linux/drivers/vfio/vfio_iommu_spapr_tce.c

1291 lines
30 KiB
C
Raw Normal View History

/*
* VFIO: IOMMU DMA mapping support for TCE on POWER
*
* Copyright (C) 2013 IBM Corp. All rights reserved.
* Author: Alexey Kardashevskiy <aik@ozlabs.ru>
*
* 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.
*
* Derived from original vfio_iommu_type1.c:
* Copyright (C) 2012 Red Hat, Inc. All rights reserved.
* Author: Alex Williamson <alex.williamson@redhat.com>
*/
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/slab.h>
#include <linux/uaccess.h>
#include <linux/err.h>
#include <linux/vfio.h>
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
#include <linux/vmalloc.h>
#include <asm/iommu.h>
#include <asm/tce.h>
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
#include <asm/mmu_context.h>
#define DRIVER_VERSION "0.1"
#define DRIVER_AUTHOR "aik@ozlabs.ru"
#define DRIVER_DESC "VFIO IOMMU SPAPR TCE"
static void tce_iommu_detach_group(void *iommu_data,
struct iommu_group *iommu_group);
static long try_increment_locked_vm(long npages)
{
long ret = 0, locked, lock_limit;
if (!current || !current->mm)
return -ESRCH; /* process exited */
if (!npages)
return 0;
down_write(&current->mm->mmap_sem);
locked = current->mm->locked_vm + npages;
lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
if (locked > lock_limit && !capable(CAP_IPC_LOCK))
ret = -ENOMEM;
else
current->mm->locked_vm += npages;
pr_debug("[%d] RLIMIT_MEMLOCK +%ld %ld/%ld%s\n", current->pid,
npages << PAGE_SHIFT,
current->mm->locked_vm << PAGE_SHIFT,
rlimit(RLIMIT_MEMLOCK),
ret ? " - exceeded" : "");
up_write(&current->mm->mmap_sem);
return ret;
}
static void decrement_locked_vm(long npages)
{
if (!current || !current->mm || !npages)
return; /* process exited */
down_write(&current->mm->mmap_sem);
if (WARN_ON_ONCE(npages > current->mm->locked_vm))
npages = current->mm->locked_vm;
current->mm->locked_vm -= npages;
pr_debug("[%d] RLIMIT_MEMLOCK -%ld %ld/%ld\n", current->pid,
npages << PAGE_SHIFT,
current->mm->locked_vm << PAGE_SHIFT,
rlimit(RLIMIT_MEMLOCK));
up_write(&current->mm->mmap_sem);
}
/*
* VFIO IOMMU fd for SPAPR_TCE IOMMU implementation
*
* This code handles mapping and unmapping of user data buffers
* into DMA'ble space using the IOMMU
*/
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct tce_iommu_group {
struct list_head next;
struct iommu_group *grp;
};
/*
* The container descriptor supports only a single group per container.
* Required by the API as the container is not supplied with the IOMMU group
* at the moment of initialization.
*/
struct tce_container {
struct mutex lock;
bool enabled;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
bool v2;
unsigned long locked_pages;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct iommu_table *tables[IOMMU_TABLE_GROUP_MAX_TABLES];
struct list_head group_list;
};
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
static long tce_iommu_unregister_pages(struct tce_container *container,
__u64 vaddr, __u64 size)
{
struct mm_iommu_table_group_mem_t *mem;
if ((vaddr & ~PAGE_MASK) || (size & ~PAGE_MASK))
return -EINVAL;
mem = mm_iommu_find(vaddr, size >> PAGE_SHIFT);
if (!mem)
return -ENOENT;
return mm_iommu_put(mem);
}
static long tce_iommu_register_pages(struct tce_container *container,
__u64 vaddr, __u64 size)
{
long ret = 0;
struct mm_iommu_table_group_mem_t *mem = NULL;
unsigned long entries = size >> PAGE_SHIFT;
if ((vaddr & ~PAGE_MASK) || (size & ~PAGE_MASK) ||
((vaddr + size) < vaddr))
return -EINVAL;
ret = mm_iommu_get(vaddr, entries, &mem);
if (ret)
return ret;
container->enabled = true;
return 0;
}
static long tce_iommu_userspace_view_alloc(struct iommu_table *tbl)
{
unsigned long cb = _ALIGN_UP(sizeof(tbl->it_userspace[0]) *
tbl->it_size, PAGE_SIZE);
unsigned long *uas;
long ret;
BUG_ON(tbl->it_userspace);
ret = try_increment_locked_vm(cb >> PAGE_SHIFT);
if (ret)
return ret;
uas = vzalloc(cb);
if (!uas) {
decrement_locked_vm(cb >> PAGE_SHIFT);
return -ENOMEM;
}
tbl->it_userspace = uas;
return 0;
}
static void tce_iommu_userspace_view_free(struct iommu_table *tbl)
{
unsigned long cb = _ALIGN_UP(sizeof(tbl->it_userspace[0]) *
tbl->it_size, PAGE_SIZE);
if (!tbl->it_userspace)
return;
vfree(tbl->it_userspace);
tbl->it_userspace = NULL;
decrement_locked_vm(cb >> PAGE_SHIFT);
}
static bool tce_page_is_contained(struct page *page, unsigned page_shift)
{
/*
* Check that the TCE table granularity is not bigger than the size of
* a page we just found. Otherwise the hardware can get access to
* a bigger memory chunk that it should.
