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
|
/*
* Copyright (C) 2001 Mike Corrigan IBM Corporation
*
* 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/stddef.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/bootmem.h>
#include <linux/seq_file.h>
#include <linux/proc_fs.h>
#include <asm/system.h>
#include <asm/paca.h>
#include <asm/iSeries/ItLpQueue.h>
#include <asm/iSeries/HvLpEvent.h>
#include <asm/iSeries/HvCallEvent.h>
#include <asm/iSeries/ItLpNaca.h>
/*
* The LpQueue is used to pass event data from the hypervisor to
* the partition. This is where I/O interrupt events are communicated.
*
* It is written to by the hypervisor so cannot end up in the BSS.
*/
struct hvlpevent_queue hvlpevent_queue __attribute__((__section__(".data")));
DEFINE_PER_CPU(unsigned long[HvLpEvent_Type_NumTypes], hvlpevent_counts);
static char *event_types[HvLpEvent_Type_NumTypes] = {
"Hypervisor",
"Machine Facilities",
"Session Manager",
"SPD I/O",
"Virtual Bus",
"PCI I/O",
"RIO I/O",
"Virtual Lan",
"Virtual I/O"
};
/* Array of LpEvent handler functions */
static LpEventHandler lpEventHandler[HvLpEvent_Type_NumTypes];
static unsigned lpEventHandlerPaths[HvLpEvent_Type_NumTypes];
static struct HvLpEvent * get_next_hvlpevent(void)
{
struct HvLpEvent * event;
event = (struct HvLpEvent *)hvlpevent_queue.xSlicCurEventPtr;
if (event->xFlags.xValid) {
/* rmb() needed only for weakly consistent machines (regatta) */
rmb();
/* Set pointer to next potential event */
hvlpevent_queue.xSlicCurEventPtr += ((event->xSizeMinus1 +
LpEventAlign) / LpEventAlign) * LpEventAlign;
/* Wrap to beginning if no room at end */
if (hvlpevent_queue.xSlicCurEventPtr >
hvlpevent_queue.xSlicLastValidEventPtr) {
hvlpevent_queue.xSlicCurEventPtr =
hvlpevent_queue.xSlicEventStackPtr;
}
} else {
event = NULL;
}
return event;
}
static unsigned long spread_lpevents = NR_CPUS;
int hvlpevent_is_pending(void)
{
struct HvLpEvent *next_event;
if (smp_processor_id() >= spread_lpevents)
return 0;
next_event = (struct HvLpEvent *)hvlpevent_queue.xSlicCurEventPtr;
return next_event->xFlags.xValid |
hvlpevent_queue.xPlicOverflowIntPending;
}
static void hvlpevent_clear_valid(struct HvLpEvent * event)
{
/* Tell the Hypervisor that we're done with this event.
* Also clear bits within this event that might look like valid bits.
* ie. on 64-byte boundaries.
*/
struct HvLpEvent *tmp;
unsigned extra = ((event->xSizeMinus1 + LpEventAlign) /
LpEventAlign) - 1;
switch (extra) {
case 3:
tmp = (struct HvLpEvent*)((char*)event + 3 * LpEventAlign);
tmp->xFlags.xValid = 0;
case 2:
tmp = (struct HvLpEvent*)((char*)event + 2 * LpEventAlign);
tmp->xFlags.xValid = 0;
case 1:
tmp = (struct HvLpEvent*)((char*)event + 1 * LpEventAlign);
tmp->xFlags.xValid = 0;
}
mb();
event->xFlags.xValid = 0;
}
void process_hvlpevents(struct pt_regs *regs)
{
struct HvLpEvent * event;
/* If we have recursed, just return */
if (!spin_trylock(&hvlpevent_queue.lock))
return;
for (;;) {
event = get_next_hvlpevent();
if (event) {
/* Call appropriate handler here, passing
* a pointer to the LpEvent. The handler
* must make a copy of the LpEvent if it
* needs it in a bottom half. (perhaps for
* an ACK)
*
* Handlers are responsible for ACK processing
*
* The Hypervisor guarantees that LpEvents will
* only be delivered with types that we have
* registered for, so no type check is necessary
* here!
*/
if (event->xType < HvLpEvent_Type_NumTypes)
__get_cpu_var(hvlpevent_counts)[event->xType]++;
if (event->xType < HvLpEvent_Type_NumTypes &&
lpEventHandler[event->xType])
lpEventHandler[event->xType](event, regs);
else
printk(KERN_INFO "Unexpected Lp Event type=%d\n", event->xType );
hvlpevent_clear_valid(event);
} else if (hvlpevent_queue.xPlicOverflowIntPending)
/*
* No more valid events. If overflow events are
* pending process them
*/
HvCallEvent_getOverflowLpEvents(hvlpevent_queue.xIndex);
else
break;
}
spin_unlock(&hvlpevent_queue.lock);
}
static int set_spread_lpevents(char *str)
{
unsigned long val = simple_strtoul(str, NULL, 0);
/*
* The parameter is the number of processors to share in processing
* lp events.
