1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
|
From: Andrew Cooper <andrew.cooper3@citrix.com>
Subject: AMD/IOMMU: Cease using a dynamic height for the IOMMU pagetables
update_paging_mode() has multiple bugs:
1) Booting with iommu=debug will cause it to inform you that that it called
without the pdev_list lock held.
2) When growing by more than a single level, it leaks the newly allocated
table(s) in the case of a further error.
Furthermore, the choice of default level for a domain has issues:
1) All HVM guests grow from 2 to 3 levels during construction because of the
position of the VRAM just below the 4G boundary, so defaulting to 2 is a
waste of effort.
2) The limit for PV guests doesn't take memory hotplug into account, and
isn't dynamic at runtime like HVM guests. This means that a PV guest may
get RAM which it can't map in the IOMMU.
The dynamic height is a property unique to AMD, and adds a substantial
quantity of complexity for what is a marginal performance improvement. Remove
the complexity by removing the dynamic height.
PV guests now get 3 or 4 levels based on any hotplug regions in the host.
This only makes a difference for hardware which previously had all RAM below
the 512G boundary, and a hotplug region above.
HVM guests now get 4 levels (which will be sufficient until 256TB guests
become a thing), because we don't currently have the information to know when
3 would be safe to use.
The overhead of this extra level is not expected to be noticeable. It costs
one page (4k) per domain, and one extra IO-TLB paging structure cache entry
which is very hot and less likely to be evicted.
This is XSA-311.
Reported-by: XXX PERSON <XXX EMAIL>3
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
--- a/xen/drivers/passthrough/amd/iommu_map.c
+++ b/xen/drivers/passthrough/amd/iommu_map.c
@@ -569,97 +569,6 @@ static int iommu_pde_from_gfn(struct dom
return 0;
}
-static int update_paging_mode(struct domain *d, unsigned long gfn)
-{
- u16 bdf;
- void *device_entry;
- unsigned int req_id, level, offset;
- unsigned long flags;
- struct pci_dev *pdev;
- struct amd_iommu *iommu = NULL;
- struct page_info *new_root = NULL;
- struct page_info *old_root = NULL;
- void *new_root_vaddr;
- unsigned long old_root_mfn;
- struct domain_iommu *hd = dom_iommu(d);
-
- if ( gfn == gfn_x(INVALID_GFN) )
- return -EADDRNOTAVAIL;
- ASSERT(!(gfn >> DEFAULT_DOMAIN_ADDRESS_WIDTH));
-
- level = hd->arch.paging_mode;
- old_root = hd->arch.root_table;
- offset = gfn >> (PTE_PER_TABLE_SHIFT * (level - 1));
-
- ASSERT(spin_is_locked(&hd->arch.mapping_lock) && is_hvm_domain(d));
-
- while ( offset >= PTE_PER_TABLE_SIZE )
- {
- /* Allocate and install a new root table.
- * Only upper I/O page table grows, no need to fix next level bits */
- new_root = alloc_amd_iommu_pgtable();
- if ( new_root == NULL )
- {
- AMD_IOMMU_DEBUG("%s Cannot allocate I/O page table\n",
- __func__);
- return -ENOMEM;
- }
-
- new_root_vaddr = __map_domain_page(new_root);
- old_root_mfn = page_to_mfn(old_root);
- set_iommu_pde_present(new_root_vaddr, old_root_mfn, level,
- !!IOMMUF_writable, !!IOMMUF_readable);
- level++;
- old_root = new_root;
- offset >>= PTE_PER_TABLE_SHIFT;
- unmap_domain_page(new_root_vaddr);
- }
-
- if ( new_root != NULL )
- {
- hd->arch.paging_mode = level;
- hd->arch.root_table = new_root;
-
- if ( !pcidevs_locked() )
- AMD_IOMMU_DEBUG("%s Try to access pdev_list "
- "without aquiring pcidevs_lock.\n", __func__);
-
- /* Update device table entries using new root table and paging mode */
- for_each_pdev( d, pdev )
- {
- bdf = PCI_BDF2(pdev->bus, pdev->devfn);
- iommu = find_iommu_for_device(pdev->seg, bdf);
- if ( !iommu )
- {
- AMD_IOMMU_DEBUG("%s Fail to find iommu.\n", __func__);
- return -ENODEV;
- }
-
- spin_lock_irqsave(&iommu->lock, flags);
- do {
- req_id = get_dma_requestor_id(pdev->seg, bdf);
- device_entry = iommu->dev_table.buffer +
- (req_id * IOMMU_DEV_TABLE_ENTRY_SIZE);
-
- /* valid = 0 only works for dom0 passthrough mode */
- amd_iommu_set_root_page_table((u32 *)device_entry,
- page_to_maddr(hd->arch.root_table),
- d->domain_id,
- hd->arch.paging_mode, 1);
-
- amd_iommu_flush_device(iommu, req_id);
- bdf += pdev->phantom_stride;
- } while ( PCI_DEVFN2(bdf) != pdev->devfn &&
- PCI_SLOT(bdf) == PCI_SLOT(pdev->devfn) );
- spin_unlock_irqrestore(&iommu->lock, flags);
- }
-
- /* For safety, invalidate all entries */
- amd_iommu_flush_all_pages(d);
- }
- return 0;
-}
-
int amd_iommu_map_page(struct domain *d, unsigned long gfn, unsigned long mfn,
unsigned int flags)
{
@@ -685,19 +594,6 @@ int amd_iommu_map_page(struct domain *d,
return rc;
}
- /* Since HVM domain is initialized with 2 level IO page table,
- * we might need a deeper page table for lager gfn now */
- if ( is_hvm_domain(d) )
- {
- if ( update_paging_mode(d, gfn) )
- {
- spin_unlock(&hd->arch.mapping_lock);
- AMD_IOMMU_DEBUG("Update page mode failed gfn = %lx\n", gfn);
- domain_crash(d);
- return -EFAULT;
- }
- }
-
if ( iommu_pde_from_gfn(d, gfn, pt_mfn, true) || (pt_mfn[1] == 0) )
{
spin_unlock(&hd->arch.mapping_lock);
--- a/xen/drivers/passthrough/amd/pci_amd_iommu.c
+++ b/xen/drivers/passthrough/amd/pci_amd_iommu.c
@@ -242,11 +242,17 @@ static int amd_iommu_domain_init(struct
{
struct domain_iommu *hd = dom_iommu(d);
- /* For pv and dom0, stick with get_paging_mode(max_page)
- * For HVM dom0, use 2 level page table at first */
- hd->arch.paging_mode = is_hvm_domain(d) ?
- IOMMU_PAGING_MODE_LEVEL_2 :
- get_paging_mode(max_page);
+ /*
+ * Choose the number of levels for the IOMMU page tables.
+ * - PV needs 3 or 4, depending on whether there is RAM (including hotplug
+ * RAM) above the 512G boundary.
+ * - HVM could in principle use 3 or 4 depending on how much guest
+ * physical address space we give it, but this isn't known yet so use 4
+ * unilaterally.
+ */
+ hd->arch.paging_mode = is_hvm_domain(d)
+ ? IOMMU_PAGING_MODE_LEVEL_4 : get_paging_mode(get_upper_mfn_bound());
+
return 0;
}
|
