1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (C) 2021 Google LLC
4 * Author: Fuad Tabba <tabba@google.com>
5 */
6
7 #include <linux/irqchip/arm-gic-v3.h>
8
9 #include <asm/kvm_asm.h>
10 #include <asm/kvm_mmu.h>
11
12 #include <hyp/adjust_pc.h>
13
14 #include <nvhe/fixed_config.h>
15
16 #include "../../sys_regs.h"
17
18 /*
19 * Copies of the host's CPU features registers holding sanitized values at hyp.
20 */
21 u64 id_aa64pfr0_el1_sys_val;
22 u64 id_aa64pfr1_el1_sys_val;
23 u64 id_aa64isar0_el1_sys_val;
24 u64 id_aa64isar1_el1_sys_val;
25 u64 id_aa64isar2_el1_sys_val;
26 u64 id_aa64mmfr0_el1_sys_val;
27 u64 id_aa64mmfr1_el1_sys_val;
28 u64 id_aa64mmfr2_el1_sys_val;
29 u64 id_aa64smfr0_el1_sys_val;
30
31 /*
32 * Inject an unknown/undefined exception to an AArch64 guest while most of its
33 * sysregs are live.
34 */
inject_undef64(struct kvm_vcpu * vcpu)35 static void inject_undef64(struct kvm_vcpu *vcpu)
36 {
37 u64 esr = (ESR_ELx_EC_UNKNOWN << ESR_ELx_EC_SHIFT);
38
39 *vcpu_pc(vcpu) = read_sysreg_el2(SYS_ELR);
40 *vcpu_cpsr(vcpu) = read_sysreg_el2(SYS_SPSR);
41
42 kvm_pend_exception(vcpu, EXCEPT_AA64_EL1_SYNC);
43
44 __kvm_adjust_pc(vcpu);
45
46 write_sysreg_el1(esr, SYS_ESR);
47 write_sysreg_el1(read_sysreg_el2(SYS_ELR), SYS_ELR);
48 write_sysreg_el2(*vcpu_pc(vcpu), SYS_ELR);
49 write_sysreg_el2(*vcpu_cpsr(vcpu), SYS_SPSR);
50 }
51
52 /*
53 * Returns the restricted features values of the feature register based on the
54 * limitations in restrict_fields.
55 * A feature id field value of 0b0000 does not impose any restrictions.
56 * Note: Use only for unsigned feature field values.
57 */
get_restricted_features_unsigned(u64 sys_reg_val,u64 restrict_fields)58 static u64 get_restricted_features_unsigned(u64 sys_reg_val,
59 u64 restrict_fields)
60 {
61 u64 value = 0UL;
62 u64 mask = GENMASK_ULL(ARM64_FEATURE_FIELD_BITS - 1, 0);
63
64 /*
65 * According to the Arm Architecture Reference Manual, feature fields
66 * use increasing values to indicate increases in functionality.
67 * Iterate over the restricted feature fields and calculate the minimum
68 * unsigned value between the one supported by the system, and what the
69 * value is being restricted to.
70 */
71 while (sys_reg_val && restrict_fields) {
72 value |= min(sys_reg_val & mask, restrict_fields & mask);
73 sys_reg_val &= ~mask;
74 restrict_fields &= ~mask;
75 mask <<= ARM64_FEATURE_FIELD_BITS;
76 }
77
78 return value;
79 }
80
81 /*
82 * Functions that return the value of feature id registers for protected VMs
83 * based on allowed features, system features, and KVM support.
