Warning with next-20210121 on s390

Warning with next-20210121 on s390

[Sven Schnelle]

there's a warning with linux-next on s390:

[   33.893818] systemd-xdg-autostart-generator[544]: Not generating service for XDG autostart app-at\x2dspi\x2ddbus\x2dbus-autostart.service, startup phases are not supported. 
[   54.613344] ------------[ cut here ]------------ 
[   54.613702] Late current task rescheduling may be lost 
[   54.613709] WARNING: CPU: 0 PID: 574 at linux/kernel/sched/core.c:628 sched_resched_local_assert_allowed+0x86/0x90 
[   54.613714] Modules linked in: 
[   54.613717] CPU: 0 PID: 574 Comm: qemu-system-s39 Not tainted 5.11.0-rc4-next-20210121 #2601 
[   54.613721] Hardware name: IBM 3906 M04 704 (z/VM 7.1.0) 
[   54.613724] Krnl PSW : 0404e00180000000 0000000000d86bea (sched_resched_local_assert_allowed+0x8a/0x90) 
[   54.613730]            R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:3 CC:2 PM:0 RI:0 EA:3 

This is because of a non-obvious merge conflict between 56e62a737028
("s390: convert to generic entry") and 845f44e8ef28 ("sched: Report local
wake up on resched blind zone within idle loop"). Can you include the attached patch
into linux-next?

Thanks,
Sven

[PATCH] s390: allow reschedule on syscall restart

[Sven Schnelle]

Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
within idle loop") from next-20210121 causes a warning because s390
doesn't call sched_resched_local_allow() when restarting a syscall.

Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
---
 arch/s390/kernel/syscall.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
index bc8e650e377d..2b39ac40f970 100644
--- a/arch/s390/kernel/syscall.c
+++ b/arch/s390/kernel/syscall.c
@@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
 		do_syscall(regs);
 		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
 			break;
+		sched_resched_local_allow();
 		local_irq_enable();
 	}
 	exit_to_user_mode();
-- 
2.17.1

Re: [PATCH] s390: allow reschedule on syscall restart

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1130 bytes --]

Hi all,

On Thu, 21 Jan 2021 15:39:26 +0100 Sven Schnelle <svens@linux.ibm.com> wrote:
>
> Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
> within idle loop") from next-20210121 causes a warning because s390
> doesn't call sched_resched_local_allow() when restarting a syscall.
> 
> Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> ---
>  arch/s390/kernel/syscall.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
> index bc8e650e377d..2b39ac40f970 100644
> --- a/arch/s390/kernel/syscall.c
> +++ b/arch/s390/kernel/syscall.c
> @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
>  		do_syscall(regs);
>  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
>  			break;
> +		sched_resched_local_allow();
>  		local_irq_enable();
>  	}
>  	exit_to_user_mode();

I add that today as a merge fixup patch to the merge of the rcu tree
(which contains commit 845f44e8ef28 ("sched: Report local wake up on
resched blind zone within idle loop") ).

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] s390: allow reschedule on syscall restart

[Paul E. McKenney]

On Fri, Jan 22, 2021 at 08:32:49AM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> On Thu, 21 Jan 2021 15:39:26 +0100 Sven Schnelle <svens@linux.ibm.com> wrote:
> >
> > Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
> > within idle loop") from next-20210121 causes a warning because s390
> > doesn't call sched_resched_local_allow() when restarting a syscall.
> > 
> > Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> > ---
> >  arch/s390/kernel/syscall.c | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
> > index bc8e650e377d..2b39ac40f970 100644
> > --- a/arch/s390/kernel/syscall.c
> > +++ b/arch/s390/kernel/syscall.c
> > @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
> >  		do_syscall(regs);
> >  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
> >  			break;
> > +		sched_resched_local_allow();
> >  		local_irq_enable();
> >  	}
> >  	exit_to_user_mode();
> 
> I add that today as a merge fixup patch to the merge of the rcu tree
> (which contains commit 845f44e8ef28 ("sched: Report local wake up on
> resched blind zone within idle loop") ).