*/
return (PAGE_SHIFT + compound_order(compound_head(page))) >= page_shift;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
static inline bool tce_groups_attached(struct tce_container *container)
{
return !list_empty(&container->group_list);
}
static long tce_iommu_find_table(struct tce_container *container,
phys_addr_t ioba, struct iommu_table **ptbl)
{
long i;
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct iommu_table *tbl = container->tables[i];
if (tbl) {
unsigned long entry = ioba >> tbl->it_page_shift;
unsigned long start = tbl->it_offset;
unsigned long end = start + tbl->it_size;
if ((start <= entry) && (entry < end)) {
*ptbl = tbl;
return i;
}
}
}
return -1;
}
static int tce_iommu_find_free_table(struct tce_container *container)
{
int i;
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
if (!container->tables[i])
return i;
}
return -ENOSPC;
}
static int tce_iommu_enable(struct tce_container *container)
{
int ret = 0;
unsigned long locked;
struct iommu_table_group *table_group;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct tce_iommu_group *tcegrp;
if (!current->mm)
return -ESRCH; /* process exited */
if (container->enabled)
return -EBUSY;
/*
* When userspace pages are mapped into the IOMMU, they are effectively
* locked memory, so, theoretically, we need to update the accounting
* of locked pages on each map and unmap. For powerpc, the map unmap
* paths can be very hot, though, and the accounting would kill
* performance, especially since it would be difficult to impossible
* to handle the accounting in real mode only.
*
* To address that, rather than precisely accounting every page, we
* instead account for a worst case on locked memory when the iommu is
* enabled and disabled. The worst case upper bound on locked memory
* is the size of the whole iommu window, which is usually relatively
* small (compared to total memory sizes) on POWER hardware.
*
* Also we don't have a nice way to fail on H_PUT_TCE due to ulimits,
* that would effectively kill the guest at random points, much better
* enforcing the limit based on the max that the guest can map.
*
* Unfortunately at the moment it counts whole tables, no matter how
* much memory the guest has. I.e. for 4GB guest and 4 IOMMU groups
* each with 2GB DMA window, 8GB will be counted here. The reason for
* this is that we cannot tell here the amount of RAM used by the guest
* as this information is only available from KVM and VFIO is
* KVM agnostic.
*
* So we do not allow enabling a container without a group attached
* as there is no way to know how much we should increment
* the locked_vm counter.
*/
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (!tce_groups_attached(container))
return -ENODEV;
tcegrp = list_first_entry(&container->group_list,
struct tce_iommu_group, next);
table_group = iommu_group_get_iommudata(tcegrp->grp);
if (!table_group)
return -ENODEV;
if (!table_group->tce32_size)
return -EPERM;
locked = table_group->tce32_size >> PAGE_SHIFT;
ret = try_increment_locked_vm(locked);
if (ret)
return ret;
container->locked_pages = locked;
container->enabled = true;
return ret;
}
static void tce_iommu_disable(struct tce_container *container)
{
if (!container->enabled)
return;
container->enabled = false;
if (!current->mm)
return;
decrement_locked_vm(container->locked_pages);
}
static void *tce_iommu_open(unsigned long arg)
{
struct tce_container *container;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if ((arg != VFIO_SPAPR_TCE_IOMMU) && (arg != VFIO_SPAPR_TCE_v2_IOMMU)) {
pr_err("tce_vfio: Wrong IOMMU type\n");
return ERR_PTR(-EINVAL);
}
container = kzalloc(sizeof(*container), GFP_KERNEL);
if (!container)
return ERR_PTR(-ENOMEM);
mutex_init(&container->lock);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
INIT_LIST_HEAD_RCU(&container->group_list);
container->v2 = arg == VFIO_SPAPR_TCE_v2_IOMMU;
return container;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
static int tce_iommu_clear(struct tce_container *container,
struct iommu_table *tbl,
unsigned long entry, unsigned long pages);
static void tce_iommu_free_table(struct iommu_table *tbl);
static void tce_iommu_release(void *iommu_data)
{
struct tce_container *container = iommu_data;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct tce_iommu_group *tcegrp;
long i;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
while (tce_groups_attached(container)) {
tcegrp = list_first_entry(&container->group_list,
struct tce_iommu_group, next);
tce_iommu_detach_group(iommu_data, tcegrp->grp);
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
/*
* If VFIO created a table, it was not disposed
* by tce_iommu_detach_group() so do it now.
