linux-kernel - Re: [PATCH v8 12/12] iommu: Use refcount for fault data access

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: <20231211152456.GB1489931@ziepe.ca>
Date:   Mon, 11 Dec 2023 11:24:56 -0400
From:   Jason Gunthorpe <jgg@...pe.ca>
To:     Lu Baolu <baolu.lu@...ux.intel.com>
Cc:     Joerg Roedel <joro@...tes.org>, Will Deacon <will@...nel.org>,
        Robin Murphy <robin.murphy@....com>,
        Kevin Tian <kevin.tian@...el.com>,
        Jean-Philippe Brucker <jean-philippe@...aro.org>,
        Nicolin Chen <nicolinc@...dia.com>,
        Yi Liu <yi.l.liu@...el.com>,
        Jacob Pan <jacob.jun.pan@...ux.intel.com>,
        Longfang Liu <liulongfang@...wei.com>,
        Yan Zhao <yan.y.zhao@...el.com>, iommu@...ts.linux.dev,
        kvm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v8 12/12] iommu: Use refcount for fault data access

On Thu, Dec 07, 2023 at 02:43:08PM +0800, Lu Baolu wrote:
> @@ -217,12 +250,9 @@ int iommu_page_response(struct device *dev,
>  	if (!ops->page_response)
>  		return -ENODEV;
>  
> -	mutex_lock(&param->lock);
> -	fault_param = param->fault_param;
> -	if (!fault_param) {
> -		mutex_unlock(&param->lock);
> +	fault_param = iopf_get_dev_fault_param(dev);
> +	if (!fault_param)
>  		return -EINVAL;
> -	}

The refcounting should work by passing around the fault_param object,
not re-obtaining it from the dev from a work.

The work should be locked to the iommu_fault_param that was active
when the work was launched.

When we get to iommu_page_response it does this:

	/* Only send response if there is a fault report pending */
	mutex_lock(&fault_param->lock);
	if (list_empty(&fault_param->faults)) {
		dev_warn_ratelimited(dev, "no pending PRQ, drop response\n");
		goto done_unlock;
	}

Which determines that the iommu_fault_param is stale and pending
free..

Also iopf_queue_remove_device() is messed up - it returns an error
code but nothing ever does anything with it :( Remove functions like
this should never fail.

Removal should be like I explained earlier:
 - Disable new PRI reception
 - Ack all outstanding PRQ to the device
 - Disable PRI on the device
 - Tear down the iopf infrastructure

So under this model if the iopf_queue_remove_device() has been called
it should be sort of a 'disassociate' action where fault_param is
still floating out there but iommu_page_response() does nothing.

IOW pass the refcount from the iommu_report_device_fault() down into
the fault handler, into the work and then into iommu_page_response()
which will ultimately put it back.

> @@ -282,22 +313,15 @@ EXPORT_SYMBOL_GPL(iommu_page_response);
>   */
>  int iopf_queue_flush_dev(struct device *dev)
>  {
> -	int ret = 0;
> -	struct iommu_fault_param *iopf_param;
> -	struct dev_iommu *param = dev->iommu;
> +	struct iommu_fault_param *iopf_param = iopf_get_dev_fault_param(dev);
>  
> -	if (!param)
> +	if (!iopf_param)
>  		return -ENODEV;

And this also seems unnecessary, it is a bug to call this after
iopf_queue_remove_device() right? Just
rcu_derefernce(param->fault_param, true) and WARN_ON NULL.

Jason