Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[cris-mirror.git] / arch / s390 / kvm / priv.c
blob46160388e9964ba201a5c61768cc83eaa19eb4bb
1 /*
2 * handling privileged instructions
4 * Copyright IBM Corp. 2008, 2013
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License (version 2 only)
8 * as published by the Free Software Foundation.
10 * Author(s): Carsten Otte <cotte@de.ibm.com>
11 * Christian Borntraeger <borntraeger@de.ibm.com>
14 #include <linux/kvm.h>
15 #include <linux/gfp.h>
16 #include <linux/errno.h>
17 #include <linux/compat.h>
18 #include <asm/asm-offsets.h>
19 #include <asm/facility.h>
20 #include <asm/current.h>
21 #include <asm/debug.h>
22 #include <asm/ebcdic.h>
23 #include <asm/sysinfo.h>
24 #include <asm/pgtable.h>
25 #include <asm/pgalloc.h>
26 #include <asm/gmap.h>
27 #include <asm/io.h>
28 #include <asm/ptrace.h>
29 #include <asm/compat.h>
30 #include <asm/sclp.h>
31 #include "gaccess.h"
32 #include "kvm-s390.h"
33 #include "trace.h"
35 /* Handle SCK (SET CLOCK) interception */
36 static int handle_set_clock(struct kvm_vcpu *vcpu)
38 int rc;
39 ar_t ar;
40 u64 op2, val;
42 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
43 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
45 op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
46 if (op2 & 7) /* Operand must be on a doubleword boundary */
47 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
48 rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
49 if (rc)
50 return kvm_s390_inject_prog_cond(vcpu, rc);
52 VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
53 kvm_s390_set_tod_clock(vcpu->kvm, val);
55 kvm_s390_set_psw_cc(vcpu, 0);
56 return 0;
59 static int handle_set_prefix(struct kvm_vcpu *vcpu)
61 u64 operand2;
62 u32 address;
63 int rc;
64 ar_t ar;
66 vcpu->stat.instruction_spx++;
68 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
69 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
71 operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
73 /* must be word boundary */
74 if (operand2 & 3)
75 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
77 /* get the value */
78 rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
79 if (rc)
80 return kvm_s390_inject_prog_cond(vcpu, rc);
82 address &= 0x7fffe000u;
85 * Make sure the new value is valid memory. We only need to check the
86 * first page, since address is 8k aligned and memory pieces are always
87 * at least 1MB aligned and have at least a size of 1MB.
89 if (kvm_is_error_gpa(vcpu->kvm, address))
90 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
92 kvm_s390_set_prefix(vcpu, address);
93 trace_kvm_s390_handle_prefix(vcpu, 1, address);
94 return 0;
97 static int handle_store_prefix(struct kvm_vcpu *vcpu)
99 u64 operand2;
100 u32 address;
101 int rc;
102 ar_t ar;
104 vcpu->stat.instruction_stpx++;
106 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
107 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
109 operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
111 /* must be word boundary */
112 if (operand2 & 3)
113 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
115 address = kvm_s390_get_prefix(vcpu);
117 /* get the value */
118 rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
119 if (rc)
120 return kvm_s390_inject_prog_cond(vcpu, rc);
122 VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
123 trace_kvm_s390_handle_prefix(vcpu, 0, address);
124 return 0;
127 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
129 u16 vcpu_id = vcpu->vcpu_id;
130 u64 ga;
131 int rc;
132 ar_t ar;
134 vcpu->stat.instruction_stap++;
136 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
137 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
139 ga = kvm_s390_get_base_disp_s(vcpu, &ar);
141 if (ga & 1)
142 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
144 rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
145 if (rc)
146 return kvm_s390_inject_prog_cond(vcpu, rc);
148 VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
149 trace_kvm_s390_handle_stap(vcpu, ga);
150 return 0;
153 static int __skey_check_enable(struct kvm_vcpu *vcpu)
155 int rc = 0;
157 trace_kvm_s390_skey_related_inst(vcpu);
158 if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
159 return rc;
161 rc = s390_enable_skey();
162 VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
163 if (!rc)
164 vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
165 return rc;
168 static int try_handle_skey(struct kvm_vcpu *vcpu)
170 int rc;
172 vcpu->stat.instruction_storage_key++;
173 rc = __skey_check_enable(vcpu);
174 if (rc)
175 return rc;
176 if (sclp.has_skey) {
177 /* with storage-key facility, SIE interprets it for us */
178 kvm_s390_retry_instr(vcpu);
179 VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
180 return -EAGAIN;
182 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
183 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
184 return 0;
187 static int handle_iske(struct kvm_vcpu *vcpu)
