[PATCH 11/25] drm/amdkfd: Centralize IOMMUv2 code and make it conditional

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

 

On 2018-02-07 06:20 AM, Christian König wrote:
> Am 07.02.2018 um 02:32 schrieb Felix Kuehling:
>> dGPUs work without IOMMUv2. Make IOMMUv2 initialization dependent on
>> ASIC information. Also allow building KFD without IOMMUv2 support.
>> This is still useful for dGPUs and prepares for enabling KFD on
>> architectures that don't support AMD IOMMUv2.
>>
>> v2:
>> * Centralize IOMMUv2 code to avoid #ifdefs in too many places
>>
>> Signed-off-by: Felix Kuehling <Felix.Kuehling at amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdkfd/Kconfig        |   2 +-
>>   drivers/gpu/drm/amd/amdkfd/Makefile       |   4 +
>>   drivers/gpu/drm/amd/amdkfd/kfd_crat.c     |  14 +-
>>   drivers/gpu/drm/amd/amdkfd/kfd_device.c   | 127 +++--------
>>   drivers/gpu/drm/amd/amdkfd/kfd_events.c   |   3 +
>>   drivers/gpu/drm/amd/amdkfd/kfd_iommu.c    | 356
>> ++++++++++++++++++++++++++++++
>>   drivers/gpu/drm/amd/amdkfd/kfd_iommu.h    |  78 +++++++
>>   drivers/gpu/drm/amd/amdkfd/kfd_priv.h     |  14 +-
>>   drivers/gpu/drm/amd/amdkfd/kfd_process.c  | 138 +-----------
>>   drivers/gpu/drm/amd/amdkfd/kfd_topology.c |  16 +-
>>   drivers/gpu/drm/amd/amdkfd/kfd_topology.h |   6 +-
>>   11 files changed, 493 insertions(+), 265 deletions(-)
>>   create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
>>   create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
>>
>> diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig
>> b/drivers/gpu/drm/amd/amdkfd/Kconfig
>> index bc5a294..5bbeb95 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/Kconfig
>> +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
>> @@ -4,6 +4,6 @@
>>     config HSA_AMD
>>       tristate "HSA kernel driver for AMD GPU devices"
>> -    depends on DRM_AMDGPU && AMD_IOMMU_V2 && X86_64
>> +    depends on DRM_AMDGPU && X86_64
>
> You still need a weak dependency on AMD_IOMMU_V2 here, in other words
> add "imply AMD_IOMMU_V2".
>
> This prevents illegal combinations like linking amdkfd into the kernel
> while amd_iommu_v2 is a module.
>
> But it should still allow to completely disable amd_iommu_v2 and
> compile amdkfd without support for it.

Thanks, that's good to know. An updated patch is attached (to avoid
resending the whole series).

Regards,
  Felix

>
> Christian.
>
>>       help
>>         Enable this if you want to use HSA features on AMD GPU devices.
>> diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile
>> b/drivers/gpu/drm/amd/amdkfd/Makefile
>> index a317e76..0d02422 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/Makefile
>> +++ b/drivers/gpu/drm/amd/amdkfd/Makefile
>> @@ -37,6 +37,10 @@ amdkfd-y    := kfd_module.o kfd_device.o
>> kfd_chardev.o kfd_topology.o \
>>           kfd_interrupt.o kfd_events.o cik_event_interrupt.o \
>>           kfd_dbgdev.o kfd_dbgmgr.o kfd_crat.o
>>   +ifneq ($(CONFIG_AMD_IOMMU_V2),)
>> +amdkfd-y += kfd_iommu.o
>> +endif
>> +
>>   amdkfd-$(CONFIG_DEBUG_FS) += kfd_debugfs.o
>>     obj-$(CONFIG_HSA_AMD)    += amdkfd.o
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
>> index 2bc2816..7493f47 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
>> @@ -22,10 +22,10 @@
>>     #include <linux/pci.h>
>>   #include <linux/acpi.h>
>> -#include <linux/amd-iommu.h>
>>   #include "kfd_crat.h"
>>   #include "kfd_priv.h"
>>   #include "kfd_topology.h"
>> +#include "kfd_iommu.h"
>>     /* GPU Processor ID base for dGPUs for which VCRAT needs to be
>> created.
>>    * GPU processor ID are expressed with Bit[31]=1.
>> @@ -1037,15 +1037,11 @@ static int kfd_create_vcrat_image_gpu(void
>> *pcrat_image,
>>       struct crat_subtype_generic *sub_type_hdr;
>>       struct crat_subtype_computeunit *cu;
>>       struct kfd_cu_info cu_info;
>> -    struct amd_iommu_device_info iommu_info;
>>       int avail_size = *size;
>>       uint32_t total_num_of_cu;
>>       int num_of_cache_entries = 0;
>>       int cache_mem_filled = 0;
>>       int ret = 0;
>> -    const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP |
>> -                     AMD_IOMMU_DEVICE_FLAG_PRI_SUP |
>> -                     AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
>>       struct kfd_local_mem_info local_mem_info;
>>         if (!pcrat_image || avail_size < VCRAT_SIZE_FOR_GPU)
>> @@ -1106,12 +1102,8 @@ static int kfd_create_vcrat_image_gpu(void
>> *pcrat_image,
>>       /* Check if this node supports IOMMU. During parsing this flag
>> will
>>        * translate to HSA_CAP_ATS_PRESENT
>>        */
>> -    iommu_info.flags = 0;
>> -    if (amd_iommu_device_info(kdev->pdev, &iommu_info) == 0) {
>> -        if ((iommu_info.flags & required_iommu_flags) ==
>> -                required_iommu_flags)
>> -            cu->hsa_capability |= CRAT_CU_FLAGS_IOMMU_PRESENT;
>> -    }
>> +    if (!kfd_iommu_check_device(kdev))
>> +        cu->hsa_capability |= CRAT_CU_FLAGS_IOMMU_PRESENT;
>>         crat_table->length += sub_type_hdr->length;
>>       crat_table->total_entries++;
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> index 83d6f41..4ac2d61 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> @@ -20,7 +20,9 @@
>>    * OTHER DEALINGS IN THE SOFTWARE.
