[PATCH v3 0/3] Replace private domain with per-group default domain

[PATCH v3 0/3] Replace private domain with per-group default domain

[Lu Baolu]

Some devices are required to use a specific type (identity or dma)
of default domain when they are used with a vendor iommu. When the
system level default domain type is different from it, the vendor
iommu driver has to request a new default domain with either
iommu_request_dma_domain_for_dev() or iommu_request_dm_for_dev()
in the add_dev() callback. Unfortunately, these two helpers only
work when the group hasn't been assigned to any other devices,
hence, some vendor iommu driver has to use a private domain if
it fails to request a new default one.

Joerg proposed an on-going proposal which makes the default domain
framework to support configuring per-group default domain during
boot process.

https://lkml.org/lkml/2020/4/14/616

Hence, there is no need to keep the private domain implementation
in the Intel IOMMU driver. This patch series aims to remove it.

Best regards,
baolu

Change log:
v2->v3:
 - Port necessary patches on the top of Joerg's new proposal.
   https://lkml.org/lkml/2020/4/14/616
   The per-group default domain proposed previously in this series
   will be deprecated due to a race concern between domain switching
   and device driver probing.

v1->v2:
 - Rename the iommu ops callback to def_domain_type

Lu Baolu (3):
  iommu/vt-d: Allow 32bit devices to uses DMA domain
  iommu/vt-d: Allow PCI sub-hierarchy to use DMA domain
  iommu/vt-d: Apply per-device dma_ops

 drivers/iommu/intel-iommu.c | 396 +++---------------------------------
 1 file changed, 26 insertions(+), 370 deletions(-)

-- 
2.17.1

[PATCH v3 1/3] iommu/vt-d: Allow 32bit devices to uses DMA domain

[Lu Baolu]

Currently, if a 32bit device initially uses an identity domain,
Intel IOMMU driver will convert it forcibly to a DMA one if its
address capability is not enough for the whole system memory.
The motivation was to overcome the overhead caused by possible
bounced buffer.

Unfortunately, this improvement has led to many problems. For
example, some 32bit devices are required to use an identity
domain, forcing them to use DMA domain will cause the device
not to work anymore. On the other hand, the VMD sub-devices
share a domain but each sub-device might have different address
capability. Forcing a VMD sub-device to use DMA domain blindly
will impact the operation of other sub-devices without any
notification. Further more, PCI aliased devices (PCI bridge
and all devices beneath it, VMD devices and various devices
quirked with pci_add_dma_alias()) must use the same domain.
Forcing one device to switch to DMA domain during runtime
will cause in-fligh DMAs for other devices to abort or target
to other memory which might cause undefind system behavior.

Cc: Daniel Drake <drake@endlessm.com>
Cc: Derrick Jonathan <jonathan.derrick@intel.com>
Cc: Jerry Snitselaar <jsnitsel@redhat.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/iommu/intel-iommu.c | 291 +-----------------------------------
 1 file changed, 1 insertion(+), 290 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index b906727f5b85..bc5c821519a5 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -355,11 +355,6 @@ static void domain_exit(struct dmar_domain *domain);
 static void domain_remove_dev_info(struct dmar_domain *domain);
 static void dmar_remove_one_dev_info(struct device *dev);
 static void __dmar_remove_one_dev_info(struct device_domain_info *info);
-static void domain_context_clear(struct intel_iommu *iommu,
-				 struct device *dev);
-static int domain_detach_iommu(struct dmar_domain *domain,
-			       struct intel_iommu *iommu);
-static bool device_is_rmrr_locked(struct device *dev);
 static int intel_iommu_attach_device(struct iommu_domain *domain,
 				     struct device *dev);
 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -1930,65 +1925,6 @@ static inline int guestwidth_to_adjustwidth(int gaw)
 	return agaw;
 }
 