84 */
85
get_pvm_id_aa64pfr0(const struct kvm_vcpu * vcpu)86 static u64 get_pvm_id_aa64pfr0(const struct kvm_vcpu *vcpu)
87 {
88 u64 set_mask = 0;
89 u64 allow_mask = PVM_ID_AA64PFR0_ALLOW;
90
91 set_mask |= get_restricted_features_unsigned(id_aa64pfr0_el1_sys_val,
92 PVM_ID_AA64PFR0_RESTRICT_UNSIGNED);
93
94 return (id_aa64pfr0_el1_sys_val & allow_mask) | set_mask;
95 }
96
get_pvm_id_aa64pfr1(const struct kvm_vcpu * vcpu)97 static u64 get_pvm_id_aa64pfr1(const struct kvm_vcpu *vcpu)
98 {
99 const struct kvm *kvm = (const struct kvm *)kern_hyp_va(vcpu->kvm);
100 u64 allow_mask = PVM_ID_AA64PFR1_ALLOW;
101
102 if (!kvm_has_mte(kvm))
103 allow_mask &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_MTE);
104
105 return id_aa64pfr1_el1_sys_val & allow_mask;
106 }
107
get_pvm_id_aa64zfr0(const struct kvm_vcpu * vcpu)108 static u64 get_pvm_id_aa64zfr0(const struct kvm_vcpu *vcpu)
109 {
110 /*
111 * No support for Scalable Vectors, therefore, hyp has no sanitized
112 * copy of the feature id register.
113 */
114 BUILD_BUG_ON(PVM_ID_AA64ZFR0_ALLOW != 0ULL);
115 return 0;
116 }
117
get_pvm_id_aa64dfr0(const struct kvm_vcpu * vcpu)118 static u64 get_pvm_id_aa64dfr0(const struct kvm_vcpu *vcpu)
119 {
120 /*
121 * No support for debug, including breakpoints, and watchpoints,
122 * therefore, pKVM has no sanitized copy of the feature id register.
123 */
124 BUILD_BUG_ON(PVM_ID_AA64DFR0_ALLOW != 0ULL);
125 return 0;
126 }
127
get_pvm_id_aa64dfr1(const struct kvm_vcpu * vcpu)128 static u64 get_pvm_id_aa64dfr1(const struct kvm_vcpu *vcpu)
129 {
130 /*
131 * No support for debug, therefore, hyp has no sanitized copy of the
132 * feature id register.
133 */
134 BUILD_BUG_ON(PVM_ID_AA64DFR1_ALLOW != 0ULL);
135 return 0;
136 }
137
get_pvm_id_aa64afr0(const struct kvm_vcpu * vcpu)138 static u64 get_pvm_id_aa64afr0(const struct kvm_vcpu *vcpu)
139 {
140 /*
141 * No support for implementation defined features, therefore, hyp has no
142 * sanitized copy of the feature id register.
143 */
144 BUILD_BUG_ON(PVM_ID_AA64AFR0_ALLOW != 0ULL);
145 return 0;
146 }
147
get_pvm_id_aa64afr1(const struct kvm_vcpu * vcpu)148 static u64 get_pvm_id_aa64afr1(const struct kvm_vcpu *vcpu)
149 {
150 /*
151 * No support for implementation defined features, therefore, hyp has no
152 * sanitized copy of the feature id register.
153 */
154 BUILD_BUG_ON(PVM_ID_AA64AFR1_ALLOW != 0ULL);
155 return 0;
156 }
157
get_pvm_id_aa64isar0(const struct kvm_vcpu * vcpu)158 static u64 get_pvm_id_aa64isar0(const struct kvm_vcpu *vcpu)
159 {
160 return id_aa64isar0_el1_sys_val & PVM_ID_AA64ISAR0_ALLOW;
161 }
162
get_pvm_id_aa64isar1(const struct kvm_vcpu * vcpu)163 static u64 get_pvm_id_aa64isar1(const struct kvm_vcpu *vcpu)
164 {
165 u64 allow_mask = PVM_ID_AA64ISAR1_ALLOW;
166
167 if (!vcpu_has_ptrauth(vcpu))
168 allow_mask &= ~(ARM64_FEATURE_MASK(ID_AA64ISAR1_EL1_APA) |
169 ARM64_FEATURE_MASK(ID_AA64ISAR1_EL1_API) |
170 ARM64_FEATURE_MASK(ID_AA64ISAR1_EL1_GPA) |
171 ARM64_FEATURE_MASK(ID_AA64ISAR1_EL1_GPI));
