exec: Don't reset euid and egid when the tracee has CAP_SETUID
[cris-mirror.git] / arch / s390 / kvm / intercept.c
blob7a27eebab28ad023069d21ae92033a06f4ab482d
1 /*
2 * in-kernel handling for sie intercepts
4 * Copyright IBM Corp. 2008, 2014
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_host.h>
15 #include <linux/errno.h>
16 #include <linux/pagemap.h>
18 #include <asm/kvm_host.h>
19 #include <asm/asm-offsets.h>
20 #include <asm/irq.h>
22 #include "kvm-s390.h"
23 #include "gaccess.h"
24 #include "trace.h"
25 #include "trace-s390.h"
28 static const intercept_handler_t instruction_handlers[256] = {
29 [0x01] = kvm_s390_handle_01,
30 [0x82] = kvm_s390_handle_lpsw,
31 [0x83] = kvm_s390_handle_diag,
32 [0xaa] = kvm_s390_handle_aa,
33 [0xae] = kvm_s390_handle_sigp,
34 [0xb2] = kvm_s390_handle_b2,
35 [0xb6] = kvm_s390_handle_stctl,
36 [0xb7] = kvm_s390_handle_lctl,
37 [0xb9] = kvm_s390_handle_b9,
38 [0xe5] = kvm_s390_handle_e5,
39 [0xeb] = kvm_s390_handle_eb,
42 u8 kvm_s390_get_ilen(struct kvm_vcpu *vcpu)
44 struct kvm_s390_sie_block *sie_block = vcpu->arch.sie_block;
45 u8 ilen = 0;
47 switch (vcpu->arch.sie_block->icptcode) {
48 case ICPT_INST:
49 case ICPT_INSTPROGI:
50 case ICPT_OPEREXC:
51 case ICPT_PARTEXEC:
52 case ICPT_IOINST:
53 /* instruction only stored for these icptcodes */
54 ilen = insn_length(vcpu->arch.sie_block->ipa >> 8);
55 /* Use the length of the EXECUTE instruction if necessary */
56 if (sie_block->icptstatus & 1) {
57 ilen = (sie_block->icptstatus >> 4) & 0x6;
58 if (!ilen)
59 ilen = 4;
61 break;
62 case ICPT_PROGI:
63 /* bit 1+2 of pgmilc are the ilc, so we directly get ilen */
64 ilen = vcpu->arch.sie_block->pgmilc & 0x6;
65 break;
67 return ilen;
70 static int handle_noop(struct kvm_vcpu *vcpu)
72 switch (vcpu->arch.sie_block->icptcode) {
73 case 0x10:
74 vcpu->stat.exit_external_request++;
75 break;
76 default:
77 break; /* nothing */
79 return 0;
82 static int handle_stop(struct kvm_vcpu *vcpu)
84 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
85 int rc = 0;
86 uint8_t flags, stop_pending;
88 vcpu->stat.exit_stop_request++;
90 /* delay the stop if any non-stop irq is pending */
91 if (kvm_s390_vcpu_has_irq(vcpu, 1))
92 return 0;
94 /* avoid races with the injection/SIGP STOP code */
95 spin_lock(&li->lock);
96 flags = li->irq.stop.flags;
97 stop_pending = kvm_s390_is_stop_irq_pending(vcpu);
98 spin_unlock(&li->lock);
100 trace_kvm_s390_stop_request(stop_pending, flags);
101 if (!stop_pending)
102 return 0;
104 if (flags & KVM_S390_STOP_FLAG_STORE_STATUS) {
105 rc = kvm_s390_vcpu_store_status(vcpu,
106 KVM_S390_STORE_STATUS_NOADDR);
107 if (rc)
108 return rc;
111 if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
112 kvm_s390_vcpu_stop(vcpu);
113 return -EOPNOTSUPP;
116 static int handle_validity(struct kvm_vcpu *vcpu)
118 int viwhy = vcpu->arch.sie_block->ipb >> 16;
