diff mbox

[5/5,v11] iommu/fsl: Freescale PAMU driver and iommu implementation.

Message ID 1364500442-20927-6-git-send-email-Varun.Sethi@freescale.com (mailing list archive)
State Superseded
Delegated to: Kumar Gala
Headers show

Commit Message

Varun Sethi March 28, 2013, 7:54 p.m. UTC
Following is a brief description of the PAMU hardware:
PAMU determines what action to take and whether to authorize the action on
the basis of the memory address, a Logical IO Device Number (LIODN), and
PAACT table (logically) indexed by LIODN and address. Hardware devices which
need to access memory must provide an LIODN in addition to the memory address.

Peripheral Access Authorization and Control Tables (PAACTs) are the primary
data structures used by PAMU. A PAACT is a table of peripheral access
authorization and control entries (PAACE).Each PAACE defines the range of
I/O bus address space that is accessible by the LIOD and the associated access
capabilities.

There are two types of PAACTs: primary PAACT (PPAACT) and secondary PAACT
(SPAACT).A given physical I/O device may be able to act as one or more
independent logical I/O devices (LIODs). Each such logical I/O device is
assigned an identifier called logical I/O device number (LIODN). A LIODN is
allocated a contiguous portion of the I/O bus address space called the DSA window
for performing DSA operations. The DSA window may optionally be divided into
multiple sub-windows, each of which may be used to map to a region in system
storage space. The first sub-window is referred to as the primary sub-window
and the remaining are called secondary sub-windows.

This patch provides the PAMU driver (fsl_pamu.c) and the corresponding IOMMU
API implementation (fsl_pamu_domain.c). The PAMU hardware driver (fsl_pamu.c)
has been derived from the work done by Ashish Kalra and Timur Tabi.

Signed-off-by: Timur Tabi <<timur@tabi.org>
Signed-off-by: Varun Sethi <Varun.Sethi@freescale.com>
---
changes in v11:
- changed iova to dma_addr_t in iova_to_phys API.
changes in v10:
- Support for new guts compatibe string for T4 & B4 devices.
- Modified comment about port ID and mentioned the errata number.
- Fixed the issue where data pointer was not freed in case of a an error.
- Pass data pointer while freeing irq.
- Whle initializing the SPAACE entry clear the valid bit.
changes in v9:
- Merged and createad a single function to delete
a device from domain list.
- Refactored the add_device API code.
- Renamed the paace and spaace init fucntions.
- Renamed functions for mapping windows and subwindows.
- Changed the MAX LIODN value to MAX value u-boot can
program.
- Hard coded maximum number of subwindows.
changes in v8:
- implemented the new API for window based IOMMUs.
changes in v7:
- Set max_subwidows in the geometry attribute.
- Add checking for maximum supported LIODN value.
- Use upper_32_bits and lower_32_bits macros while
  intializing PAMU data structures.
changes in v6:
- Simplified complex conditional statements.
- Fixed indentation issues.
- Added comments for IOMMU API implementation.
changes in v5:
- Addressed comments from Timur.
changes in v4:
- Addressed comments from Timur and Scott.
changes in v3:
- Addressed comments by Kumar Gala
- dynamic fspi allocation
- fixed alignment check in map and unmap
 arch/powerpc/sysdev/fsl_pci.h   |    5 +
 drivers/iommu/Kconfig           |    8 +
 drivers/iommu/Makefile          |    1 +
 drivers/iommu/fsl_pamu.c        | 1269 +++++++++++++++++++++++++++++++++++++++
 drivers/iommu/fsl_pamu.h        |  405 +++++++++++++
 drivers/iommu/fsl_pamu_domain.c | 1137 +++++++++++++++++++++++++++++++++++
 drivers/iommu/fsl_pamu_domain.h |   85 +++
 7 files changed, 2910 insertions(+), 0 deletions(-)
 create mode 100644 drivers/iommu/fsl_pamu.c
 create mode 100644 drivers/iommu/fsl_pamu.h
 create mode 100644 drivers/iommu/fsl_pamu_domain.c
 create mode 100644 drivers/iommu/fsl_pamu_domain.h

Comments

Yoder Stuart-B08248 April 2, 2013, 3:29 p.m. UTC | #1
> -----Original Message-----
> From: Sethi Varun-B16395
> Sent: Thursday, March 28, 2013 2:54 PM
> To: joro@8bytes.org; Yoder Stuart-B08248; Wood Scott-B07421; iommu@lists.linux-foundation.org; linuxppc-
> dev@lists.ozlabs.org; linux-kernel@vger.kernel.org; galak@kernel.crashing.org; benh@kernel.crashing.org
> Cc: Sethi Varun-B16395
> Subject: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu implementation.
> 
> Following is a brief description of the PAMU hardware:
> PAMU determines what action to take and whether to authorize the action on
> the basis of the memory address, a Logical IO Device Number (LIODN), and
> PAACT table (logically) indexed by LIODN and address. Hardware devices which
> need to access memory must provide an LIODN in addition to the memory address.
> 
> Peripheral Access Authorization and Control Tables (PAACTs) are the primary
> data structures used by PAMU. A PAACT is a table of peripheral access
> authorization and control entries (PAACE).Each PAACE defines the range of
> I/O bus address space that is accessible by the LIOD and the associated access
> capabilities.
> 
> There are two types of PAACTs: primary PAACT (PPAACT) and secondary PAACT
> (SPAACT).A given physical I/O device may be able to act as one or more
> independent logical I/O devices (LIODs). Each such logical I/O device is
> assigned an identifier called logical I/O device number (LIODN). A LIODN is
> allocated a contiguous portion of the I/O bus address space called the DSA window
> for performing DSA operations. The DSA window may optionally be divided into
> multiple sub-windows, each of which may be used to map to a region in system
> storage space. The first sub-window is referred to as the primary sub-window
> and the remaining are called secondary sub-windows.
> 
> This patch provides the PAMU driver (fsl_pamu.c) and the corresponding IOMMU
> API implementation (fsl_pamu_domain.c). The PAMU hardware driver (fsl_pamu.c)
> has been derived from the work done by Ashish Kalra and Timur Tabi.
> 
> Signed-off-by: Timur Tabi <<timur@tabi.org>
> Signed-off-by: Varun Sethi <Varun.Sethi@freescale.com>
> ---
> changes in v11:
> - changed iova to dma_addr_t in iova_to_phys API.
> changes in v10:
> - Support for new guts compatibe string for T4 & B4 devices.
> - Modified comment about port ID and mentioned the errata number.
> - Fixed the issue where data pointer was not freed in case of a an error.
> - Pass data pointer while freeing irq.
> - Whle initializing the SPAACE entry clear the valid bit.
> changes in v9:
> - Merged and createad a single function to delete
> a device from domain list.
> - Refactored the add_device API code.
> - Renamed the paace and spaace init fucntions.
> - Renamed functions for mapping windows and subwindows.
> - Changed the MAX LIODN value to MAX value u-boot can
> program.
> - Hard coded maximum number of subwindows.
> changes in v8:
> - implemented the new API for window based IOMMUs.
> changes in v7:
> - Set max_subwidows in the geometry attribute.
> - Add checking for maximum supported LIODN value.
> - Use upper_32_bits and lower_32_bits macros while
>   intializing PAMU data structures.
> changes in v6:
> - Simplified complex conditional statements.
> - Fixed indentation issues.
> - Added comments for IOMMU API implementation.
> changes in v5:
> - Addressed comments from Timur.
> changes in v4:
> - Addressed comments from Timur and Scott.
> changes in v3:
> - Addressed comments by Kumar Gala
> - dynamic fspi allocation
> - fixed alignment check in map and unmap
>  arch/powerpc/sysdev/fsl_pci.h   |    5 +
>  drivers/iommu/Kconfig           |    8 +
>  drivers/iommu/Makefile          |    1 +
>  drivers/iommu/fsl_pamu.c        | 1269 +++++++++++++++++++++++++++++++++++++++
>  drivers/iommu/fsl_pamu.h        |  405 +++++++++++++
>  drivers/iommu/fsl_pamu_domain.c | 1137 +++++++++++++++++++++++++++++++++++
>  drivers/iommu/fsl_pamu_domain.h |   85 +++
>  7 files changed, 2910 insertions(+), 0 deletions(-)
>  create mode 100644 drivers/iommu/fsl_pamu.c
>  create mode 100644 drivers/iommu/fsl_pamu.h
>  create mode 100644 drivers/iommu/fsl_pamu_domain.c
>  create mode 100644 drivers/iommu/fsl_pamu_domain.h

Ack.

Stuart
Joerg Roedel April 2, 2013, 4:18 p.m. UTC | #2
Cc'ing Alex Williamson

Alex, can you please review the iommu-group part of this patch?

My comments so far are below:

On Fri, Mar 29, 2013 at 01:24:02AM +0530, Varun Sethi wrote:
> +config FSL_PAMU
> +	bool "Freescale IOMMU support"
> +	depends on PPC_E500MC
> +	select IOMMU_API
> +	select GENERIC_ALLOCATOR
> +	help
> +	  Freescale PAMU support.

A bit lame for a help text. Can you elaborate more what PAMU is and when
it should be enabled?

> +int pamu_enable_liodn(int liodn)
> +{
> +	struct paace *ppaace;
> +
> +	ppaace = pamu_get_ppaace(liodn);
> +	if (!ppaace) {
> +		pr_err("Invalid primary paace entry\n");
> +		return -ENOENT;
> +	}
> +
> +	if (!get_bf(ppaace->addr_bitfields, PPAACE_AF_WSE)) {
> +		pr_err("liodn %d not configured\n", liodn);
> +		return -EINVAL;
> +	}
> +
> +	/* Ensure that all other stores to the ppaace complete first */
> +	mb();
> +
> +	ppaace->addr_bitfields |= PAACE_V_VALID;
> +	mb();

Why is it sufficient to set the bit in a variable when enabling liodn
but when disabling it set_bf needs to be called? This looks a bit
assymetric.

> +/* Derive the window size encoding for a particular PAACE entry */
> +static unsigned int map_addrspace_size_to_wse(phys_addr_t addrspace_size)
> +{
> +	/* Bug if not a power of 2 */
> +	BUG_ON((addrspace_size & (addrspace_size - 1)));

Please use is_power_of_2 here.

> +
> +	/* window size is 2^(WSE+1) bytes */
> +	return __ffs(addrspace_size >> PAMU_PAGE_SHIFT) + PAMU_PAGE_SHIFT - 1;

The PAMU_PAGE_SHIFT shifting and adding looks redundant.

> +	if ((win_size & (win_size - 1)) || win_size < PAMU_PAGE_SIZE) {
> +		pr_err("window size too small or not a power of two %llx\n", win_size);
> +		return -EINVAL;
> +	}
> +
> +	if (win_addr & (win_size - 1)) {
> +		pr_err("window address is not aligned with window size\n");
> +		return -EINVAL;
> +	}

Again, use is_power_of_2 instead of hand-coding.

> +	if (~stashid != 0)
> +		set_bf(paace->impl_attr, PAACE_IA_CID, stashid);
> +
> +	smp_wmb();
> +
> +	if (enable)
> +		paace->addr_bitfields |= PAACE_V_VALID;

Havn't you written a helper funtion to set this bit?

> +irqreturn_t pamu_av_isr(int irq, void *arg)
> +{
> +	struct pamu_isr_data *data = arg;
> +	phys_addr_t phys;
> +	unsigned int i, j;
> +
> +	pr_emerg("fsl-pamu: access violation interrupt\n");
> +
> +	for (i = 0; i < data->count; i++) {
> +		void __iomem *p = data->pamu_reg_base + i * PAMU_OFFSET;
> +		u32 pics = in_be32(p + PAMU_PICS);
> +
> +		if (pics & PAMU_ACCESS_VIOLATION_STAT) {
> +			pr_emerg("POES1=%08x\n", in_be32(p + PAMU_POES1));
> +			pr_emerg("POES2=%08x\n", in_be32(p + PAMU_POES2));
> +			pr_emerg("AVS1=%08x\n", in_be32(p + PAMU_AVS1));
> +			pr_emerg("AVS2=%08x\n", in_be32(p + PAMU_AVS2));
> +			pr_emerg("AVA=%016llx\n", make64(in_be32(p + PAMU_AVAH),
> +				in_be32(p + PAMU_AVAL)));
> +			pr_emerg("UDAD=%08x\n", in_be32(p + PAMU_UDAD));
> +			pr_emerg("POEA=%016llx\n", make64(in_be32(p + PAMU_POEAH),
> +				in_be32(p + PAMU_POEAL)));
> +
> +			phys = make64(in_be32(p + PAMU_POEAH),
> +				in_be32(p + PAMU_POEAL));
> +
> +			/* Assume that POEA points to a PAACE */
> +			if (phys) {
> +				u32 *paace = phys_to_virt(phys);
> +
> +				/* Only the first four words are relevant */
> +				for (j = 0; j < 4; j++)
> +					pr_emerg("PAACE[%u]=%08x\n", j, in_be32(paace + j));
> +			}
> +		}
> +	}
> +
> +	panic("\n");

A kernel panic seems like an over-reaction to an access violation.
Besides the device that caused the violation the system should still
work, no?

> +#define make64(high, low) (((u64)(high) << 32) | (low))

You redefined this make64 here.

> +static int map_subwins(int liodn, struct fsl_dma_domain *dma_domain)
> +{
> +	struct dma_window *sub_win_ptr =
> +				&dma_domain->win_arr[0];
> +	int i, ret;
> +	unsigned long rpn;
> +
> +	for (i = 0; i < dma_domain->win_cnt; i++) {
> +		if (sub_win_ptr[i].valid) {
> +			rpn = sub_win_ptr[i].paddr >>
> +				 PAMU_PAGE_SHIFT;
> +			spin_lock(&iommu_lock);

IOMMU code might run in interrupt context, so please use
spin_lock_irqsave for the iommu_lock.

> +static void detach_device(struct device *dev, struct fsl_dma_domain *dma_domain)
> +{
> +	struct device_domain_info *info;
> +	struct list_head *entry, *tmp;
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&dma_domain->domain_lock, flags);
> +	/* Remove the device from the domain device list */
> +	if (!list_empty(&dma_domain->devices)) {
> +		list_for_each_safe(entry, tmp, &dma_domain->devices) {
> +			info = list_entry(entry, struct device_domain_info, link);
> +			if (!dev || (info->dev == dev))
> +				remove_device_ref(info, dma_domain->win_cnt);
> +		}
> +	}
> +	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);

list_empty check is not needed. You can also use
list_for_each_entry_safe.

> +static void attach_device(struct fsl_dma_domain *dma_domain, int liodn, struct device *dev)
> +{
> +	struct device_domain_info *info, *old_domain_info;
> +
> +	spin_lock(&device_domain_lock);
> +	/*
> +	 * Check here if the device is already attached to domain or not.
> +	 * If the device is already attached to a domain detach it.
> +	 */
> +	old_domain_info = find_domain(dev);
> +	if (old_domain_info && old_domain_info->domain != dma_domain) {
> +		spin_unlock(&device_domain_lock);
> +		detach_device(dev, old_domain_info->domain);
> +		spin_lock(&device_domain_lock);
> +	}
> +
> +	info = kmem_cache_zalloc(iommu_devinfo_cache, GFP_KERNEL);
> +
> +	info->dev = dev;
> +	info->liodn = liodn;
> +	info->domain = dma_domain;
> +
> +	list_add(&info->link, &dma_domain->devices);
> +	/*
> +	 * In case of devices with multiple LIODNs just store
> +	 * the info for the first LIODN as all
> +	 * LIODNs share the same domain
> +	 */
> +	if (!old_domain_info)
> +		dev->archdata.iommu_domain = info;
> +	spin_unlock(&device_domain_lock);

Don't you have to tell the hardware that a device was added to a domain?
I don't see that, what I am missing?

> +static void swap_pci_ref(struct pci_dev **from, struct pci_dev *to)
> +{
> +	pci_dev_put(*from);
> +	*from = to;
> +}

Hmm, looks like this function is re-implemented in a few IOMMU drivers.
Want to use the chance to consolidate these implementations?


	Joerg
Timur Tabi April 3, 2013, 1:35 a.m. UTC | #3
On Tue, Apr 2, 2013 at 11:18 AM, Joerg Roedel <joro@8bytes.org> wrote:

> > +     panic("\n");
>
> A kernel panic seems like an over-reaction to an access violation.

We have no way to determining what code caused the violation, so we
can't just kill the process.  I agree it seems like overkill, but what
else should we do?  Does the IOMMU layer have a way for the IOMMU
driver to stop the device that caused the problem?

> Besides the device that caused the violation the system should still
> work, no?

Not really.  The PAMU was designed to add IOMMU support to legacy
devices, which have no concept of an MMU.  If the PAMU detects an
access violation, there's no way for the device to recover, because it
has no idea that a violation has occurred.  It's going to keep on
writing to bad data.

Maybe we need a mechanism where a driver can register a callback
function to handle IOMMU exceptions?

> > +     /*
> > +      * In case of devices with multiple LIODNs just store
> > +      * the info for the first LIODN as all
> > +      * LIODNs share the same domain
> > +      */
> > +     if (!old_domain_info)
> > +             dev->archdata.iommu_domain = info;
> > +     spin_unlock(&device_domain_lock);
>
> Don't you have to tell the hardware that a device was added to a domain?
> I don't see that, what I am missing?

I'm not sure I understand.  What "hardware" do you think needs to be notified?

The PAMU reads everything it needs from the PAACT, which we update.
The PAMU does not know anything about the devices that it monitors,
and the devices don't know anything about the PAMU.
Sethi Varun-B16395 April 3, 2013, 5:12 a.m. UTC | #4
> -----Original Message-----
> From: Wood Scott-B07421
> Sent: Wednesday, April 03, 2013 7:23 AM
> To: Timur Tabi
> Cc: Joerg Roedel; Sethi Varun-B16395; lkml; Kumar Gala; Yoder Stuart-
> B08248; iommu@lists.linux-foundation.org; Benjamin Herrenschmidt;
> linuxppc-dev@lists.ozlabs.org
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> implementation.
> 
> On 04/02/2013 08:35:54 PM, Timur Tabi wrote:
> > On Tue, Apr 2, 2013 at 11:18 AM, Joerg Roedel <joro@8bytes.org> wrote:
> >
> > > > +     panic("\n");
> > >
> > > A kernel panic seems like an over-reaction to an access violation.
> >
> > We have no way to determining what code caused the violation, so we
> > can't just kill the process.  I agree it seems like overkill, but what
> > else should we do?  Does the IOMMU layer have a way for the IOMMU
> > driver to stop the device that caused the problem?
> 
> At a minimum, log a message and continue.  Probably turn off the LIODN,
> at least if it continues to be noisy (otherwise we could get stuck in an
> interrupt storm as you note).  Possibly let the user know somehow,
> especially if it's a VFIO domain.
[Sethi Varun-B16395] Can definitely log the message and disable the LIODN (to avoid an interrupt storm), but
we definitely need a mechanism to inform vfio subsystem about the error. Also, disabling LIODN may not be a viable
option with the new LIODN allocation scheme (where LIODN would be associated with a domain).

> 
> Don't take down the whole kernel.  It's not just overkill; it undermines
> VFIO's efforts to make it safe for users to control devices.
> 
> > > Besides the device that caused the violation the system should still
> > > work, no?
> >
> > Not really.  The PAMU was designed to add IOMMU support to legacy
> > devices, which have no concept of an MMU.  If the PAMU detects an
> > access violation, there's no way for the device to recover, because it
> > has no idea that a violation has occurred.  It's going to keep on
> > writing to bad data.
> 
> I think that's only the case for posted writes (or devices which fail to
> take a hint and stop even after they see an I/O error).
> 
[Sethi Varun-B16395] Even in the case where the guest driver detects a failure, it may not be able to fix the problem without intervention from the VFIO subsystem.

