Build failures in -next: sys_pkey_mprotect etc. undefined

Build failures in -next: sys_pkey_mprotect etc. undefined

[Guenter Roeck]

Building h8300:allnoconfig ... failed
Building h8300:h8300h-sim_defconfig ... failed
Building c6x:dsk6455_defconfig ... failed

[ and more builds for the same architectures ]

--------------
Error log:
arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to `sys_pkey_mprotect'
arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to `sys_pkey_alloc'
arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to `sys_pkey_free'

Other architectures (frv, xtensa, score) currently fail to build for other reasons,
so I don't really know if the above are the only ones affected.

Guenter

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Arnd Bergmann]

On Monday, September 12, 2016 7:55:33 AM CEST Guenter Roeck wrote:
> Building h8300:allnoconfig ... failed
> Building h8300:h8300h-sim_defconfig ... failed
> Building c6x:dsk6455_defconfig ... failed
> 
> [ and more builds for the same architectures ]
> 
> --------------
> Error log:
> arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to `sys_pkey_mprotect'
> arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to `sys_pkey_alloc'
> arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to `sys_pkey_free'
> 
> Other architectures (frv, xtensa, score) currently fail to build for other reasons,
> so I don't really know if the above are the only ones affected.
> 
> 

I think the best solution would be to add them to kernel/sys_ni.c so we can
assign syscall numbers for all architectures regardless of whether they use it
or not (most of them will never use these AFAICT).

	Arnd

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Arnd Bergmann]

On Monday, September 12, 2016 7:55:33 AM CEST Guenter Roeck wrote:
> Building h8300:allnoconfig ... failed
> Building h8300:h8300h-sim_defconfig ... failed
> Building c6x:dsk6455_defconfig ... failed
> 
> [ and more builds for the same architectures ]
> 
> --------------
> Error log:
> arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to `sys_pkey_mprotect'
> arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to `sys_pkey_alloc'
> arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to `sys_pkey_free'
> 
> Other architectures (frv, xtensa, score) currently fail to build for other reasons,
> so I don't really know if the above are the only ones affected.
> 
> 

I think the best solution would be to add them to kernel/sys_ni.c so we can
assign syscall numbers for all architectures regardless of whether they use it
or not (most of them will never use these AFAICT).

	Arnd

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Dave Hansen]

On 09/12/2016 07:59 AM, Arnd Bergmann wrote:
> On Monday, September 12, 2016 7:55:33 AM CEST Guenter Roeck wrote:
>> Building h8300:allnoconfig ... failed
>> Building h8300:h8300h-sim_defconfig ... failed
>> Building c6x:dsk6455_defconfig ... failed
>>
>> [ and more builds for the same architectures ]
>>
>> --------------
>> Error log:
>> arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to `sys_pkey_mprotect'
>> arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to `sys_pkey_alloc'
>> arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to `sys_pkey_free'
>>
>> Other architectures (frv, xtensa, score) currently fail to build for other reasons,
>> so I don't really know if the above are the only ones affected.
> 
> I think the best solution would be to add them to kernel/sys_ni.c so we can
> assign syscall numbers for all architectures regardless of whether they use it
> or not (most of them will never use these AFAICT).

Urg, sorry about the breakage.  I'll see if I can reproduce it and get a
patch out later today that does this.

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Dave Hansen]

[-- Attachment #1: Type: text/plain, Size: 860 bytes --]

On 09/12/2016 07:55 AM, Guenter Roeck wrote:
> Building h8300:allnoconfig ... failed
> Building h8300:h8300h-sim_defconfig ... failed
> Building c6x:dsk6455_defconfig ... failed
> 
> [ and more builds for the same architectures ]
> 
> --------------
> Error log:
> arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to
> `sys_pkey_mprotect'
> arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to
> `sys_pkey_alloc'
> arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to
> `sys_pkey_free'

So, I've done what Arnd suggested, but I don't have a working
cross-compiler than can do h8300 or c6x.  Could you test the attached
patch and see if it gets things working for you?

Also, do you have a handy place to grab a cross-compiler for h8300
without having to build one?  The kernel.org ones don't seem to work for
h8300.


[-- Attachment #2: fix-pkeys-arch-breakage.patch --]
[-- Type: text/x-patch, Size: 1047 bytes --]


Guenter Roeck reported breakage on the h8300 and c6x architectures
(among others) caused by the new memory protection keys syscalls.
This patch does what Arnd suggested and adds them to
kernel/sys_ni.c.

