1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * KVM selftest s390x library code - CPU-related functions (page tables...)
4  *
5  * Copyright (C) 2019, Red Hat, Inc.
6  */
7 
8 #define _GNU_SOURCE /* for program_invocation_name */
9 
10 #include "processor.h"
11 #include "kvm_util.h"
12 #include "../kvm_util_internal.h"
13 
14 #define KVM_GUEST_PAGE_TABLE_MIN_PADDR		0x180000
15 
16 #define PAGES_PER_REGION 4
17 
18 void virt_pgd_alloc(struct kvm_vm *vm, uint32_t memslot)
19 {
20 	vm_paddr_t paddr;
21 
22 	TEST_ASSERT(vm->page_size == 4096, "Unsupported page size: 0x%x",
23 		    vm->page_size);
24 
25 	if (vm->pgd_created)
26 		return;
27 
28 	paddr = vm_phy_pages_alloc(vm, PAGES_PER_REGION,
29 				   KVM_GUEST_PAGE_TABLE_MIN_PADDR, memslot);
30 	memset(addr_gpa2hva(vm, paddr), 0xff, PAGES_PER_REGION * vm->page_size);
31 
32 	vm->pgd = paddr;
33 	vm->pgd_created = true;
34 }
35 
36 /*
37  * Allocate 4 pages for a region/segment table (ri < 4), or one page for
38  * a page table (ri == 4). Returns a suitable region/segment table entry
39  * which points to the freshly allocated pages.
40  */
41 static uint64_t virt_alloc_region(struct kvm_vm *vm, int ri, uint32_t memslot)
42 {
43 	uint64_t taddr;
44 
45 	taddr = vm_phy_pages_alloc(vm,  ri < 4 ? PAGES_PER_REGION : 1,
46 				   KVM_GUEST_PAGE_TABLE_MIN_PADDR, memslot);
47 	memset(addr_gpa2hva(vm, taddr), 0xff, PAGES_PER_REGION * vm->page_size);
48 
49 	return (taddr & REGION_ENTRY_ORIGIN)
50 		| (((4 - ri) << 2) & REGION_ENTRY_TYPE)
51 		| ((ri < 4 ? (PAGES_PER_REGION - 1) : 0) & REGION_ENTRY_LENGTH);
52 }
53 
54 /*
55  * VM Virtual Page Map
56  *
57  * Input Args:
58  *   vm - Virtual Machine
59  *   gva - VM Virtual Address
60  *   gpa - VM Physical Address
61  *   memslot - Memory region slot for new virtual translation tables
62  *
63  * Output Args: None
64  *
65  * Return: None
66  *
67  * Within the VM given by vm, creates a virtual translation for the page
68  * starting at vaddr to the page starting at paddr.
69  */
70 void virt_pg_map(struct kvm_vm *vm, uint64_t gva, uint64_t gpa,
71 		 uint32_t memslot)
72 {
73 	int ri, idx;
74 	uint64_t *entry;
75 
76 	TEST_ASSERT((gva % vm->page_size) == 0,
77 		"Virtual address not on page boundary,\n"
78 		"  vaddr: 0x%lx vm->page_size: 0x%x",
79 		gva, vm->page_size);
80 	TEST_ASSERT(sparsebit_is_set(vm->vpages_valid,
81 		(gva >> vm->page_shift)),
82 		"Invalid virtual address, vaddr: 0x%lx",
83 		gva);
84 	TEST_ASSERT((gpa % vm->page_size) == 0,
85 		"Physical address not on page boundary,\n"
86 		"  paddr: 0x%lx vm->page_size: 0x%x",
87 		gva, vm->page_size);
88 	TEST_ASSERT((gpa >> vm->page_shift) <= vm->max_gfn,
89 		"Physical address beyond beyond maximum supported,\n"
90 		"  paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
91 		gva, vm->max_gfn, vm->page_size);
92 
93 	/* Walk through region and segment tables */
94 	entry = addr_gpa2hva(vm, vm->pgd);
95 	for (ri = 1; ri <= 4; ri++) {
96 		idx = (gva >> (64 - 11 * ri)) & 0x7ffu;
97 		if (entry[idx] & REGION_ENTRY_INVALID)
98 			entry[idx] = virt_alloc_region(vm, ri, memslot);
99 		entry = addr_gpa2hva(vm, entry[idx] & REGION_ENTRY_ORIGIN);
