15 Jun
2016
15 Jun
'16
12:39 p.m.
Signed-off-by: Vineet Gupta <vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
--
2.5.0
15 Jun
15 Jun
6:34 p.m.
Hi Vineet,
Vineet Gupta wrote,
Signed-off-by: Vineet Gupta
<vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
Can you please be a little more descriptive why this patch is
useful? Will ARC support static PIE or is Scrt1.o used in other
code?
best regards
Waldemar
Sorry, had to resend with correct sender address.
16 Jun
16 Jun
6:38 a.m.
On Wednesday 15 June 2016 08:34 PM, Waldemar Brodkorb wrote:
Hi Vineet,
Vineet Gupta wrote,
This is for a dynamically linked PIE - can we have static linked PIE ? Current
approach of taking addresses of functions, in not position independent but still
works for the normal dynamically linked executables (non PIE) since the zero based
addresses are rightly resolved in final link. This is not true for PIE hence we
need a really position independent way (PC relative addr) to pass those function
addresses.
Do u want me to respin with beefed up changelog.
Signed-off-by: Vineet Gupta
<vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
Can you please be a little more descriptive why this patch is
useful? Will ARC support static PIE or is Scrt1.o used in other
code? Sorry, had to resend with correct sender address.
My address was wrong ?
7:49 p.m.
Hi Vineet,
Vineet Gupta wrote,
No, I use mutt and I am to lazy to configure send-hooks, so I
manually change my mail address before sending e-Mails :/
best regards
Waldemar
On Wednesday 15 June 2016 08:34 PM, Waldemar Brodkorb
wrote:
You can try static PIE with binutils 2.26 and up with
-Wl,--no-dynamic-linker, but my tests with ARM/MIPS failed.
So I am not sure if crt1.S is the problem or something else.
Hi Vineet,
Vineet Gupta wrote,
This is for a dynamically linked PIE - can we have static linked PIE ? Current
approach of taking addresses of functions, in not position independent but still
works for the normal dynamically linked executables (non PIE) since the zero based
addresses are rightly resolved in final link. This is not true for PIE hence we
need a really position independent way (PC relative addr) to pass those function
addresses. Signed-off-by: Vineet Gupta
<vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
Can you please be a little more descriptive why this patch is
useful? Will ARC support static PIE or is Scrt1.o used in other
code? Do u want me to respin with beefed up changelog.
Yes, please.
Sorry, had to
resend with correct sender address.
My address was wrong ? 
17 Jun
17 Jun
2:04 a.m.
On Thu, Jun 16, 2016 at 09:49:10PM +0200, Waldemar Brodkorb wrote:
Hi Vineet,
Vineet Gupta wrote,
Static PIE requires an rcrt1.o that's responsible for performing
self-relocation equivalent to what the dynamic linker does to itself.
Scrt1.o is for dynamic-linker PIE (and it's compatible with linking in
non-PIE code whether it's static or dynamic linked, but not necessary
in those cases).
Rich
On Wednesday 15 June 2016 08:34 PM, Waldemar
Brodkorb wrote:
You can try static PIE with binutils 2.26 and up with
-Wl,--no-dynamic-linker, but my tests with ARM/MIPS failed.
So I am not sure if crt1.S is the problem or something else. Hi Vineet,
Vineet Gupta wrote,
This is for a dynamically linked PIE - can we have static linked PIE ? Current
approach of taking addresses of functions, in not position independent but still
works for the normal dynamically linked executables (non PIE) since the zero based
addresses are rightly resolved in final link. This is not true for PIE hence we
need a really position independent way (PC relative addr) to pass those function
addresses. Signed-off-by: Vineet Gupta
<vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
Can you please be a little more descriptive why this patch is
useful? Will ARC support static PIE or is Scrt1.o used in other
code?
23 Jun
23 Jun
11:51 a.m.