-static int domain_init(struct dmar_domain *domain, struct intel_iommu *iommu,
-		       int guest_width)
-{
-	int adjust_width, agaw;
-	unsigned long sagaw;
-	int ret;
-
-	init_iova_domain(&domain->iovad, VTD_PAGE_SIZE, IOVA_START_PFN);
-
-	if (!intel_iommu_strict) {
-		ret = init_iova_flush_queue(&domain->iovad,
-					    iommu_flush_iova, iova_entry_free);
-		if (ret)
-			pr_info("iova flush queue initialization failed\n");
-	}
-
-	domain_reserve_special_ranges(domain);
-
-	/* calculate AGAW */
-	if (guest_width > cap_mgaw(iommu->cap))
-		guest_width = cap_mgaw(iommu->cap);
-	domain->gaw = guest_width;
-	adjust_width = guestwidth_to_adjustwidth(guest_width);
-	agaw = width_to_agaw(adjust_width);
-	sagaw = cap_sagaw(iommu->cap);
-	if (!test_bit(agaw, &sagaw)) {
-		/* hardware doesn't support it, choose a bigger one */
-		pr_debug("Hardware doesn't support agaw %d\n", agaw);
-		agaw = find_next_bit(&sagaw, 5, agaw);
-		if (agaw >= 5)
-			return -ENODEV;
-	}
-	domain->agaw = agaw;
-
-	if (ecap_coherent(iommu->ecap))
-		domain->iommu_coherency = 1;
-	else
-		domain->iommu_coherency = 0;
-
-	if (ecap_sc_support(iommu->ecap))
-		domain->iommu_snooping = 1;
-	else
-		domain->iommu_snooping = 0;
-
-	if (intel_iommu_superpage)
-		domain->iommu_superpage = fls(cap_super_page_val(iommu->cap));
-	else
-		domain->iommu_superpage = 0;
-
-	domain->nid = iommu->node;
-
-	/* always allocate the top pgd */
-	domain->pgd = (struct dma_pte *)alloc_pgtable_page(domain->nid);
-	if (!domain->pgd)
-		return -ENOMEM;
-	__iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE);
-	return 0;
-}
-
 static void domain_exit(struct dmar_domain *domain)
 {
 
@@ -2704,94 +2640,6 @@ static struct dmar_domain *dmar_insert_one_dev_info(struct intel_iommu *iommu,
 	return domain;
 }
 
-static int get_last_alias(struct pci_dev *pdev, u16 alias, void *opaque)
-{
-	*(u16 *)opaque = alias;
-	return 0;
-}
-
-static struct dmar_domain *find_or_alloc_domain(struct device *dev, int gaw)
-{
-	struct device_domain_info *info;
-	struct dmar_domain *domain = NULL;
-	struct intel_iommu *iommu;
-	u16 dma_alias;
-	unsigned long flags;
-	u8 bus, devfn;
-
-	iommu = device_to_iommu(dev, &bus, &devfn);
-	if (!iommu)
-		return NULL;
-
-	if (dev_is_pci(dev)) {
-		struct pci_dev *pdev = to_pci_dev(dev);
-
-		pci_for_each_dma_alias(pdev, get_last_alias, &dma_alias);
-
-		spin_lock_irqsave(&device_domain_lock, flags);
-		info = dmar_search_domain_by_dev_info(pci_domain_nr(pdev->bus),
-						      PCI_BUS_NUM(dma_alias),
-						      dma_alias & 0xff);
-		if (info) {
-			iommu = info->iommu;
-			domain = info->domain;
-		}
-		spin_unlock_irqrestore(&device_domain_lock, flags);
-
-		/* DMA alias already has a domain, use it */
-		if (info)
-			goto out;
-	}
-
-	/* Allocate and initialize new domain for the device */
-	domain = alloc_domain(0);
-	if (!domain)
-		return NULL;
-	if (domain_init(domain, iommu, gaw)) {
-		domain_exit(domain);
-		return NULL;
-	}
-
-out:
-	return domain;
-}
-
-static struct dmar_domain *set_domain_for_dev(struct device *dev,
-					      struct dmar_domain *domain)
-{
-	struct intel_iommu *iommu;
-	struct dmar_domain *tmp;
-	u16 req_id, dma_alias;
-	u8 bus, devfn;
-
-	iommu = device_to_iommu(dev, &bus, &devfn);
-	if (!iommu)
-		return NULL;
-
-	req_id = ((u16)bus << 8) | devfn;
-
-	if (dev_is_pci(dev)) {
-		struct pci_dev *pdev = to_pci_dev(dev);
-
-		pci_for_each_dma_alias(pdev, get_last_alias, &dma_alias);
-
-		/* register PCI DMA alias device */
-		if (req_id != dma_alias) {
-			tmp = dmar_insert_one_dev_info(iommu, PCI_BUS_NUM(dma_alias),
-					dma_alias & 0xff, NULL, domain);
-
-			if (!tmp || tmp != domain)
-				return tmp;
-		}
-	}
-
-	tmp = dmar_insert_one_dev_info(iommu, bus, devfn, dev, domain);
-	if (!tmp || tmp != domain)
-		return tmp;
-
-	return domain;
-}
-
 static int iommu_domain_identity_map(struct dmar_domain *domain,
 				     unsigned long long start,
 				     unsigned long long end)
@@ -2817,45 +2665,6 @@ static int iommu_domain_identity_map(struct dmar_domain *domain,
 				DMA_PTE_READ|DMA_PTE_WRITE);
 }
 
