Re: 32bit resctrl? (was Re: [PATCH v2] fs/resctrl: fix domid loss precision issue)

[Moger, Babu]

On 3/14/2024 10:25 AM, Reinette Chatre wrote:
> +x86 maintainers, Tony, Babu, Peter
>
> Hi Everybody,
>
> On 3/12/2024 12:53 AM, Rex Nie wrote:
> > Below statement from mkdir_mondata_subdir function will loss precision,
> > because it assigns int to 14 bits bitfield.
> > 	priv.u.domid = d->id;
> >
> > On some platforms(e.g.,x86), the max cache_id is the amount of L3 caches,
> > so it is not in the range of 0x3fff. But some platforms use higher
> > cache_id, e.g., arm uses cache_id as locator for cache MSC. This will
> > cause below issue if cache_id > 0x3fff likes:
> > /sys/fs/resctrl/mon_groups/p1/mon_data/mon_L3_1048564 # cat llc_occupancy
> > cat: read error: No such file or directory
> >
> > This is the call trace when cat llc_occupancy:
> > rdtgroup_mondata_show()
> > 	domid = md.u.domid
> > 	d = resctrl_arch_find_domain(r, domid)
> >
> > d is null here because of lossing precision
> >
> > Signed-off-by: Rex Nie <rex.nie@xxxxxxxxxxxxxxx>
> > ---
> >   fs/resctrl/internal.h | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/fs/resctrl/internal.h b/fs/resctrl/internal.h
> > index 7a6f46b4edd0..096317610949 100644
> > --- a/fs/resctrl/internal.h
> > +++ b/fs/resctrl/internal.h
> > @@ -94,7 +94,7 @@ union mon_data_bits {
> >   	struct {
> >   		unsigned int rid		: 10;
> >   		enum resctrl_event_id evtid	: 8;
> > -		unsigned int domid		: 14;
> > +		u32				domid;
> >   	} u;
> >   };
> >   
> resctrl currently supports 32bit builds. Fixing this issue* in this way

I have never bothered about 32bit builds.   Is Intel still testing 32bit 
builds?


> would first require that resctrl (the architecture independent fs part)
> depend on X86_64. Is this a change that everybody will be comfortable with?
>
> (Of course, there are other solutions available to address the issue mentioned
> in this patch that do not require depending on X86_64, but I would like
> to take this moment to understand the sentiment surrounding continuing support
> for 32bit resctrl.)

I am thinking we have bigger problem here.

The structure pointer "union mon_data_bits priv;" is created in stack 
and passed to create mondata directory. We are reading it later again in 
rdtgroup_mondata_show.

How is this pointer valid again?  Shouldn't we use static pointer or 
allocate memory for the pointer?

thanks

Babu



> Thank you.
>
> Reinette
>
> * Please note that this is not an urgent fix but instead a preparatory change
>    for future Arm support.