Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
xen: add new hypercall buffer mapping device
For passing arbitrary data from user land to the Xen hypervisor the
Xen tools today are using mlock()ed buffers. Unfortunately the kernel
might change access rights of such buffers for brief periods of time
e.g. for page migration or compaction, leading to access faults in the
hypervisor, as the hypervisor can't use the locks of the kernel.
In order to solve this problem add a new device node to the Xen privcmd
driver to easily allocate hypercall buffers via mmap(). The memory is
allocated in the kernel and just mapped into user space. Marked as
VM_IO the user mapping will not be subject to page migration et al.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
+223
-1
+1
-1
drivers/xen/Makefile
+1
-1
drivers/xen/Makefile
+210
drivers/xen/privcmd-buf.c
+210
drivers/xen/privcmd-buf.c
···
1
+
// SPDX-License-Identifier: GPL-2.0 OR MIT
2
+
3
+
/******************************************************************************
4
+
* privcmd-buf.c
5
+
*
6
+
* Mmap of hypercall buffers.
7
+
*
8
+
* Copyright (c) 2018 Juergen Gross
9
+
*/
10
+
11
+
#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
12
+
13
+
#include <linux/kernel.h>
14
+
#include <linux/module.h>
15
+
#include <linux/list.h>
16
+
#include <linux/miscdevice.h>
17
+
#include <linux/mm.h>
18
+
#include <linux/slab.h>
19
+
20
+
#include "privcmd.h"
21
+
22
+
MODULE_LICENSE("GPL");
23
+
24
+
static unsigned int limit = 64;
25
+
module_param(limit, uint, 0644);
26
+
MODULE_PARM_DESC(limit, "Maximum number of pages that may be allocated by "
27
+
"the privcmd-buf device per open file");
28
+
29
+
struct privcmd_buf_private {
30
+
struct mutex lock;
31
+
struct list_head list;
32
+
unsigned int allocated;
33
+
};
34
+
35
+
struct privcmd_buf_vma_private {
36
+
struct privcmd_buf_private *file_priv;
37
+
struct list_head list;
38
+
unsigned int users;
39
+
unsigned int n_pages;
40
+
struct page *pages[];
41
+
};
42
+
43
+
static int privcmd_buf_open(struct inode *ino, struct file *file)
44
+
{
45
+
struct privcmd_buf_private *file_priv;
46
+
47
+
file_priv = kzalloc(sizeof(*file_priv), GFP_KERNEL);
48
+
if (!file_priv)
49
+
return -ENOMEM;
50
+
51
+
mutex_init(&file_priv->lock);
52
+
INIT_LIST_HEAD(&file_priv->list);
53
+
54
+
file->private_data = file_priv;
55
+
56
+
return 0;
57
+
}
58
+
59
+
static void privcmd_buf_vmapriv_free(struct privcmd_buf_vma_private *vma_priv)
60
+
{
61
+
unsigned int i;
62
+
63
+
vma_priv->file_priv->allocated -= vma_priv->n_pages;
64
+
65
+
list_del(&vma_priv->list);
66
+
67
+
for (i = 0; i < vma_priv->n_pages; i++)
68
+
if (vma_priv->pages[i])
69
+
__free_page(vma_priv->pages[i]);
70
+
71
+
kfree(vma_priv);
72
+
}
73
+
74
+
static int privcmd_buf_release(struct inode *ino, struct file *file)
75
+
{
76
+
struct privcmd_buf_private *file_priv = file->private_data;
77
+
struct privcmd_buf_vma_private *vma_priv;
78
+
79
+
mutex_lock(&file_priv->lock);
80
+
81
+
while (!list_empty(&file_priv->list)) {
82
+
vma_priv = list_first_entry(&file_priv->list,
83
+
struct privcmd_buf_vma_private,
84
+
list);
85
+
privcmd_buf_vmapriv_free(vma_priv);
86
+
}
87
+
88
+
mutex_unlock(&file_priv->lock);
89
+
90
+
kfree(file_priv);
91
+
92
+
return 0;
93
+
}
94
+
95
+
static void privcmd_buf_vma_open(struct vm_area_struct *vma)
96
+
{
97
+
struct privcmd_buf_vma_private *vma_priv = vma->vm_private_data;
98
+
99
+
if (!vma_priv)
100
+
return;
101
+
102
+
mutex_lock(&vma_priv->file_priv->lock);
103
+
vma_priv->users++;
104
+