>>    */
>>   +#if defined(CONFIG_AMD_IOMMU_V2_MODULE) ||
>> defined(CONFIG_AMD_IOMMU_V2)
>>   #include <linux/amd-iommu.h>
>> +#endif
>>   #include <linux/bsearch.h>
>>   #include <linux/pci.h>
>>   #include <linux/slab.h>
>> @@ -28,9 +30,11 @@
>>   #include "kfd_device_queue_manager.h"
>>   #include "kfd_pm4_headers_vi.h"
>>   #include "cwsr_trap_handler_gfx8.asm"
>> +#include "kfd_iommu.h"
>>     #define MQD_SIZE_ALIGNED 768
>>   +#ifdef KFD_SUPPORT_IOMMU_V2
>>   static const struct kfd_device_info kaveri_device_info = {
>>       .asic_family = CHIP_KAVERI,
>>       .max_pasid_bits = 16,
>> @@ -41,6 +45,7 @@ static const struct kfd_device_info
>> kaveri_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = false,
>> +    .needs_iommu_device = true,
>>       .needs_pci_atomics = false,
>>   };
>>   @@ -54,8 +59,10 @@ static const struct kfd_device_info
>> carrizo_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = true,
>>       .needs_pci_atomics = false,
>>   };
>> +#endif
>>     static const struct kfd_device_info hawaii_device_info = {
>>       .asic_family = CHIP_HAWAII,
>> @@ -67,6 +74,7 @@ static const struct kfd_device_info
>> hawaii_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = false,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = false,
>>   };
>>   @@ -79,6 +87,7 @@ static const struct kfd_device_info
>> tonga_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = false,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = true,
>>   };
>>   @@ -91,6 +100,7 @@ static const struct kfd_device_info
>> tonga_vf_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = false,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = false,
>>   };
>>   @@ -103,6 +113,7 @@ static const struct kfd_device_info
>> fiji_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = true,
>>   };
>>   @@ -115,6 +126,7 @@ static const struct kfd_device_info
>> fiji_vf_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = false,
>>   };
>>   @@ -128,6 +140,7 @@ static const struct kfd_device_info
>> polaris10_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = true,
>>   };
>>   @@ -140,6 +153,7 @@ static const struct kfd_device_info
>> polaris10_vf_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = false,
>>   };
>>   @@ -152,6 +166,7 @@ static const struct kfd_device_info
>> polaris11_device_info = {
>>       .num_of_watch_points = 4,
>>       .mqd_size_aligned = MQD_SIZE_ALIGNED,
>>       .supports_cwsr = true,
>> +    .needs_iommu_device = false,
>>       .needs_pci_atomics = true,
>>   };
>>   @@ -162,6 +177,7 @@ struct kfd_deviceid {
>>   };
>>     static const struct kfd_deviceid supported_devices[] = {
>> +#ifdef KFD_SUPPORT_IOMMU_V2
>>       { 0x1304, &kaveri_device_info },    /* Kaveri */
>>       { 0x1305, &kaveri_device_info },    /* Kaveri */
>>       { 0x1306, &kaveri_device_info },    /* Kaveri */
>> @@ -189,6 +205,7 @@ static const struct kfd_deviceid
>> supported_devices[] = {
>>       { 0x9875, &carrizo_device_info },    /* Carrizo */
>>       { 0x9876, &carrizo_device_info },    /* Carrizo */
>>       { 0x9877, &carrizo_device_info },    /* Carrizo */
>> +#endif
>>       { 0x67A0, &hawaii_device_info },    /* Hawaii */
>>       { 0x67A1, &hawaii_device_info },    /* Hawaii */
>>       { 0x67A2, &hawaii_device_info },    /* Hawaii */
>> @@ -302,77 +319,6 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
>>       return kfd;
>>   }
>>   -static bool device_iommu_pasid_init(struct kfd_dev *kfd)
>> -{
>> -    const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP |
>> -                    AMD_IOMMU_DEVICE_FLAG_PRI_SUP |
>> -                    AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
>> -
>> -    struct amd_iommu_device_info iommu_info;
>> -    unsigned int pasid_limit;
>> -    int err;
>> -
>> -    err = amd_iommu_device_info(kfd->pdev, &iommu_info);
>> -    if (err < 0) {
>> -        dev_err(kfd_device,
>> -            "error getting iommu info. is the iommu enabled?\n");
>> -        return false;
>> -    }
>> -
>> -    if ((iommu_info.flags & required_iommu_flags) !=
>> required_iommu_flags) {
>> -        dev_err(kfd_device, "error required iommu flags ats %i, pri
>> %i, pasid %i\n",
>> -               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
>> -               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
>> -               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP)
>> -                                    != 0);
>> -        return false;
>> -    }
>> -
>> -    pasid_limit = min_t(unsigned int,
>> -            (unsigned int)(1 << kfd->device_info->max_pasid_bits),
>> -            iommu_info.max_pasids);
>> -
>> -    if (!kfd_set_pasid_limit(pasid_limit)) {
>> -        dev_err(kfd_device, "error setting pasid limit\n");
