summaryrefslogtreecommitdiffstats
path: root/drivers/misc/cxl/cxllib.c
blob: 2a1783f32254390e6b6beaba9cba8b768c1f1b02 (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
// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Copyright 2017 IBM Corp.
 */

#include <linux/hugetlb.h>
#include <linux/sched/mm.h>
#include <asm/pnv-pci.h>
#include <misc/cxllib.h>

#include "cxl.h"

#define CXL_INVALID_DRA                 ~0ull
#define CXL_DUMMY_READ_SIZE             128
#define CXL_DUMMY_READ_ALIGN            8
#define CXL_CAPI_WINDOW_START           0x2000000000000ull
#define CXL_CAPI_WINDOW_LOG_SIZE        48
#define CXL_XSL_CONFIG_CURRENT_VERSION  CXL_XSL_CONFIG_VERSION1


bool cxllib_slot_is_supported(struct pci_dev *dev, unsigned long flags)
{
	int rc;
	u32 phb_index;
	u64 chip_id, capp_unit_id;

	/* No flags currently supported */
	if (flags)
		return false;

	if (!cpu_has_feature(CPU_FTR_HVMODE))
		return false;

	if (!cxl_is_power9())
		return false;

	if (cxl_slot_is_switched(dev))
		return false;

	/* on p9, some pci slots are not connected to a CAPP unit */
	rc = cxl_calc_capp_routing(dev, &chip_id, &phb_index, &capp_unit_id);
	if (rc)
		return false;

	return true;
}
EXPORT_SYMBOL_GPL(cxllib_slot_is_supported);

static DEFINE_MUTEX(dra_mutex);
static u64 dummy_read_addr = CXL_INVALID_DRA;

static int allocate_dummy_read_buf(void)
{
	u64 buf, vaddr;
	size_t buf_size;

	/*
	 * Dummy read buffer is 128-byte long, aligned on a
	 * 256-byte boundary and we need the physical address.
	 */
	buf_size = CXL_DUMMY_READ_SIZE + (1ull << CXL_DUMMY_READ_ALIGN);
	buf = (u64) kzalloc(buf_size, GFP_KERNEL);
	if (!buf)
		return -ENOMEM;

	vaddr = (buf + (1ull << CXL_DUMMY_READ_ALIGN) - 1) &
					(~0ull << CXL_DUMMY_READ_ALIGN);

	WARN((vaddr + CXL_DUMMY_READ_SIZE) > (buf + buf_size),
		"Dummy read buffer alignment issue");
	dummy_read_addr = virt_to_phys((void *) vaddr);
	return 0;
}

int cxllib_get_xsl_config(struct pci_dev *dev, struct cxllib_xsl_config *cfg)
{
	int rc;
	u32 phb_index;
	u64 chip_id, capp_unit_id;

	if (!cpu_has_feature(CPU_FTR_HVMODE))
		return -EINVAL;

	mutex_lock(&dra_mutex);
	if (dummy_read_addr == CXL_INVALID_DRA) {
		rc = allocate_dummy_read_buf();
		if (rc) {
			mutex_unlock(&dra_mutex);
			return rc;
		}
	}
	mutex_unlock(&dra_mutex);

	rc = cxl_calc_capp_routing(dev, &chip_id, &phb_index, &capp_unit_id);
	if (rc)
		return rc;

	rc = cxl_get_xsl9_dsnctl(dev, capp_unit_id, &cfg->dsnctl);
	if (rc)
		return rc;

	cfg->version  = CXL_XSL_CONFIG_CURRENT_VERSION;
	cfg->log_bar_size = CXL_CAPI_WINDOW_LOG_SIZE;
	cfg->bar_addr = CXL_CAPI_WINDOW_START;
	cfg->dra = dummy_read_addr;
	return 0;
}
EXPORT_SYMBOL_GPL(cxllib_get_xsl_config);

int cxllib_switch_phb_mode(struct pci_dev *dev, enum cxllib_mode mode,
			unsigned long flags)
{
	int rc = 0;

	if (!cpu_has_feature(CPU_FTR_HVMODE))
		return -EINVAL;