*/
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
struct iommu_table *tbl = container->tables[i];
if (!tbl)
continue;
tce_iommu_clear(container, tbl, tbl->it_offset, tbl->it_size);
tce_iommu_free_table(tbl);
}
tce_iommu_disable(container);
mutex_destroy(&container->lock);
kfree(container);
}
static void tce_iommu_unuse_page(struct tce_container *container,
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
unsigned long hpa)
{
struct page *page;
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
page = pfn_to_page(hpa >> PAGE_SHIFT);
put_page(page);
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
static int tce_iommu_prereg_ua_to_hpa(unsigned long tce, unsigned long size,
unsigned long *phpa, struct mm_iommu_table_group_mem_t **pmem)
{
long ret = 0;
struct mm_iommu_table_group_mem_t *mem;
mem = mm_iommu_lookup(tce, size);
if (!mem)
return -EINVAL;
ret = mm_iommu_ua_to_hpa(mem, tce, phpa);
if (ret)
return -EINVAL;
*pmem = mem;
return 0;
}
static void tce_iommu_unuse_page_v2(struct iommu_table *tbl,
unsigned long entry)
{
struct mm_iommu_table_group_mem_t *mem = NULL;
int ret;
unsigned long hpa = 0;
unsigned long *pua = IOMMU_TABLE_USERSPACE_ENTRY(tbl, entry);
if (!pua || !current || !current->mm)
return;
ret = tce_iommu_prereg_ua_to_hpa(*pua, IOMMU_PAGE_SIZE(tbl),
&hpa, &mem);
if (ret)
pr_debug("%s: tce %lx at #%lx was not cached, ret=%d\n",
__func__, *pua, entry, ret);
if (mem)
mm_iommu_mapped_dec(mem);
*pua = 0;
}
static int tce_iommu_clear(struct tce_container *container,
struct iommu_table *tbl,
unsigned long entry, unsigned long pages)
{
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
unsigned long oldhpa;
long ret;
enum dma_data_direction direction;
for ( ; pages; --pages, ++entry) {
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
direction = DMA_NONE;
oldhpa = 0;
ret = iommu_tce_xchg(tbl, entry, &oldhpa, &direction);
if (ret)
continue;
if (direction == DMA_NONE)
continue;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (container->v2) {
tce_iommu_unuse_page_v2(tbl, entry);
continue;
}
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
tce_iommu_unuse_page(container, oldhpa);
}
return 0;
}
static int tce_iommu_use_page(unsigned long tce, unsigned long *hpa)
{
struct page *page = NULL;
enum dma_data_direction direction = iommu_tce_direction(tce);
if (get_user_pages_fast(tce & PAGE_MASK, 1,
direction != DMA_TO_DEVICE, &page) != 1)
return -EFAULT;
*hpa = __pa((unsigned long) page_address(page));
return 0;
}
static long tce_iommu_build(struct tce_container *container,
struct iommu_table *tbl,
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
unsigned long entry, unsigned long tce, unsigned long pages,
enum dma_data_direction direction)
{
long i, ret = 0;
struct page *page;
unsigned long hpa;
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
enum dma_data_direction dirtmp;
for (i = 0; i < pages; ++i) {
unsigned long offset = tce & IOMMU_PAGE_MASK(tbl) & ~PAGE_MASK;
ret = tce_iommu_use_page(tce, &hpa);
if (ret)
break;
page = pfn_to_page(hpa >> PAGE_SHIFT);
if (!tce_page_is_contained(page, tbl->it_page_shift)) {
ret = -EPERM;
break;
}
hpa |= offset;
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
dirtmp = direction;
ret = iommu_tce_xchg(tbl, entry + i, &hpa, &dirtmp);
if (ret) {
tce_iommu_unuse_page(container, hpa);
pr_err("iommu_tce: %s failed ioba=%lx, tce=%lx, ret=%ld\n",
__func__, entry << tbl->it_page_shift,
tce, ret);
break;
}
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
if (dirtmp != DMA_NONE)
tce_iommu_unuse_page(container, hpa);
tce += IOMMU_PAGE_SIZE(tbl);
}
if (ret)
tce_iommu_clear(container, tbl, entry, i);
return ret;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
static long tce_iommu_build_v2(struct tce_container *container,
struct iommu_table *tbl,
unsigned long entry, unsigned long tce, unsigned long pages,
enum dma_data_direction direction)
{
long i, ret = 0;
struct page *page;
unsigned long hpa;
enum dma_data_direction dirtmp;
for (i = 0; i < pages; ++i) {
struct mm_iommu_table_group_mem_t *mem = NULL;
unsigned long *pua = IOMMU_TABLE_USERSPACE_ENTRY(tbl,
entry + i);
ret = tce_iommu_prereg_ua_to_hpa(tce, IOMMU_PAGE_SIZE(tbl),
&hpa, &mem);
if (ret)
break;
page = pfn_to_page(hpa >> PAGE_SHIFT);