I don't have a cross-compiler for either of these architectures
handy, and kernel.org doesn't provide one either, so I have not
built this patch.  Testing would be appreciated.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: "open list:ABI/API" <linux-api@vger.kernel.org>
Cc: linux-arch@vger.kernel.org

---

 b/kernel/sys_ni.c |    5 +++++
 1 file changed, 5 insertions(+)

diff -puN kernel/sys_ni.c~fix-pkeys-arch-breakage kernel/sys_ni.c
--- a/kernel/sys_ni.c~fix-pkeys-arch-breakage	2016-09-12 12:50:47.801607239 -0700
+++ b/kernel/sys_ni.c	2016-09-12 12:50:47.804607374 -0700
@@ -250,3 +250,8 @@ cond_syscall(sys_execveat);
 
 /* membarrier */
 cond_syscall(sys_membarrier);
+
+/* memory protection keys */
+cond_syscall(sys_pkey_mprotect);
+cond_syscall(sys_pkey_alloc);
+cond_syscall(sys_pkey_free);
_

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Arnd Bergmann]

On Monday, September 12, 2016 1:09:17 PM CEST Dave Hansen wrote:
> 
> So, I've done what Arnd suggested, but I don't have a working
> cross-compiler than can do h8300 or c6x.  Could you test the attached
> patch and see if it gets things working for you?
> 
> Also, do you have a handy place to grab a cross-compiler for h8300
> without having to build one?  The kernel.org ones don't seem to work for
> h8300.
> 

I haven't tested it, but it looks good to me

Acked-by: Arnd Bergmann <arnd@arndb.de>

FWIW, you should be able to test this with an arm64 cross-compiler
too, many distros ship one.

	Arnd

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Arnd Bergmann]

On Monday, September 12, 2016 1:09:17 PM CEST Dave Hansen wrote:
> 
> So, I've done what Arnd suggested, but I don't have a working
> cross-compiler than can do h8300 or c6x.  Could you test the attached
> patch and see if it gets things working for you?
> 
> Also, do you have a handy place to grab a cross-compiler for h8300
> without having to build one?  The kernel.org ones don't seem to work for
> h8300.
> 

I haven't tested it, but it looks good to me

Acked-by: Arnd Bergmann <arnd-r2nGTMty4D4@public.gmane.org>

FWIW, you should be able to test this with an arm64 cross-compiler
too, many distros ship one.

	Arnd

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Guenter Roeck]

Hi Dave,

On Mon, Sep 12, 2016 at 01:09:17PM -0700, Dave Hansen wrote:
> On 09/12/2016 07:55 AM, Guenter Roeck wrote:
> > Building h8300:allnoconfig ... failed
> > Building h8300:h8300h-sim_defconfig ... failed
> > Building c6x:dsk6455_defconfig ... failed
> > 
> > [ and more builds for the same architectures ]
> > 
> > --------------
> > Error log:
> > arch/h8300/kernel/built-in.o:(.data+0x480): undefined reference to
> > `sys_pkey_mprotect'
> > arch/h8300/kernel/built-in.o:(.data+0x484): undefined reference to
> > `sys_pkey_alloc'
> > arch/h8300/kernel/built-in.o:(.data+0x488): undefined reference to
> > `sys_pkey_free'
> 
> So, I've done what Arnd suggested, but I don't have a working
> cross-compiler than can do h8300 or c6x.  Could you test the attached
> patch and see if it gets things working for you?
> 
Works for both h8300 and c6x. Feel free to add

Tested-by: Guenter Roeck <linux@roeck-us.net>

> Also, do you have a handy place to grab a cross-compiler for h8300
> without having to build one?  The kernel.org ones don't seem to work for
> h8300.
> 
Try http://server.roeck-us.net/toolchains/. Should be position independent.

Guenter

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Guenter Roeck]

On Mon, Sep 12, 2016 at 10:28:56PM +0200, Arnd Bergmann wrote:
> On Monday, September 12, 2016 1:09:17 PM CEST Dave Hansen wrote:
> > 
> > So, I've done what Arnd suggested, but I don't have a working
> > cross-compiler than can do h8300 or c6x.  Could you test the attached
> > patch and see if it gets things working for you?
> > 
> > Also, do you have a handy place to grab a cross-compiler for h8300
> > without having to build one?  The kernel.org ones don't seem to work for
> > h8300.
> > 
> 
> I haven't tested it, but it looks good to me
> 
> Acked-by: Arnd Bergmann <arnd@arndb.de>
> 
> FWIW, you should be able to test this with an arm64 cross-compiler
> too, many distros ship one.
> 
arm64 builds fine, though, even w/o this patch (at least for me).

Guenter

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Dave Hansen]

On 09/12/2016 01:28 PM, Arnd Bergmann wrote:
> FWIW, you should be able to test this with an arm64 cross-compiler
> too, many distros ship one.

arm64 build fine for me with and without the patch.  There must be
something else specific to h8300 and the other broken architectures.

Is there some other part of the configuration that could be hiding the
issue in my arm64 build?

Re: Build failures in -next: sys_pkey_mprotect etc. undefined

[Dave Hansen]

On 09/12/2016 01:28 PM, Arnd Bergmann wrote:
> FWIW, you should be able to test this with an arm64 cross-compiler
> too, many distros ship one.

arm64 build fine for me with and without the patch.  There must be
something else specific to h8300 and the other broken architectures.

Is there some other part of the configuration that could be hiding the
issue in my arm64 build?