy
2021-01-14 11:17             ` Szabolcs Nagy [this message]
2021-01-14 15:39               ` H.J. Lu
2021-01-15  3:36               ` H.J. Lu
2021-01-15  4:29                 ` H.J. Lu
2021-01-15 11:25                 ` Szabolcs Nagy
2021-01-15 13:43                   ` H.J. Lu
2021-01-15 14:27                     ` Szabolcs Nagy
2021-01-15 15:28                       ` H.J. Lu
2021-01-15 22:42                         ` H.J. Lu
2021-01-16  0:41                           ` H.J. Lu
2021-01-16 13:18                             ` H.J. Lu
2021-01-18 16:22                               ` Szabolcs Nagy
2021-01-12 17:22 ` [PATCH v3 3/5] elf: Make the tunable struct definition internal only Szabolcs Nagy
2021-01-13 17:38   ` Adhemerval Zanella
2021-01-12 17:22 ` [PATCH v3 4/5] elf: Avoid RELATIVE relocs in __tunables_init Szabolcs Nagy
2021-01-13 17:42   ` Adhemerval Zanella
2021-01-12 17:23 ` [PATCH v3 5/5] csu: Move static pie self relocation later [BZ #27072] Szabolcs Nagy
2021-01-12 22:55   ` H.J. Lu
2021-01-14 15:49     ` H.J. Lu
2021-01-14 15:52       ` H.J. Lu
2021-01-14 16:01         ` H.J. Lu
2021-01-14 16:26           ` H.J. Lu
2021-01-14 17:19             ` Szabolcs Nagy
2021-01-14 17:59               ` Szabolcs Nagy
2021-01-14 17:05           ` Szabolcs Nagy

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=20210114111757.GA11466@arm.com \
    --to=szabolcs.nagy@arm.com \
    --cc=hjl.tools@gmail.com \
    --cc=libc-alpha@sourceware.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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).