Re: [PATCH v5 07/17] virt: acrn: Introduce an ioctl to set vCPU registers state

[Shuo A Liu <shuo.a.liu@intel.com>]

Hi Greg,

On Mon  9.Nov'20 at 18:09:40 +0100, Greg Kroah-Hartman wrote:
>On Mon, Oct 19, 2020 at 02:17:53PM +0800, shuo.a.liu@intel.com wrote:
>> From: Shuo Liu <shuo.a.liu@intel.com>
>>
>> A virtual CPU of User VM has different context due to the different
>> registers state. ACRN userspace needs to set the virtual CPU
>> registers state (e.g. giving a initial registers state to a virtual
>> BSP of a User VM).
>>
>> HSM provides an ioctl ACRN_IOCTL_SET_VCPU_REGS to do the virtual CPU
>> registers state setting. The ioctl passes the registers state from ACRN
>> userspace to the hypervisor directly.
>>
>> 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       | 15 ++++++++
>>  drivers/virt/acrn/hypercall.h | 13 +++++++
>>  include/uapi/linux/acrn.h     | 71 +++++++++++++++++++++++++++++++++++
>>  3 files changed, 99 insertions(+)
>>
>> diff --git a/drivers/virt/acrn/hsm.c b/drivers/virt/acrn/hsm.c
>> index cbda67d4eb89..58ceb02e82db 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/io.h>
>>  #include <linux/mm.h>
>>  #include <linux/module.h>
>>  #include <linux/slab.h>
>> @@ -46,6 +47,7 @@ static long acrn_dev_ioctl(struct file *filp, unsigned int cmd,
>>  {
>>  	struct acrn_vm *vm = filp->private_data;
>>  	struct acrn_vm_creation *vm_param;
>> +	struct acrn_vcpu_regs *cpu_regs;
>>  	int ret = 0;
>>
>>  	if (vm->vmid == ACRN_INVALID_VMID && cmd != ACRN_IOCTL_CREATE_VM) {
>> @@ -97,6 +99,19 @@ static long acrn_dev_ioctl(struct file *filp, unsigned int cmd,
>>  	case ACRN_IOCTL_DESTROY_VM:
>>  		ret = acrn_vm_destroy(vm);
>>  		break;
>> +	case ACRN_IOCTL_SET_VCPU_REGS:
>> +		cpu_regs = memdup_user((void __user *)ioctl_param,
>> +				       sizeof(struct acrn_vcpu_regs));
>> +		if (IS_ERR(cpu_regs))
>> +			return PTR_ERR(cpu_regs);
>> +
>> +		ret = hcall_set_vcpu_regs(vm->vmid, virt_to_phys(cpu_regs));
>
>Why the virt_to_phys() call here?  

The hypervisor need a Guest Physical Address to access the data, so
virt_to_phys() is used here.

>And there really is no validation of
>any fields?

Yes. Because HSM driver has little knowledge to do the validation.

>
>> +		if (ret < 0)
>> +			dev_dbg(acrn_dev.this_device,
>
>Wait, a global, static device?  Where did I miss that?  That feels odd,
>why is there just one?

The device is just for dev_*() usage. The driver links all active VMs in
a global list acrn_vm_list. And also the notification interrupt
(hypervisor to Service VM) is global. So, there is only one device.

