Re: [PATCH v1 1/6] x86/resctrl: Move __resctrl_sched_in() out-of-line

[Peter Newman]

Hi Reinette,

On Thu, Apr 4, 2024 at 4:09 PM Reinette Chatre
<reinette.chatre@xxxxxxxxx> wrote:
>
> Hi Peter,
>
> On 3/25/2024 10:27 AM, Peter Newman wrote:
> > __resctrl_sched_in() is unable to dereference a struct rdtgroup pointer
> > when defined inline because rdtgroup is a private structure defined in
> > internal.h.
>
> Being inline has nothing to do with whether it can reference a struct rdtgroup
> pointer, no?

No, but it has a lot to do with whether it can de-reference a struct
rdtgroup pointer in order to obtain a CLOSID or RMID, as doing so
would pull the definitions for struct rdtgroup and struct mongroup
into an external header. Before doing so, I would want to make sure
implementing  __resctrl_sched_in() inline is actually adding value.

>
> >
> > This function is defined inline to avoid impacting context switch
> > performance for the majority of users who aren't using resctrl at all.
> > These benefits can already be realized without access to internal
> > resctrl data structures.
> >
> > The logic of performing an out-of-line call to __resctrl_sched_in() only
> > when resctrl is mounted is architecture-independent, so the inline
> > definition of resctrl_sched_in() can be moved into linux/resctrl.h.
> >
> > Signed-off-by: Peter Newman <peternewman@xxxxxxxxxx>
>
> ...
>
> > diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
> > index 0917c7f25720..8f92a87d381d 100644
> > --- a/arch/x86/kernel/process_32.c
> > +++ b/arch/x86/kernel/process_32.c
> > @@ -38,6 +38,7 @@
> >  #include <linux/io.h>
> >  #include <linux/kdebug.h>
> >  #include <linux/syscalls.h>
> > +#include <linux/resctrl.h>
> >
> >  #include <asm/ldt.h>
> >  #include <asm/processor.h>
> > @@ -51,7 +52,6 @@
> >  #include <asm/debugreg.h>
> >  #include <asm/switch_to.h>
> >  #include <asm/vm86.h>
> > -#include <asm/resctrl.h>
> >  #include <asm/proto.h>
> >
> >  #include "process.h"
> > diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
> > index 7062b84dd467..d442269bb25b 100644
> > --- a/arch/x86/kernel/process_64.c
> > +++ b/arch/x86/kernel/process_64.c
> > @@ -40,6 +40,7 @@
> >  #include <linux/ftrace.h>
> >  #include <linux/syscalls.h>
> >  #include <linux/iommu.h>
> > +#include <linux/resctrl.h>
> >
> >  #include <asm/processor.h>
> >  #include <asm/pkru.h>
>
> With a change like this we should be very careful about what is included when
> the kernel is not built with resctrl and in its current form linux/resctrl.h is
> not ready for this.
>
> If CONFIG_X86_CPU_RESCTRL is not set linux/resctrl.h should have the bare minimum,
> just like asm/resctrl.h has.
>
> > @@ -53,7 +54,6 @@
> >  #include <asm/switch_to.h>
> >  #include <asm/xen/hypervisor.h>
> >  #include <asm/vdso.h>
> > -#include <asm/resctrl.h>
> >  #include <asm/unistd.h>
> >  #include <asm/fsgsbase.h>
> >  #include <asm/fred.h>
> > diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h
> > index a365f67131ec..62d607939a73 100644
> > --- a/include/linux/resctrl.h
> > +++ b/include/linux/resctrl.h
> > @@ -304,4 +304,25 @@ void resctrl_arch_reset_rmid_all(struct rdt_resource *r, struct rdt_domain *d);
> >  extern unsigned int resctrl_rmid_realloc_threshold;
> >  extern unsigned int resctrl_rmid_realloc_limit;
> >
> > +DECLARE_STATIC_KEY_FALSE(rdt_enable_key);
> > +
> > +void __resctrl_sched_in(struct task_struct *tsk);
> > +
> > +/*
> > + * resctrl_sched_in() - Assigns the incoming task's control/monitor IDs to the
> > + *                   current CPU
> > + *
> > + * To minimize impact to the scheduler hot path, this will stay as no-op unless
> > + * running on a system supporting resctrl and the filesystem is mounted.
> > + *
> > + * Must be called with preemption disabled.
> > + */
> > +static inline void resctrl_sched_in(struct task_struct *tsk)
> > +{
> > +#ifdef CONFIG_X86_CPU_RESCTRL
> > +     if (static_branch_likely(&rdt_enable_key))
> > +             __resctrl_sched_in(tsk);
> > +#endif
> > +}
> > +
>
> include/linux/resctrl.h should rather be divided to accommodate code
> as below:
>
> #ifdef CONFIG_X86_CPU_RESCTRL
>
> static inline void resctrl_sched_in(struct task_struct *tsk)
> {
>         if (static_branch_likely(&rdt_enable_key))
>                 __resctrl_sched_in(tsk);
> }
>
> #else
>
> static inline void resctrl_sched_in(struct task_struct *tsk) {}
>
> #endif
>
> so that core code does not get anything unnecessary when CONFIG_X86_CPU_RESCTRL
> is not set.

Will do.

Thanks!
-Peter