	switch (mode) {
	case CXL_MODE_PCI:
		/*
		 * We currently don't support going back to PCI mode
		 * However, we'll turn the invalidations off, so that
		 * the firmware doesn't have to ack them and can do
		 * things like reset, etc.. with no worries.
		 * So always return EPERM (can't go back to PCI) or
		 * EBUSY if we couldn't even turn off snooping
		 */
		rc = pnv_phb_to_cxl_mode(dev, OPAL_PHB_CAPI_MODE_SNOOP_OFF);
		if (rc)
			rc = -EBUSY;
		else
			rc = -EPERM;
		break;
	case CXL_MODE_CXL:
		/* DMA only supported on TVT1 for the time being */
		if (flags != CXL_MODE_DMA_TVT1)
			return -EINVAL;
		rc = pnv_phb_to_cxl_mode(dev, OPAL_PHB_CAPI_MODE_DMA_TVT1);
		if (rc)
			return rc;
		rc = pnv_phb_to_cxl_mode(dev, OPAL_PHB_CAPI_MODE_SNOOP_ON);
		break;
	default:
		rc = -EINVAL;
	}
	return rc;
}
EXPORT_SYMBOL_GPL(cxllib_switch_phb_mode);

/*
 * When switching the PHB to capi mode, the TVT#1 entry for
 * the Partitionable Endpoint is set in bypass mode, like
 * in PCI mode.
 * Configure the device dma to use TVT#1, which is done
 * by calling dma_set_mask() with a mask large enough.
 */
int cxllib_set_device_dma(struct pci_dev *dev, unsigned long flags)
{
	int rc;

	if (flags)
		return -EINVAL;

	rc = dma_set_mask(&dev->dev, DMA_BIT_MASK(64));
	return rc;
}
EXPORT_SYMBOL_GPL(cxllib_set_device_dma);

int cxllib_get_PE_attributes(struct task_struct *task,
			     unsigned long translation_mode,
			     struct cxllib_pe_attributes *attr)
{
	struct mm_struct *mm = NULL;

	if (translation_mode != CXL_TRANSLATED_MODE &&
		translation_mode != CXL_REAL_MODE)
		return -EINVAL;

	attr->sr = cxl_calculate_sr(false,
				task == NULL,
				translation_mode == CXL_REAL_MODE,
				true);
	attr->lpid = mfspr(SPRN_LPID);
	if (task) {
		mm = get_task_mm(task);
		if (mm == NULL)
			return -EINVAL;
		/*
		 * Caller is keeping a reference on mm_users for as long
		 * as XSL uses the memory context
		 */
		attr->pid = mm->context.id;
		mmput(mm);
		attr->tid = task->thread.tidr;
	} else {
		attr->pid = 0;
		attr->tid = 0;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(cxllib_get_PE_attributes);

static int get_vma_info(struct mm_struct *mm, u64 addr,
			u64 *vma_start, u64 *vma_end,
			unsigned long *page_size)
{
	struct vm_area_struct *vma = NULL;
	int rc = 0;

	mmap_read_lock(mm);

	vma = find_vma(mm, addr);
	if (!vma) {
		rc = -EFAULT;
		goto out;
	}
	*page_size = vma_kernel_pagesize(vma);
	*vma_start = vma->vm_start;
	*vma_end = vma->vm_end;
out:
	mmap_read_unlock(mm);
	return rc;
}

int cxllib_handle_fault(struct mm_struct *mm, u64 addr, u64 size, u64 flags)
{
	int rc;
	u64 dar, vma_start, vma_end;
	unsigned long page_size;

	if (mm == NULL)
		return -EFAULT;

	/*
	 * The buffer we have to process can extend over several pages
	 * and may also cover several VMAs.
	 * We iterate over all the pages. The page size could vary
	 * between VMAs.
	 */
	rc = get_vma_info(mm, addr, &vma_start, &vma_end, &page_size);
	if (rc)
		return rc;

	for (dar = (addr & ~(page_size - 1)); dar < (addr + size);
	     dar += page_size) {
		if (dar < vma_start || dar >= vma_end) {
			/*
			 * We don't hold mm->mmap_lock while iterating, since
			 * the lock is required by one of the lower-level page
			 * fault processing functions and it could
			 * create a deadlock.
			 *
			 * It means the VMAs can be altered between 2
			 * loop iterations and we could theoretically
			 * miss a page (however unlikely). But that's
			 * not really a problem, as the driver will
			 * retry access, get another page fault on the
			 * missing page and call us again.
			 */
			rc = get_vma_info(mm, dar, &vma_start, &vma_end,
					&page_size);
			if (rc)
				return rc;
		}

		rc = cxl_handle_mm_fault(mm, flags, dar);
		if (rc)
			return -EFAULT;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(cxllib_handle_fault);