| @@ -1,1023 +1,1029 @@ | | | @@ -1,1023 +1,1029 @@ |
1 | /* $NetBSD: xbd_xenbus.c,v 1.84 2018/08/21 18:31:55 jdolecek Exp $ */ | | 1 | /* $NetBSD: xbd_xenbus.c,v 1.85 2018/08/21 18:45:16 jdolecek Exp $ */ |
2 | | | 2 | |
3 | /* | | 3 | /* |
4 | * Copyright (c) 2006 Manuel Bouyer. | | 4 | * Copyright (c) 2006 Manuel Bouyer. |
5 | * | | 5 | * |
6 | * Redistribution and use in source and binary forms, with or without | | 6 | * Redistribution and use in source and binary forms, with or without |
7 | * modification, are permitted provided that the following conditions | | 7 | * modification, are permitted provided that the following conditions |
8 | * are met: | | 8 | * are met: |
9 | * 1. Redistributions of source code must retain the above copyright | | 9 | * 1. Redistributions of source code must retain the above copyright |
10 | * notice, this list of conditions and the following disclaimer. | | 10 | * notice, this list of conditions and the following disclaimer. |
11 | * 2. Redistributions in binary form must reproduce the above copyright | | 11 | * 2. Redistributions in binary form must reproduce the above copyright |
12 | * notice, this list of conditions and the following disclaimer in the | | 12 | * notice, this list of conditions and the following disclaimer in the |
13 | * documentation and/or other materials provided with the distribution. | | 13 | * documentation and/or other materials provided with the distribution. |
14 | * | | 14 | * |
15 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR | | 15 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
16 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES | | 16 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
17 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. | | 17 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
18 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, | | 18 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
19 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT | | 19 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
20 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | | 20 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
21 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | | 21 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
22 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | | 22 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
23 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF | | 23 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
24 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | | 24 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
25 | * | | 25 | * |
26 | */ | | 26 | */ |
27 | | | 27 | |
28 | /* | | 28 | /* |
29 | * The file contains the xbd frontend code required for block-level | | 29 | * The file contains the xbd frontend code required for block-level |
30 | * communications (similar to hard disks) between two Xen domains. | | 30 | * communications (similar to hard disks) between two Xen domains. |
31 | * | | 31 | * |
32 | * We are not supposed to receive solicitations spontaneously from backend. The | | 32 | * We are not supposed to receive solicitations spontaneously from backend. The |
