1 // SPDX-License-Identifier: GPL-2.0
2 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 
4 #include "mmu_internal.h"
5 #include "tdp_iter.h"
6 #include "spte.h"
7 
8 /*
9  * Recalculates the pointer to the SPTE for the current GFN and level and
10  * reread the SPTE.
11  */
tdp_iter_refresh_sptep(struct tdp_iter * iter)12 static void tdp_iter_refresh_sptep(struct tdp_iter *iter)
13 {
14 	iter->sptep = iter->pt_path[iter->level - 1] +
15 		SPTE_INDEX((iter->gfn | iter->gfn_bits) << PAGE_SHIFT, iter->level);
16 	iter->old_spte = kvm_tdp_mmu_read_spte(iter->sptep);
17 }
18 
19 /*
20  * Return the TDP iterator to the root PT and allow it to continue its
21  * traversal over the paging structure from there.
22  */
tdp_iter_restart(struct tdp_iter * iter)23 void tdp_iter_restart(struct tdp_iter *iter)
24 {
25 	iter->yielded = false;
26 	iter->yielded_gfn = iter->next_last_level_gfn;
27 	iter->level = iter->root_level;
28 
29 	iter->gfn = gfn_round_for_level(iter->next_last_level_gfn, iter->level);
30 	tdp_iter_refresh_sptep(iter);
31 
32 	iter->valid = true;
33 }
34 
35 /*
36  * Sets a TDP iterator to walk a pre-order traversal of the paging structure
37  * rooted at root_pt, starting with the walk to translate next_last_level_gfn.
38  */
tdp_iter_start(struct tdp_iter * iter,struct kvm_mmu_page * root,int min_level,gfn_t next_last_level_gfn,gfn_t gfn_bits)39 void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
40 		    int min_level, gfn_t next_last_level_gfn, gfn_t gfn_bits)
41 {
42 	if (WARN_ON_ONCE(!root || (root->role.level < 1) ||
43 			 (root->role.level > PT64_ROOT_MAX_LEVEL) ||
44 			 (gfn_bits && next_last_level_gfn >= gfn_bits))) {
45 		iter->valid = false;
46 		return;
47 	}
48 
49 	iter->next_last_level_gfn = next_last_level_gfn;
50 	iter->gfn_bits = gfn_bits;
51 	iter->root_level = root->role.level;
52 	iter->min_level = min_level;
53 	iter->pt_path[iter->root_level - 1] = (tdp_ptep_t)root->spt;
54 	iter->as_id = kvm_mmu_page_as_id(root);
55 
56 	tdp_iter_restart(iter);
57 }
58 
59 /*
60  * Given an SPTE and its level, returns a pointer containing the host virtual
61  * address of the child page table referenced by the SPTE. Returns null if
62  * there is no such entry.
63  */
spte_to_child_pt(u64 spte,int level)64 tdp_ptep_t spte_to_child_pt(u64 spte, int level)
65 {
66 	/*
67 	 * There's no child entry if this entry isn't present or is a
68 	 * last-level entry.
69 	 */
70 	if (!is_shadow_present_pte(spte) || is_last_spte(spte, level))
71 		return NULL;
72 
73 	return (tdp_ptep_t)__va(spte_to_pfn(spte) << PAGE_SHIFT);
74 }
75 
76 /*
77  * Steps down one level in the paging structure towards the goal GFN. Returns
78  * true if the iterator was able to step down a level, false otherwise.
79  */
try_step_down(struct tdp_iter * iter)80 static bool try_step_down(struct tdp_iter *iter)
81 {
82 	tdp_ptep_t child_pt;
83 
84 	if (iter->level == iter->min_level)
85 		return false;
86 
87 	/*
88 	 * Reread the SPTE before stepping down to avoid traversing into page
89 	 * tables that are no longer linked from this entry.
90 	 */
91 	iter->old_spte = kvm_tdp_mmu_read_spte(iter->sptep);
92 
93 	child_pt = spte_to_child_pt(iter->old_spte, iter->level);
94 	if (!child_pt)
95 		return false;
96 
97 	iter->level--;
98 	iter->pt_path[iter->level - 1] = child_pt;
99 	iter->gfn = gfn_round_for_level(iter->next_last_level_gfn, iter->level);
100 	tdp_iter_refresh_sptep(iter);
101 
102 	return true;
103 }
104 
105 /*
106  * Steps to the next entry in the current page table, at the current page table
107  * level. The next entry could point to a page backing guest memory or another
108  * page table, or it could be non-present. Returns true if the iterator was
109  * able to step to the next entry in the page table, false if the iterator was
110  * already at the end of the current page table.
111  */
try_step_side(struct tdp_iter * iter)112 static bool try_step_side(struct tdp_iter *iter)
113 {
114 	/*
115 	 * Check if the iterator is already at the end of the current page
116 	 * table.
117 	 */
118 	if (SPTE_INDEX((iter->gfn | iter->gfn_bits) << PAGE_SHIFT, iter->level) ==
119 	    (SPTE_ENT_PER_PAGE - 1))
120 		return false;
121 
122 	iter->gfn += KVM_PAGES_PER_HPAGE(iter->level);
123 	iter->next_last_level_gfn = iter->gfn;
124 	iter->sptep++;
125 	iter->old_spte = kvm_tdp_mmu_read_spte(iter->sptep);
126 
127 	return true;
128 }
129 
130 /*
131  * Tries to traverse back up a level in the paging structure so that the walk
132  * can continue from the next entry in the parent page table. Returns true on a
133  * successful step up, false if already in the root page.
134  */
try_step_up(struct tdp_iter * iter)135 static bool try_step_up(struct tdp_iter *iter)
136 {
137 	if (iter->level == iter->root_level)
138 		return false;
139 
140 	iter->level++;
141 	iter->gfn = gfn_round_for_level(iter->gfn, iter->level);
142 	tdp_iter_refresh_sptep(iter);
143 
144 	return true;
145 }
146 
147 /*
148  * Step to the next SPTE in a pre-order traversal of the paging structure.
149  * To get to the next SPTE, the iterator either steps down towards the goal
150  * GFN, if at a present, non-last-level SPTE, or over to a SPTE mapping a
151  * higher GFN.
152  *
153  * The basic algorithm is as follows:
154  * 1. If the current SPTE is a non-last-level SPTE, step down into the page
155  *    table it points to.
156  * 2. If the iterator cannot step down, it will try to step to the next SPTE
157  *    in the current page of the paging structure.
158  * 3. If the iterator cannot step to the next entry in the current page, it will
159  *    try to step up to the parent paging structure page. In this case, that
160  *    SPTE will have already been visited, and so the iterator must also step
161  *    to the side again.
162  */
tdp_iter_next(struct tdp_iter * iter)163 void tdp_iter_next(struct tdp_iter *iter)
164 {
165 	if (iter->yielded) {
166 		tdp_iter_restart(iter);
167 		return;
168 	}
169 
170 	if (try_step_down(iter))
171 		return;
172 
173 	do {
174 		if (try_step_side(iter))
175 			return;
176 	} while (try_step_up(iter));
177 	iter->valid = false;
178 }
179 
180