172
173 return id_aa64isar1_el1_sys_val & allow_mask;
174 }
175
get_pvm_id_aa64isar2(const struct kvm_vcpu * vcpu)176 static u64 get_pvm_id_aa64isar2(const struct kvm_vcpu *vcpu)
177 {
178 u64 allow_mask = PVM_ID_AA64ISAR2_ALLOW;
179
180 if (!vcpu_has_ptrauth(vcpu))
181 allow_mask &= ~(ARM64_FEATURE_MASK(ID_AA64ISAR2_EL1_APA3) |
182 ARM64_FEATURE_MASK(ID_AA64ISAR2_EL1_GPA3));
183
184 return id_aa64isar2_el1_sys_val & allow_mask;
185 }
186
get_pvm_id_aa64mmfr0(const struct kvm_vcpu * vcpu)187 static u64 get_pvm_id_aa64mmfr0(const struct kvm_vcpu *vcpu)
188 {
189 u64 set_mask;
190
191 set_mask = get_restricted_features_unsigned(id_aa64mmfr0_el1_sys_val,
192 PVM_ID_AA64MMFR0_RESTRICT_UNSIGNED);
193
194 return (id_aa64mmfr0_el1_sys_val & PVM_ID_AA64MMFR0_ALLOW) | set_mask;
195 }
196
get_pvm_id_aa64mmfr1(const struct kvm_vcpu * vcpu)197 static u64 get_pvm_id_aa64mmfr1(const struct kvm_vcpu *vcpu)
198 {
199 return id_aa64mmfr1_el1_sys_val & PVM_ID_AA64MMFR1_ALLOW;
200 }
201
get_pvm_id_aa64mmfr2(const struct kvm_vcpu * vcpu)202 static u64 get_pvm_id_aa64mmfr2(const struct kvm_vcpu *vcpu)
203 {
204 return id_aa64mmfr2_el1_sys_val & PVM_ID_AA64MMFR2_ALLOW;
205 }
206
207 /* Read a sanitized cpufeature ID register by its encoding */
pvm_read_id_reg(const struct kvm_vcpu * vcpu,u32 id)208 u64 pvm_read_id_reg(const struct kvm_vcpu *vcpu, u32 id)
209 {
210 switch (id) {
211 case SYS_ID_AA64PFR0_EL1:
212 return get_pvm_id_aa64pfr0(vcpu);
213 case SYS_ID_AA64PFR1_EL1:
214 return get_pvm_id_aa64pfr1(vcpu);
215 case SYS_ID_AA64ZFR0_EL1:
216 return get_pvm_id_aa64zfr0(vcpu);
217 case SYS_ID_AA64DFR0_EL1:
218 return get_pvm_id_aa64dfr0(vcpu);
219 case SYS_ID_AA64DFR1_EL1:
220 return get_pvm_id_aa64dfr1(vcpu);
221 case SYS_ID_AA64AFR0_EL1:
222 return get_pvm_id_aa64afr0(vcpu);
223 case SYS_ID_AA64AFR1_EL1:
224 return get_pvm_id_aa64afr1(vcpu);
225 case SYS_ID_AA64ISAR0_EL1:
226 return get_pvm_id_aa64isar0(vcpu);
227 case SYS_ID_AA64ISAR1_EL1:
228 return get_pvm_id_aa64isar1(vcpu);
229 case SYS_ID_AA64ISAR2_EL1:
230 return get_pvm_id_aa64isar2(vcpu);
231 case SYS_ID_AA64MMFR0_EL1:
232 return get_pvm_id_aa64mmfr0(vcpu);
233 case SYS_ID_AA64MMFR1_EL1:
234 return get_pvm_id_aa64mmfr1(vcpu);
235 case SYS_ID_AA64MMFR2_EL1:
236 return get_pvm_id_aa64mmfr2(vcpu);
237 default:
238 /* Unhandled ID register, RAZ */
239 return 0;
240 }
241 }
242
read_id_reg(const struct kvm_vcpu * vcpu,struct sys_reg_desc const * r)243 static u64 read_id_reg(const struct kvm_vcpu *vcpu,
244 struct sys_reg_desc const *r)
245 {
246 return pvm_read_id_reg(vcpu, reg_to_encoding(r));
247 }
248
249 /* Handler to RAZ/WI sysregs */
pvm_access_raz_wi(struct kvm_vcpu * vcpu,struct sys_reg_params * p,const struct sys_reg_desc * r)250 static bool pvm_access_raz_wi(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
251 const struct sys_reg_desc *r)
252 {
253 if (!p->is_write)
254 p->regval = 0;
255
256 return true;
257 }
258
259 /*
260 * Accessor for AArch32 feature id registers.