if (!tce_page_is_contained(page, tbl->it_page_shift)) {
ret = -EPERM;
break;
}
/* Preserve offset within IOMMU page */
hpa |= tce & IOMMU_PAGE_MASK(tbl) & ~PAGE_MASK;
dirtmp = direction;
/* The registered region is being unregistered */
if (mm_iommu_mapped_inc(mem))
break;
ret = iommu_tce_xchg(tbl, entry + i, &hpa, &dirtmp);
if (ret) {
/* dirtmp cannot be DMA_NONE here */
tce_iommu_unuse_page_v2(tbl, entry + i);
pr_err("iommu_tce: %s failed ioba=%lx, tce=%lx, ret=%ld\n",
__func__, entry << tbl->it_page_shift,
tce, ret);
break;
}
if (dirtmp != DMA_NONE)
tce_iommu_unuse_page_v2(tbl, entry + i);
*pua = tce;
tce += IOMMU_PAGE_SIZE(tbl);
}
if (ret)
tce_iommu_clear(container, tbl, entry, i);
return ret;
}
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
static long tce_iommu_create_table(struct tce_container *container,
struct iommu_table_group *table_group,
int num,
__u32 page_shift,
__u64 window_size,
__u32 levels,
struct iommu_table **ptbl)
{
long ret, table_size;
table_size = table_group->ops->get_table_size(page_shift, window_size,
levels);
if (!table_size)
return -EINVAL;
ret = try_increment_locked_vm(table_size >> PAGE_SHIFT);
if (ret)
return ret;
ret = table_group->ops->create_table(table_group, num,
page_shift, window_size, levels, ptbl);
WARN_ON(!ret && !(*ptbl)->it_ops->free);
WARN_ON(!ret && ((*ptbl)->it_allocated_size != table_size));
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (!ret && container->v2) {
ret = tce_iommu_userspace_view_alloc(*ptbl);
if (ret)
(*ptbl)->it_ops->free(*ptbl);
}
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
if (ret)
decrement_locked_vm(table_size >> PAGE_SHIFT);
return ret;
}
static void tce_iommu_free_table(struct iommu_table *tbl)
{
unsigned long pages = tbl->it_allocated_size >> PAGE_SHIFT;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
tce_iommu_userspace_view_free(tbl);
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
tbl->it_ops->free(tbl);
decrement_locked_vm(pages);
}
static long tce_iommu_create_window(struct tce_container *container,
__u32 page_shift, __u64 window_size, __u32 levels,
__u64 *start_addr)
{
struct tce_iommu_group *tcegrp;
struct iommu_table_group *table_group;
struct iommu_table *tbl = NULL;
long ret, num;
num = tce_iommu_find_free_table(container);
if (num < 0)
return num;
/* Get the first group for ops::create_table */
tcegrp = list_first_entry(&container->group_list,
struct tce_iommu_group, next);
table_group = iommu_group_get_iommudata(tcegrp->grp);
if (!table_group)
return -EFAULT;
if (!(table_group->pgsizes & (1ULL << page_shift)))
return -EINVAL;
if (!table_group->ops->set_window || !table_group->ops->unset_window ||
!table_group->ops->get_table_size ||
!table_group->ops->create_table)
return -EPERM;
/* Create TCE table */
ret = tce_iommu_create_table(container, table_group, num,
page_shift, window_size, levels, &tbl);
if (ret)
return ret;
BUG_ON(!tbl->it_ops->free);
/*
* Program the table to every group.
* Groups have been tested for compatibility at the attach time.
*/
list_for_each_entry(tcegrp, &container->group_list, next) {
table_group = iommu_group_get_iommudata(tcegrp->grp);
ret = table_group->ops->set_window(table_group, num, tbl);
if (ret)
goto unset_exit;
}
container->tables[num] = tbl;
/* Return start address assigned by platform in create_table() */
*start_addr = tbl->it_offset << tbl->it_page_shift;
return 0;
unset_exit:
list_for_each_entry(tcegrp, &container->group_list, next) {
table_group = iommu_group_get_iommudata(tcegrp->grp);
table_group->ops->unset_window(table_group, num);
}
tce_iommu_free_table(tbl);
return ret;
}
static long tce_iommu_remove_window(struct tce_container *container,
__u64 start_addr)
{
struct iommu_table_group *table_group = NULL;
struct iommu_table *tbl;
struct tce_iommu_group *tcegrp;
int num;
num = tce_iommu_find_table(container, start_addr, &tbl);
if (num < 0)
return -EINVAL;
BUG_ON(!tbl->it_size);
/* Detach groups from IOMMUs */
list_for_each_entry(tcegrp, &container->group_list, next) {
table_group = iommu_group_get_iommudata(tcegrp->grp);
/*
* SPAPR TCE IOMMU exposes the default DMA window to
* the guest via dma32_window_start/size of
* VFIO_IOMMU_SPAPR_TCE_GET_INFO. Some platforms allow
* the userspace to remove this window, some do not so
* here we check for the platform capability.
