Re: [PATCH 1/1] numa: check mem or memdev in numa configuration

[Li Zhang]

On 2/17/22 2:33 PM, Igor Mammedov wrote:
> On Thu, 17 Feb 2022 13:24:08 +0100
> Li Zhang <lizhang@suse.de> wrote:
>
> > On 2/17/22 11:25 AM, Igor Mammedov wrote:
> > > On Thu, 17 Feb 2022 10:38:32 +0100
> > > Li Zhang <lizhang@suse.de> wrote:
> > >    
> > > > On 2/17/22 10:10 AM, Igor Mammedov wrote:
> > > > > On Wed, 16 Feb 2022 17:36:13 +0100
> > > > > Li Zhang <lizhang@suse.de> wrote:
> > > > >       
> > > > > > If there is no mem or memdev in numa configuration, it always
> > > > > > reports the error as the following:
> > > > > >
> > > > > > total memory for NUMA nodes (0x0) should equal RAM size (0x100000000)
> > > > > >
> > > > > > This error is confusing and the reason is that total memory of numa nodes
> > > > > > is always 0 if there is no mem or memdev in numa configuration.
> > > > > > So it's better to check mem or memdev in numa configuration.
> > > > > >
> > > > > > Signed-off-by: Li Zhang <lizhang@suse.de>
> > > > > > ---
> > > > > >     hw/core/numa.c | 5 +++++
> > > > > >     1 file changed, 5 insertions(+)
> > > > > >
> > > > > > diff --git a/hw/core/numa.c b/hw/core/numa.c
> > > > > > index 1aa05dcf42..11cbec51eb 100644
> > > > > > --- a/hw/core/numa.c
> > > > > > +++ b/hw/core/numa.c
> > > > > > @@ -132,6 +132,11 @@ static void parse_numa_node(MachineState *ms, 
> > > > > > NumaNodeOptions *node,
> > > > > >     
> > > > > >         have_memdevs = have_memdevs ? : node->has_memdev;
> > > > > >         have_mem = have_mem ? : node->has_mem;
> > > > > > +    if (!node->has_memdev && !node->has_mem) {
> > > > > > +        error_setg(errp, "numa configuration should use mem= or memdev= ");
> > > > > > +        return;
> > > > > > +    }
> > > > >
> > > > > Wouldn't this breaks memory less numa nodes?
> > > >
> > > > Yes, you are right. It will break it if there more numa nodes
> > > > than memory, and the numa nodes have no memory backends specified.
> > > >
> > > > Is it allowed for users to specify more numa nodes than memory?
> > > yep, I think we support it at least for one of the targets
> > > (but I don't remember which one(s))
> > >    
> >
> > Is it okay if I put a warning here, instead of an error and return?
> > It won't break the special case. I wonder if it is annoying to get
> > the warning.
> issuing warning in perfectly valid case (memory-less node)
> doesn't look like a good thing to do.
>
> there is already a error message,
>
>   "total memory for NUMA nodes (0x0) should equal RAM size (0x100000000)"
>
> I'd suggest to just fix this error message to be less confusing
> instead of adding dubious warning elsewhere.

OK, thanks for your suggestion.

> > Thanks
> > Li
> >
> > > >   
> > > > > I'd rather add/rephrase to original error message that memory
> > > > > should be specified explicitly for desired numa nodes.
> > > > > And I'd not mention 'mem=' since
> > > > >      docs/about/removed-features.rst:``-numa node,mem=...`` (removed in 5.1)
> > > >
> > > > Thanks for your suggestions, I will rephrase it.
> > > >   
> > > > >       
> > > > > > +
> > > > > >         if ((node->has_mem && have_memdevs) || (node->has_memdev && have_mem)) {
> > > > > >             error_setg(errp, "numa configuration should use either mem= or 
> > > > > > memdev=,"
> > > > > >                        "mixing both is not allowed");
> > > > >       
> > > >   
> > >