261 *
262 * The value of these registers is "unknown" according to the spec if AArch32
263 * isn't supported.
264 */
pvm_access_id_aarch32(struct kvm_vcpu * vcpu,struct sys_reg_params * p,const struct sys_reg_desc * r)265 static bool pvm_access_id_aarch32(struct kvm_vcpu *vcpu,
266 struct sys_reg_params *p,
267 const struct sys_reg_desc *r)
268 {
269 if (p->is_write) {
270 inject_undef64(vcpu);
271 return false;
272 }
273
274 /*
275 * No support for AArch32 guests, therefore, pKVM has no sanitized copy
276 * of AArch32 feature id registers.
277 */
278 BUILD_BUG_ON(FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_EL1),
279 PVM_ID_AA64PFR0_RESTRICT_UNSIGNED) > ID_AA64PFR0_EL1_ELx_64BIT_ONLY);
280
281 return pvm_access_raz_wi(vcpu, p, r);
282 }
283
284 /*
285 * Accessor for AArch64 feature id registers.
286 *
287 * If access is allowed, set the regval to the protected VM's view of the
288 * register and return true.
289 * Otherwise, inject an undefined exception and return false.
290 */
pvm_access_id_aarch64(struct kvm_vcpu * vcpu,struct sys_reg_params * p,const struct sys_reg_desc * r)291 static bool pvm_access_id_aarch64(struct kvm_vcpu *vcpu,
292 struct sys_reg_params *p,
293 const struct sys_reg_desc *r)
294 {
295 if (p->is_write) {
296 inject_undef64(vcpu);
297 return false;
298 }
299
300 p->regval = read_id_reg(vcpu, r);
301 return true;
302 }
303
pvm_gic_read_sre(struct kvm_vcpu * vcpu,struct sys_reg_params * p,const struct sys_reg_desc * r)304 static bool pvm_gic_read_sre(struct kvm_vcpu *vcpu,
305 struct sys_reg_params *p,
306 const struct sys_reg_desc *r)
307 {
308 /* pVMs only support GICv3. 'nuf said. */
309 if (!p->is_write)
310 p->regval = ICC_SRE_EL1_DIB | ICC_SRE_EL1_DFB | ICC_SRE_EL1_SRE;
311
312 return true;
313 }
314
315 /* Mark the specified system register as an AArch32 feature id register. */
316 #define AARCH32(REG) { SYS_DESC(REG), .access = pvm_access_id_aarch32 }
317
318 /* Mark the specified system register as an AArch64 feature id register. */
319 #define AARCH64(REG) { SYS_DESC(REG), .access = pvm_access_id_aarch64 }
320
321 /*
322 * sys_reg_desc initialiser for architecturally unallocated cpufeature ID
323 * register with encoding Op0=3, Op1=0, CRn=0, CRm=crm, Op2=op2
324 * (1 <= crm < 8, 0 <= Op2 < 8).
325 */
326 #define ID_UNALLOCATED(crm, op2) { \
327 Op0(3), Op1(0), CRn(0), CRm(crm), Op2(op2), \
328 .access = pvm_access_id_aarch64, \
329 }
330
331 /* Mark the specified system register as Read-As-Zero/Write-Ignored */
332 #define RAZ_WI(REG) { SYS_DESC(REG), .access = pvm_access_raz_wi }
333
334 /* Mark the specified system register as not being handled in hyp. */
335 #define HOST_HANDLED(REG) { SYS_DESC(REG), .access = NULL }
336
337 /*
338 * Architected system registers.
339 * Important: Must be sorted ascending by Op0, Op1, CRn, CRm, Op2
340 *
341 * NOTE: Anything not explicitly listed here is *restricted by default*, i.e.,
342 * it will lead to injecting an exception into the guest.
