Build failure in v3.12-stable-queue

Build failure in v3.12-stable-queue

[Guenter Roeck]

powerpc:defconfig, powerpc:allmodconfig:

arch/powerpc/kernel/process.c: In function 'start_thread':
arch/powerpc/kernel/process.c:1104:3: error: implicit declaration of function 'tm_reclaim_current'

Caused by "powerpc/tm: Always reclaim in start_thread() for exec() class syscalls".

This would either require d31626f70b610 ("powerpc: Don't corrupt transactional state
when using FP/VMX in kernel"), which introduces the function, or you could call
tm_reclaim() directly instead (which may be less risky).

Guenter

Re: Build failure in v3.12-stable-queue

[Jiri Slaby]

On 07/12/2016, 03:30 PM, Guenter Roeck wrote:
> powerpc:defconfig, powerpc:allmodconfig:
> 
> arch/powerpc/kernel/process.c: In function 'start_thread':
> arch/powerpc/kernel/process.c:1104:3: error: implicit declaration of
> function 'tm_reclaim_current'
> 
> Caused by "powerpc/tm: Always reclaim in start_thread() for exec() class
> syscalls".
> 
> This would either require d31626f70b610 ("powerpc: Don't corrupt
> transactional state
> when using FP/VMX in kernel"), which introduces the function, or you
> could call
> tm_reclaim() directly instead (which may be less risky).

Oh, I see.

Given d31626f70b610 is intrusive and I have no idea how to call
tm_reclaim directly and if it suffices at all, I have dropped the patch
until somebody can provide me with a backport :).

thanks,
-- 
js
suse labs

Re: Build failure in v3.12-stable-queue

[Guenter Roeck]

Hi Jiri,

On Tue, Jul 12, 2016 at 03:51:49PM +0200, Jiri Slaby wrote:
> On 07/12/2016, 03:30 PM, Guenter Roeck wrote:
> > powerpc:defconfig, powerpc:allmodconfig:
> > 
> > arch/powerpc/kernel/process.c: In function 'start_thread':
> > arch/powerpc/kernel/process.c:1104:3: error: implicit declaration of
> > function 'tm_reclaim_current'
> > 
> > Caused by "powerpc/tm: Always reclaim in start_thread() for exec() class
> > syscalls".
> > 
> > This would either require d31626f70b610 ("powerpc: Don't corrupt
> > transactional state
> > when using FP/VMX in kernel"), which introduces the function, or you
> > could call
> > tm_reclaim() directly instead (which may be less risky).
> 
> Oh, I see.
> 
> Given d31626f70b610 is intrusive and I have no idea how to call
> tm_reclaim directly and if it suffices at all, I have dropped the patch
> until somebody can provide me with a backport :).
> 

There are now more failures:

arm:allmodconfig:
i386:allmodconfig:
mips:allmodconfig:
[and more]

ERROR: "ktime_divns" [sound/core/snd-hrtimer.ko] undefined!

Caused by 'ALSA: hrtimer: Handle start/stop more properly'.
ktime_divns() isn't exported in 3.12.

Guenter

Re: Build failure in v3.12-stable-queue

[Michal Kubecek]

On Tue, Jul 12, 2016 at 02:23:05PM -0700, Guenter Roeck wrote:
> 
> There are now more failures:
> 
> arm:allmodconfig:
> i386:allmodconfig:
> mips:allmodconfig:
> [and more]
> 
> ERROR: "ktime_divns" [sound/core/snd-hrtimer.ko] undefined!
> 
> Caused by 'ALSA: hrtimer: Handle start/stop more properly'.
> ktime_divns() isn't exported in 3.12.

The symbol is exported since

   166afb64511e ("ktime: Sanitize ktime_to_us/ms conversion")

(i.e. v3.17-rc1) but it makes no sense to pick the whole commit, just
making ktime_divns exported is enough to fix the 32-bit builds.

Michal Kubecek

Re: Build failure in v3.12-stable-queue

[Jiri Slaby]

On 07/13/2016, 07:12 AM, Michal Kubecek wrote:
> On Tue, Jul 12, 2016 at 02:23:05PM -0700, Guenter Roeck wrote:
>>
>> There are now more failures:
>>
>> arm:allmodconfig:
>> i386:allmodconfig:
>> mips:allmodconfig:
>> [and more]
>>
>> ERROR: "ktime_divns" [sound/core/snd-hrtimer.ko] undefined!
>>
>> Caused by 'ALSA: hrtimer: Handle start/stop more properly'.
>> ktime_divns() isn't exported in 3.12.
> 
> The symbol is exported since
> 
>    166afb64511e ("ktime: Sanitize ktime_to_us/ms conversion")
> 
> (i.e. v3.17-rc1) but it makes no sense to pick the whole commit, just
> making ktime_divns exported is enough to fix the 32-bit builds.

Should be fixed now. Thanks!

-- 
js
suse labs

build failure in v3.12-stable-queue

[Guenter Roeck]

Hi Jiri,

I see the following build failure in v3.12-stable-queue.

Building mips:cavium_octeon_defconfig ... failed
--------------
Error log:
arch/mips/mm/tlbex.c: In function 'build_huge_handler_tail':
arch/mips/mm/tlbex.c:782:6: error: 'cpu_has_ftlb' undeclared (first use in this function)
  if (cpu_has_ftlb && flush) {
      ^
arch/mips/mm/tlbex.c:782:6: note: each undeclared identifier is reported only once for each function it appears in
In file included from include/linux/bug.h:4:0,
                 from arch/mips/mm/tlbex.c:24:
arch/mips/mm/tlbex.c:783:11: error: 'cpu_has_tlbinv' undeclared (first use in this function)
   BUG_ON(!cpu_has_tlbinv);
           ^
arch/mips/include/asm/bug.h:33:44: note: in definition of macro 'BUG_ON'
 #define BUG_ON(C) __BUG_ON((unsigned long)(C))
                                            ^
arch/mips/mm/tlbex.c:786:27: error: 'MIPS_ENTRYHI_EHINV' undeclared (first use in this function)
   uasm_i_ori(p, ptr, ptr, MIPS_ENTRYHI_EHINV);

Let me know if I need to bisect.

Thanks,
Guenter

Re: build failure in v3.12-stable-queue

[Jiri Slaby]

On 04/28/2017, 08:06 PM, Guenter Roeck wrote:
> Hi Jiri,
> 
> I see the following build failure in v3.12-stable-queue.
> 
> Building mips:cavium_octeon_defconfig ... failed

Thanks for letting me know. The current tree should have this fixed.


-- 
js
suse labs

Re: build failure in v3.12-stable-queue

[Guenter Roeck]

On Fri, Apr 28, 2017 at 09:04:50PM +0200, Jiri Slaby wrote:
> On 04/28/2017, 08:06 PM, Guenter Roeck wrote:
> > Hi Jiri,
> > 
> > I see the following build failure in v3.12-stable-queue.
> > 
> > Building mips:cavium_octeon_defconfig ... failed
> 
> Thanks for letting me know. The current tree should have this fixed.
> 
Confirmed.

Thanks,
Guenter