Re: [Patch Part2 V1 03/14] iommu/vt-d: simplify function get_domain_for_dev()

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

 




On 2014/1/8 14:48, Kai Huang wrote:
> On Wed, Jan 8, 2014 at 2:31 PM, Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx> wrote:
>>
>>
>> On 2014/1/8 14:06, Kai Huang wrote:
>>> On Wed, Jan 8, 2014 at 1:48 PM, Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx> wrote:
>>>>
>>>>
>>>> On 2014/1/8 11:06, Kai Huang wrote:
>>>>>
>>>>>
>>>>>
>>>>> On Tue, Jan 7, 2014 at 5:00 PM, Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx
>>>>> <mailto:jiang.liu@xxxxxxxxxxxxxxx>> wrote:
>>>>>
>>>>>     Function get_domain_for_dev() is a little complex, simplify it
>>>>>     by factoring out dmar_search_domain_by_dev_info() and
>>>>>     dmar_insert_dev_info().
>>>>>
>>>>>     Signed-off-by: Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx
>>>>>     <mailto:jiang.liu@xxxxxxxxxxxxxxx>>
>>>>>     ---
>>>>>      drivers/iommu/intel-iommu.c |  161
>>>>>     ++++++++++++++++++++-----------------------
>>>>>      1 file changed, 75 insertions(+), 86 deletions(-)
>>>>>
>>>>>     diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
>>>>>     index 1704e97..da65884 100644
>>>>>     --- a/drivers/iommu/intel-iommu.c
>>>>>     +++ b/drivers/iommu/intel-iommu.c
>>>>>     @@ -1957,18 +1957,70 @@ find_domain(struct pci_dev *pdev)
>>>>>             return NULL;
>>>>>      }
>>>>>
>>>>>     +static inline struct dmar_domain *
>>>>>     +dmar_search_domain_by_dev_info(int segment, int bus, int devfn)
>>>>>     +{
>>>>>     +       struct device_domain_info *info;
>>>>>     +
>>>>>     +       list_for_each_entry(info, &device_domain_list, global)
>>>>>     +               if (info->segment == segment && info->bus == bus &&
>>>>>     +                   info->devfn == devfn)
>>>>>     +                       return info->domain;
>>>>>     +
>>>>>     +       return NULL;
>>>>>     +}
>>>>>     +
>>>>>     +static int dmar_insert_dev_info(int segment, int bus, int devfn,
>>>>>     +                               struct pci_dev *dev, struct
>>>>>     dmar_domain **domp)
>>>>>     +{
>>>>>     +       struct dmar_domain *found, *domain = *domp;
>>>>>     +       struct device_domain_info *info;
>>>>>     +       unsigned long flags;
>>>>>     +
>>>>>     +       info = alloc_devinfo_mem();
>>>>>     +       if (!info)
>>>>>     +               return -ENOMEM;
>>>>>     +
>>>>>     +       info->segment = segment;
>>>>>     +       info->bus = bus;
>>>>>     +       info->devfn = devfn;
>>>>>     +       info->dev = dev;
>>>>>     +       info->domain = domain;
>>>>>     +       if (!dev)
>>>>>     +               domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES;
>>>>>     +
>>>>>     +       spin_lock_irqsave(&device_domain_lock, flags);
>>>>>     +       if (dev)
>>>>>     +               found = find_domain(dev);
>>>>>     +       else
>>>>>     +               found = dmar_search_domain_by_dev_info(segment, bus,
>>>>>     devfn);
>>>>>     +       if (found) {
>>>>>     +               spin_unlock_irqrestore(&device_domain_lock, flags);
>>>>>     +               free_devinfo_mem(info);
>>>>>     +               if (found != domain) {
>>>>>     +                       domain_exit(domain);
>>>>>     +                       *domp = found;
>>>>>     +               }
>>>>>     +       } else {
>>>>>     +               list_add(&info->link, &domain->devices);
>>>>>     +               list_add(&info->global, &device_domain_list);
>>>>>     +               if (dev)
>>>>>     +                       dev->dev.archdata.iommu = info;
>>>>>     +               spin_unlock_irqrestore(&device_domain_lock, flags);
>>>>>     +       }
>>>>>     +
>>>>>     +       return 0;
>>>>>     +}
>>>>>
>>>>>
>>>>> I am a little bit confused about the "alloc before check" sequence in
>>>>> above function. I believe the purpose of allocating the
>>>>> device_domain_info before searching the domain with spin_lock hold is to
>>>>> avoid the memory allocation in the spin_lock? In this case, why can't we
>>>>> use mutex instead of spin_lock? In my understanding, if we use mutex, we
>>>>> can first check if the domain exists or not before we really allocate
>>>>> device_domain_info, right?
>>>> Hi Kai,
>>>>         Thanks for review!
>>>>         The domain->devices list may be access in interrupt context,
>>>> so we need to protect it with spin_lock_irqsave(). Otherwise it may
>>>> case deadlock.
>>>>
>>>
>>> Thanks. That's what I am thinking. I observed lots of other iommu
>>> functions also use spin_lock.
>>>
>>>>>
>>>>> Another question is when is info->iommu field initialized? Looks it is
>>>>> not initialized here?
>>>> It's set in function iommu_support_dev_iotlb() for devices supporting
>>>> device IOTLB.
>>>
>>> Thanks. I see in iommu_support_dev_iotlb, info->iommu won't be set
>>> unless the device supports iotlb and ATS. Does this mean the
>>> info->iommu won't be used if the device doesn't support iotlb?
>>>
>>> If this is true, looks it's not convenient to find the IOMMU that
>>> handles the device via info, as it's possible that different IOMMUs
>>> share the same domain, and we don't know which IOMMU actually handles
>>> this device if we try to find it via the info->domain.
>> It's a little heavy to find info by walking the list too:)
>> Please refer to domain_get_iommu() for the way to find iommu associated
>> with a domain.
>>
> Looks domain_get_iommu just returns the first IOMMU in the bitmap, so
> it can't return the IOMMU that *really* handles the device in
> hardware. But I guess from domain's view, probably we don't care which
> IOMMU really handles the device, cause if multiple IOMMUs share the
> same domain, the IOMMUs should have the same (or very similar)
> capabilities, so referencing the first IOMMU to check some
> capabilities (that's what I see in code so far) should work for the
> domain layer API.
> 
> But looks it's safe to set info->iommu in get_domain_for_dev anyway
> (and remove it in iommu_support_dev_iotlb)? In this case it's easier
> when we want to get IOMMU from info.
> 
> Of course, it's just my opinion, and probably is wrong, it's totally
> up to you if want to do this or not. :)
Actually I'm on your side, I think it's better to initialize info->iommu
in function dmar_insert_dev_info() too. Will try that way.