mutex_unlock(&vma_priv->file_priv->lock);
105
+
}
106
+
107
+
static void privcmd_buf_vma_close(struct vm_area_struct *vma)
108
+
{
109
+
struct privcmd_buf_vma_private *vma_priv = vma->vm_private_data;
110
+
struct privcmd_buf_private *file_priv;
111
+
112
+
if (!vma_priv)
113
+
return;
114
+
115
+
file_priv = vma_priv->file_priv;
116
+
117
+
mutex_lock(&file_priv->lock);
118
+
119
+
vma_priv->users--;
120
+
if (!vma_priv->users)
121
+
privcmd_buf_vmapriv_free(vma_priv);
122
+
123
+
mutex_unlock(&file_priv->lock);
124
+
}
125
+
126
+
static vm_fault_t privcmd_buf_vma_fault(struct vm_fault *vmf)
127
+
{
128
+
pr_debug("fault: vma=%p %lx-%lx, pgoff=%lx, uv=%p\n",
129
+
vmf->vma, vmf->vma->vm_start, vmf->vma->vm_end,
130
+
vmf->pgoff, (void *)vmf->address);
131
+
132
+
return VM_FAULT_SIGBUS;
133
+
}
134
+
135
+
static const struct vm_operations_struct privcmd_buf_vm_ops = {
136
+
.open = privcmd_buf_vma_open,
137
+
.close = privcmd_buf_vma_close,
138
+
.fault = privcmd_buf_vma_fault,
139
+
};
140
+
141
+
static int privcmd_buf_mmap(struct file *file, struct vm_area_struct *vma)
142
+
{
143
+
struct privcmd_buf_private *file_priv = file->private_data;
144
+
struct privcmd_buf_vma_private *vma_priv;
145
+
unsigned long count = vma_pages(vma);
146
+
unsigned int i;
147
+
int ret = 0;
148
+
149
+
if (!(vma->vm_flags & VM_SHARED) || count > limit ||
150
+
file_priv->allocated + count > limit)
151
+
return -EINVAL;
152
+
153
+
vma_priv = kzalloc(sizeof(*vma_priv) + count * sizeof(void *),
154
+
GFP_KERNEL);
155
+
if (!vma_priv)
156
+
return -ENOMEM;
157
+
158
+
vma_priv->n_pages = count;
159
+
count = 0;
160
+
for (i = 0; i < vma_priv->n_pages; i++) {
161
+
vma_priv->pages[i] = alloc_page(GFP_KERNEL | __GFP_ZERO);
162
+
if (!vma_priv->pages[i])
163
+
break;
164
+
count++;
165
+
}
166
+
167
+
mutex_lock(&file_priv->lock);
168
+
169
+
file_priv->allocated += count;
170
+
171
+
vma_priv->file_priv = file_priv;
172
+
vma_priv->users = 1;
173
+
174
+
vma->vm_flags |= VM_IO | VM_DONTEXPAND;
175
+
vma->vm_ops = &privcmd_buf_vm_ops;
176
+
vma->vm_private_data = vma_priv;
177
+
178
+
list_add(&vma_priv->list, &file_priv->list);
179
+
180
+
if (vma_priv->n_pages != count)
181
+
ret = -ENOMEM;
182
+
else
183
+
for (i = 0; i < vma_priv->n_pages; i++) {
184
+
ret = vm_insert_page(vma, vma->vm_start + i * PAGE_SIZE,
185
+
vma_priv->pages[i]);
186
+
if (ret)
187
+
break;
188
+
}
189
+
190
+
if (ret)
191
+
privcmd_buf_vmapriv_free(vma_priv);
192
+
193
+
mutex_unlock(&file_priv->lock);
194
+
195
+
return ret;
196
+
}
197
+
198
+
const struct file_operations xen_privcmdbuf_fops = {
199
+
.owner = THIS_MODULE,
200
+
.open = privcmd_buf_open,
201
+
.release = privcmd_buf_release,
202
+
.mmap = privcmd_buf_mmap,
203
+
};
204
+
EXPORT_SYMBOL_GPL(xen_privcmdbuf_fops);
205
+
206
+
struct miscdevice xen_privcmdbuf_dev = {
207
+
.minor = MISC_DYNAMIC_MINOR,
208
+
.name = "xen/hypercall",
209
+
.fops = &xen_privcmdbuf_fops,
210
+
};
+9
drivers/xen/privcmd.c
+9
drivers/xen/privcmd.c
···
1007
1007
pr_err("Could not register Xen privcmd device\n");
1008
1008
return err;
1009
1009
}
1010
+
1011
+
err = misc_register(&xen_privcmdbuf_dev);
1012
+
if (err != 0) {
1013
+
pr_err("Could not register Xen hypercall-buf device\n");
1014
+
misc_deregister(&privcmd_dev);
1015
+
return err;
1016
+
}
1017
+
1010
1018
return 0;
1011
1019
}
1012
1020
1013
1021
static void __exit privcmd_exit(void)
1014
1022
{
1015
1023
misc_deregister(&privcmd_dev);
1024
+
misc_deregister(&xen_privcmdbuf_dev);
1016
1025
}
1017
1026
1018
1027
module_init(privcmd_init);