189 unsigned long addr;
190 unsigned char key;
191 int reg1, reg2;
192 int rc;
194 rc = try_handle_skey(vcpu);
195 if (rc)
196 return rc != -EAGAIN ? rc : 0;
198 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
200 addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
201 addr = kvm_s390_logical_to_effective(vcpu, addr);
202 addr = kvm_s390_real_to_abs(vcpu, addr);
203 addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
204 if (kvm_is_error_hva(addr))
205 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
207 down_read(&current->mm->mmap_sem);
208 rc = get_guest_storage_key(current->mm, addr, &key);
209 up_read(&current->mm->mmap_sem);
210 if (rc)
211 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
212 vcpu->run->s.regs.gprs[reg1] &= ~0xff;
213 vcpu->run->s.regs.gprs[reg1] |= key;
214 return 0;
217 static int handle_rrbe(struct kvm_vcpu *vcpu)
219 unsigned long addr;
220 int reg1, reg2;
221 int rc;
223 rc = try_handle_skey(vcpu);
224 if (rc)
225 return rc != -EAGAIN ? rc : 0;
227 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
229 addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
230 addr = kvm_s390_logical_to_effective(vcpu, addr);
231 addr = kvm_s390_real_to_abs(vcpu, addr);
232 addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
233 if (kvm_is_error_hva(addr))
234 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
236 down_read(&current->mm->mmap_sem);
237 rc = reset_guest_reference_bit(current->mm, addr);
238 up_read(&current->mm->mmap_sem);
239 if (rc < 0)
240 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
242 kvm_s390_set_psw_cc(vcpu, rc);
243 return 0;
246 #define SSKE_NQ 0x8
247 #define SSKE_MR 0x4
248 #define SSKE_MC 0x2
249 #define SSKE_MB 0x1
250 static int handle_sske(struct kvm_vcpu *vcpu)
252 unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
253 unsigned long start, end;
254 unsigned char key, oldkey;
255 int reg1, reg2;
256 int rc;
258 rc = try_handle_skey(vcpu);
259 if (rc)
260 return rc != -EAGAIN ? rc : 0;
262 if (!test_kvm_facility(vcpu->kvm, 8))
263 m3 &= ~SSKE_MB;
264 if (!test_kvm_facility(vcpu->kvm, 10))
265 m3 &= ~(SSKE_MC | SSKE_MR);
266 if (!test_kvm_facility(vcpu->kvm, 14))
267 m3 &= ~SSKE_NQ;
269 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
271 key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
272 start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
273 start = kvm_s390_logical_to_effective(vcpu, start);
274 if (m3 & SSKE_MB) {
275 /* start already designates an absolute address */
276 end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
277 } else {
278 start = kvm_s390_real_to_abs(vcpu, start);
279 end = start + PAGE_SIZE;
282 while (start != end) {
283 unsigned long addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
285 if (kvm_is_error_hva(addr))
286 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
288 down_read(&current->mm->mmap_sem);
289 rc = cond_set_guest_storage_key(current->mm, addr, key, &oldkey,
290 m3 & SSKE_NQ, m3 & SSKE_MR,
291 m3 & SSKE_MC);
292 up_read(&current->mm->mmap_sem);
293 if (rc < 0)
294 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
295 start += PAGE_SIZE;
298 if (m3 & (SSKE_MC | SSKE_MR)) {
299 if (m3 & SSKE_MB) {
300 /* skey in reg1 is unpredictable */
301 kvm_s390_set_psw_cc(vcpu, 3);
302 } else {
303 kvm_s390_set_psw_cc(vcpu, rc);
304 vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
305 vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
308 if (m3 & SSKE_MB) {
309 if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_64BIT)
310 vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
311 else
312 vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
313 end = kvm_s390_logical_to_effective(vcpu, end);
314 vcpu->run->s.regs.gprs[reg2] |= end;
316 return 0;
319 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
321 vcpu->stat.instruction_ipte_interlock++;
322 if (psw_bits(vcpu->arch.sie_block->gpsw).p)
323 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
324 wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
325 kvm_s390_retry_instr(vcpu);
326 VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
327 return 0;
330 static int handle_test_block(struct kvm_vcpu *vcpu)
332 gpa_t addr;
333 int reg2;
335 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
336 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
338 kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
339 addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
340 addr = kvm_s390_logical_to_effective(vcpu, addr);
341 if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
342 return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
343 addr = kvm_s390_real_to_abs(vcpu, addr);
345 if (kvm_is_error_gpa(vcpu->kvm, addr))
346 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
348 * We don't expect errors on modern systems, and do not care
349 * about storage keys (yet), so let's just clear the page.
