public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
* PATCH: Replace @PLT with @GOTPCREL(%rip) in call
@ 2016-03-03 13:21 H.J. Lu
  2016-03-04  0:03 ` Roland McGrath
  0 siblings, 1 reply; 4+ messages in thread
From: H.J. Lu @ 2016-03-03 13:21 UTC (permalink / raw)
  To: GNU C Library

Since both PREINIT_FUNCTION and __libc_start_main are defined externally, 
they are called via PLT.  Use "call *func@GOTPCREL(%rip)" removes the
extra branch to PLT entry.

	[BZ #19745]
	* sysdeps/x86_64/crti.S (_init): Replace @PLT with
	@GOTPCREL(%rip) in call.
	* sysdeps/x86_64/start.S (_start): Likewise.
---
 sysdeps/x86_64/crti.S   |  2 +-
 sysdeps/x86_64/start.S  |  2 +-
 sysdeps/x86_64/sysdep.h | 14 +++++++++++---
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/sysdeps/x86_64/crti.S b/sysdeps/x86_64/crti.S
index a345259..3dff141 100644
--- a/sysdeps/x86_64/crti.S
+++ b/sysdeps/x86_64/crti.S
@@ -66,7 +66,7 @@ _init:
 	movq PREINIT_FUNCTION@GOTPCREL(%rip), %rax
 	testq %rax, %rax
 	je .Lno_weak_fn
-	call PREINIT_FUNCTION@PLT
+	call *PREINIT_FUNCTION@GOTPCREL(%rip)
 .Lno_weak_fn:
 #else
 	call PREINIT_FUNCTION
diff --git a/sysdeps/x86_64/start.S b/sysdeps/x86_64/start.S
index 1374974..22b18ae 100644
--- a/sysdeps/x86_64/start.S
+++ b/sysdeps/x86_64/start.S
@@ -105,7 +105,7 @@ ENTRY (_start)
 
 	/* Call the user's main function, and exit with its value.
 	   But let the libc call main.	  */
-	call __libc_start_main@PLT
+	call *__libc_start_main@GOTPCREL(%rip)
 #else
 	/* Pass address of our own entry points to .fini and .init.  */
 	mov $__libc_csu_fini, %R8_LP

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: PATCH: Replace @PLT with @GOTPCREL(%rip) in call
  2016-03-03 13:21 PATCH: Replace @PLT with @GOTPCREL(%rip) in call H.J. Lu
@ 2016-03-04  0:03 ` Roland McGrath
  2016-03-04 23:40   ` H.J. Lu
  0 siblings, 1 reply; 4+ messages in thread
From: Roland McGrath @ 2016-03-04  0:03 UTC (permalink / raw)
  To: H.J. Lu; +Cc: GNU C Library

> --- a/sysdeps/x86_64/crti.S
> +++ b/sysdeps/x86_64/crti.S
> @@ -66,7 +66,7 @@ _init:
>  	movq PREINIT_FUNCTION@GOTPCREL(%rip), %rax
>  	testq %rax, %rax
>  	je .Lno_weak_fn
> -	call PREINIT_FUNCTION@PLT
> +	call *PREINIT_FUNCTION@GOTPCREL(%rip)

Can't this just be "call *%rax"?

> --- a/sysdeps/x86_64/start.S
> +++ b/sysdeps/x86_64/start.S
> @@ -105,7 +105,7 @@ ENTRY (_start)
>  
>  	/* Call the user's main function, and exit with its value.
>  	   But let the libc call main.	  */
> -	call __libc_start_main@PLT
> +	call *__libc_start_main@GOTPCREL(%rip)
>  #else
>  	/* Pass address of our own entry points to .fini and .init.  */
>  	mov $__libc_csu_fini, %R8_LP

In the general case, foo@PLT is preferable to foo@GOTPCREL(%rip) because it
allows for lazy resolution.  That's not worthwhile here because _start is
presumably always run and run very early.  But it merits a comment saying
why it's not following the normal pattern for PIC calls.


Thanks,
Roland

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: PATCH: Replace @PLT with @GOTPCREL(%rip) in call
  2016-03-04  0:03 ` Roland McGrath
@ 2016-03-04 23:40   ` H.J. Lu
  2016-03-04 23:46     ` Roland McGrath
  0 siblings, 1 reply; 4+ messages in thread
From: H.J. Lu @ 2016-03-04 23:40 UTC (permalink / raw)
  To: Roland McGrath; +Cc: GNU C Library

[-- Attachment #1: Type: text/plain, Size: 1157 bytes --]

On Thu, Mar 3, 2016 at 4:03 PM, Roland McGrath <roland@hack.frob.com> wrote:
>> --- a/sysdeps/x86_64/crti.S
>> +++ b/sysdeps/x86_64/crti.S
>> @@ -66,7 +66,7 @@ _init:
>>       movq PREINIT_FUNCTION@GOTPCREL(%rip), %rax
>>       testq %rax, %rax
>>       je .Lno_weak_fn
>> -     call PREINIT_FUNCTION@PLT
>> +     call *PREINIT_FUNCTION@GOTPCREL(%rip)
>
> Can't this just be "call *%rax"?

Good idea.

