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
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
|
/*
* Copyright 2014 IBM Corp.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#include <linux/interrupt.h>
#include <linux/workqueue.h>
#include <linux/sched.h>
#include <linux/wait.h>
#include <linux/slab.h>
#include <linux/pid.h>
#include <asm/cputable.h>
#include <misc/cxl-base.h>
#include "cxl.h"
#include "trace.h"
static int afu_irq_range_start(void)
{
if (cpu_has_feature(CPU_FTR_HVMODE))
return 1;
return 0;
}
static irqreturn_t schedule_cxl_fault(struct cxl_context *ctx, u64 dsisr, u64 dar)
{
ctx->dsisr = dsisr;
ctx->dar = dar;
schedule_work(&ctx->fault_work);
return IRQ_HANDLED;
}
irqreturn_t cxl_irq_psl9(int irq, struct cxl_context *ctx, struct cxl_irq_info *irq_info)
{
u64 dsisr, dar;
dsisr = irq_info->dsisr;
dar = irq_info->dar;
trace_cxl_psl9_irq(ctx, irq, dsisr, dar);
pr_devel("CXL interrupt %i for afu pe: %i DSISR: %#llx DAR: %#llx\n", irq, ctx->pe, dsisr, dar);
if (dsisr & CXL_PSL9_DSISR_An_TF) {
pr_devel("CXL interrupt: Scheduling translation fault handling for later (pe: %i)\n", ctx->pe);
return schedule_cxl_fault(ctx, dsisr, dar);
}
if (dsisr & CXL_PSL9_DSISR_An_PE)
return cxl_ops->handle_psl_slice_error(ctx, dsisr,
irq_info->errstat);
if (dsisr & CXL_PSL9_DSISR_An_AE) {
pr_devel("CXL interrupt: AFU Error 0x%016llx\n", irq_info->afu_err);
if (ctx->pending_afu_err) {
/*
* This shouldn't happen - the PSL treats these errors
* as fatal and will have reset the AFU, so there's not
* much point buffering multiple AFU errors.
* OTOH if we DO ever see a storm of these come in it's
* probably best that we log them somewhere:
*/
dev_err_ratelimited(&ctx->afu->dev, "CXL AFU Error undelivered to pe %i: 0x%016llx\n",
ctx->pe, irq_info->afu_err);
} else {
spin_lock(&ctx->lock);
ctx->afu_err = irq_info->afu_err;
ctx->pending_afu_err = 1;
spin_unlock(&ctx->lock);
wake_up_all(&ctx->wq);
}
cxl_ops->ack_irq(ctx, CXL_PSL_TFC_An_A, 0);
return IRQ_HANDLED;
}
if (dsisr & CXL_PSL9_DSISR_An_OC)
pr_devel("CXL interrupt: OS Context Warning\n");
WARN(1, "Unhandled CXL PSL IRQ\n");
return IRQ_HANDLED;
}
irqreturn_t cxl_irq_psl8(int irq, struct cxl_context *ctx, struct cxl_irq_info *irq_info)
{
u64 dsisr, dar;
dsisr = irq_info->dsisr;
dar = irq_info->dar;
trace_cxl_psl_irq(ctx, irq, dsisr, dar);
pr_devel("CXL interrupt %i for afu pe: %i DSISR: %#llx DAR: %#llx\n", irq, ctx->pe, dsisr, dar);
if (dsisr & CXL_PSL_DSISR_An_DS) {
/*
* We don't inherently need to sleep to handle this, but we do
* need to get a ref to the task's mm, which we can't do from
* irq context without the potential for a deadlock since it
* takes the task_lock. An alternate option would be to keep a
* reference to the task's mm the entire time it has cxl open,
* but to do that we need to solve the issue where we hold a
* ref to the mm, but the mm can hold a ref to the fd after an
* mmap preventing anything from being cleaned up.