-Varun
Sethi Varun-B16395 April 3, 2013, 7:01 a.m. UTC | #5
> -----Original Message-----
> From: Joerg Roedel [mailto:joro@8bytes.org]
> Sent: Tuesday, April 02, 2013 9:48 PM
> To: Sethi Varun-B16395
> Cc: Yoder Stuart-B08248; Wood Scott-B07421; iommu@lists.linux-
> foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> kernel@vger.kernel.org; galak@kernel.crashing.org;
> benh@kernel.crashing.org; Alex Williamson
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> implementation.
> 
> Cc'ing Alex Williamson
> 
> Alex, can you please review the iommu-group part of this patch?
> 
> My comments so far are below:
> 
> On Fri, Mar 29, 2013 at 01:24:02AM +0530, Varun Sethi wrote:
> > +config FSL_PAMU
> > +	bool "Freescale IOMMU support"
> > +	depends on PPC_E500MC
> > +	select IOMMU_API
> > +	select GENERIC_ALLOCATOR
> > +	help
> > +	  Freescale PAMU support.
> 
> A bit lame for a help text. Can you elaborate more what PAMU is and when
> it should be enabled?
> 
[Sethi Varun-B16395] Will update the description.

> > +int pamu_enable_liodn(int liodn)
> > +{
> > +	struct paace *ppaace;
> > +
> > +	ppaace = pamu_get_ppaace(liodn);
> > +	if (!ppaace) {
> > +		pr_err("Invalid primary paace entry\n");
> > +		return -ENOENT;
> > +	}
> > +
> > +	if (!get_bf(ppaace->addr_bitfields, PPAACE_AF_WSE)) {
> > +		pr_err("liodn %d not configured\n", liodn);
> > +		return -EINVAL;
> > +	}
> > +
> > +	/* Ensure that all other stores to the ppaace complete first */
> > +	mb();
> > +
> > +	ppaace->addr_bitfields |= PAACE_V_VALID;
> > +	mb();
> 
> Why is it sufficient to set the bit in a variable when enabling liodn but
> when disabling it set_bf needs to be called? This looks a bit assymetric.
> 
[Sethi Varun-B16395] Will make it symetric :)

> > +/* Derive the window size encoding for a particular PAACE entry */
> > +static unsigned int map_addrspace_size_to_wse(phys_addr_t
> > +addrspace_size) {
> > +	/* Bug if not a power of 2 */
> > +	BUG_ON((addrspace_size & (addrspace_size - 1)));
> 
> Please use is_power_of_2 here.
> 
> > +
> > +	/* window size is 2^(WSE+1) bytes */
> > +	return __ffs(addrspace_size >> PAMU_PAGE_SHIFT) + PAMU_PAGE_SHIFT -
> > +1;
> 
> The PAMU_PAGE_SHIFT shifting and adding looks redundant.
> 
> > +	if ((win_size & (win_size - 1)) || win_size < PAMU_PAGE_SIZE) {
> > +		pr_err("window size too small or not a power of two %llx\n",
> win_size);
> > +		return -EINVAL;
> > +	}
> > +
> > +	if (win_addr & (win_size - 1)) {
> > +		pr_err("window address is not aligned with window size\n");
> > +		return -EINVAL;
> > +	}
> 
> Again, use is_power_of_2 instead of hand-coding.
>
[Sethi Varun-B16395] ok
 
> > +	if (~stashid != 0)
> > +		set_bf(paace->impl_attr, PAACE_IA_CID, stashid);
> > +
> > +	smp_wmb();
> > +
> > +	if (enable)
> > +		paace->addr_bitfields |= PAACE_V_VALID;
> 
> Havn't you written a helper funtion to set this bit?
> 
[Sethi Varun-B16395] We already have a PAACE entry with us here so we can directly manipulate it here.

> > +irqreturn_t pamu_av_isr(int irq, void *arg) {
> > +	struct pamu_isr_data *data = arg;
> > +	phys_addr_t phys;
> > +	unsigned int i, j;
> > +
> > +	pr_emerg("fsl-pamu: access violation interrupt\n");
> > +
> > +	for (i = 0; i < data->count; i++) {
> > +		void __iomem *p = data->pamu_reg_base + i * PAMU_OFFSET;
> > +		u32 pics = in_be32(p + PAMU_PICS);
> > +
> > +		if (pics & PAMU_ACCESS_VIOLATION_STAT) {
> > +			pr_emerg("POES1=%08x\n", in_be32(p + PAMU_POES1));
> > +			pr_emerg("POES2=%08x\n", in_be32(p + PAMU_POES2));
> > +			pr_emerg("AVS1=%08x\n", in_be32(p + PAMU_AVS1));
> > +			pr_emerg("AVS2=%08x\n", in_be32(p + PAMU_AVS2));
> > +			pr_emerg("AVA=%016llx\n", make64(in_be32(p +
> PAMU_AVAH),
> > +				in_be32(p + PAMU_AVAL)));
> > +			pr_emerg("UDAD=%08x\n", in_be32(p + PAMU_UDAD));
> > +			pr_emerg("POEA=%016llx\n", make64(in_be32(p +
> PAMU_POEAH),
> > +				in_be32(p + PAMU_POEAL)));
> > +
> > +			phys = make64(in_be32(p + PAMU_POEAH),
> > +				in_be32(p + PAMU_POEAL));
> > +
> > +			/* Assume that POEA points to a PAACE */
> > +			if (phys) {
> > +				u32 *paace = phys_to_virt(phys);
> > +
> > +				/* Only the first four words are relevant */
> > +				for (j = 0; j < 4; j++)
> > +					pr_emerg("PAACE[%u]=%08x\n", j,
> in_be32(paace + j));
> > +			}
> > +		}
> > +	}
> > +
> > +	panic("\n");
> 
> A kernel panic seems like an over-reaction to an access violation.
> Besides the device that caused the violation the system should still
> work, no?
> 
[Sethi Varun-B16395] Well, if device continues to DMA outside the set aperture then it's a serious problem. We can run in to an interrupt storm (access violations).
Alternative could be to disable LIODN, but after that device DMAs would never go through. So, for the guest device is not functional.

> > +#define make64(high, low) (((u64)(high) << 32) | (low))
> 
> You redefined this make64 here.
[Sethi Varun-B16395] :(

> 
> > +static int map_subwins(int liodn, struct fsl_dma_domain *dma_domain)
> > +{
> > +	struct dma_window *sub_win_ptr =
> > +				&dma_domain->win_arr[0];
> > +	int i, ret;
> > +	unsigned long rpn;
> > +
> > +	for (i = 0; i < dma_domain->win_cnt; i++) {
> > +		if (sub_win_ptr[i].valid) {
> > +			rpn = sub_win_ptr[i].paddr >>
> > +				 PAMU_PAGE_SHIFT;
> > +			spin_lock(&iommu_lock);
> 
> IOMMU code might run in interrupt context, so please use
> spin_lock_irqsave for the iommu_lock.
[Sethi Varun-B16395] ok

> 
> > +static void detach_device(struct device *dev, struct fsl_dma_domain
> > +*dma_domain) {
> > +	struct device_domain_info *info;
> > +	struct list_head *entry, *tmp;
> > +	unsigned long flags;
> > +
> > +	spin_lock_irqsave(&dma_domain->domain_lock, flags);
> > +	/* Remove the device from the domain device list */
> > +	if (!list_empty(&dma_domain->devices)) {
> > +		list_for_each_safe(entry, tmp, &dma_domain->devices) {
> > +			info = list_entry(entry, struct device_domain_info,
> link);
> > +			if (!dev || (info->dev == dev))
> > +				remove_device_ref(info, dma_domain->win_cnt);
> > +		}
> > +	}
> > +	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
> 
> list_empty check is not needed. You can also use
> list_for_each_entry_safe.
[Sethi Varun-B16395] ok.

> 
> > +static void attach_device(struct fsl_dma_domain *dma_domain, int
> > +liodn, struct device *dev) {
> > +	struct device_domain_info *info, *old_domain_info;
> > +
> > +	spin_lock(&device_domain_lock);
> > +	/*
> > +	 * Check here if the device is already attached to domain or not.
> > +	 * If the device is already attached to a domain detach it.
> > +	 */
> > +	old_domain_info = find_domain(dev);
> > +	if (old_domain_info && old_domain_info->domain != dma_domain) {
> > +		spin_unlock(&device_domain_lock);
> > +		detach_device(dev, old_domain_info->domain);
> > +		spin_lock(&device_domain_lock);
> > +	}
> > +
> > +	info = kmem_cache_zalloc(iommu_devinfo_cache, GFP_KERNEL);
> > +
> > +	info->dev = dev;
> > +	info->liodn = liodn;
> > +	info->domain = dma_domain;
> > +
> > +	list_add(&info->link, &dma_domain->devices);
> > +	/*
> > +	 * In case of devices with multiple LIODNs just store
> > +	 * the info for the first LIODN as all
> > +	 * LIODNs share the same domain
> > +	 */
> > +	if (!old_domain_info)
> > +		dev->archdata.iommu_domain = info;
> > +	spin_unlock(&device_domain_lock);
> 
> Don't you have to tell the hardware that a device was added to a domain?
> I don't see that, what I am missing?
[Sethi Varun-B16395] Not sure I understand, once the device is attached to the domain we can do the PAMU window setup corresponding to the device LIODN (if the window information is available).

> 
> > +static void swap_pci_ref(struct pci_dev **from, struct pci_dev *to) {
> > +	pci_dev_put(*from);
> > +	*from = to;
> > +}
> 
> Hmm, looks like this function is re-implemented in a few IOMMU drivers.
> Want to use the chance to consolidate these implementations?
> 

[Sethi Varun-B16395] Will consolidate :)

-Varun
Yoder Stuart-B08248 April 3, 2013, 3:55 p.m. UTC | #6
> -----Original Message-----
> From: Sethi Varun-B16395
> Sent: Wednesday, April 03, 2013 12:12 AM
> To: Wood Scott-B07421; Timur Tabi
> Cc: Joerg Roedel; lkml; Kumar Gala; Yoder Stuart-B08248; iommu@lists.linux-foundation.org; Benjamin
> Herrenschmidt; linuxppc-dev@lists.ozlabs.org
> Subject: RE: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu implementation.
> 
> 
> 
> > -----Original Message-----
> > From: Wood Scott-B07421
> > Sent: Wednesday, April 03, 2013 7:23 AM
> > To: Timur Tabi
> > Cc: Joerg Roedel; Sethi Varun-B16395; lkml; Kumar Gala; Yoder Stuart-
> > B08248; iommu@lists.linux-foundation.org; Benjamin Herrenschmidt;
> > linuxppc-dev@lists.ozlabs.org
> > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> > implementation.
> >
> > On 04/02/2013 08:35:54 PM, Timur Tabi wrote:
> > > On Tue, Apr 2, 2013 at 11:18 AM, Joerg Roedel <joro@8bytes.org> wrote:
> > >
> > > > > +     panic("\n");
> > > >
> > > > A kernel panic seems like an over-reaction to an access violation.
> > >
> > > We have no way to determining what code caused the violation, so we
> > > can't just kill the process.  I agree it seems like overkill, but what
> > > else should we do?  Does the IOMMU layer have a way for the IOMMU
> > > driver to stop the device that caused the problem?
> >
> > At a minimum, log a message and continue.  Probably turn off the LIODN,
> > at least if it continues to be noisy (otherwise we could get stuck in an
> > interrupt storm as you note).  Possibly let the user know somehow,
> > especially if it's a VFIO domain.
> [Sethi Varun-B16395] Can definitely log the message and disable the LIODN (to avoid an interrupt storm),
> but
> we definitely need a mechanism to inform vfio subsystem about the error. Also, disabling LIODN may not
> be a viable
> option with the new LIODN allocation scheme (where LIODN would be associated with a domain).

I think for phase 1 of this, just log the error, shut down DMA as you described.
We can implement more full featured error management, like notifying vfio
or the VM somehow in the future.

Stuart
Alex Williamson April 3, 2013, 6:01 p.m. UTC | #7
On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> Cc'ing Alex Williamson
> 
> Alex, can you please review the iommu-group part of this patch?

Sure, it looks pretty reasonable.  AIUI, all PCI devices are below some
kind of host bridge that is either new and supports partitioning or old
and doesn't.  I don't know if that's a visibility or isolation
requirement, perhaps PCI ACS-ish.  In the new host bridge case, each
device gets a group.  This seems not to have any quirks for
multifunction devices though.  On AMD and Intel IOMMUs we test
multifunction device ACS support to determine whether all the functions
should be in the same group.  Is there any reason to trust multifunction
devices on PAMU?

I also find it curious what happens to the iommu group of the host
bridge.  In the partitionable case the host bridge group is removed, in
the non-partitionable case the host bridge group becomes the group for
the children, removing the host bridge.  It's unique to PAMU so far that
these host bridges are even in an iommu group (x86 only adds pci
devices), but I don't see it as necessarily wrong leaving it in either
scenario.  Does it solve some problem to remove them from the groups?
Thanks,

Alex
Sethi Varun-B16395 April 4, 2013, 1 p.m. UTC | #8
> -----Original Message-----
> From: Alex Williamson [mailto:alex.williamson@redhat.com]
> Sent: Wednesday, April 03, 2013 11:32 PM
> To: Joerg Roedel
> Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> kernel@vger.kernel.org; galak@kernel.crashing.org;
> benh@kernel.crashing.org
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> implementation.
> 
> On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > Cc'ing Alex Williamson
> >
> > Alex, can you please review the iommu-group part of this patch?
> 
> Sure, it looks pretty reasonable.  AIUI, all PCI devices are below some
> kind of host bridge that is either new and supports partitioning or old
> and doesn't.  I don't know if that's a visibility or isolation
> requirement, perhaps PCI ACS-ish.  In the new host bridge case, each
> device gets a group.  This seems not to have any quirks for multifunction
> devices though.  On AMD and Intel IOMMUs we test multifunction device ACS
> support to determine whether all the functions should be in the same
> group.  Is there any reason to trust multifunction devices on PAMU?
> 
[Sethi Varun-B16395] In the case where we can partition endpoints we can distinguish transactions based on the bus,device,function number combination. This support is available in the PCIe controller (host bridge).

> I also find it curious what happens to the iommu group of the host
> bridge.  In the partitionable case the host bridge group is removed, in
> the non-partitionable case the host bridge group becomes the group for
> the children, removing the host bridge.  It's unique to PAMU so far that
> these host bridges are even in an iommu group (x86 only adds pci
> devices), but I don't see it as necessarily wrong leaving it in either
> scenario.  Does it solve some problem to remove them from the groups?
> Thanks,
[Sethi Varun-B16395] The PCIe controller isn't a partitionable entity, it would always be owned by the host.

-Varun
Alex Williamson April 4, 2013, 3:22 p.m. UTC | #9
On Thu, 2013-04-04 at 13:00 +0000, Sethi Varun-B16395 wrote:
> 
> > -----Original Message-----
> > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > Sent: Wednesday, April 03, 2013 11:32 PM
> > To: Joerg Roedel
> > Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> > kernel@vger.kernel.org; galak@kernel.crashing.org;
> > benh@kernel.crashing.org
> > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> > implementation.
> > 
> > On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > > Cc'ing Alex Williamson
> > >
> > > Alex, can you please review the iommu-group part of this patch?
> > 
> > Sure, it looks pretty reasonable.  AIUI, all PCI devices are below some
> > kind of host bridge that is either new and supports partitioning or old
> > and doesn't.  I don't know if that's a visibility or isolation
> > requirement, perhaps PCI ACS-ish.  In the new host bridge case, each
> > device gets a group.  This seems not to have any quirks for multifunction
> > devices though.  On AMD and Intel IOMMUs we test multifunction device ACS
> > support to determine whether all the functions should be in the same
> > group.  Is there any reason to trust multifunction devices on PAMU?
> > 
> [Sethi Varun-B16395] In the case where we can partition endpoints we
> can distinguish transactions based on the bus,device,function number
> combination. This support is available in the PCIe controller (host
> bridge).

So can x86 IOMMUs, that's the visibility aspect of IOMMU groups.
Visibility alone doesn't necessarily imply that a device is isolated
though.  A multifunction PCI device that doesn't expose ACS support may
not isolate functions from each other.  For example a peer-to-peer DMA
between functions may not be translated by the upstream IOMMU.  IOMMU
groups should encompass both visibility and isolation.

> > I also find it curious what happens to the iommu group of the host
> > bridge.  In the partitionable case the host bridge group is removed, in
> > the non-partitionable case the host bridge group becomes the group for
> > the children, removing the host bridge.  It's unique to PAMU so far that
> > these host bridges are even in an iommu group (x86 only adds pci
> > devices), but I don't see it as necessarily wrong leaving it in either
> > scenario.  Does it solve some problem to remove them from the groups?
> > Thanks,
> [Sethi Varun-B16395] The PCIe controller isn't a partitionable entity,
> it would always be owned by the host.

Ownership of a device shouldn't play into the group context.  An IOMMU
group should be defined by it's visibility and isolation from other
devices.  Whether the PCIe controller is allowed to be handed to
userspace is a question for VFIO.  Thanks,

Alex
Sethi Varun-B16395 April 4, 2013, 4:35 p.m. UTC | #10
> -----Original Message-----
> From: Alex Williamson [mailto:alex.williamson@redhat.com]
> Sent: Thursday, April 04, 2013 8:52 PM
> To: Sethi Varun-B16395
> Cc: Joerg Roedel; Yoder Stuart-B08248; Wood Scott-B07421;
> iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> kernel@vger.kernel.org; galak@kernel.crashing.org;
> benh@kernel.crashing.org
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> implementation.
> 
> On Thu, 2013-04-04 at 13:00 +0000, Sethi Varun-B16395 wrote:
> >
> > > -----Original Message-----
> > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > > Sent: Wednesday, April 03, 2013 11:32 PM
> > > To: Joerg Roedel
> > > Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> > > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org;
> > > linux- kernel@vger.kernel.org; galak@kernel.crashing.org;
> > > benh@kernel.crashing.org
> > > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and
> > > iommu implementation.
> > >
> > > On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > > > Cc'ing Alex Williamson
> > > >
> > > > Alex, can you please review the iommu-group part of this patch?
> > >
> > > Sure, it looks pretty reasonable.  AIUI, all PCI devices are below
> > > some kind of host bridge that is either new and supports
> > > partitioning or old and doesn't.  I don't know if that's a
> > > visibility or isolation requirement, perhaps PCI ACS-ish.  In the
> > > new host bridge case, each device gets a group.  This seems not to
> > > have any quirks for multifunction devices though.  On AMD and Intel
> > > IOMMUs we test multifunction device ACS support to determine whether
> > > all the functions should be in the same group.  Is there any reason
> to trust multifunction devices on PAMU?
> > >
> > [Sethi Varun-B16395] In the case where we can partition endpoints we
> > can distinguish transactions based on the bus,device,function number
> > combination. This support is available in the PCIe controller (host
> > bridge).
> 
> So can x86 IOMMUs, that's the visibility aspect of IOMMU groups.
> Visibility alone doesn't necessarily imply that a device is isolated
> though.  A multifunction PCI device that doesn't expose ACS support may
> not isolate functions from each other.  For example a peer-to-peer DMA
> between functions may not be translated by the upstream IOMMU.  IOMMU
> groups should encompass both visibility and isolation.
[Sethi Varun-B16395] We can isolate the DMA access to the host based on the to the pci bus,device,function number.
I thought that was enough to put devices in to separate iommu groups. This is a PCIe controller property which allows us to partition PCIe devices. But, what I can understand from your point is that we also need to consider isolation at PCIe device level as well. I will check for the case of multifunction devices.

