[PATCH] cris: Wire up missing syscalls

[PATCH] cris: Wire up missing syscalls

[Chen Gang]

The related warnings:

    CALL    scripts/checksyscalls.sh
  <stdin>:1229:2: warning: #warning syscall sched_setattr not implemented [-Wcpp]
  <stdin>:1232:2: warning: #warning syscall sched_getattr not implemented [-Wcpp]
  <stdin>:1235:2: warning: #warning syscall renameat2 not implemented [-Wcpp]
  <stdin>:1238:2: warning: #warning syscall seccomp not implemented [-Wcpp]
  <stdin>:1241:2: warning: #warning syscall getrandom not implemented [-Wcpp]
  <stdin>:1244:2: warning: #warning syscall memfd_create not implemented [-Wcpp]
  <stdin>:1247:2: warning: #warning syscall bpf not implemented [-Wcpp]
  <stdin>:1250:2: warning: #warning syscall execveat not implemented [-Wcpp]

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
---
 arch/cris/arch-v10/kernel/entry.S   | 8 ++++++++
 arch/cris/arch-v32/kernel/entry.S   | 8 ++++++++
 arch/cris/include/uapi/asm/unistd.h | 8 ++++++++
 3 files changed, 24 insertions(+)

diff --git a/arch/cris/arch-v10/kernel/entry.S b/arch/cris/arch-v10/kernel/entry.S
index 81570fc..d94c520 100644
--- a/arch/cris/arch-v10/kernel/entry.S
+++ b/arch/cris/arch-v10/kernel/entry.S
@@ -955,6 +955,14 @@ sys_call_table:
 	.long sys_process_vm_writev
 	.long sys_kcmp			/* 350 */
 	.long sys_finit_module
+	.long _sys_sched_setattr
+	.long _sys_sched_getattr
+	.long _sys_renameat2
+	.long _sys_seccomp		/* 355 */
+	.long _sys_getrandom
+	.long _sys_memfd_create
+	.long _sys_bpf
+	.long _sys_execveat
 
         /*
          * NOTE!! This doesn't have to be exact - we just have
diff --git a/arch/cris/arch-v32/kernel/entry.S b/arch/cris/arch-v32/kernel/entry.S
index 026a0b2..62a82d8 100644
--- a/arch/cris/arch-v32/kernel/entry.S
+++ b/arch/cris/arch-v32/kernel/entry.S
@@ -875,6 +875,14 @@ sys_call_table:
 	.long sys_process_vm_writev
 	.long sys_kcmp			/* 350 */
 	.long sys_finit_module
+	.long _sys_sched_setattr
+	.long _sys_sched_getattr
+	.long _sys_renameat2
+	.long _sys_seccomp		/* 355 */
+	.long _sys_getrandom
+	.long _sys_memfd_create
+	.long _sys_bpf
+	.long _sys_execveat
 
 	/*
 	 * NOTE!! This doesn't have to be exact - we just have
diff --git a/arch/cris/include/uapi/asm/unistd.h b/arch/cris/include/uapi/asm/unistd.h
index f3287fa..062b648 100644
--- a/arch/cris/include/uapi/asm/unistd.h
+++ b/arch/cris/include/uapi/asm/unistd.h
@@ -356,5 +356,13 @@
 #define __NR_process_vm_writev	349
 #define __NR_kcmp		350
 #define __NR_finit_module	351
+#define __NR_sched_setattr	352
+#define __NR_sched_getattr	353
+#define __NR_renameat2		354
+#define __NR_seccomp		355
+#define __NR_getrandom		356
+#define __NR_memfd_create	357
+#define __NR_bpf		358
+#define __NR_execveat		359
 
 #endif /* _UAPI_ASM_CRIS_UNISTD_H_ */
-- 
1.9.3

Re: [PATCH] cris: Wire up missing syscalls

[Jesper Nilsson]

