RE: [PATCH v1 4/6] iommu/vt-d: Add VTD_FLAG_IOMMU_PROBED flag

[Tian, Kevin]

> From: Baolu Lu <baolu.lu@xxxxxxxxxxxxxxx>
> Sent: Friday, July 1, 2022 11:13 AM
> 
> On 6/30/22 4:29 PM, Tian, Kevin wrote:
> >> From: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> >> Sent: Saturday, June 25, 2022 8:52 PM
> >>
> >> In the IOMMU hot-add path, there's a need to check whether an IOMMU
> >> has been probed. Instead of checking the IOMMU pointer in the global
> >> list, it's better to allocate a flag bit in iommu->flags for this
> >> purpose.
> >
> > Sorry I didn't get the point of original check. This is the hotplug path
> > hence the caller of this function should already figure out it's a new
> > iommu before reaching this point?
> >
> 
> Either did I. It was added by below commit without any comments about
> this check.
> 
> commit ffebeb46dd34736c90ffbca1ccb0bef8f4827c44
> Author: Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx>
> Date:   Sun Nov 9 22:48:02 2014 +0800
> 
>      iommu/vt-d: Enhance intel-iommu driver to support DMAR unit hotplug
> 
>      Implement required callback functions for intel-iommu driver
>      to support DMAR unit hotplug.
> 
>      Signed-off-by: Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx>
>      Reviewed-by: Yijing Wang <wangyijing@xxxxxxxxxx>
>      Signed-off-by: Joerg Roedel <jroedel@xxxxxxx>
> 
> I went through the whole hot-add process and found this check seemed to
> be duplicate.
> 
> Hot-add process starts from dmar_device_hotplug(), it uses a rwlock to
> synchronize the hot-add paths.
> 
> 2386         down_write(&dmar_global_lock);
> 2387         if (insert)
> 2388                 ret = dmar_hotplug_insert(tmp);
> 2389         else
> 2390                 ret = dmar_hotplug_remove(tmp);
> 2391         up_write(&dmar_global_lock);
> 
> dmar_device_hotplug()
> ->dmar_hotplug_insert()
> -->dmar_parse_one_drhd()   /* the added intel_iommu is allocated here*/
> -->dmar_hp_add_drhd()	   /* the intel_iommu is about to bring up */
> --->intel_iommu_add()
> 
> The duplicate check here:
> 
>          if (g_iommus[iommu->seq_id])
>                  return 0;
> 
> All the iommu units are allocated and then initialized in the same
> synchronized path. There is no need to check a duplicate initialization.
> 
> I would like to remove this check if no objection.
> 

This matches my impression.