Re: [RFC PATCH 1/1] platform-msi: Add platform check for subdevice irq domain

From: Lu Baolu <baolu.lu@linux.intel.com>
To: Bjorn Helgaas <helgaas@kernel.org>
Cc: baolu.lu@linux.intel.com, tglx@linutronix.de,
	ashok.raj@intel.com, kevin.tian@intel.com, dave.jiang@intel.com,
	megha.dey@intel.com, alex.williamson@redhat.com,
	bhelgaas@google.com, dan.j.williams@intel.com,
	dmaengine@vger.kernel.org, eric.auger@redhat.com,
	jacob.jun.pan@intel.com, jgg@mellanox.com, jing.lin@intel.com,
	kvm@vger.kernel.org, kwankhede@nvidia.com,
	linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org,
	maz@kernel.org, mona.hossain@intel.com, netanelg@mellanox.com,
	parav@mellanox.com, pbonzini@redhat.com, rafael@kernel.org,
	samuel.ortiz@intel.com, sanjay.k.kumar@intel.com,
	shahafs@mellanox.com, tony.luck@intel.com, vkoul@kernel.org,
	yan.y.zhao@linux.intel.com, yi.l.liu@intel.com
Subject: Re: [RFC PATCH 1/1] platform-msi: Add platform check for subdevice irq domain
Date: Wed, 16 Dec 2020 09:48:03 +0800	[thread overview]
Message-ID: <f473fce5-2751-c103-4195-4d7858ac6b47@linux.intel.com> (raw)
In-Reply-To: <20201210185738.GA49060@bjorn-Precision-5520>

Hi Bjorn,

On 12/11/20 2:57 AM, Bjorn Helgaas wrote:
> On Thu, Dec 10, 2020 at 08:46:24AM +0800, Lu Baolu wrote:
>> The pci_subdevice_msi_create_irq_domain() should fail if the underlying
>> platform is not able to support IMS (Interrupt Message Storage). Otherwise,
>> the isolation of interrupt is not guaranteed.
>>
>> For x86, IMS is only supported on bare metal for now. We could enable it
>> in the virtualization environments in the future if interrupt HYPERCALL
>> domain is supported or the hardware has the capability of interrupt
>> isolation for subdevices.
> 
>> + * We want to figure out which context we are running in. But the hardware
>> + * does not introduce a reliable way (instruction, CPUID leaf, MSR, whatever)
>> + * which can be manipulated by the VMM to let the OS figure out where it runs.
>> + * So we go with the below probably_on_bare_metal() function as a replacement
>> + * for definitely_on_bare_metal() to go forward only for the very simple reason
>> + * that this is the only option we have.
>> + */
>> +static const char * const possible_vmm_vendor_name[] = {
>> +	"QEMU", "Bochs", "KVM", "Xen", "VMware", "VMW", "VMware Inc.",
>> +	"innotek GmbH", "Oracle Corporation", "Parallels", "BHYVE",
>> +	"Microsoft Corporation"
>> +};
>> +
>> +static bool probably_on_bare_metal(void)
> 
> What is the point of a function called probably_on_bare_metal()?
> *Probably*?  The caller can't really do anything with the fact that
> we're not 100% sure this gives the correct answer.  Just call it
> "on_bare_metal()" or something and accept the fact that it might be
> wrong sometimes.

Agreed. we can use on_bare_metal() and add comments and kernel messages
to let users and developers know that we're not 100% sure. People should
help to make it more accurate by reporting exceptions.

> 
> This patch goes with IMS support, which somebody else is handling, so
> I assume you don't need anything from the PCI side.

Yes. This is a followup of previous discussion.

Best regards,
baolu