>> --- a/sysdeps/x86_64/start.S
>> +++ b/sysdeps/x86_64/start.S
>> @@ -105,7 +105,7 @@ ENTRY (_start)
>>
>>       /* Call the user's main function, and exit with its value.
>>          But let the libc call main.    */
>> -     call __libc_start_main@PLT
>> +     call *__libc_start_main@GOTPCREL(%rip)
>>  #else
>>       /* Pass address of our own entry points to .fini and .init.  */
>>       mov $__libc_csu_fini, %R8_LP
>
> In the general case, foo@PLT is preferable to foo@GOTPCREL(%rip) because it
> allows for lazy resolution.  That's not worthwhile here because _start is
> presumably always run and run very early.  But it merits a comment saying
> why it's not following the normal pattern for PIC calls.
>

Here are updated patches.


-- 
H.J.

[-- Attachment #2: 0001-Replace-PLT-with-GOTPCREL-rip-in-call.patch --]
[-- Type: text/x-patch, Size: 1714 bytes --]

From 976cfff7e4538503c70413d6746ccbcf87e395e1 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Fri, 4 Mar 2016 15:28:59 -0800
Subject: [PATCH 1/2] Replace @PLT with @GOTPCREL(%rip) in call

Since __libc_start_main is called very early, lazy binding isn't relevant
here.  Use indirect branch via GOT to avoid extra branch to PLT slot.

	[BZ #19745]
	* sysdeps/x86_64/start.S (_start): __libc_start_main@PLT
	with *__libc_start_main@GOTPCREL(%rip) in call.
---
 ChangeLog              | 6 ++++++
 sysdeps/x86_64/start.S | 6 ++++--
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 143d4a6..af55529 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,9 @@
+2016-03-04  H.J. Lu  <hongjiu.lu@intel.com>
+
+	[BZ #19745]
+	* sysdeps/x86_64/start.S (_start): Replace __libc_start_main@PLT
+	with *__libc_start_main@GOTPCREL(%rip) in call.
+
 2016-03-04  Roland McGrath  <roland@hack.frob.com>
 
 	* Makefile (tests-special): Add $(objpfx)c++-types-check.out only
diff --git a/sysdeps/x86_64/start.S b/sysdeps/x86_64/start.S
index 1374974..2369b69 100644
--- a/sysdeps/x86_64/start.S
+++ b/sysdeps/x86_64/start.S
@@ -104,8 +104,10 @@ ENTRY (_start)
 	mov main@GOTPCREL(%rip), %RDI_LP
 
 	/* Call the user's main function, and exit with its value.
-	   But let the libc call main.	  */
-	call __libc_start_main@PLT
+	   But let the libc call main.  Since __libc_start_main is
+	   called very early, lazy binding isn't relevant here.  Use
+	   indirect branch via GOT to avoid extra branch to PLT slot.  */
+	call *__libc_start_main@GOTPCREL(%rip)
 #else
 	/* Pass address of our own entry points to .fini and .init.  */
 	mov $__libc_csu_fini, %R8_LP
-- 
2.5.0


[-- Attachment #3: 0002-Replace-PREINIT_FUNCTION-PLT-with-rax-in-call.patch --]
[-- Type: text/x-patch, Size: 1324 bytes --]

From 9ff9ae1b792e3d986f68a127e0b3f8679e6bea17 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Fri, 4 Mar 2016 15:36:18 -0800
Subject: [PATCH 2/2] Replace PREINIT_FUNCTION@PLT with *%rax in call

Since we have loaded address of PREINIT_FUNCTION into %rax, we can
avoid extra branch to PLT slot.

	[BZ #19745]
	* sysdeps/x86_64/crti.S (_init): Replace PREINIT_FUNCTION@PLT
	with *%rax in call.
---
 ChangeLog             | 6 ++++++
 sysdeps/x86_64/crti.S | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/ChangeLog b/ChangeLog
index af55529..f5b8d1e 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,6 +1,12 @@
 2016-03-04  H.J. Lu  <hongjiu.lu@intel.com>
 
 	[BZ #19745]
+	* sysdeps/x86_64/crti.S (_init): Replace PREINIT_FUNCTION@PLT
+	with *%rax in call.
+
+2016-03-04  H.J. Lu  <hongjiu.lu@intel.com>
+
+	[BZ #19745]
 	* sysdeps/x86_64/start.S (_start): Replace __libc_start_main@PLT
 	with *__libc_start_main@GOTPCREL(%rip) in call.
 
diff --git a/sysdeps/x86_64/crti.S b/sysdeps/x86_64/crti.S
index a345259..e9d86ed 100644
--- a/sysdeps/x86_64/crti.S
+++ b/sysdeps/x86_64/crti.S
@@ -66,7 +66,7 @@ _init:
 	movq PREINIT_FUNCTION@GOTPCREL(%rip), %rax
 	testq %rax, %rax
 	je .Lno_weak_fn
-	call PREINIT_FUNCTION@PLT
+	call *%rax
 .Lno_weak_fn:
 #else
 	call PREINIT_FUNCTION
-- 
2.5.0


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: PATCH: Replace @PLT with @GOTPCREL(%rip) in call
  2016-03-04 23:40   ` H.J. Lu
@ 2016-03-04 23:46     ` Roland McGrath
  0 siblings, 0 replies; 4+ messages in thread
From: Roland McGrath @ 2016-03-04 23:46 UTC (permalink / raw)
  To: H.J. Lu; +Cc: GNU C Library

Both look fine to me.

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2016-03-04 23:46 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-03-03 13:21 PATCH: Replace @PLT with @GOTPCREL(%rip) in call H.J. Lu
2016-03-04  0:03 ` Roland McGrath
2016-03-04 23:40   ` H.J. Lu
2016-03-04 23:46     ` Roland McGrath

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).