-static int domain_prepare_identity_map(struct device *dev,
-				       struct dmar_domain *domain,
-				       unsigned long long start,
-				       unsigned long long end)
-{
-	/* For _hardware_ passthrough, don't bother. But for software
-	   passthrough, we do it anyway -- it may indicate a memory
-	   range which is reserved in E820, so which didn't get set
-	   up to start with in si_domain */
-	if (domain == si_domain && hw_pass_through) {
-		dev_warn(dev, "Ignoring identity map for HW passthrough [0x%Lx - 0x%Lx]\n",
-			 start, end);
-		return 0;
-	}
-
-	dev_info(dev, "Setting identity map [0x%Lx - 0x%Lx]\n", start, end);
-
-	if (end < start) {
-		WARN(1, "Your BIOS is broken; RMRR ends before it starts!\n"
-			"BIOS vendor: %s; Ver: %s; Product Version: %s\n",
-			dmi_get_system_info(DMI_BIOS_VENDOR),
-			dmi_get_system_info(DMI_BIOS_VERSION),
-		     dmi_get_system_info(DMI_PRODUCT_VERSION));
-		return -EIO;
-	}
-
-	if (end >> agaw_to_width(domain->agaw)) {
-		WARN(1, "Your BIOS is broken; RMRR exceeds permitted address width (%d bits)\n"
-		     "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
-		     agaw_to_width(domain->agaw),
-		     dmi_get_system_info(DMI_BIOS_VENDOR),
-		     dmi_get_system_info(DMI_BIOS_VERSION),
-		     dmi_get_system_info(DMI_PRODUCT_VERSION));
-		return -EIO;
-	}
-
-	return iommu_domain_identity_map(domain, start, end);
-}
-
 static int md_domain_init(struct dmar_domain *domain, int guest_width);
 
 static int __init si_domain_init(int hw)
@@ -3531,98 +3340,16 @@ static unsigned long intel_alloc_iova(struct device *dev,
 	return iova_pfn;
 }
 
-static struct dmar_domain *get_private_domain_for_dev(struct device *dev)
-{
-	struct dmar_domain *domain, *tmp;
-	struct dmar_rmrr_unit *rmrr;
-	struct device *i_dev;
-	int i, ret;
-
-	/* Device shouldn't be attached by any domains. */
-	domain = find_domain(dev);
-	if (domain)
-		return NULL;
-
-	domain = find_or_alloc_domain(dev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
-	if (!domain)
-		goto out;
-
-	/* We have a new domain - setup possible RMRRs for the device */
-	rcu_read_lock();
-	for_each_rmrr_units(rmrr) {
-		for_each_active_dev_scope(rmrr->devices, rmrr->devices_cnt,
-					  i, i_dev) {
-			if (i_dev != dev)
-				continue;
-
-			ret = domain_prepare_identity_map(dev, domain,
-							  rmrr->base_address,
-							  rmrr->end_address);
-			if (ret)
-				dev_err(dev, "Mapping reserved region failed\n");
-		}
-	}
-	rcu_read_unlock();
-
-	tmp = set_domain_for_dev(dev, domain);
-	if (!tmp || domain != tmp) {
-		domain_exit(domain);
-		domain = tmp;
-	}
-
-out:
-	if (!domain)
-		dev_err(dev, "Allocating domain failed\n");
-	else
-		domain->domain.type = IOMMU_DOMAIN_DMA;
-
-	return domain;
-}
-
 /* Check if the dev needs to go through non-identity map and unmap process.*/
 static bool iommu_need_mapping(struct device *dev)
 {
-	int ret;
-
 	if (iommu_dummy(dev))
 		return false;
 
 	if (unlikely(attach_deferred(dev)))
 		do_deferred_attach(dev);
 