New subject: [PATCH v2] ARC: Enable shared crt1
Currently crt1 takes address of functions (main,_init,_fini) directly
which doesn't generate truely position independent code, but zero based
values instead. e.g.
| __start:
| ...
| add_s r2,sp,0x4
| mov_s r0, main
generates to
| 000156ec <__start>:
| ...
| 156f4: add_s r2,sp,0x4
| 156f6: mov_s r0,0x15f7c
| ...
| 00015f7c <main>:
| 15f7c: push_s blink
This works just fine for the normal (non PIE) dynamic executables since
they are loaded at address 0. However this is not true for PIE
executables. So for Scrt1 we use a true position independent way when
taking function addresses.
Cc: uclibc(a)uclibc.org <uclibc(a)uclibc.org>
Cc: devel(a)uclibc-ng.org <devel(a)uclibc-ng.org>
Cc: Cupertino Miranda <cmiranda(a)synopsys.com>
Signed-off-by: Vineet Gupta <vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
--
2.5.0
26 Jun
26 Jun
10:05 a.m.
New subject: [PATCH v2] ARC: Enable shared crt1
Hi Vineet,
Vineet Gupta wrote,
Currently crt1 takes address of functions
(main,_init,_fini) directly
which doesn't generate truely position independent code, but zero based
values instead. e.g.
| __start:
| ...
| add_s r2,sp,0x4
| mov_s r0, main
generates to
| 000156ec <__start>:
| ...
| 156f4: add_s r2,sp,0x4
| 156f6: mov_s r0,0x15f7c
| ...
| 00015f7c <main>:
| 15f7c: push_s blink
This works just fine for the normal (non PIE) dynamic executables since
they are loaded at address 0. However this is not true for PIE
executables. So for Scrt1 we use a true position independent way when
taking function addresses.
Cc: uclibc(a)uclibc.org <uclibc(a)uclibc.org>
Cc: devel(a)uclibc-ng.org <devel(a)uclibc-ng.org>
Cc: Cupertino Miranda <cmiranda(a)synopsys.com>
Signed-off-by: Vineet Gupta <vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
Applied and pushed,
thx
Waldemar
21 Jun
21 Jun
6:56 p.m.
Hi Vineet,
Vineet Gupta wrote,
ping?
(resend with better changelog)
On Wednesday 15 June 2016 08:34 PM, Waldemar Brodkorb
wrote:
Hi Vineet,
Vineet Gupta wrote,
This is for a dynamically linked PIE - can we have static linked PIE ? Current
approach of taking addresses of functions, in not position independent but still
works for the normal dynamically linked executables (non PIE) since the zero based
addresses are rightly resolved in final link. This is not true for PIE hence we
need a really position independent way (PC relative addr) to pass those function
addresses.
Do u want me to respin with beefed up changelog.
Signed-off-by: Vineet Gupta
<vgupta(a)synopsys.com>
---
libc/sysdeps/linux/arc/crt1.S | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libc/sysdeps/linux/arc/crt1.S b/libc/sysdeps/linux/arc/crt1.S
index 95c41f888850..e38c8e81b2c0 100644
--- a/libc/sysdeps/linux/arc/crt1.S
+++ b/libc/sysdeps/linux/arc/crt1.S
@@ -41,11 +41,15 @@ __start:
mov_s r5, r0 ; rltd_fini
add_s r2, sp, 4 ; argv
-
+#ifdef L_Scrt1
+ add r0, pcl, @main@pcl
+ add r3, pcl, @_init@pcl
+ add r4, pcl, @_fini@pcl
+#else
mov_s r0, main
mov_s r3, _init
mov r4, _fini
-
+#endif
and sp, sp, -8
mov r6, sp
Can you please be a little more descriptive why this patch is
useful? Will ARC support static PIE or is Scrt1.o used in other
code? Sorry, had to resend with correct sender address.
My address was wrong ?
_______________________________________________
uClibc mailing list
uClibc(a)uclibc.org
http://lists.busybox.net/mailman/listinfo/uclibc
3071
days inactive
3082
days old
7 comments
3 participants
participants (3)
-
Rich Felker -
Vineet Gupta -
Waldemar Brodkorb