>> -        return false;
>> -    }
>> -
>> -    return true;
>> -}
>> -
>> -static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, int
>> pasid)
>> -{
>> -    struct kfd_dev *dev = kfd_device_by_pci_dev(pdev);
>> -
>> -    if (dev)
>> -        kfd_process_iommu_unbind_callback(dev, pasid);
>> -}
>> -
>> -/*
>> - * This function called by IOMMU driver on PPR failure
>> - */
>> -static int iommu_invalid_ppr_cb(struct pci_dev *pdev, int pasid,
>> -        unsigned long address, u16 flags)
>> -{
>> -    struct kfd_dev *dev;
>> -
>> -    dev_warn(kfd_device,
>> -            "Invalid PPR device %x:%x.%x pasid %d address 0x%lX
>> flags 0x%X",
>> -            PCI_BUS_NUM(pdev->devfn),
>> -            PCI_SLOT(pdev->devfn),
>> -            PCI_FUNC(pdev->devfn),
>> -            pasid,
>> -            address,
>> -            flags);
>> -
>> -    dev = kfd_device_by_pci_dev(pdev);
>> -    if (!WARN_ON(!dev))
>> -        kfd_signal_iommu_event(dev, pasid, address,
>> -            flags & PPR_FAULT_WRITE, flags & PPR_FAULT_EXEC);
>> -
>> -    return AMD_IOMMU_INV_PRI_RSP_INVALID;
>> -}
>> -
>>   static void kfd_cwsr_init(struct kfd_dev *kfd)
>>   {
>>       if (cwsr_enable && kfd->device_info->supports_cwsr) {
>> @@ -462,11 +408,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>>           goto device_queue_manager_error;
>>       }
>>   -    if (!device_iommu_pasid_init(kfd)) {
>> -        dev_err(kfd_device,
>> -            "Error initializing iommuv2 for device %x:%x\n",
>> -            kfd->pdev->vendor, kfd->pdev->device);
>> -        goto device_iommu_pasid_error;
>> +    if (kfd_iommu_device_init(kfd)) {
>> +        dev_err(kfd_device, "Error initializing iommuv2\n");
>> +        goto device_iommu_error;
>>       }
>>         kfd_cwsr_init(kfd);
>> @@ -486,7 +430,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>>       goto out;
>>     kfd_resume_error:
>> -device_iommu_pasid_error:
>> +device_iommu_error:
>>       device_queue_manager_uninit(kfd->dqm);
>>   device_queue_manager_error:
>>       kfd_interrupt_exit(kfd);
>> @@ -527,11 +471,7 @@ void kgd2kfd_suspend(struct kfd_dev *kfd)
>>         kfd->dqm->ops.stop(kfd->dqm);
>>   -    kfd_unbind_processes_from_device(kfd);
>> -
>> -    amd_iommu_set_invalidate_ctx_cb(kfd->pdev, NULL);
>> -    amd_iommu_set_invalid_ppr_cb(kfd->pdev, NULL);
>> -    amd_iommu_free_device(kfd->pdev);
>> +    kfd_iommu_suspend(kfd);
>>   }
>>     int kgd2kfd_resume(struct kfd_dev *kfd)
>> @@ -546,19 +486,14 @@ int kgd2kfd_resume(struct kfd_dev *kfd)
>>   static int kfd_resume(struct kfd_dev *kfd)
>>   {
>>       int err = 0;
>> -    unsigned int pasid_limit = kfd_get_pasid_limit();
>> -
>> -    err = amd_iommu_init_device(kfd->pdev, pasid_limit);
>> -    if (err)
>> -        return -ENXIO;
>> -    amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
>> -                    iommu_pasid_shutdown_callback);
>> -    amd_iommu_set_invalid_ppr_cb(kfd->pdev,
>> -                     iommu_invalid_ppr_cb);
>>   -    err = kfd_bind_processes_to_device(kfd);
>> -    if (err)
>> -        goto processes_bind_error;
>> +    err = kfd_iommu_resume(kfd);
>> +    if (err) {
>> +        dev_err(kfd_device,
>> +            "Failed to resume IOMMU for device %x:%x\n",
>> +            kfd->pdev->vendor, kfd->pdev->device);
>> +        return err;
>> +    }
>>         err = kfd->dqm->ops.start(kfd->dqm);
>>       if (err) {
>> @@ -571,9 +506,7 @@ static int kfd_resume(struct kfd_dev *kfd)
>>       return err;
>>     dqm_start_error:
>> -processes_bind_error:
>> -    amd_iommu_free_device(kfd->pdev);
>> -
>> +    kfd_iommu_suspend(kfd);
>>       return err;
>>   }
>>   diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
>> index 93aae5c..6fb9c0d 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
>> @@ -30,6 +30,7 @@
>>   #include <linux/memory.h>
>>   #include "kfd_priv.h"
>>   #include "kfd_events.h"
>> +#include "kfd_iommu.h"
>>   #include <linux/device.h>
>>     /*
>> @@ -837,6 +838,7 @@ static void
>> lookup_events_by_type_and_signal(struct kfd_process *p,
>>       }
>>   }
>>   +#ifdef KFD_SUPPORT_IOMMU_V2
>>   void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid,
>>           unsigned long address, bool is_write_requested,
>>           bool is_execute_requested)
>> @@ -905,6 +907,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev,
>> unsigned int pasid,
>>       mutex_unlock(&p->event_mutex);
>>       kfd_unref_process(p);
>>   }
>> +#endif /* KFD_SUPPORT_IOMMU_V2 */
>>     void kfd_signal_hw_exception_event(unsigned int pasid)
>>   {
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
>> new file mode 100644
>> index 0000000..81dee34
>> --- /dev/null
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
>> @@ -0,0 +1,356 @@
>> +/*
>> + * Copyright 2018 Advanced Micro Devices, Inc.