33 | * protocol is therefore fairly simple and uses only one ring to communicate | | 33 | * protocol is therefore fairly simple and uses only one ring to communicate |
34 | * with backend: frontend posts requests to the ring then wait for their | | 34 | * with backend: frontend posts requests to the ring then wait for their |
35 | * replies asynchronously. | | 35 | * replies asynchronously. |
36 | * | | 36 | * |
37 | * xbd follows NetBSD's disk(9) convention. At any time, a LWP can schedule | | 37 | * xbd follows NetBSD's disk(9) convention. At any time, a LWP can schedule |
38 | * an operation request for the device (be it open(), read(), write(), ...). | | 38 | * an operation request for the device (be it open(), read(), write(), ...). |
39 | * Calls are typically processed that way: | | 39 | * Calls are typically processed that way: |
40 | * - initiate request: xbdread/write/open/ioctl/.. | | 40 | * - initiate request: xbdread/write/open/ioctl/.. |
41 | * - depending on operation, it is handled directly by disk(9) subsystem or | | 41 | * - depending on operation, it is handled directly by disk(9) subsystem or |
42 | * goes through physio(9) first. | | 42 | * goes through physio(9) first. |
43 | * - the request is ultimately processed by xbd_diskstart() that prepares the | | 43 | * - the request is ultimately processed by xbd_diskstart() that prepares the |
44 | * xbd requests, post them in the ring I/O queue, then signal the backend. | | 44 | * xbd requests, post them in the ring I/O queue, then signal the backend. |
45 | * | | 45 | * |
46 | * When a response is available in the queue, the backend signals the frontend | | 46 | * When a response is available in the queue, the backend signals the frontend |
47 | * via its event channel. This triggers xbd_handler(), which will link back | | 47 | * via its event channel. This triggers xbd_handler(), which will link back |
48 | * the response to its request through the request ID, and mark the I/O as | | 48 | * the response to its request through the request ID, and mark the I/O as |
49 | * completed. | | 49 | * completed. |
50 | */ | | 50 | */ |
51 | | | 51 | |
52 | #include <sys/cdefs.h> | | 52 | #include <sys/cdefs.h> |
53 | __KERNEL_RCSID(0, "$NetBSD: xbd_xenbus.c,v 1.84 2018/08/21 18:31:55 jdolecek Exp $"); | | 53 | __KERNEL_RCSID(0, "$NetBSD: xbd_xenbus.c,v 1.85 2018/08/21 18:45:16 jdolecek Exp $"); |
54 | | | 54 | |
55 | #include "opt_xen.h" | | 55 | #include "opt_xen.h" |
56 | | | 56 | |
57 | | | 57 | |
58 | #include <sys/param.h> | | 58 | #include <sys/param.h> |
59 | #include <sys/buf.h> | | 59 | #include <sys/buf.h> |
60 | #include <sys/bufq.h> | | 60 | #include <sys/bufq.h> |
61 | #include <sys/device.h> | | 61 | #include <sys/device.h> |
62 | #include <sys/disk.h> | | 62 | #include <sys/disk.h> |
63 | #include <sys/disklabel.h> | | 63 | #include <sys/disklabel.h> |
64 | #include <sys/conf.h> | | 64 | #include <sys/conf.h> |
65 | #include <sys/fcntl.h> | | 65 | #include <sys/fcntl.h> |
66 | #include <sys/kernel.h> | | 66 | #include <sys/kernel.h> |
67 | #include <sys/proc.h> | | 67 | #include <sys/proc.h> |
68 | #include <sys/systm.h> | | 68 | #include <sys/systm.h> |
69 | #include <sys/stat.h> | | 69 | #include <sys/stat.h> |
70 | #include <sys/vnode.h> | | 70 | #include <sys/vnode.h> |
71 | | | 71 | |
72 | #include <dev/dkvar.h> | | 72 | #include <dev/dkvar.h> |