On Wed, May 06, 2015 at 03:48:12PM +0200, Chen Gang wrote:
> The related warnings:
> 
>     CALL    scripts/checksyscalls.sh
>   <stdin>:1229:2: warning: #warning syscall sched_setattr not implemented [-Wcpp]
>   <stdin>:1232:2: warning: #warning syscall sched_getattr not implemented [-Wcpp]
>   <stdin>:1235:2: warning: #warning syscall renameat2 not implemented [-Wcpp]
>   <stdin>:1238:2: warning: #warning syscall seccomp not implemented [-Wcpp]
>   <stdin>:1241:2: warning: #warning syscall getrandom not implemented [-Wcpp]
>   <stdin>:1244:2: warning: #warning syscall memfd_create not implemented [-Wcpp]
>   <stdin>:1247:2: warning: #warning syscall bpf not implemented [-Wcpp]
>   <stdin>:1250:2: warning: #warning syscall execveat not implemented [-Wcpp]
> 
> Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>

Thanks and applied, sorry for the long delay, I had to rewrite the patch a bit
since the v10 and v32 does not have the same standard in prefixing underscores.
Also, to be safe I also bumped the NR_syscalls as below.

---
 arch/cris/arch-v10/kernel/entry.S   |    8 ++++++++
 arch/cris/arch-v32/kernel/entry.S   |    8 ++++++++
 arch/cris/include/asm/unistd.h      |    2 +-
 arch/cris/include/uapi/asm/unistd.h |    8 ++++++++
 4 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/arch/cris/arch-v10/kernel/entry.S b/arch/cris/arch-v10/kernel/entry.S
index 81570fc..d94c520 100644
--- a/arch/cris/arch-v10/kernel/entry.S
+++ b/arch/cris/arch-v10/kernel/entry.S
@@ -955,6 +955,14 @@ sys_call_table:
 	.long sys_process_vm_writev
 	.long sys_kcmp			/* 350 */
 	.long sys_finit_module
+	.long _sys_sched_setattr
+	.long _sys_sched_getattr
+	.long _sys_renameat2
+	.long _sys_seccomp		/* 355 */
+	.long _sys_getrandom
+	.long _sys_memfd_create
+	.long _sys_bpf
+	.long _sys_execveat
 
         /*
          * NOTE!! This doesn't have to be exact - we just have
diff --git a/arch/cris/arch-v32/kernel/entry.S b/arch/cris/arch-v32/kernel/entry.S
index 026a0b2..1c5595a 100644
--- a/arch/cris/arch-v32/kernel/entry.S
+++ b/arch/cris/arch-v32/kernel/entry.S
@@ -875,6 +875,14 @@ sys_call_table:
 	.long sys_process_vm_writev
 	.long sys_kcmp			/* 350 */
 	.long sys_finit_module
+	.long sys_sched_setattr
+	.long sys_sched_getattr
+	.long sys_renameat2
+	.long sys_seccomp		/* 355 */
+	.long sys_getrandom
+	.long sys_memfd_create
+	.long sys_bpf
+	.long sys_execveat
 
 	/*
 	 * NOTE!! This doesn't have to be exact - we just have
diff --git a/arch/cris/include/asm/unistd.h b/arch/cris/include/asm/unistd.h
index 0f40fed..9c23535 100644
--- a/arch/cris/include/asm/unistd.h
+++ b/arch/cris/include/asm/unistd.h
@@ -4,7 +4,7 @@
 #include <uapi/asm/unistd.h>
 
 
-#define NR_syscalls 360
+#define NR_syscalls 365
 
 #include <arch/unistd.h>
 
diff --git a/arch/cris/include/uapi/asm/unistd.h b/arch/cris/include/uapi/asm/unistd.h
index f3287fa..062b648 100644
--- a/arch/cris/include/uapi/asm/unistd.h
+++ b/arch/cris/include/uapi/asm/unistd.h
@@ -356,5 +356,13 @@
 #define __NR_process_vm_writev	349
 #define __NR_kcmp		350
 #define __NR_finit_module	351
+#define __NR_sched_setattr	352
+#define __NR_sched_getattr	353
+#define __NR_renameat2		354
+#define __NR_seccomp		355
+#define __NR_getrandom		356
+#define __NR_memfd_create	357
+#define __NR_bpf		358
+#define __NR_execveat		359
 
 #endif /* _UAPI_ASM_CRIS_UNISTD_H_ */
-- 
1.7.10.4


/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com

Re: [PATCH] cris: Wire up missing syscalls

[Chen Gang]