>> + *
>> + * Permission is hereby granted, free of charge, to any person
>> obtaining a
>> + * copy of this software and associated documentation files (the
>> "Software"),
>> + * to deal in the Software without restriction, including without
>> limitation
>> + * the rights to use, copy, modify, merge, publish, distribute,
>> sublicense,
>> + * and/or sell copies of the Software, and to permit persons to whom
>> the
>> + * Software is furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be
>> included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO
>> EVENT SHALL
>> + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM,
>> DAMAGES OR
>> + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
>> OTHERWISE,
>> + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
>> USE OR
>> + * OTHER DEALINGS IN THE SOFTWARE.
>> + */
>> +
>> +#include <linux/printk.h>
>> +#include <linux/device.h>
>> +#include <linux/slab.h>
>> +#include <linux/pci.h>
>> +#include <linux/amd-iommu.h>
>> +#include "kfd_priv.h"
>> +#include "kfd_dbgmgr.h"
>> +#include "kfd_topology.h"
>> +#include "kfd_iommu.h"
>> +
>> +static const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP |
>> +                    AMD_IOMMU_DEVICE_FLAG_PRI_SUP |
>> +                    AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
>> +
>> +/** kfd_iommu_check_device - Check whether IOMMU is available for
>> device
>> + */
>> +int kfd_iommu_check_device(struct kfd_dev *kfd)
>> +{
>> +    struct amd_iommu_device_info iommu_info;
>> +    int err;
>> +
>> +    if (!kfd->device_info->needs_iommu_device)
>> +        return -ENODEV;
>> +
>> +    iommu_info.flags = 0;
>> +    err = amd_iommu_device_info(kfd->pdev, &iommu_info);
>> +    if (err)
>> +        return err;
>> +
>> +    if ((iommu_info.flags & required_iommu_flags) !=
>> required_iommu_flags)
>> +        return -ENODEV;
>> +
>> +    return 0;
>> +}
>> +
>> +/** kfd_iommu_device_init - Initialize IOMMU for device
>> + */
>> +int kfd_iommu_device_init(struct kfd_dev *kfd)
>> +{
>> +    struct amd_iommu_device_info iommu_info;
>> +    unsigned int pasid_limit;
>> +    int err;
>> +
>> +    if (!kfd->device_info->needs_iommu_device)
>> +        return 0;
>> +
>> +    iommu_info.flags = 0;
>> +    err = amd_iommu_device_info(kfd->pdev, &iommu_info);
>> +    if (err < 0) {
>> +        dev_err(kfd_device,
>> +            "error getting iommu info. is the iommu enabled?\n");
>> +        return -ENODEV;
>> +    }
>> +
>> +    if ((iommu_info.flags & required_iommu_flags) !=
>> required_iommu_flags) {
>> +        dev_err(kfd_device, "error required iommu flags ats %i, pri
>> %i, pasid %i\n",
>> +               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
>> +               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
>> +               (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP)
>> +                                    != 0);
>> +        return -ENODEV;
>> +    }
>> +
>> +    pasid_limit = min_t(unsigned int,
>> +            (unsigned int)(1 << kfd->device_info->max_pasid_bits),
>> +            iommu_info.max_pasids);
>> +
>> +    if (!kfd_set_pasid_limit(pasid_limit)) {
>> +        dev_err(kfd_device, "error setting pasid limit\n");
>> +        return -EBUSY;
>> +    }
>> +
>> +    return 0;
>> +}
>> +
>> +/** kfd_iommu_bind_process_to_device - Have the IOMMU bind a process
>> + *
>> + * Binds the given process to the given device using its PASID. This
>> + * enables IOMMUv2 address translation for the process on the device.
>> + *
>> + * This function assumes that the process mutex is held.
>> + */
>> +int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd)
>> +{
>> +    struct kfd_dev *dev = pdd->dev;
>> +    struct kfd_process *p = pdd->process;
>> +    int err;
>> +
>> +    if (!dev->device_info->needs_iommu_device || pdd->bound ==
>> PDD_BOUND)
>> +        return 0;
>> +
>> +    if (unlikely(pdd->bound == PDD_BOUND_SUSPENDED)) {
>> +        pr_err("Binding PDD_BOUND_SUSPENDED pdd is unexpected!\n");
>> +        return -EINVAL;
>> +    }
>> +
>> +    err = amd_iommu_bind_pasid(dev->pdev, p->pasid, p->lead_thread);
>> +    if (!err)
>> +        pdd->bound = PDD_BOUND;
>> +
>> +    return err;
>> +}
>> +
>> +/** kfd_iommu_unbind_process - Unbind process from all devices
>> + *
>> + * This removes all IOMMU device bindings of the process. To be used
>> + * before process termination.
>> + */
>> +void kfd_iommu_unbind_process(struct kfd_process *p)
>> +{
>> +    struct kfd_process_device *pdd;
>> +
>> +    list_for_each_entry(pdd, &p->per_device_data, per_device_list)
>> +        if (pdd->bound == PDD_BOUND)
>> +            amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid);
>> +}
>> +
>> +/* Callback for process shutdown invoked by the IOMMU driver */
>> +static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, int
>> pasid)
>> +{
>> +    struct kfd_dev *dev = kfd_device_by_pci_dev(pdev);
>> +    struct kfd_process *p;
>> +    struct kfd_process_device *pdd;
>> +
>> +    if (!dev)
>> +        return;
>> +
>> +    /*
>> +     * Look for the process that matches the pasid. If there is no such
>> +     * process, we either released it in amdkfd's own notifier, or
>> there
>> +     * is a bug. Unfortunately, there is no way to tell...