-	ret = identity_mapping(dev);
-	if (ret) {
-		u64 dma_mask = *dev->dma_mask;
-
-		if (dev->coherent_dma_mask && dev->coherent_dma_mask < dma_mask)
-			dma_mask = dev->coherent_dma_mask;
-
-		if (dma_mask >= dma_direct_get_required_mask(dev))
-			return false;
-
-		/*
-		 * 32 bit DMA is removed from si_domain and fall back to
-		 * non-identity mapping.
-		 */
-		dmar_remove_one_dev_info(dev);
-		ret = iommu_request_dma_domain_for_dev(dev);
-		if (ret) {
-			struct iommu_domain *domain;
-			struct dmar_domain *dmar_domain;
-
-			domain = iommu_get_domain_for_dev(dev);
-			if (domain) {
-				dmar_domain = to_dmar_domain(domain);
-				dmar_domain->flags |= DOMAIN_FLAG_LOSE_CHILDREN;
-			}
-			dmar_remove_one_dev_info(dev);
-			get_private_domain_for_dev(dev);
-		}
-
-		dev_info(dev, "32bit DMA uses non-identity mapping\n");
-	}
-
-	return true;
+	return !identity_mapping(dev);
 }
 
 static dma_addr_t __intel_map_single(struct device *dev, phys_addr_t paddr,
@@ -5186,16 +4913,6 @@ int __init intel_iommu_init(void)
 	}
 	up_write(&dmar_global_lock);
 
-#if defined(CONFIG_X86) && defined(CONFIG_SWIOTLB)
-	/*
-	 * If the system has no untrusted device or the user has decided
-	 * to disable the bounce page mechanisms, we don't need swiotlb.
-	 * Mark this and the pre-allocated bounce pages will be released
-	 * later.
-	 */
-	if (!has_untrusted_dev() || intel_no_bounce)
-		swiotlb = 0;
-#endif
 	dma_ops = &intel_dma_ops;
 
 	init_iommu_pm_ops();
@@ -5296,12 +5013,6 @@ static void __dmar_remove_one_dev_info(struct device_domain_info *info)
 	domain_detach_iommu(domain, iommu);
 	spin_unlock_irqrestore(&iommu->lock, flags);
 
-	/* free the private domain */
-	if (domain->flags & DOMAIN_FLAG_LOSE_CHILDREN &&
-	    !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) &&
-	    list_empty(&domain->devices))
-		domain_exit(info->domain);
-
 	free_devinfo_mem(info);
 }
 
-- 
2.17.1

Re: [PATCH v3 1/3] iommu/vt-d: Allow 32bit devices to uses DMA domain

[Christoph Hellwig]

On Thu, Apr 16, 2020 at 02:23:52PM +0800, Lu Baolu wrote:
> Currently, if a 32bit device initially uses an identity domain,
> Intel IOMMU driver will convert it forcibly to a DMA one if its
> address capability is not enough for the whole system memory.
> The motivation was to overcome the overhead caused by possible
> bounced buffer.
> 
> Unfortunately, this improvement has led to many problems. For
> example, some 32bit devices are required to use an identity
> domain, forcing them to use DMA domain will cause the device
> not to work anymore. On the other hand, the VMD sub-devices
> share a domain but each sub-device might have different address
> capability. Forcing a VMD sub-device to use DMA domain blindly
> will impact the operation of other sub-devices without any
> notification. Further more, PCI aliased devices (PCI bridge
> and all devices beneath it, VMD devices and various devices
> quirked with pci_add_dma_alias()) must use the same domain.
> Forcing one device to switch to DMA domain during runtime
> will cause in-fligh DMAs for other devices to abort or target
> to other memory which might cause undefind system behavior.

