summaryrefslogtreecommitdiff
path: root/drivers/char/bsr.c
blob: 70d31aed9011e0ca46c569ecdca2423b80db6fe0 (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
336
337
338
339
340
341
342
343
344
345
346
// SPDX-License-Identifier: GPL-2.0-or-later
/* IBM POWER Barrier Synchronization Register Driver
 *
 * Copyright IBM Corporation 2008
 *
 * Author: Sonny Rao <sonnyrao@us.ibm.com>
 */

#include <linux/device.h>
#include <linux/kernel.h>
#include <linux/of.h>
#include <linux/of_address.h>
#include <linux/fs.h>
#include <linux/module.h>
#include <linux/cdev.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <asm/io.h>

/*
 This driver exposes a special register which can be used for fast
 synchronization across a large SMP machine.  The hardware is exposed
 as an array of bytes where each process will write to one of the bytes to
 indicate it has finished the current stage and this update is broadcast to
 all processors without having to bounce a cacheline between them. In
 POWER5 and POWER6 there is one of these registers per SMP,  but it is
 presented in two forms; first, it is given as a whole and then as a number
 of smaller registers which alias to parts of the single whole register.
 This can potentially allow multiple groups of processes to each have their
 own private synchronization device.

 Note that this hardware *must* be written to using *only* single byte writes.
 It may be read using 1, 2, 4, or 8 byte loads which must be aligned since
 this region is treated as cache-inhibited  processes should also use a
 full sync before and after writing to the BSR to ensure all stores and
 the BSR update have made it to all chips in the system
*/

/* This is arbitrary number, up to Power6 it's been 17 or fewer  */
#define BSR_MAX_DEVS (32)

struct bsr_dev {
	u64      bsr_addr;     /* Real address */
	u64      bsr_len;      /* length of mem region we can map */
	unsigned bsr_bytes;    /* size of the BSR reg itself */
	unsigned bsr_stride;   /* interval at which BSR repeats in the page */
	unsigned bsr_type;     /* maps to enum below */
	unsigned bsr_num;      /* bsr id number for its type */
	int      bsr_minor;

	struct list_head bsr_list;

	dev_t    bsr_dev;
	struct cdev bsr_cdev;
	struct device *bsr_device;
	char     bsr_name[32];

};

static unsigned total_bsr_devs;
static LIST_HEAD(bsr_devs);
static int bsr_major;

enum {
	BSR_8    = 0,
	BSR_16   = 1,
	BSR_64   = 2,
	BSR_128  = 3,
	BSR_4096 = 4,
	BSR_UNKNOWN = 5,
	BSR_MAX  = 6,
};

static unsigned bsr_types[BSR_MAX];

static ssize_t
bsr_size_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct bsr_dev *bsr_dev = dev_get_drvdata(dev);
	return sprintf(buf, "%u\n", bsr_dev->bsr_bytes);
}
static DEVICE_ATTR_RO(bsr_size);

static ssize_t
bsr_stride_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct bsr_dev *bsr_dev = dev_get_drvdata(dev);
	return sprintf(buf, "%u\n", bsr_dev->bsr_stride);
}
static DEVICE_ATTR_RO(bsr_stride);

static ssize_t
bsr_length_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct bsr_dev *bsr_dev = dev_get_drvdata(dev);
	return sprintf(buf, "%llu\n", bsr_dev->bsr_len);
}
static DEVICE_ATTR_RO(bsr_length);

static struct attribute *bsr_dev_attrs[] = {
	&dev_attr_bsr_size.attr,
	&dev_attr_bsr_stride.attr,
	&dev_attr_bsr_length.attr,
	NULL,
};
ATTRIBUTE_GROUPS(bsr_dev);

static const struct class bsr_class = {
	.name		= "bsr",
	.dev_groups	= bsr_dev_groups,
};

static int bsr_mmap(struct file *filp, struct vm_area_struct *vma)
{
	unsigned long size   = vma->vm_end - vma->vm_start;
	struct bsr_dev *dev = filp->private_data;
	int ret;

	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);

	/* check for the case of a small BSR device and map one 4k page for it*/
	if (dev->bsr_len < PAGE_SIZE && size == PAGE_SIZE)
		ret = remap_4k_pfn(vma, vma->vm_start, dev->bsr_addr >> 12,
				   vma->vm_page_prot);
	else if (size <= dev->bsr_len)
		ret = io_remap_pfn_range(vma, vma->vm_start,
					 dev->bsr_addr >> PAGE_SHIFT,
					 size, vma->vm_page_prot);
	else
		return -EINVAL;

	if (ret)
		return -EAGAIN;

	return 0;
}

static int bsr_open(struct inode *inode, struct file *filp)
{
	struct cdev *cdev = inode->i_cdev;
	struct bsr_dev *dev = container_of(cdev, struct bsr_dev, bsr_cdev);

	filp->private_data = dev;
	return 0;
}

static const struct file_operations bsr_fops = {
	.owner = THIS_MODULE,
	.mmap  = bsr_mmap,
	.open  = bsr_open,
	.llseek = noop_llseek,
};

static void bsr_cleanup_devs(void)
{
	struct bsr_dev *cur, *n;

	list_for_each_entry_safe(cur, n, &bsr_devs, bsr_list) {
		if (cur->bsr_device) {
			cdev_del(&cur->bsr_cdev);
			device_del(cur->bsr_device);
		}
		list_del(&cur->bsr_list);
		kfree(cur);
	}
}