On 06/03/2015 06:20 PM, Jesper Nilsson wrote:
> 
> On Wed, May 06, 2015 at 03:48:12PM +0200, Chen Gang wrote:
>> The related warnings:
>>
>>     CALL    scripts/checksyscalls.sh
>>   <stdin>:1229:2: warning: #warning syscall sched_setattr not implemented [-Wcpp]
>>   <stdin>:1232:2: warning: #warning syscall sched_getattr not implemented [-Wcpp]
>>   <stdin>:1235:2: warning: #warning syscall renameat2 not implemented [-Wcpp]
>>   <stdin>:1238:2: warning: #warning syscall seccomp not implemented [-Wcpp]
>>   <stdin>:1241:2: warning: #warning syscall getrandom not implemented [-Wcpp]
>>   <stdin>:1244:2: warning: #warning syscall memfd_create not implemented [-Wcpp]
>>   <stdin>:1247:2: warning: #warning syscall bpf not implemented [-Wcpp]
>>   <stdin>:1250:2: warning: #warning syscall execveat not implemented [-Wcpp]
>>
>> Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
> 
> Thanks and applied, sorry for the long delay, I had to rewrite the patch a bit
> since the v10 and v32 does not have the same standard in prefixing underscores.
> Also, to be safe I also bumped the NR_syscalls as below.
> 

Oh, it doesn't matter, everyone's time resources are expensive, so I
should/can understand your delay response.

I also delayed many things, e.g. I planed to wire up all syscalls for
all archs within 2015-05-31, but I delayed (2 another archs: frv and
m32r are still left, I shall try to send patches for them, next).

 - One main excuse is I have to devote much free time resources on qemu,
   which I have delayed too much to bear (now, I am still trying).

 - The other reason is frv upstream compiler can be built, but can not
   work at all, I have to analyse it firstly.

Sorry for my carelessness: use "_sys*", and keep original NR_syscalls no
touch. And also 3 additional things I guess we may need a look:

 - For v10, need we also use "sys*" instead of "_sys*"?

 - Most archs do not implement seccomp and bpf, which can pass building,
   but will return -ENOSYS during running. Need we left them still as
   warnings? (I guess, it depends on the maintainer's taste).

 - In the latest next tree, it also add additional userfaultfd syscall,
   need we add it, too?


Thanks.

> ---
>  arch/cris/arch-v10/kernel/entry.S   |    8 ++++++++
>  arch/cris/arch-v32/kernel/entry.S   |    8 ++++++++
>  arch/cris/include/asm/unistd.h      |    2 +-
>  arch/cris/include/uapi/asm/unistd.h |    8 ++++++++
>  4 files changed, 25 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/cris/arch-v10/kernel/entry.S b/arch/cris/arch-v10/kernel/entry.S
> index 81570fc..d94c520 100644
> --- a/arch/cris/arch-v10/kernel/entry.S
> +++ b/arch/cris/arch-v10/kernel/entry.S
> @@ -955,6 +955,14 @@ sys_call_table:
>  	.long sys_process_vm_writev
>  	.long sys_kcmp			/* 350 */
>  	.long sys_finit_module
> +	.long _sys_sched_setattr
> +	.long _sys_sched_getattr
> +	.long _sys_renameat2
> +	.long _sys_seccomp		/* 355 */
> +	.long _sys_getrandom
> +	.long _sys_memfd_create
> +	.long _sys_bpf
> +	.long _sys_execveat
>  
>          /*
>           * NOTE!! This doesn't have to be exact - we just have
> diff --git a/arch/cris/arch-v32/kernel/entry.S b/arch/cris/arch-v32/kernel/entry.S
> index 026a0b2..1c5595a 100644
> --- a/arch/cris/arch-v32/kernel/entry.S
> +++ b/arch/cris/arch-v32/kernel/entry.S
> @@ -875,6 +875,14 @@ sys_call_table:
>  	.long sys_process_vm_writev
>  	.long sys_kcmp			/* 350 */
>  	.long sys_finit_module
> +	.long sys_sched_setattr
> +	.long sys_sched_getattr
> +	.long sys_renameat2
> +	.long sys_seccomp		/* 355 */
> +	.long sys_getrandom
> +	.long sys_memfd_create
> +	.long sys_bpf
> +	.long sys_execveat
>  
>  	/*
>  	 * NOTE!! This doesn't have to be exact - we just have
> diff --git a/arch/cris/include/asm/unistd.h b/arch/cris/include/asm/unistd.h
> index 0f40fed..9c23535 100644
> --- a/arch/cris/include/asm/unistd.h
> +++ b/arch/cris/include/asm/unistd.h
> @@ -4,7 +4,7 @@
>  #include <uapi/asm/unistd.h>
>  
>  
> -#define NR_syscalls 360
> +#define NR_syscalls 365
>  
>  #include <arch/unistd.h>
>  
> diff --git a/arch/cris/include/uapi/asm/unistd.h b/arch/cris/include/uapi/asm/unistd.h
> index f3287fa..062b648 100644
> --- a/arch/cris/include/uapi/asm/unistd.h
> +++ b/arch/cris/include/uapi/asm/unistd.h
> @@ -356,5 +356,13 @@
>  #define __NR_process_vm_writev	349
>  #define __NR_kcmp		350
>  #define __NR_finit_module	351
> +#define __NR_sched_setattr	352
> +#define __NR_sched_getattr	353
> +#define __NR_renameat2		354
> +#define __NR_seccomp		355
> +#define __NR_getrandom		356
> +#define __NR_memfd_create	357
> +#define __NR_bpf		358
> +#define __NR_execveat		359
>  
>  #endif /* _UAPI_ASM_CRIS_UNISTD_H_ */
> 

