Re: [PATCH v4 17/17] virt: acrn: Introduce an interface for Service VM to control vCPU

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Tue, Sep 22, 2020 at 07:43:11PM +0800, shuo.a.liu@intel.com wrote:
> From: Shuo Liu <shuo.a.liu@intel.com>
> 
> ACRN supports partition mode to achieve real-time requirements. In
> partition mode, a CPU core can be dedicated to a vCPU of User VM. The
> local APIC of the dedicated CPU core can be passthrough to the User VM.
> The Service VM controls the assignment of the CPU cores.
> 
> Introduce an interface for the Service VM to remove the control of CPU
> core from hypervisor perspective so that the CPU core can be a dedicated
> CPU core of User VM.
> 
> Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
> Reviewed-by: Zhi Wang <zhi.a.wang@intel.com>
> Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>
> Cc: Zhi Wang <zhi.a.wang@intel.com>
> Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
> Cc: Yu Wang <yu1.wang@intel.com>
> Cc: Reinette Chatre <reinette.chatre@intel.com>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> ---
>  drivers/virt/acrn/hsm.c       | 50 +++++++++++++++++++++++++++++++++++
>  drivers/virt/acrn/hypercall.h | 14 ++++++++++
>  2 files changed, 64 insertions(+)
> 
> diff --git a/drivers/virt/acrn/hsm.c b/drivers/virt/acrn/hsm.c
> index aaf4e76d27b4..ef5f77a38d1f 100644
> --- a/drivers/virt/acrn/hsm.c
> +++ b/drivers/virt/acrn/hsm.c
> @@ -9,6 +9,7 @@
>   *	Yakui Zhao <yakui.zhao@intel.com>
>   */
>  
> +#include <linux/cpu.h>
>  #include <linux/io.h>
>  #include <linux/mm.h>
>  #include <linux/module.h>
> @@ -354,6 +355,47 @@ struct miscdevice acrn_dev = {
>  	.fops	= &acrn_fops,
>  };
>  
> +static ssize_t remove_cpu_store(struct device *dev,
> +				struct device_attribute *attr,
> +				const char *buf, size_t count)
> +{
> +	u64 cpu, lapicid;
> +	int ret;
> +
> +	if (kstrtoull(buf, 0, &cpu) < 0)
> +		return -EINVAL;
> +
> +	if (cpu >= num_possible_cpus() || cpu == 0 || !cpu_is_hotpluggable(cpu))
> +		return -EINVAL;
> +
> +	if (cpu_online(cpu))
> +		remove_cpu(cpu);
> +
> +	lapicid = cpu_data(cpu).apicid;
> +	dev_dbg(dev, "Try to remove cpu %lld with lapicid %lld\n", cpu, lapicid);
> +	ret = hcall_sos_remove_cpu(lapicid);
> +	if (ret < 0) {
> +		dev_err(dev, "Failed to remove cpu %lld!\n", cpu);
> +		goto fail_remove;
> +	}
> +
> +	return count;
> +
> +fail_remove:
> +	add_cpu(cpu);
> +	return ret;
> +}
> +static DEVICE_ATTR_WO(remove_cpu);
> +
> +static struct attribute *acrn_attrs[] = {
> +	&dev_attr_remove_cpu.attr,
> +	NULL
> +};
> +
> +static struct attribute_group acrn_attr_group = {
> +	.attrs = acrn_attrs,
> +};

You create a sysfs attribute without any Documentation/ABI/ update as
well?  That's not good.

And why are you trying to emulate CPU hotplug here and not using the
existing CPU hotplug mechanism?

> +
>  static int __init hsm_init(void)
>  {
>  	int ret;
> @@ -370,13 +412,21 @@ static int __init hsm_init(void)
>  		return ret;
>  	}
>  
> +	ret = sysfs_create_group(&acrn_dev.this_device->kobj, &acrn_attr_group);
> +	if (ret) {
> +		dev_warn(acrn_dev.this_device, "sysfs create failed\n");
> +		misc_deregister(&acrn_dev);
> +		return ret;
> +	}

You just raced with userspace and lost.  If you want to add attribute
files to a device, use the default attribute group list, and it will be
managed properly for you by the driver core.

Huge hint, if a driver every has to touch a kobject, or call sysfs_*,
then it is probably doing something wrong.

greg k-h