351 if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
352 return -EFAULT;
353 kvm_s390_set_psw_cc(vcpu, 0);
354 vcpu->run->s.regs.gprs[0] = 0;
355 return 0;
358 static int handle_tpi(struct kvm_vcpu *vcpu)
360 struct kvm_s390_interrupt_info *inti;
361 unsigned long len;
362 u32 tpi_data[3];
363 int rc;
364 u64 addr;
365 ar_t ar;
367 addr = kvm_s390_get_base_disp_s(vcpu, &ar);
368 if (addr & 3)
369 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
371 inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
372 if (!inti) {
373 kvm_s390_set_psw_cc(vcpu, 0);
374 return 0;
377 tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
378 tpi_data[1] = inti->io.io_int_parm;
379 tpi_data[2] = inti->io.io_int_word;
380 if (addr) {
382 * Store the two-word I/O interruption code into the
383 * provided area.
385 len = sizeof(tpi_data) - 4;
386 rc = write_guest(vcpu, addr, ar, &tpi_data, len);
387 if (rc) {
388 rc = kvm_s390_inject_prog_cond(vcpu, rc);
389 goto reinject_interrupt;
391 } else {
393 * Store the three-word I/O interruption code into
394 * the appropriate lowcore area.
396 len = sizeof(tpi_data);
397 if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
398 /* failed writes to the low core are not recoverable */
399 rc = -EFAULT;
400 goto reinject_interrupt;
404 /* irq was successfully handed to the guest */
405 kfree(inti);
406 kvm_s390_set_psw_cc(vcpu, 1);
407 return 0;
408 reinject_interrupt:
410 * If we encounter a problem storing the interruption code, the
411 * instruction is suppressed from the guest's view: reinject the
412 * interrupt.
414 if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
415 kfree(inti);
416 rc = -EFAULT;
418 /* don't set the cc, a pgm irq was injected or we drop to user space */
419 return rc ? -EFAULT : 0;
422 static int handle_tsch(struct kvm_vcpu *vcpu)
424 struct kvm_s390_interrupt_info *inti = NULL;
425 const u64 isc_mask = 0xffUL << 24; /* all iscs set */
427 /* a valid schid has at least one bit set */
428 if (vcpu->run->s.regs.gprs[1])
429 inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
430 vcpu->run->s.regs.gprs[1]);
433 * Prepare exit to userspace.
434 * We indicate whether we dequeued a pending I/O interrupt
435 * so that userspace can re-inject it if the instruction gets
436 * a program check. While this may re-order the pending I/O
437 * interrupts, this is no problem since the priority is kept
438 * intact.
440 vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
441 vcpu->run->s390_tsch.dequeued = !!inti;
442 if (inti) {
443 vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
444 vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
445 vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
446 vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
448 vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
449 kfree(inti);
450 return -EREMOTE;
453 static int handle_io_inst(struct kvm_vcpu *vcpu)
455 VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
457 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
458 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
460 if (vcpu->kvm->arch.css_support) {
462 * Most I/O instructions will be handled by userspace.
463 * Exceptions are tpi and the interrupt portion of tsch.
465 if (vcpu->arch.sie_block->ipa == 0xb236)
466 return handle_tpi(vcpu);
467 if (vcpu->arch.sie_block->ipa == 0xb235)
468 return handle_tsch(vcpu);
469 /* Handle in userspace. */
470 return -EOPNOTSUPP;
471 } else {
473 * Set condition code 3 to stop the guest from issuing channel
474 * I/O instructions.
476 kvm_s390_set_psw_cc(vcpu, 3);
477 return 0;
481 static int handle_stfl(struct kvm_vcpu *vcpu)
483 int rc;
484 unsigned int fac;
486 vcpu->stat.instruction_stfl++;
488 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
489 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
492 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
493 * into a u32 memory representation. They will remain bits 0-31.