-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed

Re: [PATCH] cris: Wire up missing syscalls

[Jesper Nilsson]

On Wed, Jun 03, 2015 at 01:36:47PM +0200, Chen Gang wrote:
> On 06/03/2015 06:20 PM, Jesper Nilsson wrote:
> > On Wed, May 06, 2015 at 03:48:12PM +0200, Chen Gang wrote:
> >> The related warnings:
> >> Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
> > Thanks and applied, sorry for the long delay, I had to rewrite the patch a bit
> > since the v10 and v32 does not have the same standard in prefixing underscores.
> > Also, to be safe I also bumped the NR_syscalls as below.
> > 
> 
> Oh, it doesn't matter, everyone's time resources are expensive, so I
> should/can understand your delay response.

Thanks for understanding.

> Sorry for my carelessness: use "_sys*", and keep original NR_syscalls no
> touch. And also 3 additional things I guess we may need a look:
> 
>  - For v10, need we also use "sys*" instead of "_sys*"?

No, the trick here is that v10 and v32 uses different standards
with regards to prefixing underscore. I'm hoping to fix that someday.

>  - Most archs do not implement seccomp and bpf, which can pass building,
>    but will return -ENOSYS during running. Need we left them still as
>    warnings? (I guess, it depends on the maintainer's taste).

Well, I don't have any strong feelings in either direction. :-)

>  - In the latest next tree, it also add additional userfaultfd syscall,
>    need we add it, too?

Hm, haven't seen that syscall, I'm guessing it's in linux-next?

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com

Re: [PATCH] cris: Wire up missing syscalls

[Hans-Peter Nilsson]

> From: Jesper Nilsson <jespern@axis.com>
> Date: Wed, 3 Jun 2015 17:04:26 +0200

> On Wed, Jun 03, 2015 at 01:36:47PM +0200, Chen Gang wrote:
> >  - For v10, need we also use "sys*" instead of "_sys*"?
> 
> No, the trick here is that v10 and v32 uses different standards
> with regards to prefixing underscore. I'm hoping to fix that someday.

Not exactly, but almost; please let me fill in JFTR:

Most people probably guessed correctly that this is related to
building the CRIS v10 kernel with the *-elf toolchain (which has
the convention that symbols are prefixed with underscores for
cris-*-elf and crisv32-*-elf) while (IIUC) the CRIS v32 kernel
is built with the *-linux-gnu toolchain (which does not have any
symbol prefix for cris-*-linux-gnu crisv32-*-linux-gnu).