100 	}
101 
102 	/* Fill in page table entry */
103 	idx = (gva >> 12) & 0x0ffu;		/* page index */
104 	if (!(entry[idx] & PAGE_INVALID))
105 		fprintf(stderr,
106 			"WARNING: PTE for gpa=0x%"PRIx64" already set!\n", gpa);
107 	entry[idx] = gpa;
108 }
109 
110 /*
111  * Address Guest Virtual to Guest Physical
112  *
113  * Input Args:
114  *   vm - Virtual Machine
115  *   gpa - VM virtual address
116  *
117  * Output Args: None
118  *
119  * Return:
120  *   Equivalent VM physical address
121  *
122  * Translates the VM virtual address given by gva to a VM physical
123  * address and then locates the memory region containing the VM
124  * physical address, within the VM given by vm.  When found, the host
125  * virtual address providing the memory to the vm physical address is
126  * returned.
127  * A TEST_ASSERT failure occurs if no region containing translated
128  * VM virtual address exists.
129  */
130 vm_paddr_t addr_gva2gpa(struct kvm_vm *vm, vm_vaddr_t gva)
131 {
132 	int ri, idx;
133 	uint64_t *entry;
134 
135 	TEST_ASSERT(vm->page_size == 4096, "Unsupported page size: 0x%x",
136 		    vm->page_size);
137 
138 	entry = addr_gpa2hva(vm, vm->pgd);
139 	for (ri = 1; ri <= 4; ri++) {
140 		idx = (gva >> (64 - 11 * ri)) & 0x7ffu;
141 		TEST_ASSERT(!(entry[idx] & REGION_ENTRY_INVALID),
142 			    "No region mapping for vm virtual address 0x%lx",
143 			    gva);
144 		entry = addr_gpa2hva(vm, entry[idx] & REGION_ENTRY_ORIGIN);
145 	}
146 
147 	idx = (gva >> 12) & 0x0ffu;		/* page index */
148 
149 	TEST_ASSERT(!(entry[idx] & PAGE_INVALID),
150 		    "No page mapping for vm virtual address 0x%lx", gva);
151 
152 	return (entry[idx] & ~0xffful) + (gva & 0xffful);
153 }
154 
155 static void virt_dump_ptes(FILE *stream, struct kvm_vm *vm, uint8_t indent,
156 			   uint64_t ptea_start)
157 {
158 	uint64_t *pte, ptea;
159 
160 	for (ptea = ptea_start; ptea < ptea_start + 0x100 * 8; ptea += 8) {
161 		pte = addr_gpa2hva(vm, ptea);
162 		if (*pte & PAGE_INVALID)
163 			continue;
164 		fprintf(stream, "%*spte @ 0x%lx: 0x%016lx\n",
165 			indent, "", ptea, *pte);
166 	}
167 }
168 
169 static void virt_dump_region(FILE *stream, struct kvm_vm *vm, uint8_t indent,
170 			     uint64_t reg_tab_addr)
171 {
172 	uint64_t addr, *entry;
173 
174 	for (addr = reg_tab_addr; addr < reg_tab_addr + 0x400 * 8; addr += 8) {
175 		entry = addr_gpa2hva(vm, addr);
176 		if (*entry & REGION_ENTRY_INVALID)
177 			continue;
178 		fprintf(stream, "%*srt%lde @ 0x%lx: 0x%016lx\n",
179 			indent, "", 4 - ((*entry & REGION_ENTRY_TYPE) >> 2),
180 			addr, *entry);
181 		if (*entry & REGION_ENTRY_TYPE) {
182 			virt_dump_region(stream, vm, indent + 2,
183 					 *entry & REGION_ENTRY_ORIGIN);
184 		} else {
185 			virt_dump_ptes(stream, vm, indent + 2,
186 				       *entry & REGION_ENTRY_ORIGIN);
187 		}
188 	}
189 }
190 
191 void virt_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
192 {
193 	if (!vm->pgd_created)
194 		return;
195 
196 	virt_dump_region(stream, vm, indent, vm->pgd);
197 }
198 
199 /*
200  * Create a VM with reasonable defaults
201  *
202  * Input Args:
203  *   vcpuid - The id of the single VCPU to add to the VM.