73 | | | 73 | |
74 | #include <uvm/uvm.h> | | 74 | #include <uvm/uvm.h> |
75 | | | 75 | |
76 | #include <xen/hypervisor.h> | | 76 | #include <xen/hypervisor.h> |
77 | #include <xen/evtchn.h> | | 77 | #include <xen/evtchn.h> |
78 | #include <xen/granttables.h> | | 78 | #include <xen/granttables.h> |
79 | #include <xen/xen-public/io/blkif.h> | | 79 | #include <xen/xen-public/io/blkif.h> |
80 | #include <xen/xen-public/io/protocols.h> | | 80 | #include <xen/xen-public/io/protocols.h> |
81 | | | 81 | |
82 | #include <xen/xenbus.h> | | 82 | #include <xen/xenbus.h> |
83 | #include "locators.h" | | 83 | #include "locators.h" |
84 | | | 84 | |
85 | #undef XBD_DEBUG | | 85 | #undef XBD_DEBUG |
86 | #ifdef XBD_DEBUG | | 86 | #ifdef XBD_DEBUG |
87 | #define DPRINTF(x) printf x; | | 87 | #define DPRINTF(x) printf x; |
88 | #else | | 88 | #else |
89 | #define DPRINTF(x) | | 89 | #define DPRINTF(x) |
90 | #endif | | 90 | #endif |
91 | | | 91 | |
92 | #define GRANT_INVALID_REF -1 | | 92 | #define GRANT_INVALID_REF -1 |
93 | | | 93 | |
94 | #define XBD_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE) | | 94 | #define XBD_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE) |
95 | #define XBD_MAX_XFER (PAGE_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST) | | 95 | #define XBD_MAX_XFER (PAGE_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST) |
96 | | | 96 | |
97 | #define XEN_BSHIFT 9 /* log2(XEN_BSIZE) */ | | 97 | #define XEN_BSHIFT 9 /* log2(XEN_BSIZE) */ |
98 | #define XEN_BSIZE (1 << XEN_BSHIFT) | | 98 | #define XEN_BSIZE (1 << XEN_BSHIFT) |
99 | | | 99 | |
100 | struct xbd_req { | | 100 | struct xbd_req { |
101 | SLIST_ENTRY(xbd_req) req_next; | | 101 | SLIST_ENTRY(xbd_req) req_next; |
102 | uint16_t req_id; /* ID passed to backend */ | | 102 | uint16_t req_id; /* ID passed to backend */ |
103 | union { | | 103 | union { |
104 | struct { | | 104 | struct { |
105 | grant_ref_t req_gntref[BLKIF_MAX_SEGMENTS_PER_REQUEST]; | | 105 | grant_ref_t req_gntref[BLKIF_MAX_SEGMENTS_PER_REQUEST]; |
106 | int req_nr_segments; /* number of segments in this request */ | | 106 | int req_nr_segments; /* number of segments in this request */ |
107 | struct buf *req_bp; /* buffer associated with this request */ | | 107 | struct buf *req_bp; /* buffer associated with this request */ |
108 | } req_rw; | | 108 | } req_rw; |
109 | struct { | | 109 | struct { |
110 | int s_error; | | 110 | int s_error; |
111 | volatile int s_done; | | 111 | volatile int s_done; |
112 | } req_sync; | | 112 | } req_sync; |
113 | } u; | | 113 | } u; |
114 | }; | | 114 | }; |
115 | #define req_gntref u.req_rw.req_gntref | | 115 | #define req_gntref u.req_rw.req_gntref |
116 | #define req_nr_segments u.req_rw.req_nr_segments | | 116 | #define req_nr_segments u.req_rw.req_nr_segments |
117 | #define req_bp u.req_rw.req_bp | | 117 | #define req_bp u.req_rw.req_bp |
118 | #define req_sync u.req_sync | | 118 | #define req_sync u.req_sync |
119 | | | 119 | |
120 | struct xbd_xenbus_softc { | | 120 | struct xbd_xenbus_softc { |
121 | struct dk_softc sc_dksc; /* Must be first in this struct */ | | 121 | struct dk_softc sc_dksc; /* Must be first in this struct */ |
122 | struct xenbus_device *sc_xbusd; | | 122 | struct xenbus_device *sc_xbusd; |
123 | | | 123 | |