> 
> > > I also find it curious what happens to the iommu group of the host
> > > bridge.  In the partitionable case the host bridge group is removed,
> > > in the non-partitionable case the host bridge group becomes the
> > > group for the children, removing the host bridge.  It's unique to
> > > PAMU so far that these host bridges are even in an iommu group (x86
> > > only adds pci devices), but I don't see it as necessarily wrong
> > > leaving it in either scenario.  Does it solve some problem to remove
> them from the groups?
> > > Thanks,
> > [Sethi Varun-B16395] The PCIe controller isn't a partitionable entity,
> > it would always be owned by the host.
> 
> Ownership of a device shouldn't play into the group context.  An IOMMU
> group should be defined by it's visibility and isolation from other
> devices.  Whether the PCIe controller is allowed to be handed to
> userspace is a question for VFIO.
[Sethi Varun-B16395] The problem is in the case, where we can't partition PCIe devices. PCIe devices share the same device group as the PCI controller. This becomes a problem while assigning the devices to the guest, as you are required to unbind all the PCIe devices including the controller from the host. PCIe controller can't be unbound from the host, so we simply delete the controller iommu_group.

-Varun
Yoder Stuart-B08248 April 4, 2013, 4:43 p.m. UTC | #11
> -----Original Message-----
> From: Alex Williamson [mailto:alex.williamson@redhat.com]
> Sent: Thursday, April 04, 2013 10:22 AM
> To: Sethi Varun-B16395
> Cc: Joerg Roedel; Yoder Stuart-B08248; Wood Scott-B07421; iommu@lists.linux-foundation.org; linuxppc-
> dev@lists.ozlabs.org; linux-kernel@vger.kernel.org; galak@kernel.crashing.org; benh@kernel.crashing.org
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu implementation.
> 
> On Thu, 2013-04-04 at 13:00 +0000, Sethi Varun-B16395 wrote:
> >
> > > -----Original Message-----
> > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > > Sent: Wednesday, April 03, 2013 11:32 PM
> > > To: Joerg Roedel
> > > Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> > > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> > > kernel@vger.kernel.org; galak@kernel.crashing.org;
> > > benh@kernel.crashing.org
> > > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> > > implementation.
> > >
> > > On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > > > Cc'ing Alex Williamson
> > > >
> > > > Alex, can you please review the iommu-group part of this patch?
> > >
> > > Sure, it looks pretty reasonable.  AIUI, all PCI devices are below some
> > > kind of host bridge that is either new and supports partitioning or old
> > > and doesn't.  I don't know if that's a visibility or isolation
> > > requirement, perhaps PCI ACS-ish.  In the new host bridge case, each
> > > device gets a group.  This seems not to have any quirks for multifunction
> > > devices though.  On AMD and Intel IOMMUs we test multifunction device ACS
> > > support to determine whether all the functions should be in the same
> > > group.  Is there any reason to trust multifunction devices on PAMU?
> > >
> > [Sethi Varun-B16395] In the case where we can partition endpoints we
> > can distinguish transactions based on the bus,device,function number
> > combination. This support is available in the PCIe controller (host
> > bridge).
> 
> So can x86 IOMMUs, that's the visibility aspect of IOMMU groups.
> Visibility alone doesn't necessarily imply that a device is isolated
> though.  A multifunction PCI device that doesn't expose ACS support may
> not isolate functions from each other.  For example a peer-to-peer DMA
> between functions may not be translated by the upstream IOMMU.  IOMMU
> groups should encompass both visibility and isolation.
> 
> > > I also find it curious what happens to the iommu group of the host
> > > bridge.  In the partitionable case the host bridge group is removed, in
> > > the non-partitionable case the host bridge group becomes the group for
> > > the children, removing the host bridge.  It's unique to PAMU so far that
> > > these host bridges are even in an iommu group (x86 only adds pci
> > > devices), but I don't see it as necessarily wrong leaving it in either
> > > scenario.  Does it solve some problem to remove them from the groups?
> > > Thanks,
> > [Sethi Varun-B16395] The PCIe controller isn't a partitionable entity,
> > it would always be owned by the host.
> 
> Ownership of a device shouldn't play into the group context.  An IOMMU
> group should be defined by it's visibility and isolation from other
> devices.  Whether the PCIe controller is allowed to be handed to
> userspace is a question for VFIO.  Thanks,

Right now the add_device() callback gets called for all platform
devices (including PCI controller) and PCI devices.  PCI controllers
are a kind of special case in that their device tree node has a
property indicating that it is DMA capable...but in fact the
PCI controller itself does not DMA, but it's the PCI endpoints
under it.

So, as you noted the bridge/controller shouldn't be in an IOMMU
group and so since the platform 'add device' code didn't special
case PCI controllers, this patch removes their group if it's there.

Stuart
Alex Williamson April 4, 2013, 4:43 p.m. UTC | #12
On Thu, 2013-04-04 at 16:35 +0000, Sethi Varun-B16395 wrote:
> 
> > -----Original Message-----
> > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > Sent: Thursday, April 04, 2013 8:52 PM
> > To: Sethi Varun-B16395
> > Cc: Joerg Roedel; Yoder Stuart-B08248; Wood Scott-B07421;
> > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> > kernel@vger.kernel.org; galak@kernel.crashing.org;
> > benh@kernel.crashing.org
> > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> > implementation.
> > 
> > On Thu, 2013-04-04 at 13:00 +0000, Sethi Varun-B16395 wrote:
> > >
> > > > -----Original Message-----
> > > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > > > Sent: Wednesday, April 03, 2013 11:32 PM
> > > > To: Joerg Roedel
> > > > Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> > > > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org;
> > > > linux- kernel@vger.kernel.org; galak@kernel.crashing.org;
> > > > benh@kernel.crashing.org
> > > > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and
> > > > iommu implementation.
> > > >
> > > > On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > > > > Cc'ing Alex Williamson
> > > > >
> > > > > Alex, can you please review the iommu-group part of this patch?
> > > >
> > > > Sure, it looks pretty reasonable.  AIUI, all PCI devices are below
> > > > some kind of host bridge that is either new and supports
> > > > partitioning or old and doesn't.  I don't know if that's a
> > > > visibility or isolation requirement, perhaps PCI ACS-ish.  In the
> > > > new host bridge case, each device gets a group.  This seems not to
> > > > have any quirks for multifunction devices though.  On AMD and Intel
> > > > IOMMUs we test multifunction device ACS support to determine whether
> > > > all the functions should be in the same group.  Is there any reason
> > to trust multifunction devices on PAMU?
> > > >
> > > [Sethi Varun-B16395] In the case where we can partition endpoints we
> > > can distinguish transactions based on the bus,device,function number
> > > combination. This support is available in the PCIe controller (host
> > > bridge).
> > 
> > So can x86 IOMMUs, that's the visibility aspect of IOMMU groups.
> > Visibility alone doesn't necessarily imply that a device is isolated
> > though.  A multifunction PCI device that doesn't expose ACS support may
> > not isolate functions from each other.  For example a peer-to-peer DMA
> > between functions may not be translated by the upstream IOMMU.  IOMMU
> > groups should encompass both visibility and isolation.
> [Sethi Varun-B16395] We can isolate the DMA access to the host based
> on the to the pci bus,device,function number.

The IOMMU can only isolate DMA that it can see.  A multifunction device
may never expose peer-to-peer DMA to the upstream device, it's
implementation specific.  The ACS flags allow that possibility to be
controlled and prevented.

> I thought that was enough to put devices in to separate iommu groups.
> This is a PCIe controller property which allows us to partition PCIe
> devices. But, what I can understand from your point is that we also
> need to consider isolation at PCIe device level as well. I will check
> for the case of multifunction devices.
> 
> > 
> > > > I also find it curious what happens to the iommu group of the host
> > > > bridge.  In the partitionable case the host bridge group is removed,
> > > > in the non-partitionable case the host bridge group becomes the
> > > > group for the children, removing the host bridge.  It's unique to
> > > > PAMU so far that these host bridges are even in an iommu group (x86
> > > > only adds pci devices), but I don't see it as necessarily wrong
> > > > leaving it in either scenario.  Does it solve some problem to remove
> > them from the groups?
> > > > Thanks,
> > > [Sethi Varun-B16395] The PCIe controller isn't a partitionable entity,
> > > it would always be owned by the host.
> > 
> > Ownership of a device shouldn't play into the group context.  An IOMMU
> > group should be defined by it's visibility and isolation from other
> > devices.  Whether the PCIe controller is allowed to be handed to
> > userspace is a question for VFIO.
> [Sethi Varun-B16395] The problem is in the case, where we can't
> partition PCIe devices. PCIe devices share the same device group as
> the PCI controller. This becomes a problem while assigning the devices
> to the guest, as you are required to unbind all the PCIe devices
> including the controller from the host. PCIe controller can't be
> unbound from the host, so we simply delete the controller iommu_group.

Unbinding devices is a VFIO implementation, it shouldn't leak into IOMMU
groups.  Also note that VFIO has a driver white list where we can have
exceptions to the rule.  I recently added pciehp to that list because
the host driver provides functionality.  Being attached to the host
driver means the device is not accessible to the user through VFIO, but
other devices in the group are.  Thanks,

Alex
Sethi Varun-B16395 April 5, 2013, 12:01 a.m. UTC | #13
> -----Original Message-----
> From: Alex Williamson [mailto:alex.williamson@redhat.com]
> Sent: Thursday, April 04, 2013 10:14 PM
> To: Sethi Varun-B16395
> Cc: Joerg Roedel; Yoder Stuart-B08248; Wood Scott-B07421;
> iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org; linux-
> kernel@vger.kernel.org; galak@kernel.crashing.org;
> benh@kernel.crashing.org
> Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and iommu
> implementation.
> 
> On Thu, 2013-04-04 at 16:35 +0000, Sethi Varun-B16395 wrote:
> >
> > > -----Original Message-----
> > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > > Sent: Thursday, April 04, 2013 8:52 PM
> > > To: Sethi Varun-B16395
> > > Cc: Joerg Roedel; Yoder Stuart-B08248; Wood Scott-B07421;
> > > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org;
> > > linux- kernel@vger.kernel.org; galak@kernel.crashing.org;
> > > benh@kernel.crashing.org
> > > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver and
> > > iommu implementation.
> > >
> > > On Thu, 2013-04-04 at 13:00 +0000, Sethi Varun-B16395 wrote:
> > > >
> > > > > -----Original Message-----
> > > > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > > > > Sent: Wednesday, April 03, 2013 11:32 PM
> > > > > To: Joerg Roedel
> > > > > Cc: Sethi Varun-B16395; Yoder Stuart-B08248; Wood Scott-B07421;
> > > > > iommu@lists.linux-foundation.org; linuxppc-dev@lists.ozlabs.org;
> > > > > linux- kernel@vger.kernel.org; galak@kernel.crashing.org;
> > > > > benh@kernel.crashing.org
> > > > > Subject: Re: [PATCH 5/5 v11] iommu/fsl: Freescale PAMU driver
> > > > > and iommu implementation.
> > > > >
> > > > > On Tue, 2013-04-02 at 18:18 +0200, Joerg Roedel wrote:
> > > > > > Cc'ing Alex Williamson
> > > > > >
> > > > > > Alex, can you please review the iommu-group part of this patch?
> > > > >
> > > > > Sure, it looks pretty reasonable.  AIUI, all PCI devices are
> > > > > below some kind of host bridge that is either new and supports
> > > > > partitioning or old and doesn't.  I don't know if that's a
> > > > > visibility or isolation requirement, perhaps PCI ACS-ish.  In
> > > > > the new host bridge case, each device gets a group.  This seems
> > > > > not to have any quirks for multifunction devices though.  On AMD
> > > > > and Intel IOMMUs we test multifunction device ACS support to
> > > > > determine whether all the functions should be in the same group.
> > > > > Is there any reason
> > > to trust multifunction devices on PAMU?
> > > > >
> > > > [Sethi Varun-B16395] In the case where we can partition endpoints
> > > > we can distinguish transactions based on the bus,device,function
> > > > number combination. This support is available in the PCIe
> > > > controller (host bridge).
> > >
> > > So can x86 IOMMUs, that's the visibility aspect of IOMMU groups.
> > > Visibility alone doesn't necessarily imply that a device is isolated
> > > though.  A multifunction PCI device that doesn't expose ACS support
> > > may not isolate functions from each other.  For example a
> > > peer-to-peer DMA between functions may not be translated by the
> > > upstream IOMMU.  IOMMU groups should encompass both visibility and
> isolation.
> > [Sethi Varun-B16395] We can isolate the DMA access to the host based
> > on the to the pci bus,device,function number.
> 
> The IOMMU can only isolate DMA that it can see.  A multifunction device
> may never expose peer-to-peer DMA to the upstream device, it's
> implementation specific.  The ACS flags allow that possibility to be
> controlled and prevented.
> 
> > I thought that was enough to put devices in to separate iommu groups.
> > This is a PCIe controller property which allows us to partition PCIe
> > devices. But, what I can understand from your point is that we also
> > need to consider isolation at PCIe device level as well. I will check
> > for the case of multifunction devices.
> >
> > >
> > > > > I also find it curious what happens to the iommu group of the
> > > > > host bridge.  In the partitionable case the host bridge group is
> > > > > removed, in the non-partitionable case the host bridge group
> > > > > becomes the group for the children, removing the host bridge.
> > > > > It's unique to PAMU so far that these host bridges are even in
> > > > > an iommu group (x86 only adds pci devices), but I don't see it
> > > > > as necessarily wrong leaving it in either scenario.  Does it
> > > > > solve some problem to remove
> > > them from the groups?
> > > > > Thanks,
> > > > [Sethi Varun-B16395] The PCIe controller isn't a partitionable
> > > > entity, it would always be owned by the host.
> > >
> > > Ownership of a device shouldn't play into the group context.  An
> > > IOMMU group should be defined by it's visibility and isolation from
> > > other devices.  Whether the PCIe controller is allowed to be handed
> > > to userspace is a question for VFIO.
> > [Sethi Varun-B16395] The problem is in the case, where we can't
> > partition PCIe devices. PCIe devices share the same device group as
> > the PCI controller. This becomes a problem while assigning the devices
> > to the guest, as you are required to unbind all the PCIe devices
> > including the controller from the host. PCIe controller can't be
> > unbound from the host, so we simply delete the controller iommu_group.
> 
> Unbinding devices is a VFIO implementation, it shouldn't leak into IOMMU
> groups.  Also note that VFIO has a driver white list where we can have
> exceptions to the rule.  I recently added pciehp to that list because the
> host driver provides functionality.  Being attached to the host driver
> means the device is not accessible to the user through VFIO, but other
> devices in the group are.  Thanks,
> 
Also, as Stuart pointed out the PCIe controller aren't the actual DMA devices (endpoints are the actual DMA devices). So, we remove the device group allocated for the PCIe controllers.

-Varun
diff mbox

Patch

diff --git a/arch/powerpc/sysdev/fsl_pci.h b/arch/powerpc/sysdev/fsl_pci.h
index c495c00..feb34f6 100644
--- a/arch/powerpc/sysdev/fsl_pci.h
+++ b/arch/powerpc/sysdev/fsl_pci.h
@@ -14,6 +14,11 @@ 
 #ifndef __POWERPC_FSL_PCI_H
 #define __POWERPC_FSL_PCI_H
 
+
+/* FSL PCI controller BRR1 register */
+#define PCI_FSL_BRR1      0xbf8
+#define PCI_FSL_BRR1_VER 0xffff
+
 #define PCIE_LTSSM	0x0404		/* PCIE Link Training and Status */
 #define PCIE_LTSSM_L0	0x16		/* L0 state */
 #define PCIE_IP_REV_2_2		0x02080202 /* PCIE IP block version Rev2.2 */
diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig
index c332fb9..3274333 100644
--- a/drivers/iommu/Kconfig
+++ b/drivers/iommu/Kconfig
@@ -17,6 +17,14 @@  config OF_IOMMU
        def_bool y
        depends on OF
 
