[PATCH v2 1/3] nvme-hwmon: Return error on kzalloc failure

[Serge Semin]

On Thu, Sep 29, 2022 at 05:53:43PM -0600, Keith Busch wrote:
> On Fri, Sep 30, 2022 at 01:46:46AM +0300, Serge Semin wrote:
> > Inability to allocate a buffer is a critical error which shouldn't be
> > tolerated since most likely the rest of the driver won't work correctly.
> > Thus instead of returning the zero status let's return the -ENOMEM error
> > if the nvme_hwmon_data structure instance couldn't be created.
> 

> Nak for this one. The hwmon is not necessary for the rest of the driver to
> function, so having the driver detach from the device seems a bit harsh.

Even if it is as you say, neither the method semantic nor the way it's
called imply that. Any failures except the allocation one are
perceived as erroneous.

> The
> driver can participate in memory reclaim, so failing on a low memory condition
> can make matters worse.

Yes it can, so can many other places in the driver utilizing kmalloc
with just GFP_KERNEL flag passed including on the same path as the
nvme_hwmon_init() execution. Kmalloc will make sure the reclaim is
either finished or executed in background anyway in all cases. Don't
really see why memory allocation failure is less worse in this case
than in many others in the same driver especially seeing as I said
above the method semantic doesn't imply the optional feature
detection. Moreover the allocated structure isn't huge at all. So
failing to allocate that would indeed mean problems with the memory
resource.

> 
> The rest looks good, though.

but you ok with kmalloc in the next line. Seems like contradicting.

@Christoph, what do you think about this?

-Sergey

> 
> > @@ -230,7 +230,7 @@ int nvme_hwmon_init(struct nvme_ctrl *ctrl)
> >  
> >  	data = kzalloc(sizeof(*data), GFP_KERNEL);
> >  	if (!data)
> > -		return 0;
> > +		return -ENOMEM;
> >  
> >  	data->ctrl = ctrl;
> >  	mutex_init(&data->read_lock);
> > --