120 vcpu->stat.exit_validity++;
121 trace_kvm_s390_intercept_validity(vcpu, viwhy);
122 KVM_EVENT(3, "validity intercept 0x%x for pid %u (kvm 0x%pK)", viwhy,
123 current->pid, vcpu->kvm);
125 /* do not warn on invalid runtime instrumentation mode */
126 WARN_ONCE(viwhy != 0x44, "kvm: unhandled validity intercept 0x%x\n",
127 viwhy);
128 return -EINVAL;
131 static int handle_instruction(struct kvm_vcpu *vcpu)
133 intercept_handler_t handler;
135 vcpu->stat.exit_instruction++;
136 trace_kvm_s390_intercept_instruction(vcpu,
137 vcpu->arch.sie_block->ipa,
138 vcpu->arch.sie_block->ipb);
139 handler = instruction_handlers[vcpu->arch.sie_block->ipa >> 8];
140 if (handler)
141 return handler(vcpu);
142 return -EOPNOTSUPP;
145 static int inject_prog_on_prog_intercept(struct kvm_vcpu *vcpu)
147 struct kvm_s390_pgm_info pgm_info = {
148 .code = vcpu->arch.sie_block->iprcc,
149 /* the PSW has already been rewound */
150 .flags = KVM_S390_PGM_FLAGS_NO_REWIND,
153 switch (vcpu->arch.sie_block->iprcc & ~PGM_PER) {
154 case PGM_AFX_TRANSLATION:
155 case PGM_ASX_TRANSLATION:
156 case PGM_EX_TRANSLATION:
157 case PGM_LFX_TRANSLATION:
158 case PGM_LSTE_SEQUENCE:
159 case PGM_LSX_TRANSLATION:
160 case PGM_LX_TRANSLATION:
161 case PGM_PRIMARY_AUTHORITY:
162 case PGM_SECONDARY_AUTHORITY:
163 case PGM_SPACE_SWITCH:
164 pgm_info.trans_exc_code = vcpu->arch.sie_block->tecmc;
165 break;
166 case PGM_ALEN_TRANSLATION:
167 case PGM_ALE_SEQUENCE:
168 case PGM_ASTE_INSTANCE:
169 case PGM_ASTE_SEQUENCE:
170 case PGM_ASTE_VALIDITY:
171 case PGM_EXTENDED_AUTHORITY:
172 pgm_info.exc_access_id = vcpu->arch.sie_block->eai;
173 break;
174 case PGM_ASCE_TYPE:
175 case PGM_PAGE_TRANSLATION:
176 case PGM_REGION_FIRST_TRANS:
177 case PGM_REGION_SECOND_TRANS:
178 case PGM_REGION_THIRD_TRANS:
179 case PGM_SEGMENT_TRANSLATION:
180 pgm_info.trans_exc_code = vcpu->arch.sie_block->tecmc;
181 pgm_info.exc_access_id = vcpu->arch.sie_block->eai;
182 pgm_info.op_access_id = vcpu->arch.sie_block->oai;
183 break;
184 case PGM_MONITOR:
185 pgm_info.mon_class_nr = vcpu->arch.sie_block->mcn;
186 pgm_info.mon_code = vcpu->arch.sie_block->tecmc;
187 break;
188 case PGM_VECTOR_PROCESSING:
189 case PGM_DATA:
190 pgm_info.data_exc_code = vcpu->arch.sie_block->dxc;
191 break;
192 case PGM_PROTECTION:
193 pgm_info.trans_exc_code = vcpu->arch.sie_block->tecmc;
194 pgm_info.exc_access_id = vcpu->arch.sie_block->eai;
195 break;
196 default:
197 break;
200 if (vcpu->arch.sie_block->iprcc & PGM_PER) {
201 pgm_info.per_code = vcpu->arch.sie_block->perc;
202 pgm_info.per_atmid = vcpu->arch.sie_block->peratmid;
203 pgm_info.per_address = vcpu->arch.sie_block->peraddr;
204 pgm_info.per_access_id = vcpu->arch.sie_block->peraid;
206 return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
210 * restore ITDB to program-interruption TDB in guest lowcore
211 * and set TX abort indication if required
213 static int handle_itdb(struct kvm_vcpu *vcpu)
215 struct kvm_s390_itdb *itdb;
216 int rc;
218 if (!IS_TE_ENABLED(vcpu) || !IS_ITDB_VALID(vcpu))
219 return 0;
220 if (current->thread.per_flags & PER_FLAG_NO_TE)
221 return 0;
222 itdb = (struct kvm_s390_itdb *)vcpu->arch.sie_block->itdba;
223 rc = write_guest_lc(vcpu, __LC_PGM_TDB, itdb, sizeof(*itdb));
224 if (rc)
225 return rc;
226 memset(itdb, 0, sizeof(*itdb));
228 return 0;
231 #define per_event(vcpu) (vcpu->arch.sie_block->iprcc & PGM_PER)
233 static int handle_prog(struct kvm_vcpu *vcpu)
235 psw_t psw;
236 int rc;
238 vcpu->stat.exit_program_interruption++;
240 if (guestdbg_enabled(vcpu) && per_event(vcpu)) {
241 kvm_s390_handle_per_event(vcpu);
242 /* the interrupt might have been filtered out completely */
243 if (vcpu->arch.sie_block->iprcc == 0)
244 return 0;
247 trace_kvm_s390_intercept_prog(vcpu, vcpu->arch.sie_block->iprcc);
248 if (vcpu->arch.sie_block->iprcc == PGM_SPECIFICATION) {
249 rc = read_guest_lc(vcpu, __LC_PGM_NEW_PSW, &psw, sizeof(psw_t));
250 if (rc)
251 return rc;
252 /* Avoid endless loops of specification exceptions */
253 if (!is_valid_psw(&psw))
254 return -EOPNOTSUPP;
256 rc = handle_itdb(vcpu);
257 if (rc)
258 return rc;
260 return inject_prog_on_prog_intercept(vcpu);
264 * handle_external_interrupt - used for external interruption interceptions
266 * This interception only occurs if the CPUSTAT_EXT_INT bit was set, or if
267 * the new PSW does not have external interrupts disabled. In the first case,
268 * we've got to deliver the interrupt manually, and in the second case, we
269 * drop to userspace to handle the situation there.
271 static int handle_external_interrupt(struct kvm_vcpu *vcpu)
273 u16 eic = vcpu->arch.sie_block->eic;
274 struct kvm_s390_irq irq;
275 psw_t newpsw;
276 int rc;
278 vcpu->stat.exit_external_interrupt++;
280 rc = read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &newpsw, sizeof(psw_t));
281 if (rc)
282 return rc;
283 /* We can not handle clock comparator or timer interrupt with bad PSW */
284 if ((eic == EXT_IRQ_CLK_COMP || eic == EXT_IRQ_CPU_TIMER) &&
285 (newpsw.mask & PSW_MASK_EXT))
286 return -EOPNOTSUPP;
288 switch (eic) {
289 case EXT_IRQ_CLK_COMP:
290 irq.type = KVM_S390_INT_CLOCK_COMP;
291 break;
292 case EXT_IRQ_CPU_TIMER:
293 irq.type = KVM_S390_INT_CPU_TIMER;
294 break;
295 case EXT_IRQ_EXTERNAL_CALL:
296 irq.type = KVM_S390_INT_EXTERNAL_CALL;
297 irq.u.extcall.code = vcpu->arch.sie_block->extcpuaddr;
298 rc = kvm_s390_inject_vcpu(vcpu, &irq);
299 /* ignore if another external call is already pending */
300 if (rc == -EBUSY)
301 return 0;
302 return rc;
303 default:
304 return -EOPNOTSUPP;
307 return kvm_s390_inject_vcpu(vcpu, &irq);
311 * Handle MOVE PAGE partial execution interception.
313 * This interception can only happen for guests with DAT disabled and
314 * addresses that are currently not mapped in the host. Thus we try to
315 * set up the mappings for the corresponding user pages here (or throw
316 * addressing exceptions in case of illegal guest addresses).
318 static int handle_mvpg_pei(struct kvm_vcpu *vcpu)
320 unsigned long srcaddr, dstaddr;
321 int reg1, reg2, rc;
323 kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
325 /* Make sure that the source is paged-in */
326 rc = guest_translate_address(vcpu, vcpu->run->s.regs.gprs[reg2],
327 reg2, &srcaddr, GACC_FETCH);
328 if (rc)
329 return kvm_s390_inject_prog_cond(vcpu, rc);
330 rc = kvm_arch_fault_in_page(vcpu, srcaddr, 0);
331 if (rc != 0)
332 return rc;
334 /* Make sure that the destination is paged-in */
335 rc = guest_translate_address(vcpu, vcpu->run->s.regs.gprs[reg1],
336 reg1, &dstaddr, GACC_STORE);
337 if (rc)
338 return kvm_s390_inject_prog_cond(vcpu, rc);
339 rc = kvm_arch_fault_in_page(vcpu, dstaddr, 1);
340 if (rc != 0)
341 return rc;
343 kvm_s390_retry_instr(vcpu);
345 return 0;
348 static int handle_partial_execution(struct kvm_vcpu *vcpu)
350 vcpu->stat.exit_pei++;
352 if (vcpu->arch.sie_block->ipa == 0xb254) /* MVPG */
353 return handle_mvpg_pei(vcpu);
354 if (vcpu->arch.sie_block->ipa >> 8 == 0xae) /* SIGP */
355 return kvm_s390_handle_sigp_pei(vcpu);
357 return -EOPNOTSUPP;
360 static int handle_operexc(struct kvm_vcpu *vcpu)
362 vcpu->stat.exit_operation_exception++;
363 trace_kvm_s390_handle_operexc(vcpu, vcpu->arch.sie_block->ipa,
364 vcpu->arch.sie_block->ipb);
366 if (vcpu->arch.sie_block->ipa == 0xb256 &&
367 test_kvm_facility(vcpu->kvm, 74))
368 return handle_sthyi(vcpu);
370 if (vcpu->arch.sie_block->ipa == 0 && vcpu->kvm->arch.user_instr0)
371 return -EOPNOTSUPP;
373 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
376 int kvm_handle_sie_intercept(struct kvm_vcpu *vcpu)
378 int rc, per_rc = 0;
380 if (kvm_is_ucontrol(vcpu->kvm))
381 return -EOPNOTSUPP;
383 switch (vcpu->arch.sie_block->icptcode) {
384 case 0x10:
385 case 0x18:
386 return handle_noop(vcpu);
387 case 0x04:
388 rc = handle_instruction(vcpu);
389 break;
390 case 0x08:
391 return handle_prog(vcpu);
392 case 0x14:
393 return handle_external_interrupt(vcpu);
394 case 0x1c:
395 return kvm_s390_handle_wait(vcpu);
396 case 0x20:
397 return handle_validity(vcpu);
398 case 0x28:
399 return handle_stop(vcpu);
400 case 0x2c:
401 rc = handle_operexc(vcpu);
402 break;
403 case 0x38:
404 rc = handle_partial_execution(vcpu);
405 break;
406 default:
407 return -EOPNOTSUPP;
410 /* process PER, also if the instrution is processed in user space */
411 if (vcpu->arch.sie_block->icptstatus & 0x02 &&
412 (!rc || rc == -EOPNOTSUPP))
413 per_rc = kvm_s390_handle_per_ifetch_icpt(vcpu);
414 return per_rc ? per_rc : rc;