linux-kernel - Re: [RFC v1 5/8] mshv: add paravirtualized IOMMU support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20210803214718.hnp3ejs35lh455fw@liuwe-devbox-debian-v2>
Date:   Tue, 3 Aug 2021 21:47:18 +0000
From:   Wei Liu <wei.liu@...nel.org>
To:     Praveen Kumar <kumarpraveen@...ux.microsoft.com>
Cc:     Wei Liu <wei.liu@...nel.org>,
        Linux on Hyper-V List <linux-hyperv@...r.kernel.org>,
        virtualization@...ts.linux-foundation.org,
        Linux Kernel List <linux-kernel@...r.kernel.org>,
        Michael Kelley <mikelley@...rosoft.com>,
        Vineeth Pillai <viremana@...ux.microsoft.com>,
        Sunil Muthuswamy <sunilmut@...rosoft.com>,
        Nuno Das Neves <nunodasneves@...ux.microsoft.com>,
        pasha.tatashin@...een.com, Joerg Roedel <joro@...tes.org>,
        Will Deacon <will@...nel.org>,
        "K. Y. Srinivasan" <kys@...rosoft.com>,
        Haiyang Zhang <haiyangz@...rosoft.com>,
        Stephen Hemminger <sthemmin@...rosoft.com>,
        Dexuan Cui <decui@...rosoft.com>,
        "open list:IOMMU DRIVERS" <iommu@...ts.linux-foundation.org>
Subject: Re: [RFC v1 5/8] mshv: add paravirtualized IOMMU support

On Wed, Aug 04, 2021 at 12:10:45AM +0530, Praveen Kumar wrote:
> On 09-07-2021 17:13, Wei Liu wrote:
> > +static void hv_iommu_domain_free(struct iommu_domain *d)
> > +{
> > +	struct hv_iommu_domain *domain = to_hv_iommu_domain(d);
> > +	unsigned long flags;
> > +	u64 status;
> > +	struct hv_input_delete_device_domain *input;
> > +
> > +	if (is_identity_domain(domain) || is_null_domain(domain))
> > +		return;
> > +
> > +	local_irq_save(flags);
> > +	input = *this_cpu_ptr(hyperv_pcpu_input_arg);
> > +	memset(input, 0, sizeof(*input));
> > +
> > +	input->device_domain= domain->device_domain;
> > +
> > +	status = hv_do_hypercall(HVCALL_DELETE_DEVICE_DOMAIN, input, NULL);
> > +
> > +	local_irq_restore(flags);
> > +
> > +	if (!hv_result_success(status))
> > +		pr_err("%s: hypercall failed, status %lld\n", __func__, status);
> 
> Is it OK to deallocate the resources, if hypercall has failed ?

It should be fine. We leak some resources in the hypervisor, but Linux
is in a rather wedged state anyway. Refusing to free up resources in
Linux does not much good.

> Do we have any specific error code EBUSY (kind of) which we need to wait upon ?
> 

I have not found a circumstance that can happen.

> > +
> > +	ida_free(&domain->hv_iommu->domain_ids, domain->device_domain.domain_id.id);
> > +
> > +	iommu_put_dma_cookie(d);
> > +
> > +	kfree(domain);
> > +}
> > +
> > +static int hv_iommu_attach_dev(struct iommu_domain *d, struct device *dev)
> > +{
> > +	struct hv_iommu_domain *domain = to_hv_iommu_domain(d);
> > +	u64 status;
> > +	unsigned long flags;
> > +	struct hv_input_attach_device_domain *input;
> > +	struct pci_dev *pdev;
> > +	struct hv_iommu_endpoint *vdev = dev_iommu_priv_get(dev);
> > +
> > +	/* Only allow PCI devices for now */
> > +	if (!dev_is_pci(dev))
> > +		return -EINVAL;
> > +
> > +	pdev = to_pci_dev(dev);
> > +
> > +	dev_dbg(dev, "Attaching (%strusted) to %d\n", pdev->untrusted ? "un" : "",
> > +		domain->device_domain.domain_id.id);
> > +
> > +	local_irq_save(flags);
> > +	input = *this_cpu_ptr(hyperv_pcpu_input_arg);
> > +	memset(input, 0, sizeof(*input));
> > +
> > +	input->device_domain = domain->device_domain;
> > +	input->device_id = hv_build_pci_dev_id(pdev);
> > +
> > +	status = hv_do_hypercall(HVCALL_ATTACH_DEVICE_DOMAIN, input, NULL);
> > +	local_irq_restore(flags);
> > +
> > +	if (!hv_result_success(status))
> > +		pr_err("%s: hypercall failed, status %lld\n", __func__, status);
> 
> Does it make sense to vdev->domain = NULL ?
>

It is already NULL -- there is no other code path that sets it and the
detach path always sets the field to NULL.

> > +	else
> > +		vdev->domain = domain;
> > +
> > +	return hv_status_to_errno(status);
> > +}
> > +
[...]
> > +static size_t hv_iommu_unmap(struct iommu_domain *d, unsigned long iova,
> > +			   size_t size, struct iommu_iotlb_gather *gather)
> > +{
> > +	size_t unmapped;
> > +	struct hv_iommu_domain *domain = to_hv_iommu_domain(d);
> > +	unsigned long flags, npages;
> > +	struct hv_input_unmap_device_gpa_pages *input;
> > +	u64 status;
> > +
> > +	unmapped = hv_iommu_del_mappings(domain, iova, size);
> > +	if (unmapped < size)
> > +		return 0;
> 
> Is there a case where unmapped > 0 && unmapped < size ?
> 

There could be such a case -- hv_iommu_del_mappings' return value is >= 0.
Is there a problem with this predicate?

Wei.