*/
if (!table_group->ops || !table_group->ops->unset_window)
return -EPERM;
table_group->ops->unset_window(table_group, num);
}
/* Free table */
tce_iommu_clear(container, tbl, tbl->it_offset, tbl->it_size);
tce_iommu_free_table(tbl);
container->tables[num] = NULL;
return 0;
}
static long tce_iommu_ioctl(void *iommu_data,
unsigned int cmd, unsigned long arg)
{
struct tce_container *container = iommu_data;
unsigned long minsz, ddwsz;
long ret;
switch (cmd) {
case VFIO_CHECK_EXTENSION:
switch (arg) {
case VFIO_SPAPR_TCE_IOMMU:
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
case VFIO_SPAPR_TCE_v2_IOMMU:
ret = 1;
break;
default:
ret = vfio_spapr_iommu_eeh_ioctl(NULL, cmd, arg);
break;
}
return (ret < 0) ? 0 : ret;
case VFIO_IOMMU_SPAPR_TCE_GET_INFO: {
struct vfio_iommu_spapr_tce_info info;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct tce_iommu_group *tcegrp;
struct iommu_table_group *table_group;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (!tce_groups_attached(container))
return -ENXIO;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
tcegrp = list_first_entry(&container->group_list,
struct tce_iommu_group, next);
table_group = iommu_group_get_iommudata(tcegrp->grp);
if (!table_group)
return -ENXIO;
minsz = offsetofend(struct vfio_iommu_spapr_tce_info,
dma32_window_size);
if (copy_from_user(&info, (void __user *)arg, minsz))
return -EFAULT;
if (info.argsz < minsz)
return -EINVAL;
info.dma32_window_start = table_group->tce32_start;
info.dma32_window_size = table_group->tce32_size;
info.flags = 0;
memset(&info.ddw, 0, sizeof(info.ddw));
if (table_group->max_dynamic_windows_supported &&
container->v2) {
info.flags |= VFIO_IOMMU_SPAPR_INFO_DDW;
info.ddw.pgsizes = table_group->pgsizes;
info.ddw.max_dynamic_windows_supported =
table_group->max_dynamic_windows_supported;
info.ddw.levels = table_group->max_levels;
}
ddwsz = offsetofend(struct vfio_iommu_spapr_tce_info, ddw);
if (info.argsz >= ddwsz)
minsz = ddwsz;
if (copy_to_user((void __user *)arg, &info, minsz))
return -EFAULT;
return 0;
}
case VFIO_IOMMU_MAP_DMA: {
struct vfio_iommu_type1_dma_map param;
struct iommu_table *tbl = NULL;
long num;
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
enum dma_data_direction direction;
if (!container->enabled)
return -EPERM;
minsz = offsetofend(struct vfio_iommu_type1_dma_map, size);
if (copy_from_user(&param, (void __user *)arg, minsz))
return -EFAULT;
if (param.argsz < minsz)
return -EINVAL;
if (param.flags & ~(VFIO_DMA_MAP_FLAG_READ |
VFIO_DMA_MAP_FLAG_WRITE))
return -EINVAL;
num = tce_iommu_find_table(container, param.iova, &tbl);
if (num < 0)
return -ENXIO;
if ((param.size & ~IOMMU_PAGE_MASK(tbl)) ||
(param.vaddr & ~IOMMU_PAGE_MASK(tbl)))
return -EINVAL;
/* iova is checked by the IOMMU API */
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
if (param.flags & VFIO_DMA_MAP_FLAG_READ) {
if (param.flags & VFIO_DMA_MAP_FLAG_WRITE)
direction = DMA_BIDIRECTIONAL;
else
direction = DMA_TO_DEVICE;
} else {
if (param.flags & VFIO_DMA_MAP_FLAG_WRITE)
direction = DMA_FROM_DEVICE;
else
return -EINVAL;
}
powerpc/iommu/powernv: Release replaced TCE At the moment writing new TCE value to the IOMMU table fails with EBUSY if there is a valid entry already. However PAPR specification allows the guest to write new TCE value without clearing it first. Another problem this patch is addressing is the use of pool locks for external IOMMU users such as VFIO. The pool locks are to protect DMA page allocator rather than entries and since the host kernel does not control what pages are in use, there is no point in pool locks and exchange()+put_page(oldtce) is sufficient to avoid possible races. This adds an exchange() callback to iommu_table_ops which does the same thing as set() plus it returns replaced TCE and DMA direction so the caller can release the pages afterwards. The exchange() receives a physical address unlike set() which receives linear mapping address; and returns a physical address as the clear() does. This implements exchange() for P5IOC2/IODA/IODA2. This adds a requirement for a platform to have exchange() implemented in order to support VFIO. This replaces iommu_tce_build() and iommu_clear_tce() with a single iommu_tce_xchg(). This makes sure that TCE permission bits are not set in TCE passed to IOMMU API as those are to be calculated by platform code from DMA direction. This moves SetPageDirty() to the IOMMU code to make it work for both VFIO ioctl interface in in-kernel TCE acceleration (when it becomes available later). Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:15 +00:00
ret = iommu_tce_put_param_check(tbl, param.iova, param.vaddr);
if (ret)
return ret;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (container->v2)
ret = tce_iommu_build_v2(container, tbl,
param.iova >> tbl->it_page_shift,
param.vaddr,
param.size >> tbl->it_page_shift,
direction);
else
ret = tce_iommu_build(container, tbl,
param.iova >> tbl->it_page_shift,
param.vaddr,
param.size >> tbl->it_page_shift,
direction);
iommu_flush_tce(tbl);
return ret;
}
case VFIO_IOMMU_UNMAP_DMA: {
struct vfio_iommu_type1_dma_unmap param;
struct iommu_table *tbl = NULL;
long num;
if (!container->enabled)
return -EPERM;
minsz = offsetofend(struct vfio_iommu_type1_dma_unmap,
size);
if (copy_from_user(&param, (void __user *)arg, minsz))
return -EFAULT;
if (param.argsz < minsz)
return -EINVAL;
/* No flag is supported now */
if (param.flags)
return -EINVAL;
num = tce_iommu_find_table(container, param.iova, &tbl);
if (num < 0)
return -ENXIO;
if (param.size & ~IOMMU_PAGE_MASK(tbl))
return -EINVAL;
ret = iommu_tce_clear_param_check(tbl, param.iova, 0,
param.size >> tbl->it_page_shift);
if (ret)
return ret;
ret = tce_iommu_clear(container, tbl,