124 | struct intrhand *sc_ih; /* Interrupt handler for this instance. */ | | 124 | struct intrhand *sc_ih; /* Interrupt handler for this instance. */ |
125 | | | 125 | |
126 | blkif_front_ring_t sc_ring; | | 126 | blkif_front_ring_t sc_ring; |
127 | | | 127 | |
128 | unsigned int sc_evtchn; | | 128 | unsigned int sc_evtchn; |
129 | | | 129 | |
130 | grant_ref_t sc_ring_gntref; | | 130 | grant_ref_t sc_ring_gntref; |
131 | | | 131 | |
132 | struct xbd_req sc_reqs[XBD_RING_SIZE]; | | 132 | struct xbd_req sc_reqs[XBD_RING_SIZE]; |
133 | SLIST_HEAD(,xbd_req) sc_xbdreq_head; /* list of free requests */ | | 133 | SLIST_HEAD(,xbd_req) sc_xbdreq_head; /* list of free requests */ |
134 | bool sc_xbdreq_wait; /* special waiting on xbd_req */ | | 134 | bool sc_xbdreq_wait; /* special waiting on xbd_req */ |
135 | | | 135 | |
136 | int sc_backend_status; /* our status with backend */ | | 136 | int sc_backend_status; /* our status with backend */ |
137 | #define BLKIF_STATE_DISCONNECTED 0 | | 137 | #define BLKIF_STATE_DISCONNECTED 0 |
138 | #define BLKIF_STATE_CONNECTED 1 | | 138 | #define BLKIF_STATE_CONNECTED 1 |
139 | #define BLKIF_STATE_SUSPENDED 2 | | 139 | #define BLKIF_STATE_SUSPENDED 2 |
140 | | | 140 | |
141 | int sc_shutdown; | | 141 | int sc_shutdown; |
142 | #define BLKIF_SHUTDOWN_RUN 0 /* no shutdown */ | | 142 | #define BLKIF_SHUTDOWN_RUN 0 /* no shutdown */ |
143 | #define BLKIF_SHUTDOWN_REMOTE 1 /* backend-initiated shutdown in progress */ | | 143 | #define BLKIF_SHUTDOWN_REMOTE 1 /* backend-initiated shutdown in progress */ |
144 | #define BLKIF_SHUTDOWN_LOCAL 2 /* locally-initiated shutdown in progress */ | | 144 | #define BLKIF_SHUTDOWN_LOCAL 2 /* locally-initiated shutdown in progress */ |
145 | | | 145 | |
146 | uint64_t sc_sectors; /* number of sectors for this device */ | | 146 | uint64_t sc_sectors; /* number of sectors for this device */ |
147 | u_long sc_secsize; /* sector size */ | | 147 | u_long sc_secsize; /* sector size */ |
148 | uint64_t sc_xbdsize; /* size of disk in DEV_BSIZE */ | | 148 | uint64_t sc_xbdsize; /* size of disk in DEV_BSIZE */ |
149 | u_long sc_info; /* VDISK_* */ | | 149 | u_long sc_info; /* VDISK_* */ |
150 | u_long sc_handle; /* from backend */ | | 150 | u_long sc_handle; /* from backend */ |
151 | int sc_cache_flush; /* backend supports BLKIF_OP_FLUSH_DISKCACHE */ | | 151 | int sc_cache_flush; /* backend supports BLKIF_OP_FLUSH_DISKCACHE */ |
152 | }; | | 152 | }; |
153 | | | 153 | |
154 | #if 0 | | 154 | #if 0 |
155 | /* too big to be on stack */ | | 155 | /* too big to be on stack */ |
156 | static multicall_entry_t rq_mcl[XBD_RING_SIZE+1]; | | 156 | static multicall_entry_t rq_mcl[XBD_RING_SIZE+1]; |
157 | static paddr_t rq_pages[XBD_RING_SIZE]; | | 157 | static paddr_t rq_pages[XBD_RING_SIZE]; |
158 | #endif | | 158 | #endif |
159 | | | 159 | |
160 | static int xbd_xenbus_match(device_t, cfdata_t, void *); | | 160 | static int xbd_xenbus_match(device_t, cfdata_t, void *); |
161 | static void xbd_xenbus_attach(device_t, device_t, void *); | | 161 | static void xbd_xenbus_attach(device_t, device_t, void *); |
162 | static int xbd_xenbus_detach(device_t, int); | | 162 | static int xbd_xenbus_detach(device_t, int); |
163 | | | 163 | |