>> +     */
>> +    p = kfd_lookup_process_by_pasid(pasid);
>> +    if (!p)
>> +        return;
>> +
>> +    pr_debug("Unbinding process %d from IOMMU\n", pasid);
>> +
>> +    mutex_lock(kfd_get_dbgmgr_mutex());
>> +
>> +    if (dev->dbgmgr && dev->dbgmgr->pasid == p->pasid) {
>> +        if (!kfd_dbgmgr_unregister(dev->dbgmgr, p)) {
>> +            kfd_dbgmgr_destroy(dev->dbgmgr);
>> +            dev->dbgmgr = NULL;
>> +        }
>> +    }
>> +
>> +    mutex_unlock(kfd_get_dbgmgr_mutex());
>> +
>> +    mutex_lock(&p->mutex);
>> +
>> +    pdd = kfd_get_process_device_data(dev, p);
>> +    if (pdd)
>> +        /* For GPU relying on IOMMU, we need to dequeue here
>> +         * when PASID is still bound.
>> +         */
>> +        kfd_process_dequeue_from_device(pdd);
>> +
>> +    mutex_unlock(&p->mutex);
>> +
>> +    kfd_unref_process(p);
>> +}
>> +
>> +/* This function called by IOMMU driver on PPR failure */
>> +static int iommu_invalid_ppr_cb(struct pci_dev *pdev, int pasid,
>> +        unsigned long address, u16 flags)
>> +{
>> +    struct kfd_dev *dev;
>> +
>> +    dev_warn(kfd_device,
>> +            "Invalid PPR device %x:%x.%x pasid %d address 0x%lX
>> flags 0x%X",
>> +            PCI_BUS_NUM(pdev->devfn),
>> +            PCI_SLOT(pdev->devfn),
>> +            PCI_FUNC(pdev->devfn),
>> +            pasid,
>> +            address,
>> +            flags);
>> +
>> +    dev = kfd_device_by_pci_dev(pdev);
>> +    if (!WARN_ON(!dev))
>> +        kfd_signal_iommu_event(dev, pasid, address,
>> +            flags & PPR_FAULT_WRITE, flags & PPR_FAULT_EXEC);
>> +
>> +    return AMD_IOMMU_INV_PRI_RSP_INVALID;
>> +}
>> +
>> +/*
>> + * Bind processes do the device that have been temporarily unbound
>> + * (PDD_BOUND_SUSPENDED) in kfd_unbind_processes_from_device.
>> + */
>> +static int kfd_bind_processes_to_device(struct kfd_dev *kfd)
>> +{
>> +    struct kfd_process_device *pdd;
>> +    struct kfd_process *p;
>> +    unsigned int temp;
>> +    int err = 0;
>> +
>> +    int idx = srcu_read_lock(&kfd_processes_srcu);
>> +
>> +    hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
>> +        mutex_lock(&p->mutex);
>> +        pdd = kfd_get_process_device_data(kfd, p);
>> +
>> +        if (WARN_ON(!pdd) || pdd->bound != PDD_BOUND_SUSPENDED) {
>> +            mutex_unlock(&p->mutex);
>> +            continue;
>> +        }
>> +
>> +        err = amd_iommu_bind_pasid(kfd->pdev, p->pasid,
>> +                p->lead_thread);
>> +        if (err < 0) {
>> +            pr_err("Unexpected pasid %d binding failure\n",
>> +                    p->pasid);
>> +            mutex_unlock(&p->mutex);
>> +            break;
>> +        }
>> +
>> +        pdd->bound = PDD_BOUND;
>> +        mutex_unlock(&p->mutex);
>> +    }
>> +
>> +    srcu_read_unlock(&kfd_processes_srcu, idx);
>> +
>> +    return err;
>> +}
>> +
>> +/*
>> + * Mark currently bound processes as PDD_BOUND_SUSPENDED. These
>> + * processes will be restored to PDD_BOUND state in
>> + * kfd_bind_processes_to_device.
>> + */
>> +static void kfd_unbind_processes_from_device(struct kfd_dev *kfd)
>> +{
>> +    struct kfd_process_device *pdd;
>> +    struct kfd_process *p;
>> +    unsigned int temp;
>> +
>> +    int idx = srcu_read_lock(&kfd_processes_srcu);
>> +
>> +    hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
>> +        mutex_lock(&p->mutex);
>> +        pdd = kfd_get_process_device_data(kfd, p);
>> +
>> +        if (WARN_ON(!pdd)) {
>> +            mutex_unlock(&p->mutex);
>> +            continue;
>> +        }
>> +
>> +        if (pdd->bound == PDD_BOUND)
>> +            pdd->bound = PDD_BOUND_SUSPENDED;
>> +        mutex_unlock(&p->mutex);
>> +    }
>> +
>> +    srcu_read_unlock(&kfd_processes_srcu, idx);
>> +}
>> +
>> +/** kfd_iommu_suspend - Prepare IOMMU for suspend
>> + *
>> + * This unbinds processes from the device and disables the IOMMU for
>> + * the device.