+config FSL_PAMU
+	bool "Freescale IOMMU support"
+	depends on PPC_E500MC
+	select IOMMU_API
+	select GENERIC_ALLOCATOR
+	help
+	  Freescale PAMU support.
+
 # MSM IOMMU support
 config MSM_IOMMU
 	bool "MSM IOMMU Support"
diff --git a/drivers/iommu/Makefile b/drivers/iommu/Makefile
index ef0e520..027d1af 100644
--- a/drivers/iommu/Makefile
+++ b/drivers/iommu/Makefile
@@ -15,3 +15,4 @@  obj-$(CONFIG_TEGRA_IOMMU_SMMU) += tegra-smmu.o
 obj-$(CONFIG_EXYNOS_IOMMU) += exynos-iommu.o
 obj-$(CONFIG_SHMOBILE_IOMMU) += shmobile-iommu.o
 obj-$(CONFIG_SHMOBILE_IPMMU) += shmobile-ipmmu.o
+obj-$(CONFIG_FSL_PAMU) += fsl_pamu.o fsl_pamu_domain.o
diff --git a/drivers/iommu/fsl_pamu.c b/drivers/iommu/fsl_pamu.c
new file mode 100644
index 0000000..e0354dc
--- /dev/null
+++ b/drivers/iommu/fsl_pamu.c
@@ -0,0 +1,1269 @@ 
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright (C) 2013 Freescale Semiconductor, Inc.
+ *
+ */
+
+#define pr_fmt(fmt)    "fsl-pamu: %s: " fmt, __func__
+
+#include <linux/init.h>
+#include <linux/iommu.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/device.h>
+#include <linux/of_platform.h>
+#include <linux/bootmem.h>
+#include <linux/genalloc.h>
+#include <asm/io.h>
+#include <asm/bitops.h>
+#include <asm/fsl_guts.h>
+
+#include "fsl_pamu.h"
+
+/* define indexes for each operation mapping scenario */
+#define OMI_QMAN        0x00
+#define OMI_FMAN        0x01
+#define OMI_QMAN_PRIV   0x02
+#define OMI_CAAM        0x03
+
+/* Handling access violations */
+#define make64(high, low) (((u64)(high) << 32) | (low))
+
+struct pamu_isr_data {
+	void __iomem *pamu_reg_base;	/* Base address of PAMU regs*/
+	unsigned int count;		/* The number of PAMUs */
+};
+
+static struct paace *ppaact;
+static struct paace *spaact;
+static struct ome *omt;
+
+/*
+ * Table for matching compatible strings, for device tree
+ * guts node, for QorIQ SOCs.
+ * "fsl,qoriq-device-config-2.0" corresponds to T4 & B4
+ * SOCs. For the older SOCs "fsl,qoriq-device-config-1.0"
+ * string would be used.
+*/
+static const struct of_device_id guts_device_ids[] = {
+	{ .compatible = "fsl,qoriq-device-config-1.0", },
+	{ .compatible = "fsl,qoriq-device-config-2.0", },
+	{}
+};
+
+/* maximum subwindows permitted per liodn */
+static u32 max_subwindow_count;
+
+/* Pool for fspi allocation */
+struct gen_pool *spaace_pool;
+
+/**
+ * pamu_get_max_subwin_cnt() - Return the maximum supported
+ * subwindow count per liodn.
+ *
+ */
+u32 pamu_get_max_subwin_cnt()
+{
+	return max_subwindow_count;
+}
+
+/**
+ * pamu_get_ppaace() - Return the primary PACCE
+ * @liodn: liodn PAACT index for desired PAACE
+ *
+ * Returns the ppace pointer upon success else return
+ * null.
+ */
+static struct paace *pamu_get_ppaace(int liodn)
+{
+	if (!ppaact || liodn >= PAACE_NUMBER_ENTRIES) {
+		pr_err("PPAACT doesn't exist\n");
+		return NULL;
+	}
+
+	return &ppaact[liodn];
+}
+
+/**
+ * pamu_enable_liodn() - Set valid bit of PACCE
+ * @liodn: liodn PAACT index for desired PAACE
+ *
+ * Returns 0 upon success else error code < 0 returned
+ */
+int pamu_enable_liodn(int liodn)
+{
+	struct paace *ppaace;
+
+	ppaace = pamu_get_ppaace(liodn);
+	if (!ppaace) {
+		pr_err("Invalid primary paace entry\n");
+		return -ENOENT;
+	}
+
+	if (!get_bf(ppaace->addr_bitfields, PPAACE_AF_WSE)) {
+		pr_err("liodn %d not configured\n", liodn);
+		return -EINVAL;
+	}
+
+	/* Ensure that all other stores to the ppaace complete first */
+	mb();
+
+	ppaace->addr_bitfields |= PAACE_V_VALID;
+	mb();
+
+	return 0;
+}
+
+/**
+ * pamu_disable_liodn() - Clears valid bit of PACCE
+ * @liodn: liodn PAACT index for desired PAACE
+ *
+ * Returns 0 upon success else error code < 0 returned
+ */
+int pamu_disable_liodn(int liodn)
+{
+	struct paace *ppaace;
+
+	ppaace = pamu_get_ppaace(liodn);
+	if (!ppaace) {
+		pr_err("Invalid primary paace entry\n");
+		return -ENOENT;
+	}
+
+	set_bf(ppaace->addr_bitfields, PAACE_AF_V, PAACE_V_INVALID);
+	mb();
+
+	return 0;
+}
+
+/* Derive the window size encoding for a particular PAACE entry */
+static unsigned int map_addrspace_size_to_wse(phys_addr_t addrspace_size)
+{
+	/* Bug if not a power of 2 */
+	BUG_ON((addrspace_size & (addrspace_size - 1)));
+
+	/* window size is 2^(WSE+1) bytes */
+	return __ffs(addrspace_size >> PAMU_PAGE_SHIFT) + PAMU_PAGE_SHIFT - 1;
+}
+
+/* Derive the PAACE window count encoding for the subwindow count */
+static unsigned int map_subwindow_cnt_to_wce(u32 subwindow_cnt)
+{
+       /* window count is 2^(WCE+1) bytes */
+       return __ffs(subwindow_cnt) - 1;
+}
+
+/*
+ * Set the PAACE type as primary and set the coherency required domain
+ * attribute
+ */
+static void pamu_init_ppaace(struct paace *ppaace)
+{
+	set_bf(ppaace->addr_bitfields, PAACE_AF_PT, PAACE_PT_PRIMARY);
+
+	set_bf(ppaace->domain_attr.to_host.coherency_required, PAACE_DA_HOST_CR,
+	       PAACE_M_COHERENCE_REQ);
+}
+
+/*
+ * Set the PAACE type as secondary and set the coherency required domain
+ * attribute.
+ */
+static void pamu_init_spaace(struct paace *spaace)
+{
+	set_bf(spaace->addr_bitfields, PAACE_AF_PT, PAACE_PT_SECONDARY);
+	set_bf(spaace->domain_attr.to_host.coherency_required, PAACE_DA_HOST_CR,
+	       PAACE_M_COHERENCE_REQ);
+}
+
+/*
+ * Return the spaace (corresponding to the secondary window index)
+ * for a particular ppaace.
+ */
+static struct paace *pamu_get_spaace(struct paace *paace, u32 wnum)
+{
+	u32 subwin_cnt;
+	struct paace *spaace = NULL;
+
+	subwin_cnt = 1UL << (get_bf(paace->impl_attr, PAACE_IA_WCE) + 1);
+
+	if (wnum < subwin_cnt)
+		spaace = &spaact[paace->fspi + wnum];
+	else
+		pr_err("secondary paace out of bounds\n");
+
+	return spaace;
+}
+
+/**
+ * pamu_get_fspi_and_allocate() - Allocates fspi index and reserves subwindows
+ *                                required for primary PAACE in the secondary
+ *                                PAACE table.
+ * @subwin_cnt: Number of subwindows to be reserved.
+ *
+ * A PPAACE entry may have a number of associated subwindows. A subwindow
+ * corresponds to a SPAACE entry in the SPAACT table. Each PAACE entry stores
+ * the index (fspi) of the first SPAACE entry in the SPAACT table. This
+ * function returns the index of the first SPAACE entry. The remaining
+ * SPAACE entries are reserved contiguously from that index.
+ *
+ * Returns a valid fspi index in the range of 0 - SPAACE_NUMBER_ENTRIES on success.
+ * If no SPAACE entry is available or the allocator can not reserve the required
+ * number of contiguous entries function returns ULONG_MAX indicating a failure.
+ *
+*/
+static unsigned long pamu_get_fspi_and_allocate(u32 subwin_cnt)
+{
+	unsigned long spaace_addr;
+
+	spaace_addr = gen_pool_alloc(spaace_pool, subwin_cnt * sizeof(struct paace));
+	if (!spaace_addr)
+		return ULONG_MAX;
+
+	return (spaace_addr - (unsigned long)spaact) / (sizeof(struct paace));
+}
+
+/* Release the subwindows reserved for a particular LIODN */
+void pamu_free_subwins(int liodn)
+{
+	struct paace *ppaace;
+	u32 subwin_cnt, size;
+
+	ppaace = pamu_get_ppaace(liodn);
+	if (!ppaace) {
+		pr_err("Invalid liodn entry\n");
+		return;
+	}
+
+	if (get_bf(ppaace->addr_bitfields, PPAACE_AF_MW)) {
+		subwin_cnt = 1UL << (get_bf(ppaace->impl_attr, PAACE_IA_WCE) + 1);
+		size = (subwin_cnt - 1) * sizeof(struct paace);
+		gen_pool_free(spaace_pool, (unsigned long)&spaact[ppaace->fspi], size);
+		set_bf(ppaace->addr_bitfields, PPAACE_AF_MW, 0);
+	}
+}
+
+/*
+ * Function used for updating stash destination for the coressponding
+ * LIODN.
+ */
+int  pamu_update_paace_stash(int liodn, u32 subwin, u32 value)
+{
+	struct paace *paace;
+
+	paace = pamu_get_ppaace(liodn);
+	if (!paace) {
+		pr_err("Invalid liodn entry\n");
+		return -ENOENT;
+	}
+	if (subwin) {
+		paace = pamu_get_spaace(paace, subwin - 1);
+		if (!paace) {
+			return -ENOENT;
+		}
+	}
+	set_bf(paace->impl_attr, PAACE_IA_CID, value);
+
+	mb();
+
+	return 0;
+}
+
+/* Disable a subwindow corresponding to the LIODN */
+int pamu_disable_spaace(int liodn, u32 subwin)
+{
+	struct paace *paace;
+
+	paace = pamu_get_ppaace(liodn);
+	if (!paace) {
+		pr_err("Invalid liodn entry\n");
+		return -ENOENT;
+	}
+	if (subwin) {
+		paace = pamu_get_spaace(paace, subwin - 1);
+		if (!paace) {
+			return -ENOENT;
+		}
+		set_bf(paace->addr_bitfields, PAACE_AF_V,
+			 PAACE_V_INVALID);
+	} else {
+		set_bf(paace->addr_bitfields, PAACE_AF_AP,
+			 PAACE_AP_PERMS_DENIED);
+	}
+
+	mb();
+
+	return 0;
+}
+
+
+/**
+ * pamu_config_paace() - Sets up PPAACE entry for specified liodn
+ *
+ * @liodn: Logical IO device number
+ * @win_addr: starting address of DSA window
+ * @win-size: size of DSA window
+ * @omi: Operation mapping index -- if ~omi == 0 then omi not defined
+ * @rpn: real (true physical) page number
+ * @stashid: cache stash id for associated cpu -- if ~stashid == 0 then
+ *	     stashid not defined
+ * @snoopid: snoop id for hardware coherency -- if ~snoopid == 0 then
+ *	     snoopid not defined
+ * @subwin_cnt: number of sub-windows
+ * @prot: window permissions
+ *
+ * Returns 0 upon success else error code < 0 returned
+ */
+int pamu_config_ppaace(int liodn, phys_addr_t win_addr, phys_addr_t win_size,
+		       u32 omi, unsigned long rpn, u32 snoopid, u32 stashid,
+		       u32 subwin_cnt, int prot)
+{
+	struct paace *ppaace;
+	unsigned long fspi;
+
+	if ((win_size & (win_size - 1)) || win_size < PAMU_PAGE_SIZE) {
+		pr_err("window size too small or not a power of two %llx\n", win_size);
+		return -EINVAL;
+	}
+
+	if (win_addr & (win_size - 1)) {
+		pr_err("window address is not aligned with window size\n");
+		return -EINVAL;
+	}
+
+	ppaace = pamu_get_ppaace(liodn);
+	if (!ppaace) {
+		return -ENOENT;
+	}
+
+	/* window size is 2^(WSE+1) bytes */
+	set_bf(ppaace->addr_bitfields, PPAACE_AF_WSE,
+		map_addrspace_size_to_wse(win_size));
+
+	pamu_init_ppaace(ppaace);
+
+	ppaace->wbah = win_addr >> (PAMU_PAGE_SHIFT + 20);
+	set_bf(ppaace->addr_bitfields, PPAACE_AF_WBAL,
+	       (win_addr >> PAMU_PAGE_SHIFT));
+
+	/* set up operation mapping if it's configured */
+	if (omi < OME_NUMBER_ENTRIES) {
+		set_bf(ppaace->impl_attr, PAACE_IA_OTM, PAACE_OTM_INDEXED);
+		ppaace->op_encode.index_ot.omi = omi;
+	} else if (~omi != 0) {
+		pr_err("bad operation mapping index: %d\n", omi);
+		return -EINVAL;
+	}
+
+	/* configure stash id */
+	if (~stashid != 0)
+		set_bf(ppaace->impl_attr, PAACE_IA_CID, stashid);
+
+	/* configure snoop id */
+	if (~snoopid != 0)
+		ppaace->domain_attr.to_host.snpid = snoopid;
+
+	if (subwin_cnt) {
+		/* The first entry is in the primary PAACE instead */
+		fspi = pamu_get_fspi_and_allocate(subwin_cnt - 1);
+		if (fspi == ULONG_MAX) {
+			pr_err("spaace indexes exhausted\n");
+			return -EINVAL;
+		}
+
+		/* window count is 2^(WCE+1) bytes */
+		set_bf(ppaace->impl_attr, PAACE_IA_WCE,
+		       map_subwindow_cnt_to_wce(subwin_cnt));
+		set_bf(ppaace->addr_bitfields, PPAACE_AF_MW, 0x1);
+		ppaace->fspi = fspi;
+	} else {
+		set_bf(ppaace->impl_attr, PAACE_IA_ATM, PAACE_ATM_WINDOW_XLATE);
+		ppaace->twbah = rpn >> 20;
+		set_bf(ppaace->win_bitfields, PAACE_WIN_TWBAL, rpn);
+		set_bf(ppaace->addr_bitfields, PAACE_AF_AP, prot);
+		set_bf(ppaace->impl_attr, PAACE_IA_WCE, 0);
+		set_bf(ppaace->addr_bitfields, PPAACE_AF_MW, 0);
+	}
+	mb();
+
+	return 0;
+}
+
+/**
+ * pamu_config_spaace() - Sets up SPAACE entry for specified subwindow
+ *
+ * @liodn:  Logical IO device number
+ * @subwin_cnt:  number of sub-windows associated with dma-window
+ * @subwin: subwindow index
+ * @subwin_size: size of subwindow
+ * @omi: Operation mapping index
+ * @rpn: real (true physical) page number
+ * @snoopid: snoop id for hardware coherency -- if ~snoopid == 0 then
+ *			  snoopid not defined
+ * @stashid: cache stash id for associated cpu
+ * @enable: enable/disable subwindow after reconfiguration
+ * @prot: sub window permissions
+ *
+ * Returns 0 upon success else error code < 0 returned
+ */
+int pamu_config_spaace(int liodn, u32 subwin_cnt, u32 subwin,
+		       phys_addr_t subwin_size, u32 omi, unsigned long rpn,
+		       u32 snoopid, u32 stashid, int enable, int prot)
+{
+	struct paace *paace;
+
+
+	/* setup sub-windows */
+	if (!subwin_cnt) {
+		pr_err("Invalid subwindow count\n");
+		return -EINVAL;
+	}
+
+	paace = pamu_get_ppaace(liodn);
+	if (subwin > 0 && subwin < subwin_cnt && paace) {
+		paace = pamu_get_spaace(paace, subwin - 1);
+
+		if (paace && !(paace->addr_bitfields & PAACE_V_VALID)) {
+			pamu_init_spaace(paace);
+			set_bf(paace->addr_bitfields, SPAACE_AF_LIODN, liodn);
+		}
+	}
+
+	if (!paace) {
+		pr_err("Invalid liodn entry\n");
+		return -ENOENT;
+	}
+
+	if (subwin_size & (subwin_size - 1) || subwin_size < PAMU_PAGE_SIZE) {
+		pr_err("subwindow size out of range, or not a power of 2\n");
+		return -EINVAL;
+	}
+
+	if (rpn == ULONG_MAX) {
+		pr_err("real page number out of range\n");
+		return -EINVAL;
+	}
+
+	/* window size is 2^(WSE+1) bytes */
+	set_bf(paace->win_bitfields, PAACE_WIN_SWSE,
+	       map_addrspace_size_to_wse(subwin_size));
+
+	set_bf(paace->impl_attr, PAACE_IA_ATM, PAACE_ATM_WINDOW_XLATE);
+	paace->twbah = rpn >> 20;
+	set_bf(paace->win_bitfields, PAACE_WIN_TWBAL, rpn);
+	set_bf(paace->addr_bitfields, PAACE_AF_AP, prot);
+
+	/* configure snoop id */
+	if (~snoopid != 0)
+		paace->domain_attr.to_host.snpid = snoopid;
+
+	/* set up operation mapping if it's configured */
+	if (omi < OME_NUMBER_ENTRIES) {
+		set_bf(paace->impl_attr, PAACE_IA_OTM, PAACE_OTM_INDEXED);
+		paace->op_encode.index_ot.omi = omi;
+	} else if (~omi != 0) {
+		pr_err("bad operation mapping index: %d\n", omi);
+		return -EINVAL;
+	}
+
+	if (~stashid != 0)
+		set_bf(paace->impl_attr, PAACE_IA_CID, stashid);
+
+	smp_wmb();
+
+	if (enable)
+		paace->addr_bitfields |= PAACE_V_VALID;
+
+	mb();
+
+	return 0;
+}
+
+/**
+* get_ome_index() - Returns the index in the operation mapping table
+*                   for device.
+* @*omi_index: pointer for storing the index value
+*
+*/
+void get_ome_index(u32 *omi_index, struct device *dev)
+{
+	if (of_device_is_compatible(dev->of_node, "fsl,qman-portal"))
+		*omi_index = OMI_QMAN;
+	if (of_device_is_compatible(dev->of_node, "fsl,qman"))
+		*omi_index = OMI_QMAN_PRIV;
+}
+
+/**
+ * get_stash_id - Returns stash destination id corresponding to a
+ *                cache type and vcpu.
+ * @stash_dest_hint: L1, L2 or L3
+ * @vcpu: vpcu target for a particular cache type.
+ *
+ * Returs stash on success or ~(u32)0 on failure.
+ *
+ */
+u32 get_stash_id(u32 stash_dest_hint, u32 vcpu)
+{
+	const u32 *prop;
+	struct device_node *node;
+	u32 cache_level;
+	int len;
+
+	/* Fastpath, exit early if L3/CPC cache is target for stashing */
+	if (stash_dest_hint == IOMMU_ATTR_CACHE_L3) {
+		node = of_find_compatible_node(NULL, NULL,
+				"fsl,p4080-l3-cache-controller");
+		if (node) {
+			prop = of_get_property(node, "cache-stash-id", 0);
+			if (!prop) {
+				pr_err("missing cache-stash-id at %s\n", node->full_name);
+				of_node_put(node);
+				return ~(u32)0;
+			}
+			of_node_put(node);
+			return be32_to_cpup(prop);
+		}
+		return ~(u32)0;
+	}
+
+	for_each_node_by_type(node, "cpu") {
+		prop = of_get_property(node, "reg", &len);
+		if (be32_to_cpup(prop) == vcpu)
+			break;
+	}
+
+	/* find the hwnode that represents the cache */
+	for (cache_level = IOMMU_ATTR_CACHE_L1; cache_level < IOMMU_ATTR_CACHE_L3; cache_level++) {
+		if (stash_dest_hint == cache_level) {
+			prop = of_get_property(node, "cache-stash-id", 0);
+			if (!prop) {
+				pr_err("missing cache-stash-id at %s\n", node->full_name);
+				of_node_put(node);
+				return ~(u32)0;
+			}
+			of_node_put(node);
+			return be32_to_cpup(prop);
+		}
+
+		prop = of_get_property(node, "next-level-cache", 0);
+		if (!prop) {
+			pr_err("can't find next-level-cache at %s\n",
+				node->full_name);
+			of_node_put(node);
+			return ~(u32)0;  /* can't traverse any further */
+		}
+		of_node_put(node);
+
+		/* advance to next node in cache hierarchy */
+		node = of_find_node_by_phandle(*prop);
+		if (!node) {