164 | static bool xbd_xenbus_suspend(device_t, const pmf_qual_t *); | | 164 | static bool xbd_xenbus_suspend(device_t, const pmf_qual_t *); |
165 | static bool xbd_xenbus_resume(device_t, const pmf_qual_t *); | | 165 | static bool xbd_xenbus_resume(device_t, const pmf_qual_t *); |
166 | | | 166 | |
167 | static int xbd_handler(void *); | | 167 | static int xbd_handler(void *); |
168 | static int xbd_diskstart(device_t, struct buf *); | | 168 | static int xbd_diskstart(device_t, struct buf *); |
169 | static void xbd_backend_changed(void *, XenbusState); | | 169 | static void xbd_backend_changed(void *, XenbusState); |
170 | static void xbd_connect(struct xbd_xenbus_softc *); | | 170 | static void xbd_connect(struct xbd_xenbus_softc *); |
171 | | | 171 | |
172 | static void xbdminphys(struct buf *); | | 172 | static void xbdminphys(struct buf *); |
173 | | | 173 | |
174 | CFATTACH_DECL3_NEW(xbd, sizeof(struct xbd_xenbus_softc), | | 174 | CFATTACH_DECL3_NEW(xbd, sizeof(struct xbd_xenbus_softc), |
175 | xbd_xenbus_match, xbd_xenbus_attach, xbd_xenbus_detach, NULL, NULL, NULL, | | 175 | xbd_xenbus_match, xbd_xenbus_attach, xbd_xenbus_detach, NULL, NULL, NULL, |
176 | DVF_DETACH_SHUTDOWN); | | 176 | DVF_DETACH_SHUTDOWN); |
177 | | | 177 | |
178 | dev_type_open(xbdopen); | | 178 | dev_type_open(xbdopen); |
179 | dev_type_close(xbdclose); | | 179 | dev_type_close(xbdclose); |
180 | dev_type_read(xbdread); | | 180 | dev_type_read(xbdread); |
181 | dev_type_write(xbdwrite); | | 181 | dev_type_write(xbdwrite); |
182 | dev_type_ioctl(xbdioctl); | | 182 | dev_type_ioctl(xbdioctl); |
183 | dev_type_strategy(xbdstrategy); | | 183 | dev_type_strategy(xbdstrategy); |
184 | dev_type_dump(xbddump); | | 184 | dev_type_dump(xbddump); |
185 | dev_type_size(xbdsize); | | 185 | dev_type_size(xbdsize); |
186 | | | 186 | |
187 | const struct bdevsw xbd_bdevsw = { | | 187 | const struct bdevsw xbd_bdevsw = { |
188 | .d_open = xbdopen, | | 188 | .d_open = xbdopen, |
189 | .d_close = xbdclose, | | 189 | .d_close = xbdclose, |
190 | .d_strategy = xbdstrategy, | | 190 | .d_strategy = xbdstrategy, |
191 | .d_ioctl = xbdioctl, | | 191 | .d_ioctl = xbdioctl, |
192 | .d_dump = xbddump, | | 192 | .d_dump = xbddump, |
193 | .d_psize = xbdsize, | | 193 | .d_psize = xbdsize, |
194 | .d_discard = nodiscard, | | 194 | .d_discard = nodiscard, |
195 | .d_flag = D_DISK | | 195 | .d_flag = D_DISK |
196 | }; | | 196 | }; |
197 | | | 197 | |
198 | const struct cdevsw xbd_cdevsw = { | | 198 | const struct cdevsw xbd_cdevsw = { |
199 | .d_open = xbdopen, | | 199 | .d_open = xbdopen, |
200 | .d_close = xbdclose, | | 200 | .d_close = xbdclose, |
201 | .d_read = xbdread, | | 201 | .d_read = xbdread, |
202 | .d_write = xbdwrite, | | 202 | .d_write = xbdwrite, |
203 | .d_ioctl = xbdioctl, | | 203 | .d_ioctl = xbdioctl, |
204 | .d_stop = nostop, | | 204 | .d_stop = nostop, |
205 | .d_tty = notty, | | 205 | .d_tty = notty, |
206 | .d_poll = nopoll, | | 206 | .d_poll = nopoll, |
207 | .d_mmap = nommap, | | 207 | .d_mmap = nommap, |
208 | .d_kqfilter = nokqfilter, | | 208 | .d_kqfilter = nokqfilter, |
209 | .d_discard = nodiscard, | | 209 | .d_discard = nodiscard, |
210 | .d_flag = D_DISK | | 210 | .d_flag = D_DISK |
211 | }; | | 211 | }; |
212 | | | 212 | |
213 | extern struct cfdriver xbd_cd; | | 213 | extern struct cfdriver xbd_cd; |
214 | | | 214 | |