>
>
>
>> +				"Failed to set regs state of VM%u!\n",
>> +				vm->vmid);
>> +		kfree(cpu_regs);
>> +		break;
>>  	default:
>>  		dev_dbg(acrn_dev.this_device, "Unknown IOCTL 0x%x!\n", cmd);
>>  		ret = -ENOTTY;
>> diff --git a/drivers/virt/acrn/hypercall.h b/drivers/virt/acrn/hypercall.h
>> index 426b66cadb1f..f29cfae08862 100644
>> --- a/drivers/virt/acrn/hypercall.h
>> +++ b/drivers/virt/acrn/hypercall.h
>> @@ -19,6 +19,7 @@
>>  #define HC_START_VM			_HC_ID(HC_ID, HC_ID_VM_BASE + 0x02)
>>  #define HC_PAUSE_VM			_HC_ID(HC_ID, HC_ID_VM_BASE + 0x03)
>>  #define HC_RESET_VM			_HC_ID(HC_ID, HC_ID_VM_BASE + 0x05)
>> +#define HC_SET_VCPU_REGS		_HC_ID(HC_ID, HC_ID_VM_BASE + 0x06)
>>
>>  /**
>>   * hcall_create_vm() - Create a User VM
>> @@ -75,4 +76,16 @@ static inline long hcall_reset_vm(u64 vmid)
>>  	return acrn_hypercall1(HC_RESET_VM, vmid);
>>  }
>>
>> +/**
>> + * hcall_set_vcpu_regs() - Set up registers of virtual CPU of a User VM
>> + * @vmid:	User VM ID
>> + * @regs_state:	Service VM GPA of registers state
>> + *
>> + * Return: 0 on success, <0 on failure
>> + */
>> +static inline long hcall_set_vcpu_regs(u64 vmid, u64 regs_state)
>> +{
>> +	return acrn_hypercall2(HC_SET_VCPU_REGS, vmid, regs_state);
>> +}
>> +
>>  #endif /* __ACRN_HSM_HYPERCALL_H */
>> diff --git a/include/uapi/linux/acrn.h b/include/uapi/linux/acrn.h
>> index 364b1a783074..1d5b82e154fb 100644
>> --- a/include/uapi/linux/acrn.h
>> +++ b/include/uapi/linux/acrn.h
>> @@ -36,6 +36,75 @@ struct acrn_vm_creation {
>>  	__u8	reserved2[8];
>>  } __attribute__((aligned(8)));
>>
>> +struct acrn_gp_regs {
>> +	__u64	rax;
>> +	__u64	rcx;
>> +	__u64	rdx;
>> +	__u64	rbx;
>> +	__u64	rsp;
>> +	__u64	rbp;
>> +	__u64	rsi;
>> +	__u64	rdi;
>> +	__u64	r8;
>> +	__u64	r9;
>> +	__u64	r10;
>> +	__u64	r11;
>> +	__u64	r12;
>> +	__u64	r13;
>> +	__u64	r14;
>> +	__u64	r15;
>> +};
>> +
>> +struct acrn_descriptor_ptr {
>> +	__u16	limit;
>> +	__u64	base;
>> +	__u16	reserved[3];
>> +} __attribute__ ((__packed__));
>> +
>> +struct acrn_regs {
>> +	struct acrn_gp_regs		gprs;
>> +	struct acrn_descriptor_ptr	gdt;
>> +	struct acrn_descriptor_ptr	idt;
>> +
>> +	__u64				rip;
>
>As these are all crossing the user/kernel boundry and then on to
>somewhere "else", you have to specify the endian of all of these, right?
>
>if not, why not?

The hypervisor and the driver only support X86_64 platform for now. So, the
endian should be certain.

>
>
>
>> +	__u64				cs_base;
>> +	__u64				cr0;
>> +	__u64				cr4;
>> +	__u64				cr3;
>> +	__u64				ia32_efer;
>> +	__u64				rflags;
>> +	__u64				reserved_64[4];
>> +
>> +	__u32				cs_ar;
>> +	__u32				cs_limit;
>> +	__u32				reserved_32[3];
>> +
>> +	__u16				cs_sel;
>> +	__u16				ss_sel;
>> +	__u16				ds_sel;
>> +	__u16				es_sel;
>> +	__u16				fs_sel;
>> +	__u16				gs_sel;
>> +	__u16				ldt_sel;
>> +	__u16				tr_sel;
>> +
>> +	__u16				reserved_16[4];
>> +};
>> +
>> +/**
>> + * struct acrn_vcpu_regs - Info of vCPU registers state
>> + * @vcpu_id:	vCPU ID
>> + * @reserved0:	Reserved
>> + * @vcpu_regs:	vCPU registers state
>> + *
>> + * This structure will be passed to hypervisor directly.
>> + */
>> +struct acrn_vcpu_regs {
>> +	__u16			vcpu_id;
>
>Endian?
>
>> +	__u16			reserved0[3];
>
>What does the reserved fields do?

To keep same layout with the hypervisor. Because the structure will be
passed to hypervisor directly.

>
>Is there a pointer to a public document for all of these structures
>somewhere?

Unfortunately, no. I have added some documents for some strutures
in the code via kernel-doc format.

>
>> +	struct acrn_regs	vcpu_regs;
>> +} __attribute__((aligned(8)));
>
>What does the alignment do here?

The hypervisor wants to access aligned data block to improve the
efficiency. Currently, the hypervisor only runs on x86_64 platform.


Thanks
shuo