> 
> -Kai
> 
>>>
>>> -Kai
>>>
>>>>
>>>>>
>>>>> -Kai
>>>>>
>>>>>     +
>>>>>      /* domain is initialized */
>>>>>      static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev,
>>>>>     int gaw)
>>>>>      {
>>>>>     -       struct dmar_domain *domain, *found = NULL;
>>>>>     +       struct dmar_domain *domain;
>>>>>             struct intel_iommu *iommu;
>>>>>             struct dmar_drhd_unit *drhd;
>>>>>     -       struct device_domain_info *info, *tmp;
>>>>>     -       struct pci_dev *dev_tmp;
>>>>>     +       struct pci_dev *bridge;
>>>>>             unsigned long flags;
>>>>>     -       int bus = 0, devfn = 0;
>>>>>     -       int segment;
>>>>>     -       int ret;
>>>>>     +       int segment, bus, devfn;
>>>>>
>>>>>             domain = find_domain(pdev);
>>>>>             if (domain)
>>>>>     @@ -1976,119 +2028,56 @@ static struct dmar_domain
>>>>>     *get_domain_for_dev(struct pci_dev *pdev, int gaw)
>>>>>
>>>>>             segment = pci_domain_nr(pdev->bus);
>>>>>
>>>>>     -       dev_tmp = pci_find_upstream_pcie_bridge(pdev);
>>>>>     -       if (dev_tmp) {
>>>>>     -               if (pci_is_pcie(dev_tmp)) {
>>>>>     -                       bus = dev_tmp->subordinate->number;
>>>>>     +       bridge = pci_find_upstream_pcie_bridge(pdev);
>>>>>     +       if (bridge) {
>>>>>     +               if (pci_is_pcie(bridge)) {
>>>>>     +                       bus = bridge->subordinate->number;
>>>>>                             devfn = 0;
>>>>>                     } else {
>>>>>     -                       bus = dev_tmp->bus->number;
>>>>>     -                       devfn = dev_tmp->devfn;
>>>>>     +                       bus = bridge->bus->number;
>>>>>     +                       devfn = bridge->devfn;
>>>>>                     }
>>>>>                     spin_lock_irqsave(&device_domain_lock, flags);
>>>>>     -               list_for_each_entry(info, &device_domain_list, global) {
>>>>>     -                       if (info->segment == segment &&
>>>>>     -                           info->bus == bus && info->devfn == devfn) {
>>>>>     -                               found = info->domain;
>>>>>     -                               break;
>>>>>     -                       }
>>>>>     -               }
>>>>>     +               domain = dmar_search_domain_by_dev_info(segment,
>>>>>     bus, devfn);
>>>>>                     spin_unlock_irqrestore(&device_domain_lock, flags);
>>>>>                     /* pcie-pci bridge already has a domain, uses it */
>>>>>     -               if (found) {
>>>>>     -                       domain = found;
>>>>>     +               if (domain)
>>>>>                             goto found_domain;
>>>>>     -               }
>>>>>             }
>>>>>
>>>>>     -       domain = alloc_domain();
>>>>>     -       if (!domain)
>>>>>     -               goto error;
>>>>>     -
>>>>>     -       /* Allocate new domain for the device */
>>>>>             drhd = dmar_find_matched_drhd_unit(pdev);
>>>>>             if (!drhd) {
>>>>>                     printk(KERN_ERR "IOMMU: can't find DMAR for device
>>>>>     %s\n",
>>>>>                             pci_name(pdev));
>>>>>     -               free_domain_mem(domain);
>>>>>                     return NULL;
>>>>>             }
>>>>>             iommu = drhd->iommu;
>>>>>
>>>>>     -       ret = iommu_attach_domain(domain, iommu);
>>>>>     -       if (ret) {
>>>>>     +       /* Allocate new domain for the device */
>>>>>     +       domain = alloc_domain();
>>>>>     +       if (!domain)
>>>>>     +               goto error;
>>>>>     +       if (iommu_attach_domain(domain, iommu)) {
>>>>>                     free_domain_mem(domain);
>>>>>                     goto error;
>>>>>             }
>>>>>     -
>>>>>             if (domain_init(domain, gaw)) {
>>>>>                     domain_exit(domain);
>>>>>                     goto error;
>>>>>             }
>>>>>
>>>>>             /* register pcie-to-pci device */
>>>>>     -       if (dev_tmp) {