If Frederic has no objections, I will fold this in with attribution.

Either way, good catch!

							Thanx, Paul

Re: [PATCH] s390: allow reschedule on syscall restart

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 448 bytes --]

Hi all,

On Fri, 22 Jan 2021 08:32:49 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> I add that today as a merge fixup patch to the merge of the rcu tree
> (which contains commit 845f44e8ef28 ("sched: Report local wake up on
> resched blind zone within idle loop") ).

That is now commit

  0932934fd95e ("sched: Report local wake up on resched blind zone within idle loop")

in the rcu tree.
-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] s390: allow reschedule on syscall restart

[Christian Borntraeger]

On 21.01.21 15:39, Sven Schnelle wrote:
> Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
> within idle loop") from next-20210121 causes a warning because s390
> doesn't call sched_resched_local_allow() when restarting a syscall.
> 
> Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> ---
>  arch/s390/kernel/syscall.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
> index bc8e650e377d..2b39ac40f970 100644
> --- a/arch/s390/kernel/syscall.c
> +++ b/arch/s390/kernel/syscall.c
> @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
>  		do_syscall(regs);
>  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
>  			break;
> +		sched_resched_local_allow();
>  		local_irq_enable();
>  	}
>  	exit_to_user_mode();

Yesterdays next now fails with


arch/s390/kernel/syscall.c: In function '__do_syscall':
arch/s390/kernel/syscall.c:165:3: error: implicit declaration of function 'sched_resched_local_allow' [-Werror=implicit-function-declaration]
  165 |   sched_resched_local_allow();
      |   ^~~~~~~~~~~~~~~~~~~~~~~~~
cc1: some warnings being treated as errors
make[2]: *** [scripts/Makefile.build:288: arch/s390/kernel/syscall.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [scripts/Makefile.build:530: arch/s390/kernel] Error 2
make[1]: *** Waiting for unfinished jobs....

Re: [PATCH] s390: allow reschedule on syscall restart

[Sven Schnelle]

Christian Borntraeger <borntraeger@de.ibm.com> writes:

> On 21.01.21 15:39, Sven Schnelle wrote:
>> Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
>> within idle loop") from next-20210121 causes a warning because s390
>> doesn't call sched_resched_local_allow() when restarting a syscall.
>> 
>> Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
>> ---
>>  arch/s390/kernel/syscall.c | 1 +
>>  1 file changed, 1 insertion(+)
>> 
>> diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
>> index bc8e650e377d..2b39ac40f970 100644
>> --- a/arch/s390/kernel/syscall.c
>> +++ b/arch/s390/kernel/syscall.c
>> @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
>>  		do_syscall(regs);
>>  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
>>  			break;
>> +		sched_resched_local_allow();
>>  		local_irq_enable();
>>  	}
>>  	exit_to_user_mode();
>
> Yesterdays next now fails with
>
>
> arch/s390/kernel/syscall.c: In function '__do_syscall':
> arch/s390/kernel/syscall.c:165:3: error: implicit declaration of function 'sched_resched_local_allow' [-Werror=implicit-function-declaration]
>   165 |   sched_resched_local_allow();
>       |   ^~~~~~~~~~~~~~~~~~~~~~~~~
> cc1: some warnings being treated as errors
> make[2]: *** [scripts/Makefile.build:288: arch/s390/kernel/syscall.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> make[1]: *** [scripts/Makefile.build:530: arch/s390/kernel] Error 2
> make[1]: *** Waiting for unfinished jobs....

Looks to me like 845f44e8ef28 ("sched: Report local wake up on resched
blind zone") was removed from linux-next. Stephen, can you remove my
commit as well? It is no longer needed.