>> + */
>> +void kfd_iommu_suspend(struct kfd_dev *kfd)
>> +{
>> +    if (!kfd->device_info->needs_iommu_device)
>> +        return;
>> +
>> +    kfd_unbind_processes_from_device(kfd);
>> +
>> +    amd_iommu_set_invalidate_ctx_cb(kfd->pdev, NULL);
>> +    amd_iommu_set_invalid_ppr_cb(kfd->pdev, NULL);
>> +    amd_iommu_free_device(kfd->pdev);
>> +}
>> +
>> +/** kfd_iommu_resume - Restore IOMMU after resume
>> + *
>> + * This reinitializes the IOMMU for the device and re-binds previously
>> + * suspended processes to the device.
>> + */
>> +int kfd_iommu_resume(struct kfd_dev *kfd)
>> +{
>> +    unsigned int pasid_limit;
>> +    int err;
>> +
>> +    if (!kfd->device_info->needs_iommu_device)
>> +        return 0;
>> +
>> +    pasid_limit = kfd_get_pasid_limit();
>> +
>> +    err = amd_iommu_init_device(kfd->pdev, pasid_limit);
>> +    if (err)
>> +        return -ENXIO;
>> +
>> +    amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
>> +                    iommu_pasid_shutdown_callback);
>> +    amd_iommu_set_invalid_ppr_cb(kfd->pdev,
>> +                     iommu_invalid_ppr_cb);
>> +
>> +    err = kfd_bind_processes_to_device(kfd);
>> +    if (err) {
>> +        amd_iommu_set_invalidate_ctx_cb(kfd->pdev, NULL);
>> +        amd_iommu_set_invalid_ppr_cb(kfd->pdev, NULL);
>> +        amd_iommu_free_device(kfd->pdev);
>> +        return err;
>> +    }
>> +
>> +    return 0;
>> +}
>> +
>> +extern bool amd_iommu_pc_supported(void);
>> +extern u8 amd_iommu_pc_get_max_banks(u16 devid);
>> +extern u8 amd_iommu_pc_get_max_counters(u16 devid);
>> +
>> +/** kfd_iommu_add_perf_counters - Add IOMMU performance counters to
>> topology
>> + */
>> +int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev)
>> +{
>> +    struct kfd_perf_properties *props;
>> +
>> +    if (!(kdev->node_props.capability & HSA_CAP_ATS_PRESENT))
>> +        return 0;
>> +
>> +    if (!amd_iommu_pc_supported())
>> +        return 0;
>> +
>> +    props = kfd_alloc_struct(props);
>> +    if (!props)
>> +        return -ENOMEM;
>> +    strcpy(props->block_name, "iommu");
>> +    props->max_concurrent = amd_iommu_pc_get_max_banks(0) *
>> +        amd_iommu_pc_get_max_counters(0); /* assume one iommu */
>> +    list_add_tail(&props->list, &kdev->perf_props);
>> +
>> +    return 0;
>> +}
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
>> b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
>> new file mode 100644
>> index 0000000..dd23d9f
>> --- /dev/null
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
>> @@ -0,0 +1,78 @@
>> +/*
>> + * Copyright 2018 Advanced Micro Devices, Inc.
>> + *
>> + * Permission is hereby granted, free of charge, to any person
>> obtaining a
>> + * copy of this software and associated documentation files (the
>> "Software"),
>> + * to deal in the Software without restriction, including without
>> limitation
>> + * the rights to use, copy, modify, merge, publish, distribute,
>> sublicense,
>> + * and/or sell copies of the Software, and to permit persons to whom
>> the
>> + * Software is furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be
>> included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO
>> EVENT SHALL
>> + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM,
>> DAMAGES OR
>> + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
>> OTHERWISE,
>> + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
>> USE OR
>> + * OTHER DEALINGS IN THE SOFTWARE.
>> + */
>> +
>> +#ifndef __KFD_IOMMU_H__
>> +#define __KFD_IOMMU_H__
>> +
>> +#if defined(CONFIG_AMD_IOMMU_V2_MODULE) || defined(CONFIG_AMD_IOMMU_V2)
>> +
>> +#define KFD_SUPPORT_IOMMU_V2
>> +
>> +int kfd_iommu_check_device(struct kfd_dev *kfd);
>> +int kfd_iommu_device_init(struct kfd_dev *kfd);
>> +
>> +int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd);
>> +void kfd_iommu_unbind_process(struct kfd_process *p);
>> +
>> +void kfd_iommu_suspend(struct kfd_dev *kfd);
>> +int kfd_iommu_resume(struct kfd_dev *kfd);
>> +
>> +int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev);
>> +
>> +#else
>> +
>> +static inline int kfd_iommu_check_device(struct kfd_dev *kfd)
>> +{
>> +    return -ENODEV;
>> +}
>> +static inline int kfd_iommu_device_init(struct kfd_dev *kfd)
>> +{
>> +    return 0;
>> +}
>> +
>> +static inline int kfd_iommu_bind_process_to_device(
>> +    struct kfd_process_device *pdd)
>> +{
>> +    return 0;
>> +}
>> +static inline void kfd_iommu_unbind_process(struct kfd_process *p)
>> +{
>> +    /* empty */
>> +}
>> +
>> +static inline void kfd_iommu_suspend(struct kfd_dev *kfd)
>> +{
>> +    /* empty */
>> +}
>> +static inline int kfd_iommu_resume(struct kfd_dev *kfd)
>> +{
>> +    return 0;
>> +}
>> +
>> +static inline int kfd_iommu_add_perf_counters(struct
>> kfd_topology_device *kdev)
>> +{
>> +    return 0;
>> +}
>> +
>> +#endif /* defined(CONFIG_AMD_IOMMU_V2) */
>> +
>> +#endif /* __KFD_IOMMU_H__ */
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
>> b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
>> index 594f853..f12eb5d 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
>> @@ -158,6 +158,7 @@ struct kfd_device_info {
>>       uint8_t num_of_watch_points;
>>       uint16_t mqd_size_aligned;
>>       bool supports_cwsr;
>> +    bool needs_iommu_device;
>>       bool needs_pci_atomics;
>>   };
>>   @@ -517,15 +518,15 @@ struct kfd_process_device {
>>       uint64_t scratch_base;
>>       uint64_t scratch_limit;
>>   -    /* Is this process/pasid bound to this device?