This commit log doesn't actually explain what you are chaning, and
as far as I can tell it just removes the code to change the domain
at run time, which seems to not actually match the subject or
description.  I'd need to look at the final code, but it seems like
this will still cause bounce buffering instead of using dynamic
mapping, which still seems like an awful idea.

Also from a purely stylistic perspective a lot of the lines seem
very short and not use up the whole 73 charaters allowed.

Re: [PATCH v3 1/3] iommu/vt-d: Allow 32bit devices to uses DMA domain

[Lu Baolu]

Hi Christoph,

On 2020/4/16 15:01, Christoph Hellwig wrote:
> On Thu, Apr 16, 2020 at 02:23:52PM +0800, Lu Baolu wrote:
>> Currently, if a 32bit device initially uses an identity domain,
>> Intel IOMMU driver will convert it forcibly to a DMA one if its
>> address capability is not enough for the whole system memory.
>> The motivation was to overcome the overhead caused by possible
>> bounced buffer.
>>
>> Unfortunately, this improvement has led to many problems. For
>> example, some 32bit devices are required to use an identity
>> domain, forcing them to use DMA domain will cause the device
>> not to work anymore. On the other hand, the VMD sub-devices
>> share a domain but each sub-device might have different address
>> capability. Forcing a VMD sub-device to use DMA domain blindly
>> will impact the operation of other sub-devices without any
>> notification. Further more, PCI aliased devices (PCI bridge
>> and all devices beneath it, VMD devices and various devices
>> quirked with pci_add_dma_alias()) must use the same domain.
>> Forcing one device to switch to DMA domain during runtime
>> will cause in-fligh DMAs for other devices to abort or target
>> to other memory which might cause undefind system behavior.
> 
> This commit log doesn't actually explain what you are chaning, and
> as far as I can tell it just removes the code to change the domain
> at run time, which seems to not actually match the subject or

This removes the domain switching in iommu_need_mapping(). Another place
where the private domain is used is intel_iommu_add_device(). Joerg's
patch set has remove that. So with domain switching in
iommu_need_mapping() removed, the private domain helpers could be
removed now. Otherwise, the compiler will complain that some functions
are defined but not used.

> description.  I'd need to look at the final code, but it seems like
> this will still cause bounce buffering instead of using dynamic
> mapping, which still seems like an awful idea.

Yes. If the user chooses to use identity domain by default through
kernel command, identity domain will be applied for all devices. For
those devices with limited addressing capability, bounce buffering will
be used when they try to access the memory beyond their address
capability. This won't cause any kernel regression as far as I can see.

Switching domain during runtime with drivers loaded will cause real
problems as I said in the commit message. That's the reason why I am
proposing to remove it. If we want to keep it, we have to make sure that
switching domain for one device should not impact other devices which
share the same domain with it. Furthermore, it's better to implement it
in the generic layer to keep device driver behavior consistent on all
architectures.

> 
> Also from a purely stylistic perspective a lot of the lines seem
> very short and not use up the whole 73 charaters allowed.
> 

Yes. I will try to use up the allowed characters.

Best regards,
baolu

Re: [PATCH v3 1/3] iommu/vt-d: Allow 32bit devices to uses DMA domain

[Christoph Hellwig]

On Thu, Apr 16, 2020 at 03:40:38PM +0800, Lu Baolu wrote:
>> description.  I'd need to look at the final code, but it seems like
>> this will still cause bounce buffering instead of using dynamic
>> mapping, which still seems like an awful idea.
>
> Yes. If the user chooses to use identity domain by default through
> kernel command, identity domain will be applied for all devices. For
> those devices with limited addressing capability, bounce buffering will
> be used when they try to access the memory beyond their address
> capability. This won't cause any kernel regression as far as I can see.
>
> Switching domain during runtime with drivers loaded will cause real
> problems as I said in the commit message. That's the reason why I am
> proposing to remove it. If we want to keep it, we have to make sure that
> switching domain for one device should not impact other devices which
> share the same domain with it. Furthermore, it's better to implement it
> in the generic layer to keep device driver behavior consistent on all
> architectures.

I don't disagree with the technical points.  What I pointed out is that

 a) the actual technical change is not in the commit log, which it
    should be
 b) that I still think taking away the ability to dynamically map
    devices in the identify domain after all the time we allowed for
    that is going to cause nasty regressions.