204  *   extra_mem_pages - The size of extra memories to add (this will
205  *                     decide how much extra space we will need to
206  *                     setup the page tables using mem slot 0)
207  *   guest_code - The vCPU's entry point
208  *
209  * Output Args: None
210  *
211  * Return:
212  *   Pointer to opaque structure that describes the created VM.
213  */
214 struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_pages,
215 				 void *guest_code)
216 {
217 	/*
218 	 * The additional amount of pages required for the page tables is:
219 	 * 1 * n / 256 + 4 * (n / 256) / 2048 + 4 * (n / 256) / 2048^2 + ...
220 	 * which is definitely smaller than (n / 256) * 2.
221 	 */
222 	uint64_t extra_pg_pages = extra_mem_pages / 256 * 2;
223 	struct kvm_vm *vm;
224 
225 	vm = vm_create(VM_MODE_DEFAULT,
226 		       DEFAULT_GUEST_PHY_PAGES + extra_pg_pages, O_RDWR);
227 
228 	kvm_vm_elf_load(vm, program_invocation_name, 0, 0);
229 	vm_vcpu_add_default(vm, vcpuid, guest_code);
230 
231 	return vm;
232 }
233 
234 /*
235  * Adds a vCPU with reasonable defaults (i.e. a stack and initial PSW)
236  *
237  * Input Args:
238  *   vcpuid - The id of the VCPU to add to the VM.
239  *   guest_code - The vCPU's entry point
240  */
241 void vm_vcpu_add_default(struct kvm_vm *vm, uint32_t vcpuid, void *guest_code)
242 {
243 	size_t stack_size =  DEFAULT_STACK_PGS * getpagesize();
244 	uint64_t stack_vaddr;
245 	struct kvm_regs regs;
246 	struct kvm_sregs sregs;
247 	struct kvm_run *run;
248 
249 	TEST_ASSERT(vm->page_size == 4096, "Unsupported page size: 0x%x",
250 		    vm->page_size);
251 
252 	stack_vaddr = vm_vaddr_alloc(vm, stack_size,
253 				     DEFAULT_GUEST_STACK_VADDR_MIN, 0, 0);
254 
255 	vm_vcpu_add(vm, vcpuid);
256 
257 	/* Setup guest registers */
258 	vcpu_regs_get(vm, vcpuid, &regs);
259 	regs.gprs[15] = stack_vaddr + (DEFAULT_STACK_PGS * getpagesize()) - 160;
260 	vcpu_regs_set(vm, vcpuid, &regs);
261 
262 	vcpu_sregs_get(vm, vcpuid, &sregs);
263 	sregs.crs[0] |= 0x00040000;		/* Enable floating point regs */
264 	sregs.crs[1] = vm->pgd | 0xf;		/* Primary region table */
265 	vcpu_sregs_set(vm, vcpuid, &sregs);
266 
267 	run = vcpu_state(vm, vcpuid);
268 	run->psw_mask = 0x0400000180000000ULL;  /* DAT enabled + 64 bit mode */
269 	run->psw_addr = (uintptr_t)guest_code;
270 }
271 
272 void vcpu_dump(FILE *stream, struct kvm_vm *vm, uint32_t vcpuid, uint8_t indent)
273 {
274 	struct vcpu *vcpu = vm->vcpu_head;
275 
276 	fprintf(stream, "%*spstate: psw: 0x%.16llx:0x%.16llx\n",
277 		indent, "", vcpu->state->psw_mask, vcpu->state->psw_addr);
278 }
279