For linux-csky upstream, I need remove them.
Signed-off-by: Guo Ren ren_guo@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
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@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
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
Hi Guo, Guo Ren wrote,
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.
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
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
-
Guo Ren -
Waldemar Brodkorb