[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v4] xen/arm: smmuv3: Add support for removing devices


  • To: Mykyta Poturai <Mykyta_Poturai@xxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: "Orzel, Michal" <michal.orzel@xxxxxxx>
  • Date: Mon, 7 Sep 2026 10:40:59 +0200
  • Arc-authentication-results: i=1; mx.microsoft.com 1; spf=pass (sender ip is 165.204.84.17) smtp.rcpttodomain=epam.com smtp.mailfrom=amd.com; dmarc=pass (p=quarantine sp=quarantine pct=100) action=none header.from=amd.com; dkim=none (message not signed); arc=none (0)
  • Arc-message-signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector10001; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=5rcyr3XZVhs0K1/6bEO+VbaLgOLMUp13wtOdx6oM/dc=; b=AWXM18w9PLRlcOJAXUEUWrvo4F4KRf/7kTR686Gv9wWyOxyRcAlHoJB5Jrx988uebn4GYRv8FRR7Pd6d2Tj5Ir/4TpXYC45qw7CS3MdajJhnyufFcwBwkr+9poDudCN89l6Uzt8bnsYA90dWkMNBNHTUW6C/cz0VVQ4lHuppL21C17gj4ZI1E/571QP11EOHc0mhrGsJ38eBsKDEbDp16JTpVrMEuenAcZ8WwlwRFI6YmpphoM9FdnNurxRsH5vlRSy9xY3oBDjm1Ng2T+0ojpKYKdj+aNOWbgnLgrz6tppMw1mhhd+8bDLTAG9c4DNFgIJqp/Ad1EUrup0LL0kPwA==
  • Arc-seal: i=1; a=rsa-sha256; s=arcselector10001; d=microsoft.com; cv=none; b=EUd2HDnD1uOEDYX48Xc+iJhXXvWyc31ksSIQ53SOsvPLdtwVpTJSNH17Qu3fggSmJznUbQBiw6bU7HtES5w+Pj+kCMdt8xcfQ9gEJY7yuQSKk/mrfZFcVABzgpll6YZUeF1ygtIeQjA9eUTu43ctq0yeBm+j3vB3x84Volw3zEt183xjkC18uljmqFlUx9Wg2cwmOOa8uUjp0h8yeQ68Oajq7Vessg8Ll5ihVk00uhaNcu/bUb+7Nb+2JFBnPv1X9W81l3hqQQIGcBizgGquxriIdkvNdn15+ERne4LA0kSEo3o6S9Ofx7ydti6BdIXdLvjtw/NfWOMZ0TtnIH14WQ==
  • Authentication-results: eu.smtp.expurgate.cloud; dkim=pass header.s=selector1 header.d=amd.com header.i="@amd.com" header.h="From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck"
  • Cc: Stefano Stabellini <sstabellini@xxxxxxxxxx>, Julien Grall <julien@xxxxxxx>, Bertrand Marquis <bertrand.marquis@xxxxxxx>, "Volodymyr Babchuk" <Volodymyr_Babchuk@xxxxxxxx>, Rahul Singh <rahul.singh@xxxxxxx>, Oleksandr Tyshchenko <Oleksandr_Tyshchenko@xxxxxxxx>
  • Delivery-date: Mon, 07 Sep 2026 08:41:19 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>