+			pr_err("Invalid node for cache hierarchy %s\n",
+				node->full_name);
+			return ~(u32)0;
+		}
+	}
+
+	pr_err("stash dest not found for %d on vcpu %d\n",
+	          stash_dest_hint, vcpu);
+	return ~(u32)0;
+}
+
+/* Identify if the PAACT table entry belongs to QMAN, BMAN or QMAN Portal */
+#define QMAN_PAACE 1
+#define QMAN_PORTAL_PAACE 2
+#define BMAN_PAACE 3
+
+/**
+ * Setup operation mapping and stash destinations for QMAN and QMAN portal.
+ * Memory accesses to QMAN and BMAN private memory need not be coherent, so
+ * clear the PAACE entry coherency attribute for them.
+ */
+static void setup_qbman_paace(struct paace *ppaace, int  paace_type)
+{
+	switch (paace_type) {
+	case QMAN_PAACE:
+		set_bf(ppaace->impl_attr, PAACE_IA_OTM, PAACE_OTM_INDEXED);
+		ppaace->op_encode.index_ot.omi = OMI_QMAN_PRIV;
+		/* setup QMAN Private data stashing for the L3 cache */
+		set_bf(ppaace->impl_attr, PAACE_IA_CID, get_stash_id(IOMMU_ATTR_CACHE_L3, 0));
+		set_bf(ppaace->domain_attr.to_host.coherency_required, PAACE_DA_HOST_CR,
+		       0);
+		break;
+	case QMAN_PORTAL_PAACE:
+		set_bf(ppaace->impl_attr, PAACE_IA_OTM, PAACE_OTM_INDEXED);
+		ppaace->op_encode.index_ot.omi = OMI_QMAN;
+		/*Set DQRR and Frame stashing for the L3 cache */
+		set_bf(ppaace->impl_attr, PAACE_IA_CID, get_stash_id(IOMMU_ATTR_CACHE_L3, 0));
+		break;
+	case BMAN_PAACE:
+		set_bf(ppaace->domain_attr.to_host.coherency_required, PAACE_DA_HOST_CR,
+		       0);
+		break;
+	}
+}
+
+/**
+ * Setup the operation mapping table for various devices. This is a static
+ * table where each table index corresponds to a particular device. PAMU uses
+ * this table to translate device transaction to appropriate corenet
+ * transaction.
+ */
+static void __init setup_omt(struct ome *omt)
+{
+	struct ome *ome;
+
+	/* Configure OMI_QMAN */
+	ome = &omt[OMI_QMAN];
+
+	ome->moe[IOE_READ_IDX] = EOE_VALID | EOE_READ;
+	ome->moe[IOE_EREAD0_IDX] = EOE_VALID | EOE_RSA;
+	ome->moe[IOE_WRITE_IDX] = EOE_VALID | EOE_WRITE;
+	ome->moe[IOE_EWRITE0_IDX] = EOE_VALID | EOE_WWSAO;
+
+	ome->moe[IOE_DIRECT0_IDX] = EOE_VALID | EOE_LDEC;
+	ome->moe[IOE_DIRECT1_IDX] = EOE_VALID | EOE_LDECPE;
+
+	/* Configure OMI_FMAN */
+	ome = &omt[OMI_FMAN];
+	ome->moe[IOE_READ_IDX]  = EOE_VALID | EOE_READI;
+	ome->moe[IOE_WRITE_IDX] = EOE_VALID | EOE_WRITE;
+
+	/* Configure OMI_QMAN private */
+	ome = &omt[OMI_QMAN_PRIV];
+	ome->moe[IOE_READ_IDX]  = EOE_VALID | EOE_READ;
+	ome->moe[IOE_WRITE_IDX] = EOE_VALID | EOE_WRITE;
+	ome->moe[IOE_EREAD0_IDX] = EOE_VALID | EOE_RSA;
+	ome->moe[IOE_EWRITE0_IDX] = EOE_VALID | EOE_WWSA;
+
+	/* Configure OMI_CAAM */
+	ome = &omt[OMI_CAAM];
+	ome->moe[IOE_READ_IDX]  = EOE_VALID | EOE_READI;
+	ome->moe[IOE_WRITE_IDX] = EOE_VALID | EOE_WRITE;
+}
+
+/*
+ * Get the maximum number of PAACT table entries
+ * and subwindows supported by PAMU
+ */
+static void get_pamu_cap_values(unsigned long pamu_reg_base)
+{
+	u32 pc_val;
+
+	pc_val = in_be32((u32 *)(pamu_reg_base + PAMU_PC3));
+	/* Maximum number of subwindows per liodn */
+	max_subwindow_count = 1 << (1 + PAMU_PC3_MWCE(pc_val));
+}
+
+/* Setup PAMU registers pointing to PAACT, SPAACT and OMT */
+int setup_one_pamu(unsigned long pamu_reg_base, unsigned long pamu_reg_size,
+	           phys_addr_t ppaact_phys, phys_addr_t spaact_phys,
+		   phys_addr_t omt_phys)
+{
+	u32 *pc;
+	struct pamu_mmap_regs *pamu_regs;
+
+	pc = (u32 *) (pamu_reg_base + PAMU_PC);
+	pamu_regs = (struct pamu_mmap_regs *)
+		(pamu_reg_base + PAMU_MMAP_REGS_BASE);
+
+	/* set up pointers to corenet control blocks */
+
+	out_be32(&pamu_regs->ppbah, upper_32_bits(ppaact_phys));
+	out_be32(&pamu_regs->ppbal, lower_32_bits(ppaact_phys));
+	ppaact_phys = ppaact_phys + PAACT_SIZE;
+	out_be32(&pamu_regs->pplah, upper_32_bits(ppaact_phys));
+	out_be32(&pamu_regs->pplal, lower_32_bits(ppaact_phys));
+
+	out_be32(&pamu_regs->spbah, upper_32_bits(spaact_phys));
+	out_be32(&pamu_regs->spbal, lower_32_bits(spaact_phys));
+	spaact_phys = spaact_phys + SPAACT_SIZE;
+	out_be32(&pamu_regs->splah, upper_32_bits(spaact_phys));
+	out_be32(&pamu_regs->splal, lower_32_bits(spaact_phys));
+
+	out_be32(&pamu_regs->obah, upper_32_bits(omt_phys));
+	out_be32(&pamu_regs->obal, lower_32_bits(omt_phys));
+	omt_phys = omt_phys + OMT_SIZE;
+	out_be32(&pamu_regs->olah, upper_32_bits(omt_phys));
+	out_be32(&pamu_regs->olal, lower_32_bits(omt_phys));
+
+	/*
+	 * set PAMU enable bit,
+	 * allow ppaact & omt to be cached
+	 * & enable PAMU access violation interrupts.
+	 */
+
+	out_be32((u32 *)(pamu_reg_base + PAMU_PICS),
+			PAMU_ACCESS_VIOLATION_ENABLE);
+	out_be32(pc, PAMU_PC_PE | PAMU_PC_OCE | PAMU_PC_SPCC | PAMU_PC_PPCC);
+	return 0;
+}
+
+/* Enable all device LIODNS */
+static void __init setup_liodns(void)
+{
+	int i, len;
+	struct paace *ppaace;
+	struct device_node *node = NULL;
+	const u32 *prop;
+
+	for_each_node_with_property(node, "fsl,liodn") {
+		prop = of_get_property(node, "fsl,liodn", &len);
+		for (i = 0; i < len / sizeof(u32); i++) {
+			int liodn;
+
+			liodn = be32_to_cpup(&prop[i]);
+			if (liodn >= PAACE_NUMBER_ENTRIES) {
+				pr_err("Invalid LIODN value %d\n", liodn);
+				continue;
+			}
+			ppaace = pamu_get_ppaace(liodn);
+			pamu_init_ppaace(ppaace);
+			/* window size is 2^(WSE+1) bytes */
+			set_bf(ppaace->addr_bitfields, PPAACE_AF_WSE, 35);
+			ppaace->wbah = 0;
+			set_bf(ppaace->addr_bitfields, PPAACE_AF_WBAL, 0);
+			set_bf(ppaace->impl_attr, PAACE_IA_ATM,
+				PAACE_ATM_NO_XLATE);
+			set_bf(ppaace->addr_bitfields, PAACE_AF_AP,
+				PAACE_AP_PERMS_ALL);
+			if (of_device_is_compatible(node, "fsl,qman-portal"))
+				setup_qbman_paace(ppaace, QMAN_PORTAL_PAACE);
+			if (of_device_is_compatible(node, "fsl,qman"))
+				setup_qbman_paace(ppaace, QMAN_PAACE);
+			if (of_device_is_compatible(node, "fsl,bman"))
+				setup_qbman_paace(ppaace, BMAN_PAACE);
+			mb();
+			pamu_enable_liodn(liodn);
+		}
+	}
+}
+
+irqreturn_t pamu_av_isr(int irq, void *arg)
+{
+	struct pamu_isr_data *data = arg;
+	phys_addr_t phys;
+	unsigned int i, j;
+
+	pr_emerg("fsl-pamu: access violation interrupt\n");
+
+	for (i = 0; i < data->count; i++) {
+		void __iomem *p = data->pamu_reg_base + i * PAMU_OFFSET;
+		u32 pics = in_be32(p + PAMU_PICS);
+
+		if (pics & PAMU_ACCESS_VIOLATION_STAT) {
+			pr_emerg("POES1=%08x\n", in_be32(p + PAMU_POES1));
+			pr_emerg("POES2=%08x\n", in_be32(p + PAMU_POES2));
+			pr_emerg("AVS1=%08x\n", in_be32(p + PAMU_AVS1));
+			pr_emerg("AVS2=%08x\n", in_be32(p + PAMU_AVS2));
+			pr_emerg("AVA=%016llx\n", make64(in_be32(p + PAMU_AVAH),
+				in_be32(p + PAMU_AVAL)));
+			pr_emerg("UDAD=%08x\n", in_be32(p + PAMU_UDAD));
+			pr_emerg("POEA=%016llx\n", make64(in_be32(p + PAMU_POEAH),
+				in_be32(p + PAMU_POEAL)));
+
+			phys = make64(in_be32(p + PAMU_POEAH),
+				in_be32(p + PAMU_POEAL));
+
+			/* Assume that POEA points to a PAACE */
+			if (phys) {
+				u32 *paace = phys_to_virt(phys);
+
+				/* Only the first four words are relevant */
+				for (j = 0; j < 4; j++)
+					pr_emerg("PAACE[%u]=%08x\n", j, in_be32(paace + j));
+			}
+		}
+	}
+
+	panic("\n");
+	/* NOTREACHED */
+
+	return IRQ_HANDLED;
+}
+
+#define LAWAR_EN		0x80000000
+#define LAWAR_TARGET_MASK	0x0FF00000
+#define LAWAR_TARGET_SHIFT	20
+#define LAWAR_SIZE_MASK		0x0000003F
+#define LAWAR_CSDID_MASK	0x000FF000
+#define LAWAR_CSDID_SHIFT	12
+
+#define LAW_SIZE_4K		0xb
+
+struct ccsr_law {
+	u32	lawbarh;	/* LAWn base address high */
+	u32	lawbarl;	/* LAWn base address low */
+	u32	lawar;		/* LAWn attributes */
+	u32	reserved;
+};
+
+#define make64(high, low) (((u64)(high) << 32) | (low))
+
+/*
+ * Create a coherence subdomain for a given memory block.
+ */
+static int __init create_csd(phys_addr_t phys, size_t size, u32 csd_port_id)
+{
+	struct device_node *np;
+	const __be32 *iprop;
+	void __iomem *lac = NULL;	/* Local Access Control registers */
+	struct ccsr_law __iomem *law;
+	void __iomem *ccm = NULL;
+	u32 __iomem *csdids;
+	unsigned int i, num_laws, num_csds;
+	u32 law_target = 0;
+	u32 csd_id = 0;
+	int ret = 0;
+
+	np = of_find_compatible_node(NULL, NULL, "fsl,corenet-law");
+	if (!np)
+		return -ENODEV;
+
+	iprop = of_get_property(np, "fsl,num-laws", NULL);
+	if (!iprop) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	num_laws = be32_to_cpup(iprop);
+	if (!num_laws) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	lac = of_iomap(np, 0);
+	if (!lac) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	/* LAW registers are at offset 0xC00 */
+	law = lac + 0xC00;
+
+	of_node_put(np);
+
+	np = of_find_compatible_node(NULL, NULL, "fsl,corenet-cf");
+	if (!np) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	iprop = of_get_property(np, "fsl,ccf-num-csdids", NULL);
+	if (!iprop) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	num_csds = be32_to_cpup(iprop);
+	if (!num_csds) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	ccm = of_iomap(np, 0);
+	if (!ccm) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	/* The undocumented CSDID registers are at offset 0x600 */
+	csdids = ccm + 0x600;
+
+	of_node_put(np);
+	np = NULL;
+
+	/* Find an unused coherence subdomain ID */
+	for (csd_id = 0; csd_id < num_csds; csd_id++) {
+		if (!csdids[csd_id])
+			break;
+	}
+
+	/* Store the Port ID in the (undocumented) proper CIDMRxx register */
+	csdids[csd_id] = csd_port_id;
+
+	/* Find the DDR LAW that maps to our buffer. */
+	for (i = 0; i < num_laws; i++) {
+		if (law[i].lawar & LAWAR_EN) {
+			phys_addr_t law_start, law_end;
+
+			law_start = make64(law[i].lawbarh, law[i].lawbarl);
+			law_end = law_start +
+				(2ULL << (law[i].lawar & LAWAR_SIZE_MASK));
+
+			if (law_start <= phys && phys < law_end) {
+				law_target = law[i].lawar & LAWAR_TARGET_MASK;
+				break;
+			}
+		}
+	}
+
+	if (i == 0 || i == num_laws) {
+		/* This should never happen*/
+		ret = -ENOENT;
+		goto error;
+	}
+
+	/* Find a free LAW entry */
+	while (law[--i].lawar & LAWAR_EN) {
+		if (i == 0) {
+			/* No higher priority LAW slots available */
+			ret = -ENOENT;
+			goto error;
+		}
+	}
+
+	law[i].lawbarh = upper_32_bits(phys);
+	law[i].lawbarl = lower_32_bits(phys);
+	wmb();
+	law[i].lawar = LAWAR_EN | law_target | (csd_id << LAWAR_CSDID_SHIFT) |
+		(LAW_SIZE_4K + get_order(size));
+	wmb();
+
+error:
+	if (ccm)
+		iounmap(ccm);
+
+	if (lac)
+		iounmap(lac);
+
+	if (np)
+		of_node_put(np);
+
+	return ret;
+}
+
+/*
+ * Table of SVRs and the corresponding PORT_ID values. Port ID corresponds to a 
+ * bit map of snoopers for a given range of memory mapped by a LAW.
+ *
+ * All future CoreNet-enabled SOCs will have this erratum(A-004510) fixed, so this
+ * table should never need to be updated.  SVRs are guaranteed to be unique, so
+ * there is no worry that a future SOC will inadvertently have one of these
+ * values.
+ */
+static const struct {
+	u32 svr;
+	u32 port_id;
+} port_id_map[] = {
+	{0x82100010, 0xFF000000},	/* P2040 1.0 */
+	{0x82100011, 0xFF000000},	/* P2040 1.1 */
+	{0x82100110, 0xFF000000},	/* P2041 1.0 */
+	{0x82100111, 0xFF000000},	/* P2041 1.1 */
+	{0x82110310, 0xFF000000},	/* P3041 1.0 */
+	{0x82110311, 0xFF000000},	/* P3041 1.1 */
+	{0x82010020, 0xFFF80000},	/* P4040 2.0 */
+	{0x82000020, 0xFFF80000},	/* P4080 2.0 */
+	{0x82210010, 0xFC000000},       /* P5010 1.0 */
+	{0x82210020, 0xFC000000},       /* P5010 2.0 */
+	{0x82200010, 0xFC000000},	/* P5020 1.0 */
+	{0x82050010, 0xFF800000},	/* P5021 1.0 */
+	{0x82040010, 0xFF800000},	/* P5040 1.0 */
+};
+
+#define SVR_SECURITY	0x80000	/* The Security (E) bit */
+
+static int __init fsl_pamu_probe(struct platform_device *pdev)
+{
+	void __iomem *pamu_regs = NULL;
+	struct ccsr_guts __iomem *guts_regs = NULL;
+	u32 pamubypenr, pamu_counter;
+	unsigned long pamu_reg_off;
+	unsigned long pamu_reg_base;
+	struct pamu_isr_data *data = NULL;
+	struct device_node *guts_node;
+	u64 size;
+	struct page *p;
+	int ret = 0;
+	int irq;
+	phys_addr_t ppaact_phys;
+	phys_addr_t spaact_phys;
+	phys_addr_t omt_phys;
+	size_t mem_size = 0;
+	unsigned int order = 0;
+	u32 csd_port_id = 0;
+	unsigned i;
+	/*
+	 * enumerate all PAMUs and allocate and setup PAMU tables
+	 * for each of them,
+	 * NOTE : All PAMUs share the same LIODN tables.
+	 */
+
+	pamu_regs = of_iomap(pdev->dev.of_node, 0);
+	if (!pamu_regs) {
+		dev_err(&pdev->dev, "ioremap of PAMU node failed\n");
+		return -ENOMEM;
+	}
+	of_get_address(pdev->dev.of_node, 0, &size, NULL);
+
+	irq = irq_of_parse_and_map(pdev->dev.of_node, 0);
+	if (irq == NO_IRQ) {
+		dev_warn(&pdev->dev, "no interrupts listed in PAMU node\n");
+		goto error;
+	}
+
+	data = kzalloc(sizeof(struct pamu_isr_data), GFP_KERNEL);
+	if (!data) {
+		dev_err(&pdev->dev, "PAMU isr data memory allocation failed\n");
+		ret = -ENOMEM;
+		goto error;
+	}
+	data->pamu_reg_base = pamu_regs;
+	data->count = size / PAMU_OFFSET;
+
+	/* The ISR needs access to the regs, so we won't iounmap them */
+	ret = request_irq(irq, pamu_av_isr, 0, "pamu", data);
+	if (ret < 0) {
+		dev_err(&pdev->dev, "error %i installing ISR for irq %i\n",
+			ret, irq);
+		goto error;
+	}
+
+	guts_node = of_find_matching_node(NULL, guts_device_ids);
+	if (!guts_node) {
+		dev_err(&pdev->dev, "could not find GUTS node %s\n",
+			pdev->dev.of_node->full_name);
+		ret = -ENODEV;
+		goto error;
+	}
+
+	guts_regs = of_iomap(guts_node, 0);
+	of_node_put(guts_node);
+	if (!guts_regs) {
+		dev_err(&pdev->dev, "ioremap of GUTS node failed\n");
+		ret = -ENODEV;
+		goto error;
+	}
+
+	/* read in the PAMU capability registers */
+	get_pamu_cap_values((unsigned long)pamu_regs);
+	/*
+	 * To simplify the allocation of a coherency domain, we allocate the
+	 * PAACT and the OMT in the same memory buffer.  Unfortunately, this
+	 * wastes more memory compared to allocating the buffers separately.
+	 */
+	/* Determine how much memory we need */
+	mem_size = (PAGE_SIZE << get_order(PAACT_SIZE)) +
+		(PAGE_SIZE << get_order(SPAACT_SIZE)) +
+		(PAGE_SIZE << get_order(OMT_SIZE));
+	order = get_order(mem_size);
+
+	p = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
+	if (!p) {
+		dev_err(&pdev->dev, "unable to allocate PAACT/SPAACT/OMT block\n");
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	ppaact = page_address(p);
+	ppaact_phys = page_to_phys(p);
+
+	/* Make sure the memory is naturally aligned */
+	if (ppaact_phys & ((PAGE_SIZE << order) - 1)) {
+		dev_err(&pdev->dev, "PAACT/OMT block is unaligned\n");
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	spaact = (void *)ppaact + (PAGE_SIZE << get_order(PAACT_SIZE));
+	omt = (void *)spaact + (PAGE_SIZE << get_order(SPAACT_SIZE));
+
+	dev_dbg(&pdev->dev, "ppaact virt=%p phys=0x%llx\n", ppaact,
+		(unsigned long long) ppaact_phys);
+
+	/* Check to see if we need to implement the work-around on this SOC */
+
+	/* Determine the Port ID for our coherence subdomain */
+	for (i = 0; i < ARRAY_SIZE(port_id_map); i++) {
+		if (port_id_map[i].svr == (mfspr(SPRN_SVR) & ~SVR_SECURITY)) {
+			csd_port_id = port_id_map[i].port_id;
+			dev_dbg(&pdev->dev, "found matching SVR %08x\n",
+				port_id_map[i].svr);
+			break;
+		}
+	}
+
+	if (csd_port_id) {
+		dev_dbg(&pdev->dev, "creating coherency subdomain at address "
+			"0x%llx, size %zu, port id 0x%08x", ppaact_phys,
+			mem_size, csd_port_id);
+
+		ret = create_csd(ppaact_phys, mem_size, csd_port_id);
+		if (ret) {
+			dev_err(&pdev->dev, "could not create coherence "
+				"subdomain\n");
+			return ret;
+		}
+	}
+
+	spaact_phys = virt_to_phys(spaact);
+	omt_phys = virt_to_phys(omt);
+
+	spaace_pool = gen_pool_create(ilog2(sizeof(struct paace)), -1);
+	if (!spaace_pool) {
+		ret = -ENOMEM;
+		dev_err(&pdev->dev, "PAMU : failed to allocate spaace gen pool\n");
+		goto error;
+	}
+
+	ret = gen_pool_add(spaace_pool, (unsigned long)spaact, SPAACT_SIZE, -1);
+	if (ret)
+		goto error_genpool;
+
+	pamubypenr = in_be32(&guts_regs->pamubypenr);
+
+	for (pamu_reg_off = 0, pamu_counter = 0x80000000; pamu_reg_off < size;
+	     pamu_reg_off += PAMU_OFFSET, pamu_counter >>= 1) {
+
+		pamu_reg_base = (unsigned long) pamu_regs + pamu_reg_off;