495 fac = *vcpu->kvm->arch.model.fac_list >> 32;
496 rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
497 &fac, sizeof(fac));
498 if (rc)
499 return rc;
500 VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
501 trace_kvm_s390_handle_stfl(vcpu, fac);
502 return 0;
505 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
506 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
507 #define PSW_ADDR_24 0x0000000000ffffffUL
508 #define PSW_ADDR_31 0x000000007fffffffUL
510 int is_valid_psw(psw_t *psw)
512 if (psw->mask & PSW_MASK_UNASSIGNED)
513 return 0;
514 if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
515 if (psw->addr & ~PSW_ADDR_31)
516 return 0;
518 if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
519 return 0;
520 if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_EA)
521 return 0;
522 if (psw->addr & 1)
523 return 0;
524 return 1;
527 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
529 psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
530 psw_compat_t new_psw;
531 u64 addr;
532 int rc;
533 ar_t ar;
535 if (gpsw->mask & PSW_MASK_PSTATE)
536 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
538 addr = kvm_s390_get_base_disp_s(vcpu, &ar);
539 if (addr & 7)
540 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
542 rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
543 if (rc)
544 return kvm_s390_inject_prog_cond(vcpu, rc);
545 if (!(new_psw.mask & PSW32_MASK_BASE))
546 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
547 gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
548 gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
549 gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
550 if (!is_valid_psw(gpsw))
551 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
552 return 0;
555 static int handle_lpswe(struct kvm_vcpu *vcpu)
557 psw_t new_psw;
558 u64 addr;
559 int rc;
560 ar_t ar;
562 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
563 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
565 addr = kvm_s390_get_base_disp_s(vcpu, &ar);
566 if (addr & 7)
567 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
568 rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
569 if (rc)
570 return kvm_s390_inject_prog_cond(vcpu, rc);
571 vcpu->arch.sie_block->gpsw = new_psw;
572 if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
573 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
574 return 0;
577 static int handle_stidp(struct kvm_vcpu *vcpu)
579 u64 stidp_data = vcpu->kvm->arch.model.cpuid;
580 u64 operand2;
581 int rc;
582 ar_t ar;
584 vcpu->stat.instruction_stidp++;
586 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
587 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
589 operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
591 if (operand2 & 7)
592 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
594 rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
595 if (rc)
596 return kvm_s390_inject_prog_cond(vcpu, rc);
598 VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
599 return 0;
602 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
604 int cpus = 0;
605 int n;
607 cpus = atomic_read(&vcpu->kvm->online_vcpus);
609 /* deal with other level 3 hypervisors */
610 if (stsi(mem, 3, 2, 2))
611 mem->count = 0;
612 if (mem->count < 8)
613 mem->count++;
614 for (n = mem->count - 1; n > 0 ; n--)
615 memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
617 memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
618 mem->vm[0].cpus_total = cpus;
619 mem->vm[0].cpus_configured = cpus;
620 mem->vm[0].cpus_standby = 0;
621 mem->vm[0].cpus_reserved = 0;
622 mem->vm[0].caf = 1000;
623 memcpy(mem->vm[0].name, "KVMguest", 8);
624 ASCEBC(mem->vm[0].name, 8);
625 memcpy(mem->vm[0].cpi, "KVM/Linux ", 16);
626 ASCEBC(mem->vm[0].cpi, 16);
629 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, ar_t ar,
630 u8 fc, u8 sel1, u16 sel2)
632 vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
633 vcpu->run->s390_stsi.addr = addr;
634 vcpu->run->s390_stsi.ar = ar;
635 vcpu->run->s390_stsi.fc = fc;
636 vcpu->run->s390_stsi.sel1 = sel1;
637 vcpu->run->s390_stsi.sel2 = sel2;
640 static int handle_stsi(struct kvm_vcpu *vcpu)
642 int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
643 int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
644 int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
645 unsigned long mem = 0;
646 u64 operand2;
647 int rc = 0;
648 ar_t ar;
650 vcpu->stat.instruction_stsi++;
651 VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
653 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
654 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
656 if (fc > 3) {
657 kvm_s390_set_psw_cc(vcpu, 3);
658 return 0;
661 if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
662 || vcpu->run->s.regs.gprs[1] & 0xffff0000)
663 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
665 if (fc == 0) {
666 vcpu->run->s.regs.gprs[0] = 3 << 28;
667 kvm_s390_set_psw_cc(vcpu, 0);
668 return 0;
671 operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
673 if (operand2 & 0xfff)