343 */
344 static const struct sys_reg_desc pvm_sys_reg_descs[] = {
345 /* Cache maintenance by set/way operations are restricted. */
346
347 /* Debug and Trace Registers are restricted. */
348
349 /* AArch64 mappings of the AArch32 ID registers */
350 /* CRm=1 */
351 AARCH32(SYS_ID_PFR0_EL1),
352 AARCH32(SYS_ID_PFR1_EL1),
353 AARCH32(SYS_ID_DFR0_EL1),
354 AARCH32(SYS_ID_AFR0_EL1),
355 AARCH32(SYS_ID_MMFR0_EL1),
356 AARCH32(SYS_ID_MMFR1_EL1),
357 AARCH32(SYS_ID_MMFR2_EL1),
358 AARCH32(SYS_ID_MMFR3_EL1),
359
360 /* CRm=2 */
361 AARCH32(SYS_ID_ISAR0_EL1),
362 AARCH32(SYS_ID_ISAR1_EL1),
363 AARCH32(SYS_ID_ISAR2_EL1),
364 AARCH32(SYS_ID_ISAR3_EL1),
365 AARCH32(SYS_ID_ISAR4_EL1),
366 AARCH32(SYS_ID_ISAR5_EL1),
367 AARCH32(SYS_ID_MMFR4_EL1),
368 AARCH32(SYS_ID_ISAR6_EL1),
369
370 /* CRm=3 */
371 AARCH32(SYS_MVFR0_EL1),
372 AARCH32(SYS_MVFR1_EL1),
373 AARCH32(SYS_MVFR2_EL1),
374 ID_UNALLOCATED(3,3),
375 AARCH32(SYS_ID_PFR2_EL1),
376 AARCH32(SYS_ID_DFR1_EL1),
377 AARCH32(SYS_ID_MMFR5_EL1),
378 ID_UNALLOCATED(3,7),
379
380 /* AArch64 ID registers */
381 /* CRm=4 */
382 AARCH64(SYS_ID_AA64PFR0_EL1),
383 AARCH64(SYS_ID_AA64PFR1_EL1),
384 ID_UNALLOCATED(4,2),
385 ID_UNALLOCATED(4,3),
386 AARCH64(SYS_ID_AA64ZFR0_EL1),
387 ID_UNALLOCATED(4,5),
388 ID_UNALLOCATED(4,6),
389 ID_UNALLOCATED(4,7),
390 AARCH64(SYS_ID_AA64DFR0_EL1),
391 AARCH64(SYS_ID_AA64DFR1_EL1),
392 ID_UNALLOCATED(5,2),
393 ID_UNALLOCATED(5,3),
394 AARCH64(SYS_ID_AA64AFR0_EL1),
395 AARCH64(SYS_ID_AA64AFR1_EL1),
396 ID_UNALLOCATED(5,6),
397 ID_UNALLOCATED(5,7),
398 AARCH64(SYS_ID_AA64ISAR0_EL1),
399 AARCH64(SYS_ID_AA64ISAR1_EL1),
400 AARCH64(SYS_ID_AA64ISAR2_EL1),
401 ID_UNALLOCATED(6,3),
402 ID_UNALLOCATED(6,4),
403 ID_UNALLOCATED(6,5),
404 ID_UNALLOCATED(6,6),
405 ID_UNALLOCATED(6,7),
406 AARCH64(SYS_ID_AA64MMFR0_EL1),
407 AARCH64(SYS_ID_AA64MMFR1_EL1),
408 AARCH64(SYS_ID_AA64MMFR2_EL1),
409 ID_UNALLOCATED(7,3),
410 ID_UNALLOCATED(7,4),
411 ID_UNALLOCATED(7,5),
412 ID_UNALLOCATED(7,6),
413 ID_UNALLOCATED(7,7),
414
415 /* Scalable Vector Registers are restricted. */
416
417 RAZ_WI(SYS_ERRIDR_EL1),
418 RAZ_WI(SYS_ERRSELR_EL1),
419 RAZ_WI(SYS_ERXFR_EL1),
420 RAZ_WI(SYS_ERXCTLR_EL1),
421 RAZ_WI(SYS_ERXSTATUS_EL1),
422 RAZ_WI(SYS_ERXADDR_EL1),
423 RAZ_WI(SYS_ERXMISC0_EL1),
424 RAZ_WI(SYS_ERXMISC1_EL1),
425
426 /* Performance Monitoring Registers are restricted. */
427
428 /* Limited Ordering Regions Registers are restricted. */
429
430 HOST_HANDLED(SYS_ICC_SGI1R_EL1),
431 HOST_HANDLED(SYS_ICC_ASGI1R_EL1),
432 HOST_HANDLED(SYS_ICC_SGI0R_EL1),