*/
if (( val > 0) && (val <= NR_CPUS)) {
spread_lpevents = val;
printk("lpevent processing spread over %ld processors\n", val);
} else {
printk("invalid spread_lpevents %ld\n", val);
}
return 1;
}
__setup("spread_lpevents=", set_spread_lpevents);
void setup_hvlpevent_queue(void)
{
void *eventStack;
/*
* Allocate a page for the Event Stack. The Hypervisor needs the
* absolute real address, so we subtract out the KERNELBASE and add
* in the absolute real address of the kernel load area.
*/
eventStack = alloc_bootmem_pages(LpEventStackSize);
memset(eventStack, 0, LpEventStackSize);
/* Invoke the hypervisor to initialize the event stack */
HvCallEvent_setLpEventStack(0, eventStack, LpEventStackSize);
hvlpevent_queue.xSlicEventStackPtr = (char *)eventStack;
hvlpevent_queue.xSlicCurEventPtr = (char *)eventStack;
hvlpevent_queue.xSlicLastValidEventPtr = (char *)eventStack +
(LpEventStackSize - LpEventMaxSize);
hvlpevent_queue.xIndex = 0;
}
/* Register a handler for an LpEvent type */
int HvLpEvent_registerHandler(HvLpEvent_Type eventType, LpEventHandler handler)
{
if (eventType < HvLpEvent_Type_NumTypes) {
lpEventHandler[eventType] = handler;
return 0;
}
return 1;
}
EXPORT_SYMBOL(HvLpEvent_registerHandler);
int HvLpEvent_unregisterHandler(HvLpEvent_Type eventType)
{
might_sleep();
if (eventType < HvLpEvent_Type_NumTypes) {
if (!lpEventHandlerPaths[eventType]) {
lpEventHandler[eventType] = NULL;
/*
* We now sleep until all other CPUs have scheduled.
* This ensures that the deletion is seen by all
* other CPUs, and that the deleted handler isn't
* still running on another CPU when we return.
*/
synchronize_rcu();
return 0;
}
}
return 1;
}
EXPORT_SYMBOL(HvLpEvent_unregisterHandler);
/*
* lpIndex is the partition index of the target partition.
* needed only for VirtualIo, VirtualLan and SessionMgr. Zero
* indicates to use our partition index - for the other types.
*/
int HvLpEvent_openPath(HvLpEvent_Type eventType, HvLpIndex lpIndex)
{
if ((eventType < HvLpEvent_Type_NumTypes) &&
lpEventHandler[eventType]) {
if (lpIndex == 0)
lpIndex = itLpNaca.xLpIndex;
HvCallEvent_openLpEventPath(lpIndex, eventType);
++lpEventHandlerPaths[eventType];
return 0;
}
return 1;
}
int HvLpEvent_closePath(HvLpEvent_Type eventType, HvLpIndex lpIndex)
{
if ((eventType < HvLpEvent_Type_NumTypes) &&
lpEventHandler[eventType] &&
lpEventHandlerPaths[eventType]) {
if (lpIndex == 0)
lpIndex = itLpNaca.xLpIndex;
HvCallEvent_closeLpEventPath(lpIndex, eventType);
--lpEventHandlerPaths[eventType];
return 0;
}
return 1;
}
static int proc_lpevents_show(struct seq_file *m, void *v)
{
int cpu, i;
unsigned long sum;
static unsigned long cpu_totals[NR_CPUS];
/* FIXME: do we care that there's no locking here? */
sum = 0;
for_each_online_cpu(cpu) {
cpu_totals[cpu] = 0;
for (i = 0; i < HvLpEvent_Type_NumTypes; i++) {
cpu_totals[cpu] += per_cpu(hvlpevent_counts, cpu)[i];
}
sum += cpu_totals[cpu];
}
seq_printf(m, "LpEventQueue 0\n");
seq_printf(m, " events processed:\t%lu\n", sum);
for (i = 0; i < HvLpEvent_Type_NumTypes; ++i) {
sum = 0;
for_each_online_cpu(cpu) {
sum += per_cpu(hvlpevent_counts, cpu)[i];
}
seq_printf(m, " %-20s %10lu\n", event_types[i], sum);
}
seq_printf(m, "\n events processed by processor:\n");
for_each_online_cpu(cpu) {
seq_printf(m, " CPU%02d %10lu\n", cpu, cpu_totals[cpu]);
}
return 0;
}
static int proc_lpevents_open(struct inode *inode, struct file *file)
{
return single_open(file, proc_lpevents_show, NULL);
}
static struct file_operations proc_lpevents_operations = {
.open = proc_lpevents_open,
.read = seq_read,
.llseek = seq_lseek,
.release = single_release,
};
static int __init proc_lpevents_init(void)
{
struct proc_dir_entry *e;
e = create_proc_entry("iSeries/lpevents", S_IFREG|S_IRUGO, NULL);
if (e)
e->proc_fops = &proc_lpevents_operations;
return 0;
}
__initcall(proc_lpevents_init);
|