674 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
676 switch (fc) {
677 case 1: /* same handling for 1 and 2 */
678 case 2:
679 mem = get_zeroed_page(GFP_KERNEL);
680 if (!mem)
681 goto out_no_data;
682 if (stsi((void *) mem, fc, sel1, sel2))
683 goto out_no_data;
684 break;
685 case 3:
686 if (sel1 != 2 || sel2 != 2)
687 goto out_no_data;
688 mem = get_zeroed_page(GFP_KERNEL);
689 if (!mem)
690 goto out_no_data;
691 handle_stsi_3_2_2(vcpu, (void *) mem);
692 break;
695 rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
696 if (rc) {
697 rc = kvm_s390_inject_prog_cond(vcpu, rc);
698 goto out;
700 if (vcpu->kvm->arch.user_stsi) {
701 insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
702 rc = -EREMOTE;
704 trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
705 free_page(mem);
706 kvm_s390_set_psw_cc(vcpu, 0);
707 vcpu->run->s.regs.gprs[0] = 0;
708 return rc;
709 out_no_data:
710 kvm_s390_set_psw_cc(vcpu, 3);
711 out:
712 free_page(mem);
713 return rc;
716 static const intercept_handler_t b2_handlers[256] = {
717 [0x02] = handle_stidp,
718 [0x04] = handle_set_clock,
719 [0x10] = handle_set_prefix,
720 [0x11] = handle_store_prefix,
721 [0x12] = handle_store_cpu_address,
722 [0x14] = kvm_s390_handle_vsie,
723 [0x21] = handle_ipte_interlock,
724 [0x29] = handle_iske,
725 [0x2a] = handle_rrbe,
726 [0x2b] = handle_sske,
727 [0x2c] = handle_test_block,
728 [0x30] = handle_io_inst,
729 [0x31] = handle_io_inst,
730 [0x32] = handle_io_inst,
731 [0x33] = handle_io_inst,
732 [0x34] = handle_io_inst,
733 [0x35] = handle_io_inst,
734 [0x36] = handle_io_inst,
735 [0x37] = handle_io_inst,
736 [0x38] = handle_io_inst,
737 [0x39] = handle_io_inst,
738 [0x3a] = handle_io_inst,
739 [0x3b] = handle_io_inst,
740 [0x3c] = handle_io_inst,
741 [0x50] = handle_ipte_interlock,
742 [0x5f] = handle_io_inst,
743 [0x74] = handle_io_inst,
744 [0x76] = handle_io_inst,
745 [0x7d] = handle_stsi,
746 [0xb1] = handle_stfl,
747 [0xb2] = handle_lpswe,
750 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
752 intercept_handler_t handler;
755 * A lot of B2 instructions are priviledged. Here we check for
756 * the privileged ones, that we can handle in the kernel.
757 * Anything else goes to userspace.
759 handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
760 if (handler)
761 return handler(vcpu);
763 return -EOPNOTSUPP;
766 static int handle_epsw(struct kvm_vcpu *vcpu)
768 int reg1, reg2;
770 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
772 /* This basically extracts the mask half of the psw. */
773 vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
774 vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
775 if (reg2) {
776 vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
777 vcpu->run->s.regs.gprs[reg2] |=
778 vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
780 return 0;
783 #define PFMF_RESERVED 0xfffc0101UL
784 #define PFMF_SK 0x00020000UL
785 #define PFMF_CF 0x00010000UL
786 #define PFMF_UI 0x00008000UL
787 #define PFMF_FSC 0x00007000UL
788 #define PFMF_NQ 0x00000800UL
789 #define PFMF_MR 0x00000400UL
790 #define PFMF_MC 0x00000200UL
791 #define PFMF_KEY 0x000000feUL
793 static int handle_pfmf(struct kvm_vcpu *vcpu)
795 bool mr = false, mc = false, nq;
796 int reg1, reg2;
797 unsigned long start, end;
798 unsigned char key;
800 vcpu->stat.instruction_pfmf++;
802 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
804 if (!test_kvm_facility(vcpu->kvm, 8))
805 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
807 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
808 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
810 if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
811 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
813 /* Only provide non-quiescing support if enabled for the guest */
814 if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
815 !test_kvm_facility(vcpu->kvm, 14))
816 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
818 /* Only provide conditional-SSKE support if enabled for the guest */
819 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
820 test_kvm_facility(vcpu->kvm, 10)) {
821 mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
822 mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
825 nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
826 key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
827 start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
828 start = kvm_s390_logical_to_effective(vcpu, start);
830 if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
831 if (kvm_s390_check_low_addr_prot_real(vcpu, start))
832 return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
835 switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
836 case 0x00000000:
837 /* only 4k frames specify a real address */
838 start = kvm_s390_real_to_abs(vcpu, start);
839 end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
840 break;
841 case 0x00001000:
842 end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
843 break;
844 case 0x00002000:
845 /* only support 2G frame size if EDAT2 is available and we are
846 not in 24-bit addressing mode */