+		setup_one_pamu(pamu_reg_base, pamu_reg_off, ppaact_phys,
+				 spaact_phys, omt_phys);
+		/* Disable PAMU bypass for this PAMU */
+		pamubypenr &= ~pamu_counter;
+	}
+
+	setup_omt(omt);
+
+	/* Enable all relevant PAMU(s) */
+	out_be32(&guts_regs->pamubypenr, pamubypenr);
+
+	iounmap(guts_regs);
+
+	/* Enable DMA for the LIODNs in the device tree*/
+
+	setup_liodns();
+
+	return 0;
+
+error_genpool:
+	gen_pool_destroy(spaace_pool);
+
+error:
+	if (irq != NO_IRQ)
+		free_irq(irq, data);
+
+	if (data) {
+		memset(data, 0, sizeof(struct pamu_isr_data));
+		kfree(data);
+	}
+
+	if (pamu_regs)
+		iounmap(pamu_regs);
+
+	if (guts_regs)
+		iounmap(guts_regs);
+
+	if (ppaact)
+		free_pages((unsigned long)ppaact, order);
+
+	ppaact = NULL;
+
+	return ret;
+}
+
+static const struct of_device_id fsl_of_pamu_ids[] = {
+	{
+		.compatible = "fsl,p4080-pamu",
+	},
+	{
+		.compatible = "fsl,pamu",
+	},
+	{},
+};
+
+static struct platform_driver fsl_of_pamu_driver = {
+	.driver = {
+		.name = "fsl-of-pamu",
+		.owner = THIS_MODULE,
+	},
+	.probe = fsl_pamu_probe,
+};
+
+static __init int fsl_pamu_init(void)
+{
+	struct platform_device *pdev = NULL;
+	struct device_node *np;
+	int ret;
+
+	/*
+	 * The normal OF process calls the probe function at some
+	 * indeterminate later time, after most drivers have loaded.  This is
+	 * too late for us, because PAMU clients (like the Qman driver)
+	 * depend on PAMU being initialized early.
+	 *
+	 * So instead, we "manually" call our probe function by creating the
+	 * platform devices ourselves.
+	 */
+
+	/*
+	 * We assume that there is only one PAMU node in the device tree.  A
+	 * single PAMU node represents all of the PAMU devices in the SOC
+	 * already.   Everything else already makes that assumption, and the
+	 * binding for the PAMU nodes doesn't allow for any parent-child
+	 * relationships anyway.  In other words, support for more than one
+	 * PAMU node would require significant changes to a lot of code.
+	 */
+
+	np = of_find_compatible_node(NULL, NULL, "fsl,pamu");
+	if (!np) {
+		pr_err("fsl-pamu: could not find a PAMU node\n");
+		return -ENODEV;
+	}
+
+	ret = platform_driver_register(&fsl_of_pamu_driver);
+	if (ret) {
+		pr_err("fsl-pamu: could not register driver (err=%i)\n", ret);
+		goto error_driver_register;
+	}
+
+	pdev = platform_device_alloc("fsl-of-pamu", 0);
+	if (!pdev) {
+		pr_err("fsl-pamu: could not allocate device %s\n",
+		       np->full_name);
+		ret = -ENOMEM;
+		goto error_device_alloc;
+	}
+	pdev->dev.of_node = of_node_get(np);
+
+	ret = pamu_domain_init();
+	if (ret)
+		goto error_device_add;
+
+	ret = platform_device_add(pdev);
+	if (ret) {
+		pr_err("fsl-pamu: could not add device %s (err=%i)\n",
+		       np->full_name, ret);
+		goto error_device_add;
+	}
+
+	return 0;
+
+error_device_add:
+	of_node_put(pdev->dev.of_node);
+	pdev->dev.of_node = NULL;
+
+	platform_device_put(pdev);
+
+error_device_alloc:
+	platform_driver_unregister(&fsl_of_pamu_driver);
+
+error_driver_register:
+	of_node_put(np);
+
+	return ret;
+}
+arch_initcall(fsl_pamu_init);
diff --git a/drivers/iommu/fsl_pamu.h b/drivers/iommu/fsl_pamu.h
new file mode 100644
index 0000000..83cbd26
--- /dev/null
+++ b/drivers/iommu/fsl_pamu.h
@@ -0,0 +1,405 @@ 
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright (C) 2013 Freescale Semiconductor, Inc.
+ *
+ */
+
+#ifndef __FSL_PAMU_H
+#define __FSL_PAMU_H
+
+/* Bit Field macros
+ *	v = bit field variable; m = mask, m##_SHIFT = shift, x = value to load
+ */
+#define set_bf(v, m, x)		(v = ((v) & ~(m)) | (((x) << (m##_SHIFT)) & (m)))
+#define get_bf(v, m)		(((v) & (m)) >> (m##_SHIFT))
+
+/* PAMU CCSR space */
+#define PAMU_PGC 0x00000000     /* Allows all peripheral accesses */
+#define PAMU_PE 0x40000000      /* enable PAMU                    */
+
+/* PAMU_OFFSET to the next pamu space in ccsr */
+#define PAMU_OFFSET 0x1000
+
+#define PAMU_MMAP_REGS_BASE 0
+
+struct pamu_mmap_regs {
+	u32 ppbah;
+	u32 ppbal;
+	u32 pplah;
+	u32 pplal;
+	u32 spbah;
+	u32 spbal;
+	u32 splah;
+	u32 splal;
+	u32 obah;
+	u32 obal;
+	u32 olah;
+	u32 olal;
+};
+
+/* PAMU Error Registers */
+#define PAMU_POES1 0x0040
+#define PAMU_POES2 0x0044
+#define PAMU_POEAH 0x0048
+#define PAMU_POEAL 0x004C
+#define PAMU_AVS1  0x0050
+#define PAMU_AVS1_AV    0x1
+#define PAMU_AVS1_OTV   0x6
+#define PAMU_AVS1_APV   0x78
+#define PAMU_AVS1_WAV   0x380
+#define PAMU_AVS1_LAV   0x1c00
+#define PAMU_AVS1_GCV   0x2000
+#define PAMU_AVS1_PDV   0x4000
+#define PAMU_AV_MASK    (PAMU_AVS1_AV | PAMU_AVS1_OTV | PAMU_AVS1_APV | PAMU_AVS1_WAV \
+			| PAMU_AVS1_LAV | PAMU_AVS1_GCV | PAMU_AVS1_PDV)
+#define PAMU_AVS1_LIODN_SHIFT 16
+#define PAMU_LAV_LIODN_NOT_IN_PPAACT 0x400
+
+#define PAMU_AVS2  0x0054
+#define PAMU_AVAH  0x0058
+#define PAMU_AVAL  0x005C
+#define PAMU_EECTL 0x0060
+#define PAMU_EEDIS 0x0064
+#define PAMU_EEINTEN 0x0068
+#define PAMU_EEDET 0x006C
+#define PAMU_EEATTR 0x0070
+#define PAMU_EEAHI 0x0074
+#define PAMU_EEALO 0x0078
+#define PAMU_EEDHI 0X007C
+#define PAMU_EEDLO 0x0080
+#define PAMU_EECC  0x0084
+#define PAMU_UDAD  0x0090
+
+/* PAMU Revision Registers */
+#define PAMU_PR1 0x0BF8
+#define PAMU_PR2 0x0BFC
+
+/* PAMU Capabilities Registers */
+#define PAMU_PC1 0x0C00
+#define PAMU_PC2 0x0C04
+#define PAMU_PC3 0x0C08
+#define PAMU_PC4 0x0C0C
+
+/* PAMU Control Register */
+#define PAMU_PC 0x0C10
+
+/* PAMU control defs */
+#define PAMU_CONTROL 0x0C10
+#define PAMU_PC_PGC 0x80000000  /* PAMU gate closed bit */
+#define PAMU_PC_PE   0x40000000 /* PAMU enable bit */
+#define PAMU_PC_SPCC 0x00000010 /* sPAACE cache enable */
+#define PAMU_PC_PPCC 0x00000001 /* pPAACE cache enable */
+#define PAMU_PC_OCE  0x00001000 /* OMT cache enable */
+
+#define PAMU_PFA1 0x0C14
+#define PAMU_PFA2 0x0C18
+
+#define PAMU_PC2_MLIODN(X) ((X) >> 16)
+#define PAMU_PC3_MWCE(X) (((X) >> 21) & 0xf)
+
+/* PAMU Interrupt control and Status Register */
+#define PAMU_PICS 0x0C1C
+#define PAMU_ACCESS_VIOLATION_STAT   0x8
+#define PAMU_ACCESS_VIOLATION_ENABLE 0x4
+
+/* PAMU Debug Registers */
+#define PAMU_PD1 0x0F00
+#define PAMU_PD2 0x0F04
+#define PAMU_PD3 0x0F08
+#define PAMU_PD4 0x0F0C
+
+#define PAACE_AP_PERMS_DENIED  0x0
+#define PAACE_AP_PERMS_QUERY   0x1
+#define PAACE_AP_PERMS_UPDATE  0x2
+#define PAACE_AP_PERMS_ALL     0x3
+
+#define PAACE_DD_TO_HOST       0x0
+#define PAACE_DD_TO_IO         0x1
+#define PAACE_PT_PRIMARY       0x0
+#define PAACE_PT_SECONDARY     0x1
+#define PAACE_V_INVALID        0x0
+#define PAACE_V_VALID          0x1
+#define PAACE_MW_SUBWINDOWS    0x1
+
+#define PAACE_WSE_4K           0xB
+#define PAACE_WSE_8K           0xC
+#define PAACE_WSE_16K          0xD
+#define PAACE_WSE_32K          0xE
+#define PAACE_WSE_64K          0xF
+#define PAACE_WSE_128K         0x10
+#define PAACE_WSE_256K         0x11
+#define PAACE_WSE_512K         0x12
+#define PAACE_WSE_1M           0x13
+#define PAACE_WSE_2M           0x14
+#define PAACE_WSE_4M           0x15
+#define PAACE_WSE_8M           0x16
+#define PAACE_WSE_16M          0x17
+#define PAACE_WSE_32M          0x18
+#define PAACE_WSE_64M          0x19
+#define PAACE_WSE_128M         0x1A
+#define PAACE_WSE_256M         0x1B
+#define PAACE_WSE_512M         0x1C
+#define PAACE_WSE_1G           0x1D
+#define PAACE_WSE_2G           0x1E
+#define PAACE_WSE_4G           0x1F
+
+#define PAACE_DID_PCI_EXPRESS_1 0x00
+#define PAACE_DID_PCI_EXPRESS_2 0x01
+#define PAACE_DID_PCI_EXPRESS_3 0x02
+#define PAACE_DID_PCI_EXPRESS_4 0x03
+#define PAACE_DID_LOCAL_BUS     0x04
+#define PAACE_DID_SRIO          0x0C
+#define PAACE_DID_MEM_1         0x10
+#define PAACE_DID_MEM_2         0x11
+#define PAACE_DID_MEM_3         0x12
+#define PAACE_DID_MEM_4         0x13
+#define PAACE_DID_MEM_1_2       0x14
+#define PAACE_DID_MEM_3_4       0x15
+#define PAACE_DID_MEM_1_4       0x16
+#define PAACE_DID_BM_SW_PORTAL  0x18
+#define PAACE_DID_PAMU          0x1C
+#define PAACE_DID_CAAM          0x21
+#define PAACE_DID_QM_SW_PORTAL  0x3C
+#define PAACE_DID_CORE0_INST    0x80
+#define PAACE_DID_CORE0_DATA    0x81
+#define PAACE_DID_CORE1_INST    0x82
+#define PAACE_DID_CORE1_DATA    0x83
+#define PAACE_DID_CORE2_INST    0x84
+#define PAACE_DID_CORE2_DATA    0x85
+#define PAACE_DID_CORE3_INST    0x86
+#define PAACE_DID_CORE3_DATA    0x87
+#define PAACE_DID_CORE4_INST    0x88
+#define PAACE_DID_CORE4_DATA    0x89
+#define PAACE_DID_CORE5_INST    0x8A
+#define PAACE_DID_CORE5_DATA    0x8B
+#define PAACE_DID_CORE6_INST    0x8C
+#define PAACE_DID_CORE6_DATA    0x8D
+#define PAACE_DID_CORE7_INST    0x8E
+#define PAACE_DID_CORE7_DATA    0x8F
+#define PAACE_DID_BROADCAST     0xFF
+
+#define PAACE_ATM_NO_XLATE      0x00
+#define PAACE_ATM_WINDOW_XLATE  0x01
+#define PAACE_ATM_PAGE_XLATE    0x02
+#define PAACE_ATM_WIN_PG_XLATE  \
+                (PAACE_ATM_WINDOW_XLATE | PAACE_ATM_PAGE_XLATE)
+#define PAACE_OTM_NO_XLATE      0x00
+#define PAACE_OTM_IMMEDIATE     0x01
+#define PAACE_OTM_INDEXED       0x02
+#define PAACE_OTM_RESERVED      0x03
+
+#define PAACE_M_COHERENCE_REQ   0x01
+
+#define PAACE_PID_0             0x0
+#define PAACE_PID_1             0x1
+#define PAACE_PID_2             0x2
+#define PAACE_PID_3             0x3
+#define PAACE_PID_4             0x4
+#define PAACE_PID_5             0x5
+#define PAACE_PID_6             0x6
+#define PAACE_PID_7             0x7
+
+#define PAACE_TCEF_FORMAT0_8B   0x00
+#define PAACE_TCEF_FORMAT1_RSVD 0x01
+/*
+ * Hard coded value for the PAACT size to accomodate
+ * maximum LIODN value generated by u-boot.
+ */
+#define PAACE_NUMBER_ENTRIES    0x500
+/* Hard coded value for the SPAACT size */
+#define SPAACE_NUMBER_ENTRIES	0x800
+
+#define	OME_NUMBER_ENTRIES      16
+
+/* PAACE Bit Field Defines */
+#define PPAACE_AF_WBAL			0xfffff000
+#define PPAACE_AF_WBAL_SHIFT		12
+#define PPAACE_AF_WSE			0x00000fc0
+#define PPAACE_AF_WSE_SHIFT		6
+#define PPAACE_AF_MW			0x00000020
+#define PPAACE_AF_MW_SHIFT		5
+
+#define SPAACE_AF_LIODN			0xffff0000
+#define SPAACE_AF_LIODN_SHIFT		16
+
+#define PAACE_AF_AP			0x00000018
+#define PAACE_AF_AP_SHIFT		3
+#define PAACE_AF_DD			0x00000004
+#define PAACE_AF_DD_SHIFT		2
+#define PAACE_AF_PT			0x00000002
+#define PAACE_AF_PT_SHIFT		1
+#define PAACE_AF_V			0x00000001
+#define PAACE_AF_V_SHIFT		0
+
+#define PAACE_DA_HOST_CR		0x80
+#define PAACE_DA_HOST_CR_SHIFT		7
+
+#define PAACE_IA_CID			0x00FF0000
+#define PAACE_IA_CID_SHIFT		16
+#define PAACE_IA_WCE			0x000000F0
+#define PAACE_IA_WCE_SHIFT		4
+#define PAACE_IA_ATM			0x0000000C
+#define PAACE_IA_ATM_SHIFT		2
+#define PAACE_IA_OTM			0x00000003
+#define PAACE_IA_OTM_SHIFT		0
+
+#define PAACE_WIN_TWBAL			0xfffff000
+#define PAACE_WIN_TWBAL_SHIFT		12
+#define PAACE_WIN_SWSE			0x00000fc0
+#define PAACE_WIN_SWSE_SHIFT		6
+
+/* PAMU Data Structures */
+/* primary / secondary paact structure */
+struct paace {
+	/* PAACE Offset 0x00 */
+	u32 wbah;				/* only valid for Primary PAACE */
+	u32 addr_bitfields;		/* See P/S PAACE_AF_* */
+
+	/* PAACE Offset 0x08 */
+	/* Interpretation of first 32 bits dependent on DD above */
+	union {
+		struct {
+			/* Destination ID, see PAACE_DID_* defines */
+			u8 did;
+			/* Partition ID */
+			u8 pid;
+			/* Snoop ID */
+			u8 snpid;
+			/* coherency_required : 1 reserved : 7 */
+			u8 coherency_required; /* See PAACE_DA_* */
+		} to_host;
+		struct {
+			/* Destination ID, see PAACE_DID_* defines */
+			u8  did;
+			u8  reserved1;
+			u16 reserved2;
+		} to_io;
+	} domain_attr;
+
+	/* Implementation attributes + window count + address & operation translation modes */
+	u32 impl_attr;			/* See PAACE_IA_* */
+
+	/* PAACE Offset 0x10 */
+	/* Translated window base address */
+	u32 twbah;
+	u32 win_bitfields;			/* See PAACE_WIN_* */
+
+	/* PAACE Offset 0x18 */
+	/* first secondary paace entry */
+	u32 fspi;				/* only valid for Primary PAACE */
+	union {
+		struct {
+			u8 ioea;
+			u8 moea;
+			u8 ioeb;
+			u8 moeb;
+		} immed_ot;
+		struct {
+			u16 reserved;
+			u16 omi;
+		} index_ot;
+	} op_encode;
+
+	/* PAACE Offsets 0x20-0x38 */
+	u32 reserved[8];			/* not currently implemented */
+};
+
+/* OME : Operation mapping entry
+ * MOE : Mapped Operation Encodings
+ * The operation mapping table is table containing operation mapping entries (OME).
+ * The index of a particular OME is programmed in the PAACE entry for translation
+ * in bound I/O operations corresponding to an LIODN. The OMT is used for translation
+ * specifically in case of the indexed translation mode. Each OME contains a 128
+ * byte mapped operation encoding (MOE), where each byte represents an MOE.
+ */
+#define NUM_MOE 128
+struct ome {
+	u8 moe[NUM_MOE];
+} __attribute__((packed));
+
+#define PAACT_SIZE              (sizeof(struct paace) * PAACE_NUMBER_ENTRIES)
+#define SPAACT_SIZE              (sizeof(struct paace) * SPAACE_NUMBER_ENTRIES)
+#define OMT_SIZE                (sizeof(struct ome) * OME_NUMBER_ENTRIES)
+
+#define PAMU_PAGE_SHIFT 12
+#define PAMU_PAGE_SIZE  4096ULL
+
+#define IOE_READ        0x00
+#define IOE_READ_IDX    0x00
+#define IOE_WRITE       0x81
+#define IOE_WRITE_IDX   0x01
+#define IOE_EREAD0      0x82    /* Enhanced read type 0 */
+#define IOE_EREAD0_IDX  0x02    /* Enhanced read type 0 */
+#define IOE_EWRITE0     0x83    /* Enhanced write type 0 */
+#define IOE_EWRITE0_IDX 0x03    /* Enhanced write type 0 */
+#define IOE_DIRECT0     0x84    /* Directive type 0 */
+#define IOE_DIRECT0_IDX 0x04    /* Directive type 0 */
+#define IOE_EREAD1      0x85    /* Enhanced read type 1 */
+#define IOE_EREAD1_IDX  0x05    /* Enhanced read type 1 */
+#define IOE_EWRITE1     0x86    /* Enhanced write type 1 */
+#define IOE_EWRITE1_IDX 0x06    /* Enhanced write type 1 */
+#define IOE_DIRECT1     0x87    /* Directive type 1 */
+#define IOE_DIRECT1_IDX 0x07    /* Directive type 1 */
+#define IOE_RAC         0x8c    /* Read with Atomic clear */
+#define IOE_RAC_IDX     0x0c    /* Read with Atomic clear */
+#define IOE_RAS         0x8d    /* Read with Atomic set */
+#define IOE_RAS_IDX     0x0d    /* Read with Atomic set */
+#define IOE_RAD         0x8e    /* Read with Atomic decrement */
+#define IOE_RAD_IDX     0x0e    /* Read with Atomic decrement */
+#define IOE_RAI         0x8f    /* Read with Atomic increment */
+#define IOE_RAI_IDX     0x0f    /* Read with Atomic increment */
+
+#define EOE_READ        0x00
+#define EOE_WRITE       0x01
+#define EOE_RAC         0x0c    /* Read with Atomic clear */
+#define EOE_RAS         0x0d    /* Read with Atomic set */
+#define EOE_RAD         0x0e    /* Read with Atomic decrement */
+#define EOE_RAI         0x0f    /* Read with Atomic increment */
+#define EOE_LDEC        0x10    /* Load external cache */
+#define EOE_LDECL       0x11    /* Load external cache with stash lock */
+#define EOE_LDECPE      0x12    /* Load external cache with preferred exclusive */
+#define EOE_LDECPEL     0x13    /* Load external cache with preferred exclusive and lock */
+#define EOE_LDECFE      0x14    /* Load external cache with forced exclusive */
+#define EOE_LDECFEL     0x15    /* Load external cache with forced exclusive and lock */
+#define EOE_RSA         0x16    /* Read with stash allocate */
+#define EOE_RSAU        0x17    /* Read with stash allocate and unlock */
+#define EOE_READI       0x18    /* Read with invalidate */
+#define EOE_RWNITC      0x19    /* Read with no intention to cache */
+#define EOE_WCI         0x1a    /* Write cache inhibited */
+#define EOE_WWSA        0x1b    /* Write with stash allocate */
+#define EOE_WWSAL       0x1c    /* Write with stash allocate and lock */
+#define EOE_WWSAO       0x1d    /* Write with stash allocate only */
+#define EOE_WWSAOL      0x1e    /* Write with stash allocate only and lock */
+#define EOE_VALID       0x80
+
+/* Function prototypes */
+int pamu_domain_init(void);
+int pamu_enable_liodn(int liodn);
+int pamu_disable_liodn(int liodn);
+void pamu_free_subwins(int liodn);
+int pamu_config_ppaace(int liodn, phys_addr_t win_addr, phys_addr_t win_size,
+		       u32 omi, unsigned long rpn, u32 snoopid, uint32_t stashid,