*/
pr_devel("Scheduling segment miss handling for later pe: %i\n", ctx->pe);
return schedule_cxl_fault(ctx, dsisr, dar);
}
if (dsisr & CXL_PSL_DSISR_An_M)
pr_devel("CXL interrupt: PTE not found\n");
if (dsisr & CXL_PSL_DSISR_An_P)
pr_devel("CXL interrupt: Storage protection violation\n");
if (dsisr & CXL_PSL_DSISR_An_A)
pr_devel("CXL interrupt: AFU lock access to write through or cache inhibited storage\n");
if (dsisr & CXL_PSL_DSISR_An_S)
pr_devel("CXL interrupt: Access was afu_wr or afu_zero\n");
if (dsisr & CXL_PSL_DSISR_An_K)
pr_devel("CXL interrupt: Access not permitted by virtual page class key protection\n");
if (dsisr & CXL_PSL_DSISR_An_DM) {
/*
* In some cases we might be able to handle the fault
* immediately if hash_page would succeed, but we still need
* the task's mm, which as above we can't get without a lock
*/
pr_devel("Scheduling page fault handling for later pe: %i\n", ctx->pe);
return schedule_cxl_fault(ctx, dsisr, dar);
}
if (dsisr & CXL_PSL_DSISR_An_ST)
WARN(1, "CXL interrupt: Segment Table PTE not found\n");
if (dsisr & CXL_PSL_DSISR_An_UR)
pr_devel("CXL interrupt: AURP PTE not found\n");
if (dsisr & CXL_PSL_DSISR_An_PE)
return cxl_ops->handle_psl_slice_error(ctx, dsisr,
irq_info->errstat);
if (dsisr & CXL_PSL_DSISR_An_AE) {
pr_devel("CXL interrupt: AFU Error 0x%016llx\n", irq_info->afu_err);
if (ctx->pending_afu_err) {
/*
* This shouldn't happen - the PSL treats these errors
* as fatal and will have reset the AFU, so there's not
* much point buffering multiple AFU errors.
* OTOH if we DO ever see a storm of these come in it's
* probably best that we log them somewhere:
*/
dev_err_ratelimited(&ctx->afu->dev, "CXL AFU Error "
"undelivered to pe %i: 0x%016llx\n",
ctx->pe, irq_info->afu_err);
} else {
spin_lock(&ctx->lock);
ctx->afu_err = irq_info->afu_err;
ctx->pending_afu_err = true;
spin_unlock(&ctx->lock);
wake_up_all(&ctx->wq);
}
cxl_ops->ack_irq(ctx, CXL_PSL_TFC_An_A, 0);
return IRQ_HANDLED;
}
if (dsisr & CXL_PSL_DSISR_An_OC)
pr_devel("CXL interrupt: OS Context Warning\n");
WARN(1, "Unhandled CXL PSL IRQ\n");
return IRQ_HANDLED;
}
static irqreturn_t cxl_irq_afu(int irq, void *data)
{
struct cxl_context *ctx = data;
irq_hw_number_t hwirq = irqd_to_hwirq(irq_get_irq_data(irq));
int irq_off, afu_irq = 0;
__u16 range;
int r;
/*
* Look for the interrupt number.
* On bare-metal, we know range 0 only contains the PSL
* interrupt so we could start counting at range 1 and initialize
* afu_irq at 1.
* In a guest, range 0 also contains AFU interrupts, so it must
* be counted for. Therefore we initialize afu_irq at 0 to take into
* account the PSL interrupt.
*
* For code-readability, it just seems easier to go over all
* the ranges on bare-metal and guest. The end result is the same.
*/
for (r = 0; r < CXL_IRQ_RANGES; r++) {
irq_off = hwirq - ctx->irqs.offset[r];
range = ctx->irqs.range[r];
if (irq_off >= 0 && irq_off < range) {
afu_irq += irq_off;
break;
}
afu_irq += range;
}
if (unlikely(r >= CXL_IRQ_RANGES)) {
WARN(1, "Received AFU IRQ out of range for pe %i (virq %i hwirq %lx)\n",
ctx->pe, irq, hwirq);
return IRQ_HANDLED;
}
trace_cxl_afu_irq(ctx, afu_irq, irq, hwirq);
pr_devel("Received AFU interrupt %i for pe: %i (virq %i hwirq %lx)\n",
afu_irq, ctx->pe, irq, hwirq);
if (unlikely(!ctx->irq_bitmap)) {
WARN(1, "Received AFU IRQ for context with no IRQ bitmap\n");
return IRQ_HANDLED;
}
spin_lock(&ctx->lock);
set_bit(afu_irq - 1, ctx->irq_bitmap);
ctx->pending_irq = true;
spin_unlock(&ctx->lock);
wake_up_all(&ctx->wq);
return IRQ_HANDLED;
}
unsigned int cxl_map_irq(struct cxl *adapter, irq_hw_number_t hwirq,
irq_handler_t handler, void *cookie, const char *name)
{
unsigned int virq;
int result;
/* IRQ Domain? */
virq = irq_create_mapping(NULL, hwirq);
if (!virq) {
dev_warn(&adapter->dev, "cxl_map_irq: irq_create_mapping failed\n");
return 0;
}
if (cxl_ops->setup_irq)
cxl_ops->setup_irq(adapter, hwirq, virq);
pr_devel("hwirq %#lx mapped to virq %u\n", hwirq, virq);
result = request_irq(virq, handler, 0, name, cookie);
if (result) {
dev_warn(&adapter->dev,
|