linux-kernel - Re: [PATCH 08/12] iommufd/selftest: Put iopf enablement in domain attach path

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 for Android: free password hash cracker in your pocket

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

Message-ID: <6754e308-d2ee-4389-9dcd-c2a5bbb16c48@linux.intel.com>
Date: Fri, 21 Feb 2025 09:31:48 +0800
From: Baolu Lu <baolu.lu@...ux.intel.com>
To: Jason Gunthorpe <jgg@...pe.ca>
Cc: Joerg Roedel <joro@...tes.org>, Will Deacon <will@...nel.org>,
 Robin Murphy <robin.murphy@....com>, Kevin Tian <kevin.tian@...el.com>,
 Dave Jiang <dave.jiang@...el.com>, Vinod Koul <vkoul@...nel.org>,
 Zhangfei Gao <zhangfei.gao@...aro.org>, Zhou Wang <wangzhou1@...ilicon.com>,
 iommu@...ts.linux.dev, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 08/12] iommufd/selftest: Put iopf enablement in domain
 attach path

On 2/21/25 02:00, Jason Gunthorpe wrote:
> On Thu, Feb 20, 2025 at 03:03:21PM +0800, Baolu Lu wrote:
>> On 2/20/25 09:02, Jason Gunthorpe wrote:
>>> On Fri, Feb 14, 2025 at 02:11:00PM +0800, Lu Baolu wrote:
>>>> @@ -197,11 +201,19 @@ static int mock_domain_nop_attach(struct iommu_domain *domain,
>>>>    	if (domain->dirty_ops && (mdev->flags & MOCK_FLAGS_DEVICE_NO_DIRTY))
>>>>    		return -EINVAL;
>>>> +	return mock_dev_enable_iopf(dev, domain);
>>>> +}
>>> This isn't going to work for a replace type operation? Maybe like:
>>>
>>> if (old_domain->iopf_handler && !domain->iopf_handler)
>>> 	return mock_dev_disable_iopf(dev, domain);
>>> if (old_domain->iopf_handler && domain->iopf_handler)
>>>           return 0;
>>> return mock_dev_enable_iopf(dev, domain);
>>>
>>> ?
>> The iommufd mock device driver appears not to support replacement.
> That's not technically a choice the driver gets to have ..

Yes.

> 
>> The
>> replacement operation on this driver is likely handled as follows:
>>
>> - attach domain_a attach blocking_domain attach domain_b
> Nothing actually does that though?

Ah! you are right.

This driver allows a new domain to be set even if a domain already
exists. This is different from what vt-d driver does, which removes the
old domain first. So perhaps we need to enhance it in two ways:

- Store the existing domain (a.k.a. old domain).
- Handle iopf enablement, taking the existing domain into account.

How about below change?

diff --git a/drivers/iommu/iommufd/selftest.c 
b/drivers/iommu/iommufd/selftest.c
index a6b12cee7b00..5ffbd4e3f372 100644
--- a/drivers/iommu/iommufd/selftest.c
+++ b/drivers/iommu/iommufd/selftest.c
@@ -168,6 +168,7 @@ struct mock_dev {
         int id;
         u32 cache[MOCK_DEV_CACHE_NUM];
         unsigned int iopf_refcount;
+       struct iommu_domain *domain;
  };

  static inline struct mock_dev *to_mock_dev(struct device *dev)
@@ -197,17 +198,28 @@ static int mock_domain_nop_attach(struct 
iommu_domain *domain,
                                   struct device *dev)
  {
         struct mock_dev *mdev = to_mock_dev(dev);
+       int ret;

         if (domain->dirty_ops && (mdev->flags & 
MOCK_FLAGS_DEVICE_NO_DIRTY))
                 return -EINVAL;

-       return mock_dev_enable_iopf(dev, domain);
+       if (mdev->domain)
+               mock_dev_disable_iopf(dev, mdev->domain);
+
+       ret = mock_dev_enable_iopf(dev, domain);
+       if (ret)
+               return ret;
+
+       mdev->domain = domain;
+       return 0;
  }

Thanks,
baolu