+		       u32 subwin_cnt, int prot);
+int pamu_config_spaace(int liodn, u32 subwin_cnt, u32 subwin_addr,
+		       phys_addr_t subwin_size, u32 omi, unsigned long rpn,
+		       uint32_t snoopid, u32 stashid, int enable, int prot);
+
+u32 get_stash_id(u32 stash_dest_hint, u32 vcpu);
+void get_ome_index(u32 *omi_index, struct device *dev);
+int  pamu_update_paace_stash(int liodn, u32 subwin, u32 value);
+int pamu_disable_spaace(int liodn, u32 subwin);
+u32 pamu_get_max_subwin_cnt(void);
+
+#endif  /* __FSL_PAMU_H */
diff --git a/drivers/iommu/fsl_pamu_domain.c b/drivers/iommu/fsl_pamu_domain.c
new file mode 100644
index 0000000..16b728f
--- /dev/null
+++ b/drivers/iommu/fsl_pamu_domain.c
@@ -0,0 +1,1137 @@ 
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright (C) 2013 Freescale Semiconductor, Inc.
+ * Author: Varun Sethi <varun.sethi@freescale.com>
+ *
+ */
+
+#define pr_fmt(fmt)    "fsl-pamu-domain: %s: " fmt, __func__
+
+#include <linux/init.h>
+#include <linux/iommu.h>
+#include <linux/notifier.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/device.h>
+#include <linux/of_platform.h>
+#include <linux/bootmem.h>
+#include <linux/err.h>
+#include <asm/io.h>
+#include <asm/bitops.h>
+
+#include <asm/pci-bridge.h>
+#include <sysdev/fsl_pci.h>
+
+#include "fsl_pamu_domain.h"
+
+/*
+ * Global spinlock that needs to be held while
+ * configuring PAMU.
+ */
+static DEFINE_SPINLOCK(iommu_lock);
+
+static struct kmem_cache *fsl_pamu_domain_cache;
+static struct kmem_cache *iommu_devinfo_cache;
+static DEFINE_SPINLOCK(device_domain_lock);
+
+static int __init iommu_init_mempool(void)
+{
+
+	fsl_pamu_domain_cache = kmem_cache_create("fsl_pamu_domain",
+					 sizeof(struct fsl_dma_domain),
+					 0,
+					 SLAB_HWCACHE_ALIGN,
+
+					 NULL);
+	if (!fsl_pamu_domain_cache) {
+		pr_err("Couldn't create fsl iommu_domain cache\n");
+		return -ENOMEM;
+	}
+
+	iommu_devinfo_cache = kmem_cache_create("iommu_devinfo",
+					 sizeof(struct device_domain_info),
+					 0,
+					 SLAB_HWCACHE_ALIGN,
+					 NULL);
+	if (!iommu_devinfo_cache) {
+		pr_err("Couldn't create devinfo cache\n");
+		kmem_cache_destroy(fsl_pamu_domain_cache);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static phys_addr_t get_phys_addr(struct fsl_dma_domain *dma_domain, dma_addr_t iova)
+{
+	u32 win_cnt = dma_domain->win_cnt;
+	struct dma_window *win_ptr =
+				&dma_domain->win_arr[0];
+	struct iommu_domain_geometry *geom;
+
+	geom = &dma_domain->iommu_domain->geometry;
+
+	if (!win_cnt || !dma_domain->geom_size) {
+		pr_err("Number of windows/geometry not configured for the domain\n");
+		return 0;
+	}
+
+	if (win_cnt > 1) {
+		u64 subwin_size;
+		dma_addr_t subwin_iova;
+		u32 wnd;
+
+		subwin_size = dma_domain->geom_size >> ilog2(win_cnt);
+		subwin_iova = iova & ~(subwin_size - 1);
+		wnd = (subwin_iova - geom->aperture_start) >> ilog2(subwin_size);
+		win_ptr = &dma_domain->win_arr[wnd];
+	}
+
+	if (win_ptr->valid)
+		return (win_ptr->paddr + (iova & (win_ptr->size - 1)));
+
+	return 0;
+}
+
+static int map_subwins(int liodn, struct fsl_dma_domain *dma_domain)
+{
+	struct dma_window *sub_win_ptr =
+				&dma_domain->win_arr[0];
+	int i, ret;
+	unsigned long rpn;
+
+	for (i = 0; i < dma_domain->win_cnt; i++) {
+		if (sub_win_ptr[i].valid) {
+			rpn = sub_win_ptr[i].paddr >>
+				 PAMU_PAGE_SHIFT;
+			spin_lock(&iommu_lock);
+			ret = pamu_config_spaace(liodn, dma_domain->win_cnt, i,
+						 sub_win_ptr[i].size,
+						 ~(u32)0,
+						 rpn,
+						 dma_domain->snoop_id,
+						 dma_domain->stash_id,
+						 (i > 0) ? 1 : 0,
+						 sub_win_ptr[i].prot);
+			spin_unlock(&iommu_lock);
+			if (ret) {
+				pr_err("PAMU SPAACE configuration failed for liodn %d\n",
+					 liodn);
+				return ret;
+			}
+		}
+	}
+
+	return ret;
+}
+
+static int map_win(int liodn, struct fsl_dma_domain *dma_domain)
+{
+	int ret;
+	struct dma_window *wnd = &dma_domain->win_arr[0];
+	phys_addr_t wnd_addr = dma_domain->iommu_domain->geometry.aperture_start;
+
+	spin_lock(&iommu_lock);
+	ret = pamu_config_ppaace(liodn, wnd_addr,
+				 wnd->size,
+				 ~(u32)0,
+				 wnd->paddr >> PAMU_PAGE_SHIFT,
+				 dma_domain->snoop_id, dma_domain->stash_id,
+				 0, wnd->prot);
+	spin_unlock(&iommu_lock);
+	if (ret)
+		pr_err("PAMU PAACE configuration failed for liodn %d\n",
+			liodn);
+
+	return ret;
+}
+
+/* Map the DMA window corresponding to the LIODN */
+static int map_liodn(int liodn, struct fsl_dma_domain *dma_domain)
+{
+	if (dma_domain->win_cnt > 1)
+		return map_subwins(liodn, dma_domain);
+	else
+		return map_win(liodn, dma_domain);
+
+}
+
+/* Update window/subwindow mapping for the LIODN */
+static int update_liodn(int liodn, struct fsl_dma_domain *dma_domain, u32 wnd_nr)
+{
+	int ret;
+	struct dma_window *wnd = &dma_domain->win_arr[wnd_nr];
+
+	spin_lock(&iommu_lock);
+	if (dma_domain->win_cnt > 1) {
+		ret = pamu_config_spaace(liodn, dma_domain->win_cnt, wnd_nr,
+					 wnd->size,
+					 ~(u32)0,
+					 wnd->paddr >> PAMU_PAGE_SHIFT,
+					 dma_domain->snoop_id,
+					 dma_domain->stash_id,
+					 (wnd_nr > 0) ? 1 : 0,
+					 wnd->prot);
+		if (ret)
+			pr_err("Subwindow reconfiguration failed for liodn %d\n", liodn);
+	} else {
+		phys_addr_t wnd_addr;
+
+		wnd_addr = dma_domain->iommu_domain->geometry.aperture_start;
+
+		ret = pamu_config_ppaace(liodn, wnd_addr,
+					 wnd->size,
+					 ~(u32)0,
+					 wnd->paddr >> PAMU_PAGE_SHIFT,
+					dma_domain->snoop_id, dma_domain->stash_id,
+					0, wnd->prot);
+		if (ret)
+			pr_err("Window reconfiguration failed for liodn %d\n", liodn);
+	}
+
+	spin_unlock(&iommu_lock);
+
+	return ret;
+}
+
+static int update_liodn_stash(int liodn, struct fsl_dma_domain *dma_domain,
+				 u32 val)
+{
+	int ret = 0, i;
+
+	spin_lock(&iommu_lock);
+	if (!dma_domain->win_cnt) {
+		ret = pamu_update_paace_stash(liodn, 0, val);
+		if (ret) {
+			pr_err("Failed to update PAACE field for liodn %d\n ", liodn);
+			spin_unlock(&iommu_lock);
+			return ret;
+		}
+	} else {
+		for (i = 0; i < dma_domain->win_cnt; i++) {
+			ret = pamu_update_paace_stash(liodn, i, val);
+			if (ret) {
+				pr_err("Failed to update SPAACE %d field for liodn %d\n ", i, liodn);
+				spin_unlock(&iommu_lock);
+				return ret;
+			}
+		}
+	}
+	spin_unlock(&iommu_lock);
+
+	return ret;
+}
+
+/* Set the geometry parameters for a LIODN */
+static int pamu_set_liodn(int liodn, struct device *dev,
+			   struct fsl_dma_domain *dma_domain,
+			   struct iommu_domain_geometry *geom_attr,
+			   u32 win_cnt)
+{
+	phys_addr_t window_addr, window_size;
+	phys_addr_t subwin_size;
+	int ret = 0, i;
+	u32 omi_index = ~(u32)0;
+
+	/*
+	 * Configure the omi_index at the geometry setup time.
+	 * This is a static value which depends on the type of
+	 * device and would not change thereafter.
+	 */
+	get_ome_index(&omi_index, dev);
+
+	window_addr = geom_attr->aperture_start;
+	window_size = dma_domain->geom_size;
+
+	spin_lock(&iommu_lock);
+	ret = pamu_disable_liodn(liodn);
+	if (!ret)
+		ret = pamu_config_ppaace(liodn, window_addr, window_size, omi_index,
+					 0, dma_domain->snoop_id,
+					 dma_domain->stash_id, win_cnt, 0);
+	spin_unlock(&iommu_lock);
+	if (ret) {
+		pr_err("PAMU PAACE configuration failed for liodn %d, win_cnt =%d\n", liodn, win_cnt);
+		return ret;
+	}
+
+	if (win_cnt > 1) {
+		subwin_size = window_size >> ilog2(win_cnt);
+		for (i = 0; i < win_cnt; i++) {
+			spin_lock(&iommu_lock);
+			ret = pamu_disable_spaace(liodn, i);
+			if (!ret)
+				ret = pamu_config_spaace(liodn, win_cnt, i,
+							 subwin_size, omi_index,
+							 0, dma_domain->snoop_id,
+							 dma_domain->stash_id,
+							 0, 0);
+			spin_unlock(&iommu_lock);
+			if (ret) {
+				pr_err("PAMU SPAACE configuration failed for liodn %d\n", liodn);
+				return ret;
+			}
+		}
+	}
+
+	return ret;
+}
+
+static int check_size(u64 size, dma_addr_t iova)
+{
+	/*
+	 * Size must be a power of two and at least be equal
+	 * to PAMU page size.
+	 */
+	if ((size & (size - 1)) || size < PAMU_PAGE_SIZE) {
+		pr_err("%s: size too small or not a power of two\n", __func__);
+		return -EINVAL;
+	}
+
+	/* iova must be page size aligned*/
+	if (iova & (size - 1)) {
+		pr_err("%s: address is not aligned with window size\n", __func__);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static struct fsl_dma_domain *iommu_alloc_dma_domain(void)
+{
+	struct fsl_dma_domain *domain;
+
+	domain = kmem_cache_zalloc(fsl_pamu_domain_cache, GFP_KERNEL);
+	if (!domain)
+		return NULL;
+
+	domain->stash_id = ~(u32)0;
+	domain->snoop_id = ~(u32)0;
+	domain->win_cnt = pamu_get_max_subwin_cnt();
+	domain->geom_size = 0;
+
+	INIT_LIST_HEAD(&domain->devices);
+
+	spin_lock_init(&domain->domain_lock);
+
+	return domain;
+}
+
+static inline struct device_domain_info *find_domain(struct device *dev)
+{
+	return dev->archdata.iommu_domain;
+}
+
+static void remove_device_ref(struct device_domain_info *info, u32 win_cnt)
+{
+	list_del(&info->link);
+	spin_lock(&iommu_lock);
+	if (win_cnt > 1)
+		pamu_free_subwins(info->liodn);
+	pamu_disable_liodn(info->liodn);
+	spin_unlock(&iommu_lock);
+	spin_lock(&device_domain_lock);
+	info->dev->archdata.iommu_domain = NULL;
+	kmem_cache_free(iommu_devinfo_cache, info);
+	spin_unlock(&device_domain_lock);
+}
+
+static void detach_device(struct device *dev, struct fsl_dma_domain *dma_domain)
+{
+	struct device_domain_info *info;
+	struct list_head *entry, *tmp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	/* Remove the device from the domain device list */
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_safe(entry, tmp, &dma_domain->devices) {
+			info = list_entry(entry, struct device_domain_info, link);
+			if (!dev || (info->dev == dev))
+				remove_device_ref(info, dma_domain->win_cnt);
+		}
+	}
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+}
+
+static void attach_device(struct fsl_dma_domain *dma_domain, int liodn, struct device *dev)
+{
+	struct device_domain_info *info, *old_domain_info;
+
+	spin_lock(&device_domain_lock);
+	/*
+	 * Check here if the device is already attached to domain or not.
+	 * If the device is already attached to a domain detach it.
+	 */
+	old_domain_info = find_domain(dev);
+	if (old_domain_info && old_domain_info->domain != dma_domain) {
+		spin_unlock(&device_domain_lock);
+		detach_device(dev, old_domain_info->domain);
+		spin_lock(&device_domain_lock);
+	}
+
+	info = kmem_cache_zalloc(iommu_devinfo_cache, GFP_KERNEL);
+
+	info->dev = dev;
+	info->liodn = liodn;
+	info->domain = dma_domain;
+
+	list_add(&info->link, &dma_domain->devices);
+	/*
+	 * In case of devices with multiple LIODNs just store
+	 * the info for the first LIODN as all
+	 * LIODNs share the same domain
+	 */
+	if (!old_domain_info)
+		dev->archdata.iommu_domain = info;
+	spin_unlock(&device_domain_lock);
+
+}
+
+static phys_addr_t fsl_pamu_iova_to_phys(struct iommu_domain *domain,
+					    dma_addr_t iova)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+
+	if ((iova < domain->geometry.aperture_start) ||
+		iova > (domain->geometry.aperture_end))
+		return 0;
+
+	return get_phys_addr(dma_domain, iova);
+}
+
+static int fsl_pamu_domain_has_cap(struct iommu_domain *domain,
+				      unsigned long cap)
+{
+	return cap == IOMMU_CAP_CACHE_COHERENCY;
+}
+
+static void fsl_pamu_domain_destroy(struct iommu_domain *domain)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+
+	domain->priv = NULL;
+
+	/* remove all the devices from the device list */
+	detach_device(NULL, dma_domain);
+
+	dma_domain->enabled = 0;
+	dma_domain->mapped = 0;
+
+	kmem_cache_free(fsl_pamu_domain_cache, dma_domain);
+}
+
+static int fsl_pamu_domain_init(struct iommu_domain *domain)
+{
+	struct fsl_dma_domain *dma_domain;
+
+	dma_domain = iommu_alloc_dma_domain();
+	if (!dma_domain) {
+		pr_err("dma_domain allocation failed\n");
+		return -ENOMEM;
+	}
+	domain->priv = dma_domain;
+	dma_domain->iommu_domain = domain;
+	/* defaul geometry 64 GB i.e. maximum system address */
+	domain->geometry.aperture_start = 0;
+	domain->geometry.aperture_end = 1ULL << 36;
+	domain->geometry.force_aperture = true;
+
+	return 0;
+}
+
+/* Configure geometry settings for all LIODNs associated with domain */
+static int pamu_set_domain_geometry(struct fsl_dma_domain *dma_domain,
+				    struct iommu_domain_geometry *geom_attr,
+				    u32 win_cnt)
+{
+	struct device_domain_info *info;
+	int ret = 0;
+
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_entry(info, &dma_domain->devices, link) {
+			ret = pamu_set_liodn(info->liodn, info->dev, dma_domain,
+					      geom_attr, win_cnt);
+			if (ret)
+				break;
+		}
+	}
+
+	return ret;
+}
+
+/* Update stash destination for all LIODNs associated with the domain */
+static int update_domain_stash(struct fsl_dma_domain *dma_domain, u32 val)
+{
+	struct device_domain_info *info;
+	int ret = 0;
+
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_entry(info, &dma_domain->devices, link) {
+			ret = update_liodn_stash(info->liodn, dma_domain, val);
+			if (ret)
+				break;
+		}
+	}
+
+	return ret;
+}
+
+/* Update domain mappings for all LIODNs associated with the domain */
+static int update_domain_mapping(struct fsl_dma_domain *dma_domain, u32 wnd_nr)
+{
+	struct device_domain_info *info;
+	int ret = 0;
+
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_entry(info, &dma_domain->devices, link) {
+			ret = update_liodn(info->liodn, dma_domain, wnd_nr);
+			if (ret)
+				break;
+		}
+	}
+	return ret;
+}
+
+static int disable_domain_win(struct fsl_dma_domain *dma_domain, u32 wnd_nr)
+{
+	struct device_domain_info *info;
+	int ret = 0;
+
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_entry(info, &dma_domain->devices, link) {
+			if (dma_domain->win_cnt == 1 && dma_domain->enabled) {
+				ret = pamu_disable_liodn(info->liodn);
+				if (!ret)
+					dma_domain->enabled = 0;
+			} else {
+				ret = pamu_disable_spaace(info->liodn, wnd_nr);
+			}
+		}
+	}
+
+	return ret;
+}
+
+static void fsl_pamu_window_disable(struct iommu_domain *domain, u32 wnd_nr)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	if (!dma_domain->win_arr) {
+		pr_err("Number of windows not configured\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return;
+	}
+
+	if (wnd_nr >= dma_domain->win_cnt) {
+		pr_err("Invalid window index\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return;
+	}
+
+	if (dma_domain->win_arr[wnd_nr].valid) {
+		ret = disable_domain_win(dma_domain, wnd_nr);
+		if (!ret) {
+			dma_domain->win_arr[wnd_nr].valid = 0;
+			dma_domain->mapped--;
+		}
+	}
+
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+}
+
+static int fsl_pamu_window_enable(struct iommu_domain *domain, u32 wnd_nr,
+				  phys_addr_t paddr, u64 size, int prot)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	struct dma_window *wnd;
+	int pamu_prot = 0;
+	int ret;
+	unsigned long flags;
+	u64 win_size;
+
+	if (prot & IOMMU_READ)
+		pamu_prot |= PAACE_AP_PERMS_QUERY;
+	if (prot & IOMMU_WRITE)
+		pamu_prot |= PAACE_AP_PERMS_UPDATE;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	if (!dma_domain->win_arr) {
+		pr_err("Number of windows not configured\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -ENODEV;
+	}
+
+	if (wnd_nr >= dma_domain->win_cnt) {
+		pr_err("Invalid window index\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -EINVAL;
+	}
+
+	win_size = dma_domain->geom_size >> ilog2(dma_domain->win_cnt);
+	if (size > win_size) {
+		pr_err("Invalid window size \n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -EINVAL;
+	}
+
+	if (dma_domain->win_cnt == 1) {
+		if (dma_domain->enabled) {
+			pr_err("Disable the window before updating the mapping\n");
+			spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+			return -EBUSY;
+		}
+
+		ret = check_size(size, domain->geometry.aperture_start);
+		if (ret) {
+			pr_err("Aperture start not aligned to the size\n");
+			spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+			return -EINVAL;
+		}
+	}
+
+	wnd = &dma_domain->win_arr[wnd_nr];