847 if (!test_kvm_facility(vcpu->kvm, 78) ||
848 psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_24BIT)
849 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
850 end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
851 break;
852 default:
853 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
856 while (start != end) {
857 unsigned long useraddr;
859 /* Translate guest address to host address */
860 useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
861 if (kvm_is_error_hva(useraddr))
862 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
864 if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
865 if (clear_user((void __user *)useraddr, PAGE_SIZE))
866 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
869 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
870 int rc = __skey_check_enable(vcpu);
872 if (rc)
873 return rc;
874 down_read(&current->mm->mmap_sem);
875 rc = cond_set_guest_storage_key(current->mm, useraddr,
876 key, NULL, nq, mr, mc);
877 up_read(&current->mm->mmap_sem);
878 if (rc < 0)
879 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
882 start += PAGE_SIZE;
884 if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
885 if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_64BIT) {
886 vcpu->run->s.regs.gprs[reg2] = end;
887 } else {
888 vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
889 end = kvm_s390_logical_to_effective(vcpu, end);
890 vcpu->run->s.regs.gprs[reg2] |= end;
893 return 0;
896 static int handle_essa(struct kvm_vcpu *vcpu)
898 /* entries expected to be 1FF */
899 int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
900 unsigned long *cbrlo;
901 struct gmap *gmap;
902 int i;
904 VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
905 gmap = vcpu->arch.gmap;
906 vcpu->stat.instruction_essa++;
907 if (!vcpu->kvm->arch.use_cmma)
908 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
910 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
911 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
913 if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
914 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
916 /* Retry the ESSA instruction */
917 kvm_s390_retry_instr(vcpu);
918 vcpu->arch.sie_block->cbrlo &= PAGE_MASK; /* reset nceo */
919 cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
920 down_read(&gmap->mm->mmap_sem);
921 for (i = 0; i < entries; ++i)
922 __gmap_zap(gmap, cbrlo[i]);
923 up_read(&gmap->mm->mmap_sem);
924 return 0;
927 static const intercept_handler_t b9_handlers[256] = {
928 [0x8a] = handle_ipte_interlock,
929 [0x8d] = handle_epsw,
930 [0x8e] = handle_ipte_interlock,
931 [0x8f] = handle_ipte_interlock,
932 [0xab] = handle_essa,
933 [0xaf] = handle_pfmf,
936 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
938 intercept_handler_t handler;
940 /* This is handled just as for the B2 instructions. */
941 handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
942 if (handler)
943 return handler(vcpu);
945 return -EOPNOTSUPP;
948 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
950 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
951 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
952 int reg, rc, nr_regs;
953 u32 ctl_array[16];
954 u64 ga;
955 ar_t ar;
957 vcpu->stat.instruction_lctl++;
959 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
960 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
962 ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
964 if (ga & 3)
965 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
967 VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
968 trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
970 nr_regs = ((reg3 - reg1) & 0xf) + 1;
971 rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
972 if (rc)
973 return kvm_s390_inject_prog_cond(vcpu, rc);
974 reg = reg1;
975 nr_regs = 0;
976 do {
977 vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
978 vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
979 if (reg == reg3)
980 break;
981 reg = (reg + 1) % 16;
982 } while (1);
983 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
984 return 0;
987 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
989 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
990 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
991 int reg, rc, nr_regs;
992 u32 ctl_array[16];
993 u64 ga;
994 ar_t ar;
996 vcpu->stat.instruction_stctl++;
998 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
999 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1001 ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1003 if (ga & 3)
1004 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1006 VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1007 trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1009 reg = reg1;
1010 nr_regs = 0;
1011 do {
1012 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1013 if (reg == reg3)
1014 break;
1015 reg = (reg + 1) % 16;
1016 } while (1);
1017 rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1018 return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1021 static int handle_lctlg(struct kvm_vcpu *vcpu)
1023 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1024 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1025 int reg, rc, nr_regs;
1026 u64 ctl_array[16];
1027 u64 ga;
1028 ar_t ar;
1030 vcpu->stat.instruction_lctlg++;
1032 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1033 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1035 ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1037 if (ga & 7)
1038 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1040 VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1041 trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1043 nr_regs = ((reg3 - reg1) & 0xf) + 1;
1044 rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1045 if (rc)
1046 return kvm_s390_inject_prog_cond(vcpu, rc);
1047 reg = reg1;
1048 nr_regs = 0;
1049 do {
1050 vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1051 if (reg == reg3)
1052 break;
1053 reg = (reg + 1) % 16;
1054 } while (1);
1055 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1056 return 0;
1059 static int handle_stctg(struct kvm_vcpu *vcpu)
1061 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1062 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1063 int reg, rc, nr_regs;
1064 u64 ctl_array[16];
1065 u64 ga;
1066 ar_t ar;
1068 vcpu->stat.instruction_stctg++;
1070 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1071 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1073 ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1075 if (ga & 7)
1076 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1078 VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1079 trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1081 reg = reg1;
1082 nr_regs = 0;
1083 do {
1084 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1085 if (reg == reg3)
1086 break;
1087 reg = (reg + 1) % 16;
1088 } while (1);
1089 rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1090 return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1093 static const intercept_handler_t eb_handlers[256] = {
1094 [0x2f] = handle_lctlg,
1095 [0x25] = handle_stctg,
1098 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1100 intercept_handler_t handler;
1102 handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
1103 if (handler)
1104 return handler(vcpu);
1105 return -EOPNOTSUPP;
1108 static int handle_tprot(struct kvm_vcpu *vcpu)
1110 u64 address1, address2;
1111 unsigned long hva, gpa;
1112 int ret = 0, cc = 0;
1113 bool writable;
1114 ar_t ar;
1116 vcpu->stat.instruction_tprot++;
1118 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1119 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1121 kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1123 /* we only handle the Linux memory detection case:
1124 * access key == 0
1125 * everything else goes to userspace. */
1126 if (address2 & 0xf0)
1127 return -EOPNOTSUPP;
1128 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1129 ipte_lock(vcpu);
1130 ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1131 if (ret == PGM_PROTECTION) {
1132 /* Write protected? Try again with read-only... */
1133 cc = 1;
1134 ret = guest_translate_address(vcpu, address1, ar, &gpa,
1135 GACC_FETCH);
1137 if (ret) {
1138 if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1139 ret = kvm_s390_inject_program_int(vcpu, ret);
1140 } else if (ret > 0) {
1141 /* Translation not available */
1142 kvm_s390_set_psw_cc(vcpu, 3);
1143 ret = 0;
1145 goto out_unlock;
1148 hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1149 if (kvm_is_error_hva(hva)) {
1150 ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1151 } else {
1152 if (!writable)
1153 cc = 1; /* Write not permitted ==> read-only */
1154 kvm_s390_set_psw_cc(vcpu, cc);
1155 /* Note: CC2 only occurs for storage keys (not supported yet) */
1157 out_unlock:
1158 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1159 ipte_unlock(vcpu);
1160 return ret;
1163 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1165 /* For e5xx... instructions we only handle TPROT */
1166 if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1167 return handle_tprot(vcpu);
1168 return -EOPNOTSUPP;
1171 static int handle_sckpf(struct kvm_vcpu *vcpu)
1173 u32 value;
1175 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1176 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1178 if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1179 return kvm_s390_inject_program_int(vcpu,
1180 PGM_SPECIFICATION);
1182 value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1183 vcpu->arch.sie_block->todpr = value;
1185 return 0;
1188 static int handle_ptff(struct kvm_vcpu *vcpu)
1190 /* we don't emulate any control instructions yet */
1191 kvm_s390_set_psw_cc(vcpu, 3);
1192 return 0;
1195 static const intercept_handler_t x01_handlers[256] = {
1196 [0x04] = handle_ptff,
1197 [0x07] = handle_sckpf,
1200 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1202 intercept_handler_t handler;
1204 handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1205 if (handler)
1206 return handler(vcpu);
1207 return -EOPNOTSUPP;