Re: [PATCH v3 1/3] iommu/vt-d: Allow 32bit devices to uses DMA domain

[Lu Baolu]

Hi Christoph,

On 2020/4/17 14:50, Christoph Hellwig wrote:
> On Thu, Apr 16, 2020 at 03:40:38PM +0800, Lu Baolu wrote:
>>> description.  I'd need to look at the final code, but it seems like
>>> this will still cause bounce buffering instead of using dynamic
>>> mapping, which still seems like an awful idea.
>>
>> Yes. If the user chooses to use identity domain by default through
>> kernel command, identity domain will be applied for all devices. For
>> those devices with limited addressing capability, bounce buffering will
>> be used when they try to access the memory beyond their address
>> capability. This won't cause any kernel regression as far as I can see.
>>
>> Switching domain during runtime with drivers loaded will cause real
>> problems as I said in the commit message. That's the reason why I am
>> proposing to remove it. If we want to keep it, we have to make sure that
>> switching domain for one device should not impact other devices which
>> share the same domain with it. Furthermore, it's better to implement it
>> in the generic layer to keep device driver behavior consistent on all
>> architectures.
> 
> I don't disagree with the technical points.  What I pointed out is that
> 
>   a) the actual technical change is not in the commit log, which it
>      should be

Sorry! I should make the commit message more comprehensive.

>   b) that I still think taking away the ability to dynamically map
>      devices in the identify domain after all the time we allowed for
>      that is going to cause nasty regressions.
> 

This change just asks Intel IOMMU driver to use the default domain
specified by the generic default domain framework, just like what other
vendor iommu drivers do. I understand that some users wants to use DMA
domain for some specific devices when the default domain type is
identity, and vice versa, use identity domain for some devices while
default one is DMA. I think Sai's patch series posted at

https://www.spinics.net/lists/iommu/msg41680.html

is a good start. It changes the default domain with all device drivers
unbound in the generic layer, hence every vendor iommu driver could
benefit from it.

Best regards,
baolu

[PATCH v3 2/3] iommu/vt-d: Allow PCI sub-hierarchy to use DMA domain

[Lu Baolu]

Before commit fa954e6831789 ("iommu/vt-d: Delegate the dma domain
to upper layer"), Intel IOMMU started off with all devices in the
identity domain, and took them out later if it found they couldn't
access all of memory. This required devices behind a PCI bridge to
use a DMA domain at the beginning because all PCI devices behind
the bridge use the same source-id in their transactions and the
domain couldn't be changed at run-time.

Intel IOMMU driver is now aligned with the default domain framework,
there's no need to keep this requirement anymore.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/iommu/intel-iommu.c | 25 -------------------------
 1 file changed, 25 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index bc5c821519a5..f556fd89c7d2 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -2857,31 +2857,6 @@ static int device_def_domain_type(struct device *dev)
 
 		if ((iommu_identity_mapping & IDENTMAP_GFX) && IS_GFX_DEVICE(pdev))
 			return IOMMU_DOMAIN_IDENTITY;
-
-		/*
-		 * We want to start off with all devices in the 1:1 domain, and
-		 * take them out later if we find they can't access all of memory.
-		 *
-		 * However, we can't do this for PCI devices behind bridges,
-		 * because all PCI devices behind the same bridge will end up
-		 * with the same source-id on their transactions.
-		 *
-		 * Practically speaking, we can't change things around for these
-		 * devices at run-time, because we can't be sure there'll be no
-		 * DMA transactions in flight for any of their siblings.
-		 *
-		 * So PCI devices (unless they're on the root bus) as well as
-		 * their parent PCI-PCI or PCIe-PCI bridges must be left _out_ of
-		 * the 1:1 domain, just in _case_ one of their siblings turns out
-		 * not to be able to map all of memory.
-		 */
-		if (!pci_is_pcie(pdev)) {
-			if (!pci_is_root_bus(pdev->bus))
-				return IOMMU_DOMAIN_DMA;
-			if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI)
-				return IOMMU_DOMAIN_DMA;
-		} else if (pci_pcie_type(pdev) == PCI_EXP_TYPE_PCI_BRIDGE)
-			return IOMMU_DOMAIN_DMA;
 	}
 
 	return 0;
-- 
2.17.1

[PATCH v3 3/3] iommu/vt-d: Apply per-device dma_ops

[Lu Baolu]