>>>>>     -               info = alloc_devinfo_mem();
>>>>>     -               if (!info) {
>>>>>     +       if (bridge) {
>>>>>     +               if (dmar_insert_dev_info(segment, bus, devfn, NULL,
>>>>>     &domain)) {
>>>>>                             domain_exit(domain);
>>>>>                             goto error;
>>>>>                     }
>>>>>     -               info->segment = segment;
>>>>>     -               info->bus = bus;
>>>>>     -               info->devfn = devfn;
>>>>>     -               info->dev = NULL;
>>>>>     -               info->domain = domain;
>>>>>     -               /* This domain is shared by devices under p2p bridge */
>>>>>     -               domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES;
>>>>>     -
>>>>>     -               /* pcie-to-pci bridge already has a domain, uses it */
>>>>>     -               found = NULL;
>>>>>     -               spin_lock_irqsave(&device_domain_lock, flags);
>>>>>     -               list_for_each_entry(tmp, &device_domain_list, global) {
>>>>>     -                       if (tmp->segment == segment &&
>>>>>     -                           tmp->bus == bus && tmp->devfn == devfn) {
>>>>>     -                               found = tmp->domain;
>>>>>     -                               break;
>>>>>     -                       }
>>>>>     -               }
>>>>>     -               if (found) {
>>>>>     -                       spin_unlock_irqrestore(&device_domain_lock,
>>>>>     flags);
>>>>>     -                       free_devinfo_mem(info);
>>>>>     -                       domain_exit(domain);
>>>>>     -                       domain = found;
>>>>>     -               } else {
>>>>>     -                       list_add(&info->link, &domain->devices);
>>>>>     -                       list_add(&info->global, &device_domain_list);
>>>>>     -                       spin_unlock_irqrestore(&device_domain_lock,
>>>>>     flags);
>>>>>     -               }
>>>>>             }
>>>>>
>>>>>      found_domain:
>>>>>     -       info = alloc_devinfo_mem();
>>>>>     -       if (!info)
>>>>>     -               goto error;
>>>>>     -       info->segment = segment;
>>>>>     -       info->bus = pdev->bus->number;
>>>>>     -       info->devfn = pdev->devfn;
>>>>>     -       info->dev = pdev;
>>>>>     -       info->domain = domain;
>>>>>     -       spin_lock_irqsave(&device_domain_lock, flags);
>>>>>     -       /* somebody is fast */
>>>>>     -       found = find_domain(pdev);
>>>>>     -       if (found != NULL) {
>>>>>     -               spin_unlock_irqrestore(&device_domain_lock, flags);
>>>>>     -               if (found != domain) {
>>>>>     -                       domain_exit(domain);
>>>>>     -                       domain = found;
>>>>>     -               }
>>>>>     -               free_devinfo_mem(info);
>>>>>     +       if (dmar_insert_dev_info(segment, pdev->bus->number,
>>>>>     pdev->devfn,
>>>>>     +                                pdev, &domain) == 0)
>>>>>                     return domain;
>>>>>     -       }
>>>>>     -       list_add(&info->link, &domain->devices);
>>>>>     -       list_add(&info->global, &device_domain_list);
>>>>>     -       pdev->dev.archdata.iommu = info;
>>>>>     -       spin_unlock_irqrestore(&device_domain_lock, flags);
>>>>>     -       return domain;
>>>>>      error:
>>>>>             /* recheck it here, maybe others set it */
>>>>>             return find_domain(pdev);
>>>>>     --
>>>>>     1.7.10.4
>>>>>
>>>>>     _______________________________________________
>>>>>     iommu mailing list
>>>>>     iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx
>>>>>     <mailto:iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx>
>>>>>     https://lists.linuxfoundation.org/mailman/listinfo/iommu
>>>>>
>>>>>
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [DMA Engine]     [Linux Coverity]     [Linux USB]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Greybus]

  Powered by Linux