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
|
/*
* Block driver for s390 storage class memory.
*
* Copyright IBM Corp. 2012
* Author(s): Sebastian Ott <sebott@linux.vnet.ibm.com>
*/
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/blkdev.h>
#include <linux/genhd.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <asm/eadm.h>
#include "scm_blk.h"
static unsigned int write_cluster_size = 64;
module_param(write_cluster_size, uint, S_IRUGO);
MODULE_PARM_DESC(write_cluster_size,
"Number of pages used for contiguous writes.");
#define CLUSTER_SIZE (write_cluster_size * PAGE_SIZE)
void __scm_free_rq_cluster(struct scm_request *scmrq)
{
int i;
if (!scmrq->cluster.buf)
return;
for (i = 0; i < 2 * write_cluster_size; i++)
free_page((unsigned long) scmrq->cluster.buf[i]);
kfree(scmrq->cluster.buf);
}
int __scm_alloc_rq_cluster(struct scm_request *scmrq)
{
int i;
scmrq->cluster.buf = kzalloc(sizeof(void *) * 2 * write_cluster_size,
GFP_KERNEL);
if (!scmrq->cluster.buf)
return -ENOMEM;
for (i = 0; i < 2 * write_cluster_size; i++) {
scmrq->cluster.buf[i] = (void *) get_zeroed_page(GFP_DMA);
if (!scmrq->cluster.buf[i])
return -ENOMEM;
}
INIT_LIST_HEAD(&scmrq->cluster.list);
return 0;
}
void scm_request_cluster_init(struct scm_request *scmrq)
{
scmrq->cluster.state = CLUSTER_NONE;
}
static bool clusters_intersect(struct scm_request *A, struct scm_request *B)
{
unsigned long firstA, lastA, firstB, lastB;
firstA = ((u64) blk_rq_pos(A->request) << 9) / CLUSTER_SIZE;
lastA = (((u64) blk_rq_pos(A->request) << 9) +
blk_rq_bytes(A->request) - 1) / CLUSTER_SIZE;
firstB = ((u64) blk_rq_pos(B->request) << 9) / CLUSTER_SIZE;
lastB = (((u64) blk_rq_pos(B->request) << 9) +
blk_rq_bytes(B->request) - 1) / CLUSTER_SIZE;
return (firstB <= lastA && firstA <= lastB);
}
bool scm_reserve_cluster(struct scm_request *scmrq)
{
struct scm_blk_dev *bdev = scmrq->bdev;
struct scm_request *iter;
if (write_cluster_size == 0)
return true;
spin_lock(&bdev->lock);
list_for_each_entry(iter, &bdev->cluster_list, cluster.list) {
if (clusters_intersect(scmrq, iter) &&
(rq_data_dir(scmrq->request) == WRITE ||
rq_data_dir(iter->request) == WRITE)) {
spin_unlock(&bdev->lock);
return false;
}
}
list_add(&scmrq->cluster.list, &bdev->cluster_list);
spin_unlock(&bdev->lock);
return true;
}
void scm_release_cluster(struct scm_request *scmrq)
{
struct scm_blk_dev *bdev = scmrq->bdev;
unsigned long flags;
if (write_cluster_size == 0)
return;
spin_lock_irqsave(&bdev->lock, flags);
list_del(&scmrq->cluster.list);
spin_unlock_irqrestore(&bdev->lock, flags);
}
void scm_blk_dev_cluster_setup(struct scm_blk_dev *bdev)
{
INIT_LIST_HEAD(&bdev->cluster_list);
blk_queue_io_opt(bdev->rq, CLUSTER_SIZE);
}
static void scm_prepare_cluster_request(struct scm_request *scmrq)
{
struct scm_blk_dev *bdev = scmrq->bdev;
struct scm_device *scmdev = bdev->gendisk->private_data;
struct request *req = scmrq->request;
struct aidaw *aidaw = scmrq->aidaw;
struct msb *msb = &scmrq->aob->msb[0];
struct req_iterator iter;
struct bio_vec *bv;
int i = 0;
u64 addr;
switch (scmrq->cluster.state) {
case CLUSTER_NONE:
scmrq->cluster.state = CLUSTER_READ;
/* fall through */
case CLUSTER_READ:
scmrq->aob->request.msb_count = 1;
msb->bs = MSB_BS_4K;
msb->oc = MSB_OC_READ;
msb->flags = MSB_FLAG_IDA;
msb->data_addr = (u64) aidaw;
msb->blk_count = write_cluster_size;
addr = scmdev->address + ((u64) blk_rq_pos(req) << 9);
msb->scm_addr = round_down(addr, CLUSTER_SIZE);
if (msb->scm_addr !=
round_down(addr + (u64) blk_rq_bytes(req) - 1,
CLUSTER_SIZE))
msb->blk_count = 2 * write_cluster_size;
for (i = 0; i < msb->blk_count; i++) {
aidaw->data_addr = (u64) scmrq->cluster.buf[i];
aidaw++;
}
break;
case CLUSTER_WRITE:
msb->oc = MSB_OC_WRITE;
for (addr = msb->scm_addr;
addr < scmdev->address + ((u64) blk_rq_pos(req) << 9);
addr += PAGE_SIZE) {
aidaw->data_addr = (u64) scmrq->cluster.buf[i];
aidaw++;
i++;
}
rq_for_each_segment(bv, req, iter) {
aidaw->data_addr = (u64) page_address(bv->bv_page);
aidaw++;
i++;
}
for (; i < msb->blk_count; i++) {
aidaw->data_addr = (u64) scmrq->cluster.buf[i];
aidaw++;
}
break;
}
}
bool scm_need_cluster_request(struct scm_request *scmrq)
{
if (rq_data_dir(scmrq->request) == READ)
return false;
return blk_rq_bytes(scmrq->request) < CLUSTER_SIZE;
}
/* Called with queue lock held. */
void scm_initiate_cluster_request(struct scm_request *scmrq)
{
scm_prepare_cluster_request(scmrq);
if (scm_start_aob(scmrq->aob))
scm_request_requeue(scmrq);
}
bool scm_test_cluster_request(struct scm_request *scmrq)
{
return scmrq->cluster.state != CLUSTER_NONE;
}
void scm_cluster_request_irq(struct scm_request *scmrq)
{
struct scm_blk_dev *bdev = scmrq->bdev;
unsigned long flags;
switch (scmrq->cluster.state) {
case CLUSTER_NONE:
BUG();
break;
case CLUSTER_READ:
if (scmrq->error) {
scm_request_finish(scmrq);
break;
}
scmrq->cluster.state = CLUSTER_WRITE;
spin_lock_irqsave(&bdev->rq_lock, flags);
scm_initiate_cluster_request(scmrq);
spin_unlock_irqrestore(&bdev->rq_lock, flags);
break;
case CLUSTER_WRITE:
scm_request_finish(scmrq);
break;
}
}
bool scm_cluster_size_valid(void)
{
return write_cluster_size == 0 || write_cluster_size == 32 ||
write_cluster_size == 64 || write_cluster_size == 128;
}
|