Current Intel IOMMU driver sets the system level dma_ops. This
causes each dma API to go through the IOMMU driver even the
devices are using identity mapped domains. This sets per-device
dma_ops only if a device is using a DMA domain. Otherwise, use
the default system level dma_ops for direct dma.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/iommu/intel-iommu.c | 82 ++++++++++++-------------------------
 1 file changed, 26 insertions(+), 56 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index f556fd89c7d2..ac6f94573d1f 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -2720,17 +2720,6 @@ static int __init si_domain_init(int hw)
 	return 0;
 }
 
-static int identity_mapping(struct device *dev)
-{
-	struct device_domain_info *info;
-
-	info = dev->archdata.iommu;
-	if (info)
-		return (info->domain == si_domain);
-
-	return 0;
-}
-
 static int domain_add_dev_info(struct dmar_domain *domain, struct device *dev)
 {
 	struct dmar_domain *ndomain;
@@ -3315,18 +3304,6 @@ static unsigned long intel_alloc_iova(struct device *dev,
 	return iova_pfn;
 }
 
-/* Check if the dev needs to go through non-identity map and unmap process.*/
-static bool iommu_need_mapping(struct device *dev)
-{
-	if (iommu_dummy(dev))
-		return false;
-
-	if (unlikely(attach_deferred(dev)))
-		do_deferred_attach(dev);
-
-	return !identity_mapping(dev);
-}
-
 static dma_addr_t __intel_map_single(struct device *dev, phys_addr_t paddr,
 				     size_t size, int dir, u64 dma_mask)
 {
@@ -3340,6 +3317,9 @@ static dma_addr_t __intel_map_single(struct device *dev, phys_addr_t paddr,
 
 	BUG_ON(dir == DMA_NONE);
 
+	if (unlikely(attach_deferred(dev)))
+		do_deferred_attach(dev);
+
 	domain = find_domain(dev);
 	if (!domain)
 		return DMA_MAPPING_ERROR;
@@ -3391,20 +3371,15 @@ static dma_addr_t intel_map_page(struct device *dev, struct page *page,
 				 enum dma_data_direction dir,
 				 unsigned long attrs)
 {
-	if (iommu_need_mapping(dev))
-		return __intel_map_single(dev, page_to_phys(page) + offset,
-				size, dir, *dev->dma_mask);
-	return dma_direct_map_page(dev, page, offset, size, dir, attrs);
+	return __intel_map_single(dev, page_to_phys(page) + offset,
+				  size, dir, *dev->dma_mask);
 }
 
 static dma_addr_t intel_map_resource(struct device *dev, phys_addr_t phys_addr,
 				     size_t size, enum dma_data_direction dir,
 				     unsigned long attrs)
 {
-	if (iommu_need_mapping(dev))
-		return __intel_map_single(dev, phys_addr, size, dir,
-				*dev->dma_mask);
-	return dma_direct_map_resource(dev, phys_addr, size, dir, attrs);
+	return __intel_map_single(dev, phys_addr, size, dir, *dev->dma_mask);
 }
 
 static void intel_unmap(struct device *dev, dma_addr_t dev_addr, size_t size)
@@ -3455,17 +3430,13 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
 			     size_t size, enum dma_data_direction dir,
 			     unsigned long attrs)
 {
-	if (iommu_need_mapping(dev))
-		intel_unmap(dev, dev_addr, size);
-	else
-		dma_direct_unmap_page(dev, dev_addr, size, dir, attrs);
+	intel_unmap(dev, dev_addr, size);
 }
 
 static void intel_unmap_resource(struct device *dev, dma_addr_t dev_addr,
 		size_t size, enum dma_data_direction dir, unsigned long attrs)
 {
-	if (iommu_need_mapping(dev))
-		intel_unmap(dev, dev_addr, size);
+	intel_unmap(dev, dev_addr, size);
 }
 
 static void *intel_alloc_coherent(struct device *dev, size_t size,
@@ -3475,8 +3446,8 @@ static void *intel_alloc_coherent(struct device *dev, size_t size,
 	struct page *page = NULL;
 	int order;
 
