26 Mar
2018
26 Mar
'18
6:30 a.m.
For linux-csky upstream, I need remove them.
Signed-off-by: Guo Ren <ren_guo(a)c-sky.com>
---
extra/Configs/Config.csky | 1 -
libc/sysdeps/linux/csky/Makefile.arch | 2 +-
2 files changed, 1 insertion(+), 2 deletions(-)
diff --git a/extra/Configs/Config.csky b/extra/Configs/Config.csky
index 3173ce9..f5dc683 100644
--- a/extra/Configs/Config.csky
+++ b/extra/Configs/Config.csky
@@ -6,7 +6,6 @@ config FORCE_OPTIONS_FOR_ARCH
bool
default y
select ARCH_ANY_ENDIAN
- select ARCH_HAS_DEPRECATED_SYSCALLS
select ARCH_USE_MMU
select ARCH_HAS_MMU
diff --git a/libc/sysdeps/linux/csky/Makefile.arch
b/libc/sysdeps/linux/csky/Makefile.arch
index 704b36a..a3157c2 100644
--- a/libc/sysdeps/linux/csky/Makefile.arch
+++ b/libc/sysdeps/linux/csky/Makefile.arch
@@ -1,6 +1,6 @@
CSRC-y := clone.c __syscall_error.c cacheflush.c
SSRC-y := __longjmp.S setjmp.S
-SSRC-y += libc-read_tp.S vfork.S csky_clone.S
+SSRC-y += libc-read_tp.S csky_clone.S
--
2.7.4
26 Mar
26 Mar
6:30 a.m.
New subject: [PATCH 2/2] csky: bugfix remove uneeded bkpt from crti.S
LD need word-align to merge every _init into one func.
If it's not word-align, ld will pad zero into the free-space.
That's "bkpt".
The "bkpt" is OK when there is no jtag server connected.
Because linux kernel exception could handle it.
Signed-off-by: Guo Ren <ren_guo(a)c-sky.com>
---
libc/sysdeps/linux/csky/crti.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libc/sysdeps/linux/csky/crti.S b/libc/sysdeps/linux/csky/crti.S
index e7b045f..a2434c0 100644
--- a/libc/sysdeps/linux/csky/crti.S
+++ b/libc/sysdeps/linux/csky/crti.S
@@ -12,9 +12,11 @@
ENTRY(_init)
subi sp, 8
stw lr, (sp, 4)
+ nop /* LD need word-align to compile into one func */
.section .fini
ENTRY(_fini)
subi sp, 8
stw lr, (sp, 4)
+ nop
--
2.7.4
27 Mar
27 Mar
3:41 a.m.
Hi Waldemar,
On Mon, Mar 26, 2018 at 09:25:03PM +0200, Waldemar Brodkorb wrote:
Only openrisc has no vfork.S.
So new architectures have no vfork anymore?
ref: linux-4.16-rc5/include/uapi/asm-generic/unistd.h:838-900
#ifdef __ARCH_WANT_SYSCALL_DEPRECATED
...
#define __NR_vfork 1071
...
#endif /* __ARCH_WANT_SYSCALL_DEPRECATED */
If there is no __NR_vfork in unistd.h, vfork.S is empty.
eg nds32 vfork.S:
#ifdef __NR_vfork
...
#endif
eg arc vfork.S:
Are you sure it could compile with the linux-4.16-rc5? There is no
__NR_vfork at all.
For csky, we don't have historical burdens, we are new born.
Best Regards
Guo Ren
28 Mar
28 Mar
6:31 p.m.
Hi Guo,
Guo Ren wrote,
Hi Waldemar,
On Mon, Mar 26, 2018 at 09:25:03PM +0200, Waldemar Brodkorb wrote:
Thanks. Would it be better if we wait with your uClibc-ng patches
until a Linux kernel is released with c-sky support?
I have now working configurations in my regression testing script,
so I find any issues when changing common code.
best regards
Waldemar
Only openrisc has no vfork.S.
So new architectures have no vfork anymore?
ref: linux-4.16-rc5/include/uapi/asm-generic/unistd.h:838-900
#ifdef __ARCH_WANT_SYSCALL_DEPRECATED
...
#define __NR_vfork 1071
...
#endif /* __ARCH_WANT_SYSCALL_DEPRECATED */
If there is no __NR_vfork in unistd.h, vfork.S is empty.
eg nds32 vfork.S:
#ifdef __NR_vfork
...
#endif
eg arc vfork.S:
Are you sure it could compile with the linux-4.16-rc5? There is no
__NR_vfork at all.
For csky, we don't have historical burdens, we are new born.
29 Mar
29 Mar
5:35 a.m.
On Wed, Mar 28, 2018 at 08:31:34PM +0200, Waldemar Brodkorb wrote:
Thanks. Would it be better if we wait with your
uClibc-ng patches
until a Linux kernel is released with c-sky support?
I have now working configurations in my regression testing script,
so I find any issues when changing common code.
Sure, I'll resend you patchset together when Linux kernel accept csky.
Just drop this patch and forcus on your work :)
Best Regards
Guo Ren
2221
days inactive
2224
days old
5 comments
2 participants
participants (2)
-
Guo Ren -
Waldemar Brodkorb