param.iova >> tbl->it_page_shift,
param.size >> tbl->it_page_shift);
iommu_flush_tce(tbl);
return ret;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
case VFIO_IOMMU_SPAPR_REGISTER_MEMORY: {
struct vfio_iommu_spapr_register_memory param;
if (!container->v2)
break;
minsz = offsetofend(struct vfio_iommu_spapr_register_memory,
size);
if (copy_from_user(&param, (void __user *)arg, minsz))
return -EFAULT;
if (param.argsz < minsz)
return -EINVAL;
/* No flag is supported now */
if (param.flags)
return -EINVAL;
mutex_lock(&container->lock);
ret = tce_iommu_register_pages(container, param.vaddr,
param.size);
mutex_unlock(&container->lock);
return ret;
}
case VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY: {
struct vfio_iommu_spapr_register_memory param;
if (!container->v2)
break;
minsz = offsetofend(struct vfio_iommu_spapr_register_memory,
size);
if (copy_from_user(&param, (void __user *)arg, minsz))
return -EFAULT;
if (param.argsz < minsz)
return -EINVAL;
/* No flag is supported now */
if (param.flags)
return -EINVAL;
mutex_lock(&container->lock);
ret = tce_iommu_unregister_pages(container, param.vaddr,
param.size);
mutex_unlock(&container->lock);
return ret;
}
case VFIO_IOMMU_ENABLE:
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (container->v2)
break;
mutex_lock(&container->lock);
ret = tce_iommu_enable(container);
mutex_unlock(&container->lock);
return ret;
case VFIO_IOMMU_DISABLE:
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (container->v2)
break;
mutex_lock(&container->lock);
tce_iommu_disable(container);
mutex_unlock(&container->lock);
return 0;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
case VFIO_EEH_PE_OP: {
struct tce_iommu_group *tcegrp;
ret = 0;
list_for_each_entry(tcegrp, &container->group_list, next) {
ret = vfio_spapr_iommu_eeh_ioctl(tcegrp->grp,
cmd, arg);
if (ret)
return ret;
}
return ret;
}
case VFIO_IOMMU_SPAPR_TCE_CREATE: {
struct vfio_iommu_spapr_tce_create create;
if (!container->v2)
break;
if (!tce_groups_attached(container))
return -ENXIO;
minsz = offsetofend(struct vfio_iommu_spapr_tce_create,
start_addr);
if (copy_from_user(&create, (void __user *)arg, minsz))
return -EFAULT;
if (create.argsz < minsz)
return -EINVAL;
if (create.flags)
return -EINVAL;
mutex_lock(&container->lock);
ret = tce_iommu_create_window(container, create.page_shift,
create.window_size, create.levels,
&create.start_addr);
mutex_unlock(&container->lock);
if (!ret && copy_to_user((void __user *)arg, &create, minsz))
ret = -EFAULT;
return ret;
}
case VFIO_IOMMU_SPAPR_TCE_REMOVE: {
struct vfio_iommu_spapr_tce_remove remove;
if (!container->v2)
break;
if (!tce_groups_attached(container))
return -ENXIO;
minsz = offsetofend(struct vfio_iommu_spapr_tce_remove,
start_addr);
if (copy_from_user(&remove, (void __user *)arg, minsz))
return -EFAULT;
if (remove.argsz < minsz)
return -EINVAL;
if (remove.flags)
return -EINVAL;
mutex_lock(&container->lock);
ret = tce_iommu_remove_window(container, remove.start_addr);
mutex_unlock(&container->lock);
return ret;
}
}
return -ENOTTY;
}
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
static void tce_iommu_release_ownership(struct tce_container *container,
struct iommu_table_group *table_group)
{
int i;
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct iommu_table *tbl = container->tables[i];
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
if (!tbl)
continue;
tce_iommu_clear(container, tbl, tbl->it_offset, tbl->it_size);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
tce_iommu_userspace_view_free(tbl);
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
if (tbl->it_map)
iommu_release_ownership(tbl);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
container->tables[i] = NULL;
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
}
}
static int tce_iommu_take_ownership(struct tce_container *container,
struct iommu_table_group *table_group)
{
int i, j, rc = 0;
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
struct iommu_table *tbl = table_group->tables[i];
if (!tbl || !tbl->it_map)
continue;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
rc = tce_iommu_userspace_view_alloc(tbl);
if (!rc)
rc = iommu_take_ownership(tbl);
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
if (rc) {
for (j = 0; j < i; ++j)
iommu_release_ownership(
table_group->tables[j]);
return rc;
}
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i)
container->tables[i] = table_group->tables[i];
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
return 0;
}
static void tce_iommu_release_ownership_ddw(struct tce_container *container,
struct iommu_table_group *table_group)
{
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
long i;
if (!table_group->ops->unset_window) {
WARN_ON_ONCE(1);
return;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i)
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
table_group->ops->unset_window(table_group, i);
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
table_group->ops->release_ownership(table_group);
}
static long tce_iommu_take_ownership_ddw(struct tce_container *container,
struct iommu_table_group *table_group)
{
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
long i, ret = 0;
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
struct iommu_table *tbl = NULL;
if (!table_group->ops->create_table || !table_group->ops->set_window ||
!table_group->ops->release_ownership) {
WARN_ON_ONCE(1);
return -EFAULT;
}
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
table_group->ops->take_ownership(table_group);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
/*
* If it the first group attached, check if there is
* a default DMA window and create one if none as
* the userspace expects it to exist.
