Re: [RFC PATCH 1/8] KVM: Document KVM_MAP_MEMORY ioctl

From: "Huang, Kai" <kai.huang@intel.com>
To: Isaku Yamahata <isaku.yamahata@linux.intel.com>
Cc: "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
	"Yamahata, Isaku" <isaku.yamahata@intel.com>,
	"federico.parola@polito.it" <federico.parola@polito.it>,
	"pbonzini@redhat.com" <pbonzini@redhat.com>,
	"dmatlack@google.com" <dmatlack@google.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"isaku.yamahata@gmail.com" <isaku.yamahata@gmail.com>,
	"michael.roth@amd.com" <michael.roth@amd.com>,
	"seanjc@google.com" <seanjc@google.com>
Subject: Re: [RFC PATCH 1/8] KVM: Document KVM_MAP_MEMORY ioctl
Date: Fri, 8 Mar 2024 13:20:55 +1300	[thread overview]
Message-ID: <35141245-ce1a-4315-8597-3df4f66168f8@intel.com> (raw)
In-Reply-To: <20240307203340.GI368614@ls.amr.corp.intel.com>

>>>   
>>> +4.143 KVM_MAP_MEMORY
>>> +------------------------
>>> +
>>> +:Capability: KVM_CAP_MAP_MEMORY
>>> +:Architectures: none
>>> +:Type: vcpu ioctl
>>
>> I think "vcpu ioctl" means theoretically it can be called on multiple vcpus.
>>
>> What happens in that case?
> 
> Each vcpu can handle the ioctl simaltaneously.  

Not sure whether it is implied, but should we document it can be called 
simultaneously?

Also, I believe this is only supposed to be called before VM starts to 
run?  I think we should document that too.

This is userspace ABI, we need to be explicit on how it is supposed to 
be called from userspace.

Btw, I believe there should be some justification in the changelog why 
this should be a vcpu ioctl().

[...]

>>> +:Parameters: struct kvm_memory_mapping(in/out)
>>> +:Returns: 0 on success, <0 on error
>>> +
>>> +KVM_MAP_MEMORY populates guest memory without running vcpu.
>>> +
>>> +::
>>> +
>>> +  struct kvm_memory_mapping {
>>> +	__u64 base_gfn;
>>> +	__u64 nr_pages;
>>> +	__u64 flags;
>>> +	__u64 source;
>>> +  };
>>> +
>>> +  /* For kvm_memory_mapping:: flags */
>>> +  #define KVM_MEMORY_MAPPING_FLAG_WRITE         _BITULL(0)
>>> +  #define KVM_MEMORY_MAPPING_FLAG_EXEC          _BITULL(1)
>>> +  #define KVM_MEMORY_MAPPING_FLAG_USER          _BITULL(2)
>>
>> I am not sure what's the good of having "FLAG_USER"?
>>
>> This ioctl is called from userspace, thus I think we can just treat this always
>> as user-fault?
> 
> The point is how to emulate kvm page fault as if vcpu caused the kvm page
> fault.  Not we call the ioctl as user context.

Sorry I don't quite follow.  What's wrong if KVM just append the #PF 
USER error bit before it calls into the fault handler?

My question is, since this is ABI, you have to tell how userspace is 
supposed to use this.  Maybe I am missing something, but I don't see how 
USER should be used here.