-	if (!iommu_need_mapping(dev))
-		return dma_direct_alloc(dev, size, dma_handle, flags, attrs);
+	if (unlikely(attach_deferred(dev)))
+		do_deferred_attach(dev);
 
 	size = PAGE_ALIGN(size);
 	order = get_order(size);
@@ -3511,9 +3482,6 @@ static void intel_free_coherent(struct device *dev, size_t size, void *vaddr,
 	int order;
 	struct page *page = virt_to_page(vaddr);
 
-	if (!iommu_need_mapping(dev))
-		return dma_direct_free(dev, size, vaddr, dma_handle, attrs);
-
 	size = PAGE_ALIGN(size);
 	order = get_order(size);
 
@@ -3531,9 +3499,6 @@ static void intel_unmap_sg(struct device *dev, struct scatterlist *sglist,
 	struct scatterlist *sg;
 	int i;
 
-	if (!iommu_need_mapping(dev))
-		return dma_direct_unmap_sg(dev, sglist, nelems, dir, attrs);
-
 	for_each_sg(sglist, sg, nelems, i) {
 		nrpages += aligned_nrpages(sg_dma_address(sg), sg_dma_len(sg));
 	}
@@ -3557,8 +3522,9 @@ static int intel_map_sg(struct device *dev, struct scatterlist *sglist, int nele
 	struct intel_iommu *iommu;
 
 	BUG_ON(dir == DMA_NONE);
-	if (!iommu_need_mapping(dev))
-		return dma_direct_map_sg(dev, sglist, nelems, dir, attrs);
+
+	if (unlikely(attach_deferred(dev)))
+		do_deferred_attach(dev);
 
 	domain = find_domain(dev);
 	if (!domain)
@@ -3605,8 +3571,6 @@ static int intel_map_sg(struct device *dev, struct scatterlist *sglist, int nele
 
 static u64 intel_get_required_mask(struct device *dev)
 {
-	if (!iommu_need_mapping(dev))
-		return dma_direct_get_required_mask(dev);
 	return DMA_BIT_MASK(32);
 }
 
@@ -4888,8 +4852,6 @@ int __init intel_iommu_init(void)
 	}
 	up_write(&dmar_global_lock);
 
-	dma_ops = &intel_dma_ops;
-
 	init_iommu_pm_ops();
 
 	down_read(&dmar_global_lock);
@@ -5479,11 +5441,6 @@ static struct iommu_device *intel_iommu_probe_device(struct device *dev)
 	if (translation_pre_enabled(iommu))
 		dev->archdata.iommu = DEFER_DEVICE_DOMAIN_INFO;
 
-	if (device_needs_bounce(dev)) {
-		dev_info(dev, "Use Intel IOMMU bounce page dma_ops\n");
-		set_dma_ops(dev, &bounce_dma_ops);
-	}
-
 	return &iommu->iommu;
 }
 
@@ -5498,7 +5455,19 @@ static void intel_iommu_release_device(struct device *dev)
 
 	dmar_remove_one_dev_info(dev);
 
+	set_dma_ops(dev, NULL);
+}
+
+static void intel_iommu_probe_finalize(struct device *dev)
+{
+	struct iommu_domain *domain;
+
+	domain = iommu_get_domain_for_dev(dev);
 	if (device_needs_bounce(dev))
+		set_dma_ops(dev, &bounce_dma_ops);
+	else if (domain && domain->type == IOMMU_DOMAIN_DMA)
+		set_dma_ops(dev, &intel_dma_ops);
+	else
 		set_dma_ops(dev, NULL);
 }
 
@@ -5830,6 +5799,7 @@ const struct iommu_ops intel_iommu_ops = {
 	.unmap			= intel_iommu_unmap,
 	.iova_to_phys		= intel_iommu_iova_to_phys,
 	.probe_device		= intel_iommu_probe_device,
+	.probe_finalize		= intel_iommu_probe_finalize,
 	.release_device		= intel_iommu_release_device,
 	.get_resv_regions	= intel_iommu_get_resv_regions,
 	.put_resv_regions	= generic_iommu_put_resv_regions,
-- 
2.17.1