*/
if (!tce_groups_attached(container) && !container->tables[0]) {
ret = tce_iommu_create_table(container,
table_group,
0, /* window number */
IOMMU_PAGE_SHIFT_4K,
table_group->tce32_size,
1, /* default levels */
&tbl);
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
if (ret)
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
goto release_exit;
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
else
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
container->tables[0] = tbl;
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
/* Set all windows to the new group */
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
tbl = container->tables[i];
if (!tbl)
continue;
/* Set the default window to a new group */
ret = table_group->ops->set_window(table_group, i, tbl);
if (ret)
goto release_exit;
}
return 0;
release_exit:
for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i)
table_group->ops->unset_window(table_group, i);
table_group->ops->release_ownership(table_group);
vfio: powerpc/spapr: powerpc/powernv/ioda2: Use DMA windows API in ownership control Before the IOMMU user (VFIO) would take control over the IOMMU table belonging to a specific IOMMU group. This approach did not allow sharing tables between IOMMU groups attached to the same container. This introduces a new IOMMU ownership flavour when the user can not just control the existing IOMMU table but remove/create tables on demand. If an IOMMU implements take/release_ownership() callbacks, this lets the user have full control over the IOMMU group. When the ownership is taken, the platform code removes all the windows so the caller must create them. Before returning the ownership back to the platform code, VFIO unprograms and removes all the tables it created. This changes IODA2's onwership handler to remove the existing table rather than manipulating with the existing one. From now on, iommu_take_ownership() and iommu_release_ownership() are only called from the vfio_iommu_spapr_tce driver. Old-style ownership is still supported allowing VFIO to run on older P5IOC2 and IODA IO controllers. No change in userspace-visible behaviour is expected. Since it recreates TCE tables on each ownership change, related kernel traces will appear more often. This adds a pnv_pci_ioda2_setup_default_config() which is called when PE is being configured at boot time and when the ownership is passed from VFIO to the platform code. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:23 +00:00
return ret;
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
}
static int tce_iommu_attach_group(void *iommu_data,
struct iommu_group *iommu_group)
{
int ret;
struct tce_container *container = iommu_data;
struct iommu_table_group *table_group;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
struct tce_iommu_group *tcegrp = NULL;
mutex_lock(&container->lock);
/* pr_debug("tce_vfio: Attaching group #%u to iommu %p\n",
iommu_group_id(iommu_group), iommu_group); */
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
table_group = iommu_group_get_iommudata(iommu_group);
if (tce_groups_attached(container) && (!table_group->ops ||
!table_group->ops->take_ownership ||
!table_group->ops->release_ownership)) {
ret = -EBUSY;
goto unlock_exit;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
/* Check if new group has the same iommu_ops (i.e. compatible) */
list_for_each_entry(tcegrp, &container->group_list, next) {
struct iommu_table_group *table_group_tmp;
if (tcegrp->grp == iommu_group) {
pr_warn("tce_vfio: Group %d is already attached\n",
iommu_group_id(iommu_group));
ret = -EBUSY;
goto unlock_exit;
}
table_group_tmp = iommu_group_get_iommudata(tcegrp->grp);
if (table_group_tmp->ops->create_table !=
table_group->ops->create_table) {
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
pr_warn("tce_vfio: Group %d is incompatible with group %d\n",
iommu_group_id(iommu_group),
iommu_group_id(tcegrp->grp));
ret = -EPERM;
goto unlock_exit;
}
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
tcegrp = kzalloc(sizeof(*tcegrp), GFP_KERNEL);
if (!tcegrp) {
ret = -ENOMEM;
goto unlock_exit;
}
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
if (!table_group->ops || !table_group->ops->take_ownership ||
!table_group->ops->release_ownership)
ret = tce_iommu_take_ownership(container, table_group);
else
ret = tce_iommu_take_ownership_ddw(container, table_group);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (!ret) {
tcegrp->grp = iommu_group;
list_add(&tcegrp->next, &container->group_list);
}
unlock_exit:
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (ret && tcegrp)
kfree(tcegrp);
mutex_unlock(&container->lock);
return ret;
}
static void tce_iommu_detach_group(void *iommu_data,
struct iommu_group *iommu_group)
{
struct tce_container *container = iommu_data;
struct iommu_table_group *table_group;
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
bool found = false;
struct tce_iommu_group *tcegrp;
mutex_lock(&container->lock);
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
list_for_each_entry(tcegrp, &container->group_list, next) {
if (tcegrp->grp == iommu_group) {
found = true;
break;
}