>> (amd_iommu_bind_pasid) */
>> -    enum kfd_pdd_bound bound;
>> -
>>       /* Flag used to tell the pdd has dequeued from the dqm.
>>        * This is used to prevent dev->dqm->ops.process_termination()
>> from
>>        * being called twice when it is already called in IOMMU callback
>>        * function.
>>        */
>>       bool already_dequeued;
>> +
>> +    /* Is this process/pasid bound to this device?
>> (amd_iommu_bind_pasid) */
>> +    enum kfd_pdd_bound bound;
>>   };
>>     #define qpd_to_pdd(x) container_of(x, struct kfd_process_device,
>> qpd)
>> @@ -590,6 +591,10 @@ struct kfd_process {
>>       bool signal_event_limit_reached;
>>   };
>>   +#define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */
>> +extern DECLARE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE);
>> +extern struct srcu_struct kfd_processes_srcu;
>> +
>>   /**
>>    * Ioctl function type.
>>    *
>> @@ -617,9 +622,6 @@ void kfd_unref_process(struct kfd_process *p);
>>     struct kfd_process_device *kfd_bind_process_to_device(struct
>> kfd_dev *dev,
>>                           struct kfd_process *p);
>> -int kfd_bind_processes_to_device(struct kfd_dev *dev);
>> -void kfd_unbind_processes_from_device(struct kfd_dev *dev);
>> -void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned
>> int pasid);
>>   struct kfd_process_device *kfd_get_process_device_data(struct
>> kfd_dev *dev,
>>                               struct kfd_process *p);
>>   struct kfd_process_device *kfd_create_process_device_data(struct
>> kfd_dev *dev,
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> index 4ff5f0f..e9aee76 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> @@ -35,16 +35,16 @@ struct mm_struct;
>>     #include "kfd_priv.h"
>>   #include "kfd_dbgmgr.h"
>> +#include "kfd_iommu.h"
>>     /*
>>    * List of struct kfd_process (field kfd_process).
>>    * Unique/indexed by mm_struct*
>>    */
>> -#define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */
>> -static DEFINE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE);
>> +DEFINE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE);
>>   static DEFINE_MUTEX(kfd_processes_mutex);
>>   -DEFINE_STATIC_SRCU(kfd_processes_srcu);
>> +DEFINE_SRCU(kfd_processes_srcu);
>>     static struct workqueue_struct *kfd_process_wq;
>>   @@ -173,14 +173,8 @@ static void kfd_process_wq_release(struct
>> work_struct *work)
>>   {
>>       struct kfd_process *p = container_of(work, struct kfd_process,
>>                            release_work);
>> -    struct kfd_process_device *pdd;
>>   -    pr_debug("Releasing process (pasid %d) in workqueue\n",
>> p->pasid);
>> -
>> -    list_for_each_entry(pdd, &p->per_device_data, per_device_list) {
>> -        if (pdd->bound == PDD_BOUND)
>> -            amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid);
>> -    }
>> +    kfd_iommu_unbind_process(p);
>>         kfd_process_destroy_pdds(p);
>>   @@ -429,133 +423,13 @@ struct kfd_process_device
>> *kfd_bind_process_to_device(struct kfd_dev *dev,
>>           return ERR_PTR(-ENOMEM);
>>       }
>>   -    if (pdd->bound == PDD_BOUND) {
>> -        return pdd;
>> -    } else if (unlikely(pdd->bound == PDD_BOUND_SUSPENDED)) {
>> -        pr_err("Binding PDD_BOUND_SUSPENDED pdd is unexpected!\n");
>> -        return ERR_PTR(-EINVAL);
>> -    }
>> -
>> -    err = amd_iommu_bind_pasid(dev->pdev, p->pasid, p->lead_thread);
>> -    if (err < 0)
>> +    err = kfd_iommu_bind_process_to_device(pdd);
>> +    if (err)
>>           return ERR_PTR(err);
>>   -    pdd->bound = PDD_BOUND;
>> -
>>       return pdd;
>>   }
>>   -/*
>> - * Bind processes do the device that have been temporarily unbound
>> - * (PDD_BOUND_SUSPENDED) in kfd_unbind_processes_from_device.
>> - */
>> -int kfd_bind_processes_to_device(struct kfd_dev *dev)
>> -{
>> -    struct kfd_process_device *pdd;
>> -    struct kfd_process *p;
>> -    unsigned int temp;
>> -    int err = 0;
>> -
>> -    int idx = srcu_read_lock(&kfd_processes_srcu);
>> -
>> -    hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
>> -        mutex_lock(&p->mutex);
>> -        pdd = kfd_get_process_device_data(dev, p);
>> -
>> -        if (WARN_ON(!pdd) || pdd->bound != PDD_BOUND_SUSPENDED) {
>> -            mutex_unlock(&p->mutex);
>> -            continue;
>> -        }
>> -
>> -        err = amd_iommu_bind_pasid(dev->pdev, p->pasid,
>> -                p->lead_thread);
>> -        if (err < 0) {
>> -            pr_err("Unexpected pasid %d binding failure\n",
>> -                    p->pasid);
>> -            mutex_unlock(&p->mutex);
>> -            break;
>> -        }
>> -
>> -        pdd->bound = PDD_BOUND;
>> -        mutex_unlock(&p->mutex);
>> -    }
>> -
>> -    srcu_read_unlock(&kfd_processes_srcu, idx);
>> -
>> -    return err;
>> -}
>> -
>> -/*
>> - * Mark currently bound processes as PDD_BOUND_SUSPENDED. These
>> - * processes will be restored to PDD_BOUND state in
>> - * kfd_bind_processes_to_device.
