Re: [PATCH] virt: acrn: Fix vCPU removing code build error

[Greg Kroah-Hartman]

On Fri, Feb 12, 2021 at 06:58:53PM +0800, Shuo A Liu wrote:
> Hi Greg,
> 
> On Fri 12.Feb'21 at  8:52:33 +0100, Greg Kroah-Hartman wrote:
> > On Fri, Feb 12, 2021 at 12:57:24PM +0800, shuo.a.liu@xxxxxxxxx wrote:
> > > From: Shuo Liu <shuo.a.liu@xxxxxxxxx>
> > > 
> > > vCPU removing code depends on CONFIG_HOTPLUG_CPU as it uses remove_cpu()
> > > and add_cpu(). Make the vCPU removing interface building with
> > > CONFIG_HOTPLUG_CPU.
> > > 
> > > ../drivers/virt/acrn/hsm.c: In function ‘remove_cpu_store’:
> > > ../drivers/virt/acrn/hsm.c:389:3: error: implicit declaration of function ‘remove_cpu’; [-Werror=implicit-function-declaration]
> > >    remove_cpu(cpu);
> > > 
> > > ../drivers/virt/acrn/hsm.c:402:2: error: implicit declaration of function ‘add_cpu’; [-Werror=implicit-function-declaration]
> > >    add_cpu(cpu);
> > > 
> > > Fixes: 279dcf693ac7 ("virt: acrn: Introduce an interface for Service VM to control vCPU")
> > > Reported-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> > > Acked-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx> # build-tested
> > > Signed-off-by: Shuo Liu <shuo.a.liu@xxxxxxxxx>
> > > ---
> > >  drivers/virt/acrn/hsm.c | 4 ++++
> > >  1 file changed, 4 insertions(+)
> > > 
> > > diff --git a/drivers/virt/acrn/hsm.c b/drivers/virt/acrn/hsm.c
> > > index 1f6b7c54a1a4..e340788aacdf 100644
> > > --- a/drivers/virt/acrn/hsm.c
> > > +++ b/drivers/virt/acrn/hsm.c
> > > @@ -372,6 +372,7 @@ static int acrn_dev_release(struct inode *inode, struct file *filp)
> > >  	return 0;
> > >  }
> > > 
> > > +#ifdef CONFIG_HOTPLUG_CPU
> > >  static ssize_t remove_cpu_store(struct device *dev,
> > >  				struct device_attribute *attr,
> > >  				const char *buf, size_t count)
> > > @@ -403,9 +404,12 @@ static ssize_t remove_cpu_store(struct device *dev,
> > >  	return ret;
> > >  }
> > >  static DEVICE_ATTR_WO(remove_cpu);
> > > +#endif
> > > 
> > >  static struct attribute *acrn_attrs[] = {
> > > +#ifdef CONFIG_HOTPLUG_CPU
> > >  	&dev_attr_remove_cpu.attr,
> > > +#endif
> > >  	NULL
> > >  };
> > > 
> > > 
> > 
> > Shouldn't the real solution for this be that remove_cpu() and add_cpu()
> > have function prototypes for when this is not enabled in the kernel
> > build?
> 
> Something like this in linux/cpu.h?
> 
> diff --git a/include/linux/cpu.h b/include/linux/cpu.h
> index 3aaa0687e8df..94a578a96202 100644
> --- a/include/linux/cpu.h
> +++ b/include/linux/cpu.h
> @@ -108,6 +108,8 @@ static inline void cpu_maps_update_done(void)
> {
> }
> 
> +static inline int add_cpu(unsigned int cpu) { return 0;}
> +
> #endif /* CONFIG_SMP */
> extern struct bus_type cpu_subsys;
> 
> @@ -137,6 +139,7 @@ static inline int  cpus_read_trylock(void) { return
> true; }
> static inline void lockdep_assert_cpus_held(void) { }
> static inline void cpu_hotplug_disable(void) { }
> static inline void cpu_hotplug_enable(void) { }
> +static inline int remove_cpu(unsigned int cpu) { return -EPERM; }
> static inline void smp_shutdown_nonboot_cpus(unsigned int primary_cpu)
> { }
> #endif /* !CONFIG_HOTPLUG_CPU */
> 
> 
> > 
> > Putting #ifdef in .c files like this is not a good idea at all.
> > 
> > Then, at runtime, you can determine if you need to create this sysfs
> > file or not, as you do not want to expose it to userspace if the kernel
> > can not handle it, right?
> 
> Right. I don't want to expose the sysfs to userspace if the kernel built
> w/o CONFIG_HOTPLUG_CPU. But how to implement that if #ifdef is not used?
> misc_register() creates sysfs with .groups (acrn_attr_groups)
> unconditionally, then userspace can see the interface even it doesn't
> work.

Use the is_visible() callback for your attribute group.

thanks,

greg k-h