433 { SYS_DESC(SYS_ICC_SRE_EL1), .access = pvm_gic_read_sre, },
434
435 HOST_HANDLED(SYS_CCSIDR_EL1),
436 HOST_HANDLED(SYS_CLIDR_EL1),
437 HOST_HANDLED(SYS_CSSELR_EL1),
438 HOST_HANDLED(SYS_CTR_EL0),
439
440 /* Performance Monitoring Registers are restricted. */
441
442 /* Activity Monitoring Registers are restricted. */
443
444 HOST_HANDLED(SYS_CNTP_TVAL_EL0),
445 HOST_HANDLED(SYS_CNTP_CTL_EL0),
446 HOST_HANDLED(SYS_CNTP_CVAL_EL0),
447
448 /* Performance Monitoring Registers are restricted. */
449 };
450
451 /*
452 * Checks that the sysreg table is unique and in-order.
453 *
454 * Returns 0 if the table is consistent, or 1 otherwise.
455 */
kvm_check_pvm_sysreg_table(void)456 int kvm_check_pvm_sysreg_table(void)
457 {
458 unsigned int i;
459
460 for (i = 1; i < ARRAY_SIZE(pvm_sys_reg_descs); i++) {
461 if (cmp_sys_reg(&pvm_sys_reg_descs[i-1], &pvm_sys_reg_descs[i]) >= 0)
462 return 1;
463 }
464
465 return 0;
466 }
467
468 /*
469 * Handler for protected VM MSR, MRS or System instruction execution.
470 *
471 * Returns true if the hypervisor has handled the exit, and control should go
472 * back to the guest, or false if it hasn't, to be handled by the host.
473 */
kvm_handle_pvm_sysreg(struct kvm_vcpu * vcpu,u64 * exit_code)474 bool kvm_handle_pvm_sysreg(struct kvm_vcpu *vcpu, u64 *exit_code)
475 {
476 const struct sys_reg_desc *r;
477 struct sys_reg_params params;
478 unsigned long esr = kvm_vcpu_get_esr(vcpu);
479 int Rt = kvm_vcpu_sys_get_rt(vcpu);
480
481 params = esr_sys64_to_params(esr);
482 params.regval = vcpu_get_reg(vcpu, Rt);
483
484 r = find_reg(¶ms, pvm_sys_reg_descs, ARRAY_SIZE(pvm_sys_reg_descs));
485
486 /* Undefined (RESTRICTED). */
487 if (r == NULL) {
488 inject_undef64(vcpu);
489 return true;
490 }
491
492 /* Handled by the host (HOST_HANDLED) */
493 if (r->access == NULL)
494 return false;
495
496 /* Handled by hyp: skip instruction if instructed to do so. */
497 if (r->access(vcpu, ¶ms, r))
498 __kvm_skip_instr(vcpu);
499
500 if (!params.is_write)
501 vcpu_set_reg(vcpu, Rt, params.regval);
502
503 return true;
504 }
505
506 /*
507 * Handler for protected VM restricted exceptions.
508 *
509 * Inject an undefined exception into the guest and return true to indicate that
510 * the hypervisor has handled the exit, and control should go back to the guest.
511 */
kvm_handle_pvm_restricted(struct kvm_vcpu * vcpu,u64 * exit_code)512 bool kvm_handle_pvm_restricted(struct kvm_vcpu *vcpu, u64 *exit_code)
513 {
514 inject_undef64(vcpu);
515 return true;
516 }
517