>> - */
>> -void kfd_unbind_processes_from_device(struct kfd_dev *dev)
>> -{
>> -    struct kfd_process_device *pdd;
>> -    struct kfd_process *p;
>> -    unsigned int temp;
>> -
>> -    int idx = srcu_read_lock(&kfd_processes_srcu);
>> -
>> -    hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
>> -        mutex_lock(&p->mutex);
>> -        pdd = kfd_get_process_device_data(dev, p);
>> -
>> -        if (WARN_ON(!pdd)) {
>> -            mutex_unlock(&p->mutex);
>> -            continue;
>> -        }
>> -
>> -        if (pdd->bound == PDD_BOUND)
>> -            pdd->bound = PDD_BOUND_SUSPENDED;
>> -        mutex_unlock(&p->mutex);
>> -    }
>> -
>> -    srcu_read_unlock(&kfd_processes_srcu, idx);
>> -}
>> -
>> -void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned
>> int pasid)
>> -{
>> -    struct kfd_process *p;
>> -    struct kfd_process_device *pdd;
>> -
>> -    /*
>> -     * Look for the process that matches the pasid. If there is no such
>> -     * process, we either released it in amdkfd's own notifier, or
>> there
>> -     * is a bug. Unfortunately, there is no way to tell...
>> -     */
>> -    p = kfd_lookup_process_by_pasid(pasid);
>> -    if (!p)
>> -        return;
>> -
>> -    pr_debug("Unbinding process %d from IOMMU\n", pasid);
>> -
>> -    mutex_lock(kfd_get_dbgmgr_mutex());
>> -
>> -    if (dev->dbgmgr && dev->dbgmgr->pasid == p->pasid) {
>> -        if (!kfd_dbgmgr_unregister(dev->dbgmgr, p)) {
>> -            kfd_dbgmgr_destroy(dev->dbgmgr);
>> -            dev->dbgmgr = NULL;
>> -        }
>> -    }
>> -
>> -    mutex_unlock(kfd_get_dbgmgr_mutex());
>> -
>> -    mutex_lock(&p->mutex);
>> -
>> -    pdd = kfd_get_process_device_data(dev, p);
>> -    if (pdd)
>> -        /* For GPU relying on IOMMU, we need to dequeue here
>> -         * when PASID is still bound.
>> -         */
>> -        kfd_process_dequeue_from_device(pdd);
>> -
>> -    mutex_unlock(&p->mutex);
>> -
>> -    kfd_unref_process(p);
>> -}
>> -
>>   struct kfd_process_device *kfd_get_first_process_device_data(
>>                           struct kfd_process *p)
>>   {
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
>> index 7783250..2506155 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
>> @@ -35,6 +35,7 @@
>>   #include "kfd_crat.h"
>>   #include "kfd_topology.h"
>>   #include "kfd_device_queue_manager.h"
>> +#include "kfd_iommu.h"
>>     /* topology_device_list - Master list of all topology devices */
>>   static struct list_head topology_device_list;
>> @@ -875,19 +876,8 @@ static void find_system_memory(const struct
>> dmi_header *dm,
>>    */
>>   static int kfd_add_perf_to_topology(struct kfd_topology_device *kdev)
>>   {
>> -    struct kfd_perf_properties *props;
>> -
>> -    if (amd_iommu_pc_supported()) {
>> -        props = kfd_alloc_struct(props);
>> -        if (!props)
>> -            return -ENOMEM;
>> -        strcpy(props->block_name, "iommu");
>> -        props->max_concurrent = amd_iommu_pc_get_max_banks(0) *
>> -            amd_iommu_pc_get_max_counters(0); /* assume one iommu */
>> -        list_add_tail(&props->list, &kdev->perf_props);
>> -    }
>> -
>> -    return 0;
>> +    /* These are the only counters supported so far */
>> +    return kfd_iommu_add_perf_counters(kdev);
>>   }
>>     /* kfd_add_non_crat_information - Add information that is not
>> currently
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
>> b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
>> index 53fca1f..c0be2be 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
>> @@ -25,7 +25,7 @@
>>     #include <linux/types.h>
>>   #include <linux/list.h>
>> -#include "kfd_priv.h"
>> +#include "kfd_crat.h"
>>     #define KFD_TOPOLOGY_PUBLIC_NAME_SIZE 128
>>   @@ -183,8 +183,4 @@ struct kfd_topology_device
>> *kfd_create_topology_device(
>>           struct list_head *device_list);
>>   void kfd_release_topology_device_list(struct list_head *device_list);
>>   -extern bool amd_iommu_pc_supported(void);
>> -extern u8 amd_iommu_pc_get_max_banks(u16 devid);
>> -extern u8 amd_iommu_pc_get_max_counters(u16 devid);
>> -
>>   #endif /* __KFD_TOPOLOGY_H__ */
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0011-drm-amdkfd-Centralize-IOMMUv2-code-and-make-it-condi.patch
Type: text/x-patch
Size: 34246 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20180207/7b353c52/attachment-0001.bin>
[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux