diff mbox series

[v8,05/16] sched/core: Allow sched_setattr() to use the current policy

Message ID 20190402104153.25404-6-patrick.bellasi@arm.com (mailing list archive)
State Not Applicable, archived
Headers show
Series Add utilization clamping support | expand

Commit Message

Patrick Bellasi April 2, 2019, 10:41 a.m. UTC 
The sched_setattr() syscall mandates that a policy is always specified.
This requires to always know which policy a task will have when
attributes are configured and this makes it impossible to add more
generic task attributes valid across different scheduling policies.
Reading the policy before setting generic tasks attributes is racy since
we cannot be sure it is not changed concurrently.

Introduce the required support to change generic task attributes without
affecting the current task policy. This is done by adding an attribute flag
(SCHED_FLAG_KEEP_POLICY) to enforce the usage of the current policy.

Add support for the SETPARAM_POLICY policy, which is already used by the
sched_setparam() POSIX syscall, to the sched_setattr() non-POSIX
syscall.

Signed-off-by: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
---
 include/uapi/linux/sched.h |  6 +++++-
 kernel/sched/core.c        | 11 ++++++++++-
 2 files changed, 15 insertions(+), 2 deletions(-)

Comments

Peter Zijlstra May 8, 2019, 7:21 p.m. UTC | #1 
On Tue, Apr 02, 2019 at 11:41:41AM +0100, Patrick Bellasi wrote:
> diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
> index 22627f80063e..075c610adf45 100644
> --- a/include/uapi/linux/sched.h
> +++ b/include/uapi/linux/sched.h
> @@ -40,6 +40,8 @@
>  /* SCHED_ISO: reserved but not implemented yet */
>  #define SCHED_IDLE		5
>  #define SCHED_DEADLINE		6
> +/* Must be the last entry: used to sanity check attr.policy values */
> +#define SCHED_POLICY_MAX	SCHED_DEADLINE

This is a wee bit sad to put in a uapi header; but yeah, where else :/

Another option would be something like:

enum {
	SCHED_NORMAL = 0,
	SCHED_FIFO = 1,
	SCHED_RR = 2,
	SCHED_BATCH = 3,
	/* SCHED_ISO = 4, reserved */
	SCHED_IDLE = 5,
	SCHED_DEADLINE = 6,
	SCHED_POLICY_NR
};

>  /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
>  #define SCHED_RESET_ON_FORK     0x40000000
> @@ -50,9 +52,11 @@
>  #define SCHED_FLAG_RESET_ON_FORK	0x01
>  #define SCHED_FLAG_RECLAIM		0x02
>  #define SCHED_FLAG_DL_OVERRUN		0x04
> +#define SCHED_FLAG_KEEP_POLICY		0x08
>  
>  #define SCHED_FLAG_ALL	(SCHED_FLAG_RESET_ON_FORK	| \
>  			 SCHED_FLAG_RECLAIM		| \
> -			 SCHED_FLAG_DL_OVERRUN)
> +			 SCHED_FLAG_DL_OVERRUN		| \
> +			 SCHED_FLAG_KEEP_POLICY)
>  
>  #endif /* _UAPI_LINUX_SCHED_H */
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index d368ac26b8aa..20efb32e1a7e 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -4907,8 +4907,17 @@ SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
>  	if (retval)
>  		return retval;
>  
> -	if ((int)attr.sched_policy < 0)
> +	/*
> +	 * A valid policy is always required from userspace, unless
> +	 * SCHED_FLAG_KEEP_POLICY is set and the current policy
> +	 * is enforced for this call.
> +	 */
> +	if (attr.sched_policy > SCHED_POLICY_MAX &&
> +	    !(attr.sched_flags & SCHED_FLAG_KEEP_POLICY)) {
>  		return -EINVAL;
> +	}

And given I just looked at those darn SCHED_* things, I now note the
above does 'funny' things when passed: attr.policy=4.

> +	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
> +		attr.sched_policy = SETPARAM_POLICY;
>  
>  	rcu_read_lock();
>  	retval = -ESRCH;
Patrick Bellasi May 9, 2019, 9:18 a.m. UTC | #2 
On 08-May 21:21, Peter Zijlstra wrote:
> On Tue, Apr 02, 2019 at 11:41:41AM +0100, Patrick Bellasi wrote:
> > diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
> > index 22627f80063e..075c610adf45 100644
> > --- a/include/uapi/linux/sched.h
> > +++ b/include/uapi/linux/sched.h
> > @@ -40,6 +40,8 @@
> >  /* SCHED_ISO: reserved but not implemented yet */
> >  #define SCHED_IDLE		5
> >  #define SCHED_DEADLINE		6
> > +/* Must be the last entry: used to sanity check attr.policy values */
> > +#define SCHED_POLICY_MAX	SCHED_DEADLINE
> 
> This is a wee bit sad to put in a uapi header; but yeah, where else :/
> 
> Another option would be something like:
> 
> enum {
> 	SCHED_NORMAL = 0,
> 	SCHED_FIFO = 1,
> 	SCHED_RR = 2,
> 	SCHED_BATCH = 3,
> 	/* SCHED_ISO = 4, reserved */
> 	SCHED_IDLE = 5,
> 	SCHED_DEADLINE = 6,
> 	SCHED_POLICY_NR
> };

I just wanted to minimize the changes by keeping the same structure...
If you prefer the above I can add a refactoring patch just to update
existing definitions before adding this patch...

> 
> >  /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
> >  #define SCHED_RESET_ON_FORK     0x40000000
> > @@ -50,9 +52,11 @@
> >  #define SCHED_FLAG_RESET_ON_FORK	0x01
> >  #define SCHED_FLAG_RECLAIM		0x02
> >  #define SCHED_FLAG_DL_OVERRUN		0x04
> > +#define SCHED_FLAG_KEEP_POLICY		0x08
> >  
> >  #define SCHED_FLAG_ALL	(SCHED_FLAG_RESET_ON_FORK	| \
> >  			 SCHED_FLAG_RECLAIM		| \
> > -			 SCHED_FLAG_DL_OVERRUN)
> > +			 SCHED_FLAG_DL_OVERRUN		| \
> > +			 SCHED_FLAG_KEEP_POLICY)
> >  
> >  #endif /* _UAPI_LINUX_SCHED_H */
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index d368ac26b8aa..20efb32e1a7e 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -4907,8 +4907,17 @@ SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
> >  	if (retval)
> >  		return retval;
> >  
> > -	if ((int)attr.sched_policy < 0)
> > +	/*
> > +	 * A valid policy is always required from userspace, unless
> > +	 * SCHED_FLAG_KEEP_POLICY is set and the current policy
> > +	 * is enforced for this call.
> > +	 */
> > +	if (attr.sched_policy > SCHED_POLICY_MAX &&
> > +	    !(attr.sched_flags & SCHED_FLAG_KEEP_POLICY)) {
> >  		return -EINVAL;
> > +	}
> 
> And given I just looked at those darn SCHED_* things, I now note the
> above does 'funny' things when passed: attr.policy=4.

... and maybe factor in the same refactoring patch a check on
SCHED_ISO being not yet supported.

> 
> > +	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
> > +		attr.sched_policy = SETPARAM_POLICY;
> >  
> >  	rcu_read_lock();
> >  	retval = -ESRCH;
Peter Zijlstra May 9, 2019, 11:55 a.m. UTC | #3 
On Thu, May 09, 2019 at 10:18:07AM +0100, Patrick Bellasi wrote:
> On 08-May 21:21, Peter Zijlstra wrote:
> > On Tue, Apr 02, 2019 at 11:41:41AM +0100, Patrick Bellasi wrote:
> > > diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
> > > index 22627f80063e..075c610adf45 100644
> > > --- a/include/uapi/linux/sched.h
> > > +++ b/include/uapi/linux/sched.h
> > > @@ -40,6 +40,8 @@
> > >  /* SCHED_ISO: reserved but not implemented yet */
> > >  #define SCHED_IDLE		5
> > >  #define SCHED_DEADLINE		6
> > > +/* Must be the last entry: used to sanity check attr.policy values */
> > > +#define SCHED_POLICY_MAX	SCHED_DEADLINE
> > 
> > This is a wee bit sad to put in a uapi header; but yeah, where else :/
> > 
> > Another option would be something like:
> > 
> > enum {
> > 	SCHED_NORMAL = 0,
> > 	SCHED_FIFO = 1,
> > 	SCHED_RR = 2,
> > 	SCHED_BATCH = 3,
> > 	/* SCHED_ISO = 4, reserved */
> > 	SCHED_IDLE = 5,
> > 	SCHED_DEADLINE = 6,
> > 	SCHED_POLICY_NR
> > };
> 
> I just wanted to minimize the changes by keeping the same structure...
> If you prefer the above I can add a refactoring patch just to update
> existing definitions before adding this patch...

Right; I've no idea really. The thing that started all this was adding
that define to UAPI. Maybe we can do without it and instead put in a
comment to check sched_setattr() any time we add a new policy and just
hard code the thing.
Patrick Bellasi May 9, 2019, 2:59 p.m. UTC | #4 
On 08-May 21:21, Peter Zijlstra wrote:
> On Tue, Apr 02, 2019 at 11:41:41AM +0100, Patrick Bellasi wrote:
> > diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
> > index 22627f80063e..075c610adf45 100644
> > --- a/include/uapi/linux/sched.h
> > +++ b/include/uapi/linux/sched.h
> > @@ -40,6 +40,8 @@
> >  /* SCHED_ISO: reserved but not implemented yet */
> >  #define SCHED_IDLE		5
> >  #define SCHED_DEADLINE		6
> > +/* Must be the last entry: used to sanity check attr.policy values */
> > +#define SCHED_POLICY_MAX	SCHED_DEADLINE
> 
> This is a wee bit sad to put in a uapi header; but yeah, where else :/
> 
> Another option would be something like:
> 
> enum {
> 	SCHED_NORMAL = 0,
> 	SCHED_FIFO = 1,
> 	SCHED_RR = 2,
> 	SCHED_BATCH = 3,
> 	/* SCHED_ISO = 4, reserved */
> 	SCHED_IDLE = 5,
> 	SCHED_DEADLINE = 6,
> 	SCHED_POLICY_NR
> };
> 
> >  /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
> >  #define SCHED_RESET_ON_FORK     0x40000000
> > @@ -50,9 +52,11 @@
> >  #define SCHED_FLAG_RESET_ON_FORK	0x01
> >  #define SCHED_FLAG_RECLAIM		0x02
> >  #define SCHED_FLAG_DL_OVERRUN		0x04
> > +#define SCHED_FLAG_KEEP_POLICY		0x08
> >  
> >  #define SCHED_FLAG_ALL	(SCHED_FLAG_RESET_ON_FORK	| \
> >  			 SCHED_FLAG_RECLAIM		| \
> > -			 SCHED_FLAG_DL_OVERRUN)
> > +			 SCHED_FLAG_DL_OVERRUN		| \
> > +			 SCHED_FLAG_KEEP_POLICY)
> >  
> >  #endif /* _UAPI_LINUX_SCHED_H */
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index d368ac26b8aa..20efb32e1a7e 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -4907,8 +4907,17 @@ SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
> >  	if (retval)
> >  		return retval;
> >  
> > -	if ((int)attr.sched_policy < 0)
> > +	/*
> > +	 * A valid policy is always required from userspace, unless
> > +	 * SCHED_FLAG_KEEP_POLICY is set and the current policy
> > +	 * is enforced for this call.
> > +	 */
> > +	if (attr.sched_policy > SCHED_POLICY_MAX &&
> > +	    !(attr.sched_flags & SCHED_FLAG_KEEP_POLICY)) {
> >  		return -EINVAL;
> > +	}
> 
> And given I just looked at those darn SCHED_* things, I now note the
> above does 'funny' things when passed: attr.policy=4.

Looking better at the code, I see now that we don't really need that
check anymore. Indeed, v8 introduced the support to change policy
specific and independent attributes at the same time. Thus:

1. the policy validity is already checked in:

     sched_setattr()
       sched_setattr()
         __sched_setscheduler()
            valid_policy()

   which knows how to deal with attr.policy=4 (i.e. -EINVAL)

2. when we pass in SCHED_FLAG_KEEP_POLICY we force the current policy
   by setting attr.sched_policy = SETPARAM_POLICY, so we just need a
   non negative policy being defined (usually 0 by default).

Thus, I'll remove the new #define and update the check above to be just:

	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
		attr.sched_policy = SETPARAM_POLICY;
	else if ((int)attr.sched_policy < 0)
		return -EINVAL;

which should cover the additional case:

   you can syscall with just SCHED_FLAG_KEEP_POLICY set if you want to
   change only cross-policy attributes.

> > +	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
> > +		attr.sched_policy = SETPARAM_POLICY;
> >  
> >  	rcu_read_lock();
> >  	retval = -ESRCH;
diff mbox series

Patch

diff --git a/include/uapi/linux/sched.h b/include/uapi/linux/sched.h
index 22627f80063e..075c610adf45 100644
--- a/include/uapi/linux/sched.h
+++ b/include/uapi/linux/sched.h
@@ -40,6 +40,8 @@ 
 /* SCHED_ISO: reserved but not implemented yet */
 #define SCHED_IDLE		5
 #define SCHED_DEADLINE		6
+/* Must be the last entry: used to sanity check attr.policy values */
+#define SCHED_POLICY_MAX	SCHED_DEADLINE
 
 /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */
 #define SCHED_RESET_ON_FORK     0x40000000
@@ -50,9 +52,11 @@ 
 #define SCHED_FLAG_RESET_ON_FORK	0x01
 #define SCHED_FLAG_RECLAIM		0x02
 #define SCHED_FLAG_DL_OVERRUN		0x04
+#define SCHED_FLAG_KEEP_POLICY		0x08
 
 #define SCHED_FLAG_ALL	(SCHED_FLAG_RESET_ON_FORK	| \
 			 SCHED_FLAG_RECLAIM		| \
-			 SCHED_FLAG_DL_OVERRUN)
+			 SCHED_FLAG_DL_OVERRUN		| \
+			 SCHED_FLAG_KEEP_POLICY)
 
 #endif /* _UAPI_LINUX_SCHED_H */
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index d368ac26b8aa..20efb32e1a7e 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4907,8 +4907,17 @@  SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
 	if (retval)
 		return retval;
 
-	if ((int)attr.sched_policy < 0)
+	/*
+	 * A valid policy is always required from userspace, unless
+	 * SCHED_FLAG_KEEP_POLICY is set and the current policy
+	 * is enforced for this call.
+	 */
+	if (attr.sched_policy > SCHED_POLICY_MAX &&
+	    !(attr.sched_flags & SCHED_FLAG_KEEP_POLICY)) {
 		return -EINVAL;
+	}
+	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
+		attr.sched_policy = SETPARAM_POLICY;
 
 	rcu_read_lock();
 	retval = -ESRCH;