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
if (!found) {
pr_warn("tce_vfio: detaching unattached group #%u\n",
iommu_group_id(iommu_group));
goto unlock_exit;
}
vfio: powerpc/spapr: Register memory and define IOMMU v2 The existing implementation accounts the whole DMA window in the locked_vm counter. This is going to be worse with multiple containers and huge DMA windows. Also, real-time accounting would requite additional tracking of accounted pages due to the page size difference - IOMMU uses 4K pages and system uses 4K or 64K pages. Another issue is that actual pages pinning/unpinning happens on every DMA map/unmap request. This does not affect the performance much now as we spend way too much time now on switching context between guest/userspace/host but this will start to matter when we add in-kernel DMA map/unmap acceleration. This introduces a new IOMMU type for SPAPR - VFIO_SPAPR_TCE_v2_IOMMU. New IOMMU deprecates VFIO_IOMMU_ENABLE/VFIO_IOMMU_DISABLE and introduces 2 new ioctls to register/unregister DMA memory - VFIO_IOMMU_SPAPR_REGISTER_MEMORY and VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - which receive user space address and size of a memory region which needs to be pinned/unpinned and counted in locked_vm. New IOMMU splits physical pages pinning and TCE table update into 2 different operations. It requires: 1) guest pages to be registered first 2) consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. When a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. The new memory registration ioctls are not supported by VFIO_SPAPR_TCE_IOMMU. Dynamic DMA window and in-kernel acceleration will require memory to be preregistered in order to work. The accounting is done per the user process. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. In order to support memory pre-registration, we need a way to track the use of every registered memory region and only allow unregistration if a region is not in use anymore. So we need a way to tell from what region the just cleared TCE was from. This adds a userspace view of the TCE table into iommu_table struct. It contains userspace address, one per TCE entry. The table is only allocated when the ownership over an IOMMU group is taken which means it is only used from outside of the powernv code (such as VFIO). As v2 IOMMU supports IODA2 and pre-IODA2 IOMMUs (which do not support DDW API), this creates a default DMA window for IODA2 for consistency. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:25 +00:00
list_del(&tcegrp->next);
kfree(tcegrp);
table_group = iommu_group_get_iommudata(iommu_group);
BUG_ON(!table_group);
vfio: powerpc/spapr/iommu/powernv/ioda2: Rework IOMMU ownership control This adds tce_iommu_take_ownership() and tce_iommu_release_ownership which call in a loop iommu_take_ownership()/iommu_release_ownership() for every table on the group. As there is just one now, no change in behaviour is expected. At the moment the iommu_table struct has a set_bypass() which enables/ disables DMA bypass on IODA2 PHB. This is exposed to POWERPC IOMMU code which calls this callback when external IOMMU users such as VFIO are about to get over a PHB. The set_bypass() callback is not really an iommu_table function but IOMMU/PE function. This introduces a iommu_table_group_ops struct and adds take_ownership()/release_ownership() callbacks to it which are called when an external user takes/releases control over the IOMMU. This replaces set_bypass() with ownership callbacks as it is not necessarily just bypass enabling, it can be something else/more so let's give it more generic name. The callbacks is implemented for IODA2 only. Other platforms (P5IOC2, IODA1) will use the old iommu_take_ownership/iommu_release_ownership API. The following patches will replace iommu_take_ownership/ iommu_release_ownership calls in IODA2 with full IOMMU table release/ create. As we here and touching bypass control, this removes pnv_pci_ioda2_setup_bypass_pe() as it does not do much more compared to pnv_pci_ioda2_set_bypass. This moves tce_bypass_base initialization to pnv_pci_ioda2_setup_dma_pe. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> [aw: for the vfio related changes] Acked-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-06-05 06:35:10 +00:00
if (!table_group->ops || !table_group->ops->release_ownership)
tce_iommu_release_ownership(container, table_group);
else
tce_iommu_release_ownership_ddw(container, table_group);
unlock_exit:
mutex_unlock(&container->lock);
}
const struct vfio_iommu_driver_ops tce_iommu_driver_ops = {
.name = "iommu-vfio-powerpc",
.owner = THIS_MODULE,
.open = tce_iommu_open,
.release = tce_iommu_release,
.ioctl = tce_iommu_ioctl,
.attach_group = tce_iommu_attach_group,
.detach_group = tce_iommu_detach_group,
};
static int __init tce_iommu_init(void)
{
return vfio_register_iommu_driver(&tce_iommu_driver_ops);
}
static void __exit tce_iommu_cleanup(void)
{
vfio_unregister_iommu_driver(&tce_iommu_driver_ops);
}
module_init(tce_iommu_init);
module_exit(tce_iommu_cleanup);
MODULE_VERSION(DRIVER_VERSION);
MODULE_LICENSE("GPL v2");
MODULE_AUTHOR(DRIVER_AUTHOR);
MODULE_DESCRIPTION(DRIVER_DESC);