One Might Imagine that there'd be build machinery in place in
the kernel such that such a symbol prefix issue could be hidden,
like a header to include and using e.g. S(symbol) in all
assembly files.  (Actually some time ago I thought there already
was, but Jesper corrected me.)

brgds, H-P

Re: [PATCH] cris: Wire up missing syscalls

[Jesper Nilsson]

On Wed, Jun 03, 2015 at 05:04:26PM +0200, Jesper Nilsson wrote:
> On Wed, Jun 03, 2015 at 01:36:47PM +0200, Chen Gang wrote:
> >  - For v10, need we also use "sys*" instead of "_sys*"?
> 
> No, the trick here is that v10 and v32 uses different standards
> with regards to prefixing underscore. I'm hoping to fix that someday.

... and as your autotest found, this was a real problem.
I'll push a new commit to drop the underscore for v10 also.

BR,

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com

Re: [PATCH] cris: Wire up missing syscalls

[Chen Gang]

On 06/04/2015 03:20 PM, Jesper Nilsson wrote:
> On Wed, Jun 03, 2015 at 05:04:26PM +0200, Jesper Nilsson wrote:
>> On Wed, Jun 03, 2015 at 01:36:47PM +0200, Chen Gang wrote:
>>>  - For v10, need we also use "sys*" instead of "_sys*"?
>>
>> No, the trick here is that v10 and v32 uses different standards
>> with regards to prefixing underscore. I'm hoping to fix that someday.
> 
> ... and as your autotest found, this was a real problem.
> I'll push a new commit to drop the underscore for v10 also.
> 

Thank you for your work.

And "kbuild test robot" found it (not me).  :-)

Thanks.
-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed

Re: [PATCH] cris: Wire up missing syscalls

[Chen Gang]

On 06/04/2015 03:20 PM, Jesper Nilsson wrote:
> On Wed, Jun 03, 2015 at 05:04:26PM +0200, Jesper Nilsson wrote:
>> On Wed, Jun 03, 2015 at 01:36:47PM +0200, Chen Gang wrote:
>>>  - For v10, need we also use "sys*" instead of "_sys*"?
>>
>> No, the trick here is that v10 and v32 uses different standards
>> with regards to prefixing underscore. I'm hoping to fix that someday.
> 
> ... and as your autotest found, this was a real problem.
> I'll push a new commit to drop the underscore for v10 also.
> 

Thank you for your work.

And "kbuild test robot" found it (not me).  :-)

-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed

Re: [PATCH] cris: Wire up missing syscalls

[Chen Gang]

On 6/3/15 23:23, Hans-Peter Nilsson wrote:
> 
> Most people probably guessed correctly that this is related to
> building the CRIS v10 kernel with the *-elf toolchain (which has
> the convention that symbols are prefixed with underscores for
> cris-*-elf and crisv32-*-elf) while (IIUC) the CRIS v32 kernel
> is built with the *-linux-gnu toolchain (which does not have any
> symbol prefix for cris-*-linux-gnu crisv32-*-linux-gnu).
> 

What you said above is helpful to me. Originally I often use *-elf for
cross compiling linux kernel, it is incorrect for some archs (e.g. frv,
m32r, and our cris).

After use *-linux, the upstream frv toolchain can work well (for me, it
is the last arch which has missing syscalls). :-)


Thanks.
-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed

Re: [PATCH] cris: Wire up missing syscalls

[Chen Gang]

On 6/3/15 23:23, Hans-Peter Nilsson wrote:
> 
> Most people probably guessed correctly that this is related to
> building the CRIS v10 kernel with the *-elf toolchain (which has
> the convention that symbols are prefixed with underscores for
> cris-*-elf and crisv32-*-elf) while (IIUC) the CRIS v32 kernel
> is built with the *-linux-gnu toolchain (which does not have any
> symbol prefix for cris-*-linux-gnu crisv32-*-linux-gnu).
> 

What you said above is helpful to me. Originally I often use *-elf for
cross compiling linux kernel, it is incorrect for some archs.

After use *-linux, the upstream frv toolchain can work well (for me, it
is the last arch which has missing syscalls). :-)


Thanks.
-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed