summaryrefslogtreecommitdiff
path: root/arch/powerpc/kvm/book3s_hv_ras.c
blob: b11043b23c185be659ae82f0648eb1c343342128 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
/*
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License, version 2, as
 * published by the Free Software Foundation.
 *
 * Copyright 2012 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
 */

#include <linux/types.h>
#include <linux/string.h>
#include <linux/kvm.h>
#include <linux/kvm_host.h>
#include <linux/kernel.h>
#include <asm/opal.h>
#include <asm/mce.h>
#include <asm/machdep.h>
#include <asm/cputhreads.h>
#include <asm/hmi.h>
#include <asm/kvm_ppc.h>

/* SRR1 bits for machine check on POWER7 */
#define SRR1_MC_LDSTERR		(1ul << (63-42))
#define SRR1_MC_IFETCH_SH	(63-45)
#define SRR1_MC_IFETCH_MASK	0x7
#define SRR1_MC_IFETCH_SLBPAR		2	/* SLB parity error */
#define SRR1_MC_IFETCH_SLBMULTI		3	/* SLB multi-hit */
#define SRR1_MC_IFETCH_SLBPARMULTI	4	/* SLB parity + multi-hit */
#define SRR1_MC_IFETCH_TLBMULTI		5	/* I-TLB multi-hit */

/* DSISR bits for machine check on POWER7 */
#define DSISR_MC_DERAT_MULTI	0x800		/* D-ERAT multi-hit */
#define DSISR_MC_TLB_MULTI	0x400		/* D-TLB multi-hit */
#define DSISR_MC_SLB_PARITY	0x100		/* SLB parity error */
#define DSISR_MC_SLB_MULTI	0x080		/* SLB multi-hit */
#define DSISR_MC_SLB_PARMULTI	0x040		/* SLB parity + multi-hit */

/* POWER7 SLB flush and reload */
static void reload_slb(struct kvm_vcpu *vcpu)
{
	struct slb_shadow *slb;
	unsigned long i, n;

	/* First clear out SLB */
	asm volatile("slbmte %0,%0; slbia" : : "r" (0));

	/* Do they have an SLB shadow buffer registered? */
	slb = vcpu->arch.slb_shadow.pinned_addr;
	if (!slb)
		return;

	/* Sanity check */
	n = min_t(u32, be32_to_cpu(slb->persistent), SLB_MIN_SIZE);
	if ((void *) &slb->save_area[n] > vcpu->arch.slb_shadow.pinned_end)
		return;

	/* Load up the SLB from that */
	for (i = 0; i < n; ++i) {
		unsigned long rb = be64_to_cpu(slb->save_area[i].esid);
		unsigned long rs = be64_to_cpu(slb->save_area[i].vsid);

		rb = (rb & ~0xFFFul) | i;	/* insert entry number */
		asm volatile("slbmte %0,%1" : : "r" (rs), "r" (rb));
	}
}

/*
 * On POWER7, see if we can handle a machine check that occurred inside
 * the guest in real mode, without switching to the host partition.
 *
 * Returns: 0 => exit guest, 1 => deliver machine check to guest
 */
static long kvmppc_realmode_mc_power7(struct kvm_vcpu *vcpu)
{
	unsigned long srr1 = vcpu->arch.shregs.msr;
	struct machine_check_event mce_evt;
	long handled = 1;

	if (srr1 & SRR1_MC_LDSTERR) {
		/* error on load/store */
		unsigned long dsisr = vcpu->arch.shregs.dsisr;

		if (dsisr & (DSISR_MC_SLB_PARMULTI | DSISR_MC_SLB_MULTI |
			     DSISR_MC_SLB_PARITY | DSISR_MC_DERAT_MULTI)) {
			/* flush and reload SLB; flushes D-ERAT too */
			reload_slb(vcpu);
			dsisr &= ~(DSISR_MC_SLB_PARMULTI | DSISR_MC_SLB_MULTI |
				   DSISR_MC_SLB_PARITY | DSISR_MC_DERAT_MULTI);
		}
		if (dsisr & DSISR_MC_TLB_MULTI) {
			tlbiel_all_lpid(vcpu->kvm->arch.radix);
			dsisr &= ~DSISR_MC_TLB_MULTI;
		}
		/* Any other errors we don't understand? */
		if (dsisr & 0xffffffffUL)
			handled = 0;
	}

	switch ((srr1 >> SRR1_MC_IFETCH_SH) & SRR1_MC_IFETCH_MASK) {
	case 0:
		break;
	case SRR1_MC_IFETCH_SLBPAR:
	case SRR1_MC_IFETCH_SLBMULTI:
	case SRR1_MC_IFETCH_SLBPARMULTI:
		reload_slb(vcpu);
		break;
	case SRR1_MC_IFETCH_TLBMULTI:
		tlbiel_all_lpid(vcpu->kvm->arch.radix);
		break;
	default:
		handled = 0;
	}

	/*
	 * See if we have already handled the condition in the linux host.
	 * We assume that if the condition is recovered then linux host
	 * will have generated an error log event that we will pick
	 * up and log later.
	 * Don't release mce event now. We will queue up the event so that
	 * we can log the MCE event info on host console.
	 */
	if (!get_mce_event(&mce_evt, MCE_EVENT_DONTRELEASE))
		goto out;

	if (mce_evt.version == MCE_V1 &&
	    (mce_evt.severity == MCE_SEV_NO_ERROR ||
	     mce_evt.disposition == MCE_DISPOSITION_RECOVERED))
		handled = 1;

out:
	/*
	 * For guest that supports FWNMI capability, hook the MCE event into
	 * vcpu structure. We are going to exit the guest with KVM_EXIT_NMI
	 * exit reason. On our way to exit we will pull this event from vcpu
	 * structure and print it from thread 0 of the core/subcore.
	 *
	 * For guest that does not support FWNMI capability (old QEMU):
	 * We are now going enter guest either through machine check
	 * interrupt (for unhandled errors) or will continue from
	 * current HSRR0 (for handled errors) in guest. Hence
	 * queue up the event so that we can log it from host console later.
	 */
	if (vcpu->kvm->arch.fwnmi_enabled) {
		/*
		 * Hook up the mce event on to vcpu structure.
		 * First clear the old event.
		 */
		memset(&vcpu->arch.mce_evt, 0, sizeof(vcpu->arch.mce_evt));
		if (get_mce_event(&mce_evt, MCE_EVENT_RELEASE)) {
			vcpu->arch.mce_evt = mce_evt;
		}
	} else
		machine_check_queue_event();

	return handled;
}

long kvmppc_realmode_machine_check(struct kvm_vcpu *vcpu)
{
	return kvmppc_realmode_mc_power7(vcpu);
}

/* Check if dynamic split is in force and return subcore size accordingly. */
static inline int kvmppc_cur_subcore_size(void)
{
	if (local_paca->kvm_hstate.kvm_split_mode)
		return local_paca->kvm_hstate.kvm_split_mode->subcore_size;

	return threads_per_subcore;
}

void kvmppc_subcore_enter_guest(void)
{
	int thread_id, subcore_id;

	thread_id = cpu_thread_in_core(local_paca->paca_index);
	subcore_id = thread_id / kvmppc_cur_subcore_size();

	local_paca->sibling_subcore_state->in_guest[subcore_id] = 1;
}

void kvmppc_subcore_exit_guest(void)
{
	int thread_id, subcore_id;

	thread_id = cpu_thread_in_core(local_paca->paca_index);
	subcore_id = thread_id / kvmppc_cur_subcore_size();

	local_paca->sibling_subcore_state->in_guest[subcore_id] = 0;
}

static bool kvmppc_tb_resync_required(void)
{
	if (test_and_set_bit(CORE_TB_RESYNC_REQ_BIT,
				&local_paca->sibling_subcore_state->flags))
		return false;

	return true;
}

static void kvmppc_tb_resync_done(void)
{
	clear_bit(CORE_TB_RESYNC_REQ_BIT,
			&local_paca->sibling_subcore_state->flags);
}