On 25-Aug-26 09:39, Mykyta Poturai wrote:
> Allow for removing devices from SMMUv3. arm_smmu_deassign_dev handles
> most of the work by disabling ATS and zeroing STEs. Additionally, unset
> the dt_device_is_protected flag and free no longer needed smmu_master.
> Free iommu_fwspec for PCI devices only, for DT devices it is handled by
> generic IOMMU layer.
> 
> Rework dt_device_set_protected to accept a boolean parameter, update
> callsites.
> 
> Signed-off-by: Mykyta Poturai <mykyta_poturai@xxxxxxxx>
> ---
> 
> Tested on QEMU with SRIOV series[1] by repeatedly enabling/disabling
> VFs.
> 
> [1]: https://patchew.org/Xen/cover.1772806036.git.mykyta._5Fpoturai@xxxxxxxx/
> 
> V3->V4:
> * s/u8/uint8_t/
> * assert pcidevs_locked
> * assert deassignment is only reachable by PCI
> * fix build error with HAS_PCI=n
> 
> V2->V3:
> * free fwspec for pci devices
> * remove testing note from commit message
> 
> V1->V2:
> * check for phantom functions
> * simplify pci/dt device split
> * improve error handling
> * don't try to free master for unprotected devices
> * rework dt_device_set_protected
> ---
>  xen/drivers/passthrough/arm/ipmmu-vmsa.c |  2 +-
>  xen/drivers/passthrough/arm/smmu-v3.c    | 77 +++++++++++++++++++++++-
>  xen/drivers/passthrough/arm/smmu.c       |  4 +-
>  xen/include/xen/device_tree.h            |  5 +-
>  4 files changed, 82 insertions(+), 6 deletions(-)
> 
> diff --git a/xen/drivers/passthrough/arm/ipmmu-vmsa.c 
> b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> index fa9ab9cb13..0648f9b407 100644
> --- a/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> +++ b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> @@ -1367,7 +1367,7 @@ static int ipmmu_add_device(u8 devfn, struct device 
> *dev)
>          }
>  
>          /* Let Xen know that the master device is protected by an IOMMU. */
> -        dt_device_set_protected(dev_to_dt(dev));
> +        dt_device_set_protected(dev_to_dt(dev), true);
>      }
>  #ifdef CONFIG_HAS_PCI
>      if ( dev_is_pci(dev) )
> diff --git a/xen/drivers/passthrough/arm/smmu-v3.c 
> b/xen/drivers/passthrough/arm/smmu-v3.c
> index bf153227db..3578f36259 100644
> --- a/xen/drivers/passthrough/arm/smmu-v3.c
> +++ b/xen/drivers/passthrough/arm/smmu-v3.c
> @@ -1493,6 +1493,80 @@ static int arm_smmu_assign_dev(struct domain *d, u8 
> devfn, struct device *dev,
>  static int arm_smmu_deassign_dev(struct domain *d, uint8_t devfn,
>                                struct device *dev);
>  
> +static int arm_smmu_remove_device(uint8_t devfn, struct device *dev)
> +{
> +     struct arm_smmu_master *master;
> +     struct iommu_fwspec *fwspec;
> +     struct domain *d = NULL;
> +
> +     fwspec = dev_iommu_fwspec_get(dev);
> +     if ( !fwspec )
> +             return -ENODEV;
> +
> +     master = dev_iommu_priv_get(dev);
> +     if ( !master )
> +             return -ENODEV;
> +
> +#ifdef CONFIG_HAS_PCI
> +     if ( dev_is_pci(dev) )
> +     {
> +             struct pci_dev *pdev = dev_to_pci(dev);
> +
> +             /* Ignore calls for phantom functions */
> +             if ( devfn != pdev->devfn )
> +                     return 0;
> +
> +             ASSERT(pcidevs_locked());
> +
> +             d = pdev->domain;
> +     }
> +     else
> +#endif
> +     {
> +             if ( !dt_device_is_protected(dev_to_dt(dev)) )
> +             {
> +                     dev_err(dev, "Not added to SMMUv3\n");
> +                     return -ENODEV;
> +             }
> +
> +             dt_device_set_protected(dev_to_dt(dev), false);
> +             if ( master->domain && master->domain->d )
> +                     d = master->domain->d;
Given the comment below and `ASSERT(dev_is_pci(dev))`, `d` can never be != NULL
in DT case, so these two lines can be dropped.

> +     }
> +
> +     if ( d )
> +     {
> +             int ret;
> +
> +             /*
> +              * For DT devices, iommu_remove_dt_device() returns -EBUSY if 
> the
> +              * device is still assigned, so d is always NULL on the DT path.
> +              */
> +             ASSERT(dev_is_pci(dev));
> +
> +             ret = arm_smmu_deassign_dev(d, devfn, dev);
> +             /* This should never fail because we already checked the domain 
> */
> +             ASSERT(!ret);
You checked the domain but the assignment takes place before
`arm_smmu_attach_dev()` can fail in .assign_device, so I think it's still
possible and the ASSERT should be dropped.

> +     }
> +
> +     arm_smmu_disable_pasid(master);
> +
> +     dev_info(dev, "Removed master device (SMMUv3 %s StreamIds %u)\n",
> +              dev_name(fwspec->iommu_dev), fwspec->num_ids);
> +
> +     xfree(master);
> +     dev_iommu_priv_set(dev, NULL);
> +
> +     /*
> +      * For DT devices the fwspec is freed by iommu subsystem, but for PCI
> +      * devices we need to free it here
> +      */
> +     if ( IS_ENABLED(CONFIG_HAS_PCI) && dev_is_pci(dev) )
No need for IS_ENABLED here.

> +         iommu_fwspec_free(dev);
Spaces but should be tabs.

~Michal




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.