+	if (!wnd->valid) {
+		wnd->paddr = paddr;
+		wnd->size = size;
+		wnd->prot = pamu_prot;
+
+		ret = update_domain_mapping(dma_domain, wnd_nr);
+		if (!ret) {
+			wnd->valid = 1;
+			dma_domain->mapped++;
+		}
+	} else {
+		pr_err("Disable the window before updating the mapping\n");
+		ret = -EBUSY;
+	}
+
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return ret;
+}
+
+/*
+ * Attach the LIODN to the DMA domain and configure the geometry
+ * and window mappings.
+ */
+static int handle_attach_device(struct fsl_dma_domain *dma_domain,
+				 struct device *dev, const u32 *liodn,
+				 int num)
+{
+	unsigned long flags;
+	struct iommu_domain *domain = dma_domain->iommu_domain;
+	int ret = 0;
+	int i;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	for (i = 0; i < num; i++) {
+
+		/* Ensure that LIODN value is valid */
+		if (liodn[i] >= PAACE_NUMBER_ENTRIES) {
+			pr_err("Invalid liodn %d, attach device failed for %s\n",
+				liodn[i], dev->of_node->full_name);
+			ret = -EINVAL;
+			break;
+		}
+
+		attach_device(dma_domain, liodn[i], dev);
+		/*
+		 * Check if geometry has already been configured
+		 * for the domain. If yes, set the geometry for
+		 * the LIODN.
+		 */
+		if (dma_domain->win_cnt) {
+			u32 win_cnt = dma_domain->win_cnt > 1 ? dma_domain->win_cnt : 0;
+			ret = pamu_set_liodn(liodn[i], dev, dma_domain,
+					      &domain->geometry,
+					      win_cnt);
+			if (ret)
+				break;
+			if (dma_domain->mapped) {
+				/*
+				 * Create window/subwindow mapping for
+				 * the LIODN.
+				 */
+				ret = map_liodn(liodn[i], dma_domain);
+				if (ret)
+					break;
+			}
+		}
+	}
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return ret;
+}
+
+static int fsl_pamu_attach_device(struct iommu_domain *domain,
+				  struct device *dev)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	const u32 *liodn;
+	u32 liodn_cnt;
+	int len, ret = 0;
+	struct pci_dev *pdev = NULL;
+	struct pci_controller *pci_ctl;
+
+	/*
+	 * Use LIODN of the PCI controller while attaching a
+	 * PCI device.
+	 */
+	if (dev->bus == &pci_bus_type) {
+		pdev = to_pci_dev(dev);
+		pci_ctl = pci_bus_to_host(pdev->bus);
+		/*
+		 * make dev point to pci controller device
+		 * so we can get the LIODN programmed by
+		 * u-boot.
+		 */
+		dev = pci_ctl->parent;
+	}
+
+	liodn = of_get_property(dev->of_node, "fsl,liodn", &len);
+	if (liodn) {
+		liodn_cnt = len / sizeof(u32);
+		ret = handle_attach_device(dma_domain, dev,
+					 liodn, liodn_cnt);
+	} else {
+		pr_err("missing fsl,liodn property at %s\n",
+		          dev->of_node->full_name);
+			ret = -EINVAL;
+	}
+
+	return ret;
+}
+
+static void fsl_pamu_detach_device(struct iommu_domain *domain,
+				      struct device *dev)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	const u32 *prop;
+	int len;
+	struct pci_dev *pdev = NULL;
+	struct pci_controller *pci_ctl;
+
+	/*
+	 * Use LIODN of the PCI controller while detaching a
+	 * PCI device.
+	 */
+	if (dev->bus == &pci_bus_type) {
+		pdev = to_pci_dev(dev);
+		pci_ctl = pci_bus_to_host(pdev->bus);
+		/*
+		 * make dev point to pci controller device
+		 * so we can get the LIODN programmed by
+		 * u-boot.
+		 */
+		dev = pci_ctl->parent;
+	}
+
+	prop = of_get_property(dev->of_node, "fsl,liodn", &len);
+	if (prop)
+		detach_device(dev, dma_domain);
+	else
+		pr_err("missing fsl,liodn property at %s\n",
+		          dev->of_node->full_name);
+}
+
+static  int configure_domain_geometry(struct iommu_domain *domain, void *data)
+{
+	struct iommu_domain_geometry *geom_attr = data;
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	dma_addr_t geom_size;
+	unsigned long flags;
+
+	geom_size = geom_attr->aperture_end - geom_attr->aperture_start;
+	/*
+	 * Sanity check the geometry size. Also, we do not support
+	 * DMA outside of the geometry.
+	 */
+	if (check_size(geom_size, geom_attr->aperture_start) ||
+		!geom_attr->force_aperture) {
+			pr_err("Invalid PAMU geometry attributes\n");
+			return -EINVAL;
+		}
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	if (dma_domain->enabled) {
+		pr_err("Can't set geometry attributes as domain is active\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return  -EBUSY;
+	}
+
+	/* Copy the domain geometry information */
+	memcpy(&domain->geometry, geom_attr,
+	       sizeof(struct iommu_domain_geometry));
+	dma_domain->geom_size = geom_size;
+
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return 0;
+}
+
+/* Set the domain stash attribute */
+static int configure_domain_stash(struct fsl_dma_domain *dma_domain, void *data)
+{
+	struct iommu_stash_attribute *stash_attr = data;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+
+	memcpy(&dma_domain->dma_stash, stash_attr,
+		 sizeof(struct iommu_stash_attribute));
+
+	dma_domain->stash_id = get_stash_id(stash_attr->cache,
+					    stash_attr->cpu);
+	if (dma_domain->stash_id == ~(u32)0) {
+		pr_err("Invalid stash attributes\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -EINVAL;
+	}
+
+	ret = update_domain_stash(dma_domain, dma_domain->stash_id);
+
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return ret;
+}
+
+/* Configure domain dma state i.e. enable/disable DMA*/
+static int configure_domain_dma_state(struct fsl_dma_domain *dma_domain, bool enable)
+{
+	struct device_domain_info *info;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+
+	if (enable && !dma_domain->mapped) {
+		pr_err("Can't enable DMA domain without valid mapping\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -ENODEV;
+	}
+
+	dma_domain->enabled = enable;
+	if (!list_empty(&dma_domain->devices)) {
+		list_for_each_entry(info, &dma_domain->devices,
+					 link) {
+			ret = (enable) ? pamu_enable_liodn(info->liodn) :
+				pamu_disable_liodn(info->liodn);
+			if (ret)
+				pr_err("Unable to set dma state for liodn %d",
+					 info->liodn);
+		}
+	}
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return 0;
+}
+
+static int fsl_pamu_set_domain_attr(struct iommu_domain *domain,
+				 enum iommu_attr attr_type, void *data)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	int ret = 0;
+
+
+	switch (attr_type) {
+	case DOMAIN_ATTR_GEOMETRY:
+		ret = configure_domain_geometry(domain, data);
+		break;
+	case DOMAIN_ATTR_PAMU_STASH:
+		ret = configure_domain_stash(dma_domain, data);
+		break;
+	case DOMAIN_ATTR_PAMU_ENABLE:
+		ret = configure_domain_dma_state(dma_domain, *(int *)data);
+		break;
+	default:
+		pr_err("Unsupported attribute type\n");
+		ret = -EINVAL;
+		break;
+	};
+
+	return ret;
+}
+
+static int fsl_pamu_get_domain_attr(struct iommu_domain *domain,
+				 enum iommu_attr attr_type, void *data)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	int ret = 0;
+
+
+	switch (attr_type) {
+	case DOMAIN_ATTR_PAMU_STASH:
+		memcpy((struct iommu_stash_attribute *) data, &dma_domain->dma_stash,
+				 sizeof(struct iommu_stash_attribute));
+		break;
+	case DOMAIN_ATTR_PAMU_ENABLE:
+		*(int *)data = dma_domain->enabled;
+		break;
+	case DOMAIN_ATTR_FSL_PAMUV1:
+		*(int *)data = DOMAIN_ATTR_FSL_PAMUV1;
+		break;
+	default:
+		pr_err("Unsupported attribute type\n");
+		ret = -EINVAL;
+		break;
+	};
+
+	return ret;
+}
+
+static void swap_pci_ref(struct pci_dev **from, struct pci_dev *to)
+{
+	pci_dev_put(*from);
+	*from = to;
+}
+
+static struct iommu_group *get_device_iommu_group(struct device *dev)
+{
+	struct iommu_group *group;
+
+	group = iommu_group_get(dev);
+	if (!group)
+		group = iommu_group_alloc();
+
+	return group;
+}
+
+static  bool check_pci_ctl_endpt_part(struct pci_controller *pci_ctl)
+{
+	u32 version;
+
+	/* Check the PCI controller version number by readding BRR1 register */
+	version = in_be32(pci_ctl->cfg_addr + (PCI_FSL_BRR1 >> 2));
+	version &= PCI_FSL_BRR1_VER;
+	/* If PCI controller version is >= 0x204 we can partition endpoints*/
+	if (version >= 0x204)
+		return 1;
+
+	return 0;
+}
+
+static struct iommu_group *get_peer_pci_device_group(struct pci_dev *pdev)
+{
+	struct iommu_group *group = NULL;
+
+	/* check if this is the first device on the bus*/
+	if (pdev->bus_list.next == pdev->bus_list.prev) {
+		struct pci_bus *bus = pdev->bus->parent;
+		/* Traverese the parent bus list to get
+		 * pdev & dev for the sibling device.
+		 */
+		while (bus) {
+			if (!list_empty(&bus->devices)) {
+				pdev = container_of(bus->devices.next,
+					            struct pci_dev, bus_list);
+				group = iommu_group_get(&pdev->dev);
+				break;
+			} else
+				bus = bus->parent;
+		}
+	} else {
+		/*
+		 * Get the pdev & dev for the sibling device
+		 */
+		pdev = container_of(pdev->bus_list.prev,
+				    struct pci_dev, bus_list);
+		group = iommu_group_get(&pdev->dev);
+	}
+
+	return group;
+}
+
+static struct iommu_group *get_pci_device_group(struct pci_dev *pdev)
+{
+	struct iommu_group *group = NULL;
+	struct pci_dev *bridge, *dma_pdev = NULL;
+	struct pci_controller *pci_ctl;
+	bool pci_endpt_partioning;
+
+	pci_ctl = pci_bus_to_host(pdev->bus);
+	pci_endpt_partioning = check_pci_ctl_endpt_part(pci_ctl);
+	/* We can partition PCIe devices so assign device group to the device */
+	if (pci_endpt_partioning) {
+		bridge = pci_find_upstream_pcie_bridge(pdev);
+		if (bridge) {
+			if (pci_is_pcie(bridge))
+				dma_pdev = pci_get_domain_bus_and_slot(
+						pci_domain_nr(pdev->bus),
+						bridge->subordinate->number, 0);
+			if (!dma_pdev)
+				dma_pdev = pci_dev_get(bridge);
+		} else
+			dma_pdev = pci_dev_get(pdev);
+		/* Account for quirked devices */
+		swap_pci_ref(&dma_pdev, pci_get_dma_source(dma_pdev));
+		group = get_device_iommu_group(&pdev->dev);
+		pci_dev_put(pdev);
+		/*
+		 * PCIe controller is not a paritionable entity
+		 * free the controller device iommu_group.
+		 */
+		if (pci_ctl->parent->iommu_group)
+			iommu_group_remove_device(pci_ctl->parent);
+	} else {
+		/*
+		 * All devices connected to the controller will share the
+		 * PCI controllers device group. If this is the first
+		 * device to be probed for the pci controller, copy the
+		 * device group information from the PCI controller device
+		 * node and remove the PCI controller iommu group.
+		 * For subsequent devices, the iommu group information can
+		 * be obtained from sibling devices (i.e. from the bus_devices
+		 * link list).
+		 */
+		if (pci_ctl->parent->iommu_group) {
+			group = get_device_iommu_group(pci_ctl->parent);
+			iommu_group_remove_device(pci_ctl->parent);
+		} else
+			group = get_peer_pci_device_group(pdev);
+	}
+
+	return group;
+}
+
+static int fsl_pamu_add_device(struct device *dev)
+{
+	struct iommu_group *group = NULL;
+	struct pci_dev *pdev;
+	int ret;
+
+	/*
+	 * For platform devices we allocate a separate group for
+	 * each of the devices.
+	 */
+	if (dev->bus == &pci_bus_type) {
+		pdev = to_pci_dev(dev);
+		/* Don't create device groups for virtual PCI bridges */
+		if (pdev->subordinate)
+			return 0;
+
+		group = get_pci_device_group(pdev);
+
+	} else
+		group = get_device_iommu_group(dev);
+
+	if (!group || IS_ERR(group))
+		return PTR_ERR(group);
+
+	ret = iommu_group_add_device(group, dev);
+
+	iommu_group_put(group);
+	return ret;
+}
+
+static void fsl_pamu_remove_device(struct device *dev)
+{
+	iommu_group_remove_device(dev);
+}
+
+static int fsl_pamu_set_windows(struct iommu_domain *domain, u32 w_count)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&dma_domain->domain_lock, flags);
+	/* Ensure domain is inactive i.e. DMA should be disabled for the domain */
+	if (dma_domain->enabled) {
+		pr_err("Can't set geometry attributes as domain is active\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return  -EBUSY;
+	}
+
+	/* Ensure that the geometry has been set for the domain */
+	if (!dma_domain->geom_size) {
+		pr_err("Please configure geometry before setting the number of windows\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -EINVAL;
+	}
+
+	/*
+	 * Ensure we have valid window count i.e. it should be less than
+	 * maximum permissible limit and should be a power of two.
+	 */
+	if (w_count > pamu_get_max_subwin_cnt() || (w_count & (w_count - 1))) {
+		pr_err("Invalid window count\n");
+		spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+		return -EINVAL;
+	}
+
+	ret = pamu_set_domain_geometry(dma_domain, &domain->geometry,
+				((w_count > 1) ? w_count : 0));
+	if (!ret) {
+		if (dma_domain->win_arr)
+			kfree(dma_domain->win_arr);
+		dma_domain->win_arr = kzalloc(sizeof(struct dma_window) *
+							  w_count, GFP_KERNEL);
+		if (!dma_domain->win_arr) {
+			spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+			return -ENOMEM;
+		}
+		dma_domain->win_cnt = w_count;
+	}
+	spin_unlock_irqrestore(&dma_domain->domain_lock, flags);
+
+	return ret;
+}
+
+static u32 fsl_pamu_get_windows(struct iommu_domain *domain)
+{
+	struct fsl_dma_domain *dma_domain = domain->priv;
+
+	return dma_domain->win_cnt;
+}
+
+static struct iommu_ops fsl_pamu_ops = {
+	.domain_init	= fsl_pamu_domain_init,
+	.domain_destroy = fsl_pamu_domain_destroy,
+	.attach_dev	= fsl_pamu_attach_device,
+	.detach_dev	= fsl_pamu_detach_device,
+	.domain_window_enable = fsl_pamu_window_enable,
+	.domain_window_disable = fsl_pamu_window_disable,
+	.domain_get_windows = fsl_pamu_get_windows,
+	.domain_set_windows = fsl_pamu_set_windows,
+	.iova_to_phys	= fsl_pamu_iova_to_phys,
+	.domain_has_cap = fsl_pamu_domain_has_cap,
+	.domain_set_attr = fsl_pamu_set_domain_attr,
+	.domain_get_attr = fsl_pamu_get_domain_attr,
+	.add_device	= fsl_pamu_add_device,
+	.remove_device	= fsl_pamu_remove_device,
+};
+
+int pamu_domain_init()
+{
+	int ret = 0;
+
+	ret = iommu_init_mempool();
+	if (ret)
+		return ret;
+
+	bus_set_iommu(&platform_bus_type, &fsl_pamu_ops);
+	bus_set_iommu(&pci_bus_type, &fsl_pamu_ops);
+
+	return ret;
+}
diff --git a/drivers/iommu/fsl_pamu_domain.h b/drivers/iommu/fsl_pamu_domain.h
new file mode 100644
index 0000000..52dede7
--- /dev/null
+++ b/drivers/iommu/fsl_pamu_domain.h
@@ -0,0 +1,85 @@ 
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright (C) 2013 Freescale Semiconductor, Inc.
+ *
+ */
+
+#ifndef __FSL_PAMU_DOMAIN_H
+#define __FSL_PAMU_DOMAIN_H
+
+#include "fsl_pamu.h"
+
+struct dma_window {
+	phys_addr_t paddr;
+	u64 size;
+	int valid;
+	int prot;
+};
+
+struct fsl_dma_domain {
+	/*
+	 * Indicates the geometry size for the domain.
+	 * This would be set when the geometry is
+	 * configured for the domain.
+	 */
+	dma_addr_t			geom_size;
+	/*
+	 * Number of windows assocaited with this domain.
+	 * During domain initialization, it is set to the
+	 * the maximum number of subwindows allowed for a LIODN.
+	 * Minimum value for this is 1 indicating a single PAMU
+	 * window, without any sub windows. Value can be set/
+	 * queried by set_attr/get_attr API for DOMAIN_ATTR_WINDOWS.
+	 * Value can only be set once the geometry has been configured.
+	 */
+	u32				win_cnt;
+	/*
+	 * win_arr contains information of the configured
+	 * windows for a domain. This is allocated only
+	 * when the number of windows for the domain are
+	 * set.
+	 */
+	struct dma_window		*win_arr;
+	/* list of devices associated with the domain */
+	struct list_head		devices;
+	/* dma_domain states:
+	 * mapped - A particular mapping has been created
+	 * within the configured geometry.
+	 * enabled - DMA has been enabled for the given
+	 * domain. This translates to setting of the
+	 * valid bit for the primary PAACE in the PAMU
+	 * PAACT table. Domain geometry should be set and
+	 * it must have a valid mapping before DMA can be
+	 * enabled for it.
+	 *
+	 */
+	int				mapped;
+	int				enabled;
+	/* stash_id obtained from the stash attribute details */
+	u32				stash_id;
+	struct iommu_stash_attribute	dma_stash;
+	u32				snoop_id;
+	struct iommu_domain		*iommu_domain;
+	spinlock_t			domain_lock;
+};
+
+/* domain-device relationship */
+struct device_domain_info {
+	struct list_head link;	/* link to domain siblings */
+	struct device *dev;
+	u32 liodn;
+	struct fsl_dma_domain *domain; /* pointer to domain */
+};
+#endif  /* __FSL_PAMU_DOMAIN_H */