Thanks
Sven

Re: [PATCH] s390: allow reschedule on syscall restart

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1787 bytes --]

Hi all,

On Tue, 26 Jan 2021 07:59:51 +0100 Christian Borntraeger <borntraeger@de.ibm.com> wrote:
>
> On 21.01.21 15:39, Sven Schnelle wrote:
> > Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
> > within idle loop") from next-20210121 causes a warning because s390
> > doesn't call sched_resched_local_allow() when restarting a syscall.
> > 
> > Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> > ---
> >  arch/s390/kernel/syscall.c | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
> > index bc8e650e377d..2b39ac40f970 100644
> > --- a/arch/s390/kernel/syscall.c
> > +++ b/arch/s390/kernel/syscall.c
> > @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
> >  		do_syscall(regs);
> >  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
> >  			break;
> > +		sched_resched_local_allow();
> >  		local_irq_enable();
> >  	}
> >  	exit_to_user_mode();  
> 
> Yesterdays next now fails with
> 
> 
> arch/s390/kernel/syscall.c: In function '__do_syscall':
> arch/s390/kernel/syscall.c:165:3: error: implicit declaration of function 'sched_resched_local_allow' [-Werror=implicit-function-declaration]
>   165 |   sched_resched_local_allow();
>       |   ^~~~~~~~~~~~~~~~~~~~~~~~~
> cc1: some warnings being treated as errors
> make[2]: *** [scripts/Makefile.build:288: arch/s390/kernel/syscall.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> make[1]: *** [scripts/Makefile.build:530: arch/s390/kernel] Error 2
> make[1]: *** Waiting for unfinished jobs....

I have now removed the merge fix up for tomorrow.  It seems that the
commits that needed it have been removed :-(

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] s390: allow reschedule on syscall restart

[Paul E. McKenney]

On Tue, Jan 26, 2021 at 09:25:53PM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> On Tue, 26 Jan 2021 07:59:51 +0100 Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> >
> > On 21.01.21 15:39, Sven Schnelle wrote:
> > > Commit 845f44e8ef28 ("sched: Report local wake up on resched blind zone
> > > within idle loop") from next-20210121 causes a warning because s390
> > > doesn't call sched_resched_local_allow() when restarting a syscall.
> > > 
> > > Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> > > ---
> > >  arch/s390/kernel/syscall.c | 1 +
> > >  1 file changed, 1 insertion(+)
> > > 
> > > diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c
> > > index bc8e650e377d..2b39ac40f970 100644
> > > --- a/arch/s390/kernel/syscall.c
> > > +++ b/arch/s390/kernel/syscall.c
> > > @@ -162,6 +162,7 @@ void noinstr __do_syscall(struct pt_regs *regs, int per_trap)
> > >  		do_syscall(regs);
> > >  		if (!test_pt_regs_flag(regs, PIF_SYSCALL_RESTART))
> > >  			break;
> > > +		sched_resched_local_allow();
> > >  		local_irq_enable();
> > >  	}
> > >  	exit_to_user_mode();  
> > 
> > Yesterdays next now fails with
> > 
> > 
> > arch/s390/kernel/syscall.c: In function '__do_syscall':
> > arch/s390/kernel/syscall.c:165:3: error: implicit declaration of function 'sched_resched_local_allow' [-Werror=implicit-function-declaration]
> >   165 |   sched_resched_local_allow();
> >       |   ^~~~~~~~~~~~~~~~~~~~~~~~~
> > cc1: some warnings being treated as errors
> > make[2]: *** [scripts/Makefile.build:288: arch/s390/kernel/syscall.o] Error 1
> > make[2]: *** Waiting for unfinished jobs....
> > make[1]: *** [scripts/Makefile.build:530: arch/s390/kernel] Error 2
> > make[1]: *** Waiting for unfinished jobs....
> 
> I have now removed the merge fix up for tomorrow.  It seems that the
> commits that needed it have been removed :-(

Review comments mean that they need rework, apologies for the hassle!

							Thanx, Paul