/*
 * kvmppc_realmode_hmi_handler() is called only by primary thread during
 * guest exit path.
 *
 * There are multiple reasons why HMI could occur, one of them is
 * Timebase (TB) error. If this HMI is due to TB error, then TB would
 * have been in stopped state. The opal hmi handler Will fix it and
 * restore the TB value with host timebase value. For HMI caused due
 * to non-TB errors, opal hmi handler will not touch/restore TB register
 * and hence there won't be any change in TB value.
 *
 * Since we are not sure about the cause of this HMI, we can't be sure
 * about the content of TB register whether it holds guest or host timebase
 * value. Hence the idea is to resync the TB on every HMI, so that we
 * know about the exact state of the TB value. Resync TB call will
 * restore TB to host timebase.
 *
 * Things to consider:
 * - On TB error, HMI interrupt is reported on all the threads of the core
 *   that has encountered TB error irrespective of split-core mode.
 * - The very first thread on the core that get chance to fix TB error
 *   would rsync the TB with local chipTOD value.
 * - The resync TB is a core level action i.e. it will sync all the TBs
 *   in that core independent of split-core mode. This means if we trigger
 *   TB sync from a thread from one subcore, it would affect TB values of
 *   sibling subcores of the same core.
 *
 * All threads need to co-ordinate before making opal hmi handler.
 * All threads will use sibling_subcore_state->in_guest[] (shared by all
 * threads in the core) in paca which holds information about whether
 * sibling subcores are in Guest mode or host mode. The in_guest[] array
 * is of size MAX_SUBCORE_PER_CORE=4, indexed using subcore id to set/unset
 * subcore status. Only primary threads from each subcore is responsible
 * to set/unset its designated array element while entering/exiting the
 * guset.
 *
 * After invoking opal hmi handler call, one of the thread (of entire core)
 * will need to resync the TB. Bit 63 from subcore state bitmap flags
 * (sibling_subcore_state->flags) will be used to co-ordinate between
 * primary threads to decide who takes up the responsibility.
 *
 * This is what we do:
 * - Primary thread from each subcore tries to set resync required bit[63]
 *   of paca->sibling_subcore_state->flags.
 * - The first primary thread that is able to set the flag takes the
 *   responsibility of TB resync. (Let us call it as thread leader)
 * - All other threads which are in host will call
 *   wait_for_subcore_guest_exit() and wait for in_guest[0-3] from
 *   paca->sibling_subcore_state to get cleared.
 * - All the primary thread will clear its subcore status from subcore
 *   state in_guest[] array respectively.
 * - Once all primary threads clear in_guest[0-3], all of them will invoke
 *   opal hmi handler.
 * - Now all threads will wait for TB resync to complete by invoking
 *   wait_for_tb_resync() except the thread leader.
 * - Thread leader will do a TB resync by invoking opal_resync_timebase()
 *   call and the it will clear the resync required bit.
 * - All other threads will now come out of resync wait loop and proceed
 *   with individual execution.
 * - On return of this function, primary thread will signal all
 *   secondary threads to proceed.
 * - All secondary threads will eventually call opal hmi handler on
 *   their exit path.
 *
 * Returns 1 if the timebase offset should be applied, 0 if not.
 */

long kvmppc_realmode_hmi_handler(void)
{
	bool resync_req;

	__this_cpu_inc(irq_stat.hmi_exceptions);

	if (hmi_handle_debugtrig(NULL) >= 0)
		return 1;

	/*
	 * By now primary thread has already completed guest->host
	 * partition switch but haven't signaled secondaries yet.
	 * All the secondary threads on this subcore is waiting
	 * for primary thread to signal them to go ahead.
	 *
	 * For threads from subcore which isn't in guest, they all will
	 * wait until all other subcores on this core exit the guest.
	 *
	 * Now set the resync required bit. If you are the first to
	 * set this bit then kvmppc_tb_resync_required() function will
	 * return true. For rest all other subcores
	 * kvmppc_tb_resync_required() will return false.
	 *
	 * If resync_req == true, then this thread is responsible to
	 * initiate TB resync after hmi handler has completed.
	 * All other threads on this core will wait until this thread
	 * clears the resync required bit flag.
	 */
	resync_req = kvmppc_tb_resync_required();

	/* Reset the subcore status to indicate it has exited guest */
	kvmppc_subcore_exit_guest();

	/*
	 * Wait for other subcores on this core to exit the guest.
	 * All the primary threads and threads from subcore that are
	 * not in guest will wait here until all subcores are out
	 * of guest context.
	 */
	wait_for_subcore_guest_exit();

	/*
	 * At this point we are sure that primary threads from each
	 * subcore on this core have completed guest->host partition
	 * switch. Now it is safe to call HMI handler.
	 */
	if (ppc_md.hmi_exception_early)
		ppc_md.hmi_exception_early(NULL);

	/*
	 * Check if this thread is responsible to resync TB.
	 * All other threads will wait until this thread completes the
	 * TB resync.
	 */
	if (resync_req) {
		opal_resync_timebase();
		/* Reset TB resync req bit */
		kvmppc_tb_resync_done();
	} else {
		wait_for_tb_resync();
	}
	return 0;
}