static int bsr_add_node(struct device_node *bn)
{
	int bsr_stride_len, bsr_bytes_len, num_bsr_devs;
	const u32 *bsr_stride;
	const u32 *bsr_bytes;
	unsigned i;
	int ret = -ENODEV;

	bsr_stride = of_get_property(bn, "ibm,lock-stride", &bsr_stride_len);
	bsr_bytes  = of_get_property(bn, "ibm,#lock-bytes", &bsr_bytes_len);

	if (!bsr_stride || !bsr_bytes ||
	    (bsr_stride_len != bsr_bytes_len)) {
		printk(KERN_ERR "bsr of-node has missing/incorrect property\n");
		return ret;
	}

	num_bsr_devs = bsr_bytes_len / sizeof(u32);

	for (i = 0 ; i < num_bsr_devs; i++) {
		struct bsr_dev *cur = kzalloc(sizeof(struct bsr_dev),
					      GFP_KERNEL);
		struct resource res;
		int result;

		if (!cur) {
			printk(KERN_ERR "Unable to alloc bsr dev\n");
			ret = -ENOMEM;
			goto out_err;
		}

		result = of_address_to_resource(bn, i, &res);
		if (result < 0) {
			printk(KERN_ERR "bsr of-node has invalid reg property, skipping\n");
			kfree(cur);
			continue;
		}

		cur->bsr_minor  = i + total_bsr_devs;
		cur->bsr_addr   = res.start;
		cur->bsr_len    = resource_size(&res);
		cur->bsr_bytes  = bsr_bytes[i];
		cur->bsr_stride = bsr_stride[i];
		cur->bsr_dev    = MKDEV(bsr_major, i + total_bsr_devs);

		/* if we have a bsr_len of > 4k and less then PAGE_SIZE (64k pages) */
		/* we can only map 4k of it, so only advertise the 4k in sysfs */
		if (cur->bsr_len > 4096 && cur->bsr_len < PAGE_SIZE)
			cur->bsr_len = 4096;

		switch(cur->bsr_bytes) {
		case 8:
			cur->bsr_type = BSR_8;
			break;
		case 16:
			cur->bsr_type = BSR_16;
			break;
		case 64:
			cur->bsr_type = BSR_64;
			break;
		case 128:
			cur->bsr_type = BSR_128;
			break;
		case 4096:
			cur->bsr_type = BSR_4096;
			break;
		default:
			cur->bsr_type = BSR_UNKNOWN;
		}

		cur->bsr_num = bsr_types[cur->bsr_type];
		snprintf(cur->bsr_name, 32, "bsr%d_%d",
			 cur->bsr_bytes, cur->bsr_num);

		cdev_init(&cur->bsr_cdev, &bsr_fops);
		result = cdev_add(&cur->bsr_cdev, cur->bsr_dev, 1);
		if (result) {
			kfree(cur);
			goto out_err;
		}

		cur->bsr_device = device_create(&bsr_class, NULL, cur->bsr_dev,
						cur, "%s", cur->bsr_name);
		if (IS_ERR(cur->bsr_device)) {
			printk(KERN_ERR "device_create failed for %s\n",
			       cur->bsr_name);
			cdev_del(&cur->bsr_cdev);
			kfree(cur);
			goto out_err;
		}

		bsr_types[cur->bsr_type] = cur->bsr_num + 1;
		list_add_tail(&cur->bsr_list, &bsr_devs);
	}

	total_bsr_devs += num_bsr_devs;

	return 0;

 out_err:

	bsr_cleanup_devs();
	return ret;
}

static int bsr_create_devs(struct device_node *bn)
{
	int ret;

	while (bn) {
		ret = bsr_add_node(bn);
		if (ret) {
			of_node_put(bn);
			return ret;
		}
		bn = of_find_compatible_node(bn, NULL, "ibm,bsr");
	}
	return 0;
}

static int __init bsr_init(void)
{
	struct device_node *np;
	dev_t bsr_dev;
	int ret = -ENODEV;

	np = of_find_compatible_node(NULL, NULL, "ibm,bsr");
	if (!np)
		goto out_err;

	ret = class_register(&bsr_class);
	if (ret)
		goto out_err_1;

	ret = alloc_chrdev_region(&bsr_dev, 0, BSR_MAX_DEVS, "bsr");
	bsr_major = MAJOR(bsr_dev);
	if (ret < 0) {
		printk(KERN_ERR "alloc_chrdev_region() failed for bsr\n");
		goto out_err_2;
	}

	ret = bsr_create_devs(np);
	if (ret < 0) {
		np = NULL;
		goto out_err_3;
	}

	return 0;

 out_err_3:
	unregister_chrdev_region(bsr_dev, BSR_MAX_DEVS);

 out_err_2:
	class_unregister(&bsr_class);

 out_err_1:
	of_node_put(np);

 out_err:

	return ret;
}

static void __exit  bsr_exit(void)
{

	bsr_cleanup_devs();

	class_unregister(&bsr_class);

	if (bsr_major)
		unregister_chrdev_region(MKDEV(bsr_major, 0), BSR_MAX_DEVS);
}

module_init(bsr_init);
module_exit(bsr_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Sonny Rao <sonnyrao@us.ibm.com>");