215 | static struct dkdriver xbddkdriver = { | | 215 | static struct dkdriver xbddkdriver = { |
216 | .d_strategy = xbdstrategy, | | 216 | .d_strategy = xbdstrategy, |
217 | .d_minphys = xbdminphys, | | 217 | .d_minphys = xbdminphys, |
218 | .d_open = xbdopen, | | 218 | .d_open = xbdopen, |
219 | .d_close = xbdclose, | | 219 | .d_close = xbdclose, |
220 | .d_diskstart = xbd_diskstart, | | 220 | .d_diskstart = xbd_diskstart, |
221 | }; | | 221 | }; |
222 | | | 222 | |
223 | static int | | 223 | static int |
224 | xbd_xenbus_match(device_t parent, cfdata_t match, void *aux) | | 224 | xbd_xenbus_match(device_t parent, cfdata_t match, void *aux) |
225 | { | | 225 | { |
226 | struct xenbusdev_attach_args *xa = aux; | | 226 | struct xenbusdev_attach_args *xa = aux; |
227 | | | 227 | |
228 | if (strcmp(xa->xa_type, "vbd") != 0) | | 228 | if (strcmp(xa->xa_type, "vbd") != 0) |
229 | return 0; | | 229 | return 0; |
230 | | | 230 | |
231 | if (match->cf_loc[XENBUSCF_ID] != XENBUSCF_ID_DEFAULT && | | 231 | if (match->cf_loc[XENBUSCF_ID] != XENBUSCF_ID_DEFAULT && |
232 | match->cf_loc[XENBUSCF_ID] != xa->xa_id) | | 232 | match->cf_loc[XENBUSCF_ID] != xa->xa_id) |
233 | return 0; | | 233 | return 0; |
234 | | | 234 | |
235 | return 1; | | 235 | return 1; |
236 | } | | 236 | } |
237 | | | 237 | |
238 | static void | | 238 | static void |
239 | xbd_xenbus_attach(device_t parent, device_t self, void *aux) | | 239 | xbd_xenbus_attach(device_t parent, device_t self, void *aux) |
240 | { | | 240 | { |
241 | struct xbd_xenbus_softc *sc = device_private(self); | | 241 | struct xbd_xenbus_softc *sc = device_private(self); |
242 | struct xenbusdev_attach_args *xa = aux; | | 242 | struct xenbusdev_attach_args *xa = aux; |
243 | blkif_sring_t *ring; | | 243 | blkif_sring_t *ring; |
244 | RING_IDX i; | | 244 | RING_IDX i; |
245 | #ifdef XBD_DEBUG | | 245 | #ifdef XBD_DEBUG |
246 | char **dir, *val; | | 246 | char **dir, *val; |
247 | int dir_n = 0; | | 247 | int dir_n = 0; |
248 | char id_str[20]; | | 248 | char id_str[20]; |
249 | int err; | | 249 | int err; |
250 | #endif | | 250 | #endif |
251 | | | 251 | |
252 | config_pending_incr(self); | | 252 | config_pending_incr(self); |
253 | aprint_normal(": Xen Virtual Block Device Interface\n"); | | 253 | aprint_normal(": Xen Virtual Block Device Interface\n"); |
254 | | | 254 | |
255 | dk_init(&sc->sc_dksc, self, DKTYPE_ESDI); | | 255 | dk_init(&sc->sc_dksc, self, DKTYPE_ESDI); |
256 | disk_init(&sc->sc_dksc.sc_dkdev, device_xname(self), &xbddkdriver); | | 256 | disk_init(&sc->sc_dksc.sc_dkdev, device_xname(self), &xbddkdriver); |
257 | | | 257 | |
258 | #ifdef XBD_DEBUG | | 258 | #ifdef XBD_DEBUG |
259 | printf("path: %s\n", xa->xa_xbusd->xbusd_path); | | 259 | printf("path: %s\n", xa->xa_xbusd->xbusd_path); |
260 | snprintf(id_str, sizeof(id_str), "%d", xa->xa_id); | | 260 | snprintf(id_str, sizeof(id_str), "%d", xa->xa_id); |
261 | err = xenbus_directory(NULL, "device/vbd", id_str, &dir_n, &dir); | | 261 | err = xenbus_directory(NULL, "device/vbd", id_str, &dir_n, &dir); |
262 | if (err) { | | 262 | if (err) { |
263 | aprint_error_dev(self, "xenbus_directory err %d\n", err); | | 263 | aprint_error_dev(self, "xenbus_directory err %d\n", err); |
264 | } else { | | 264 | } else { |
265 | printf("%s/\n", xa->xa_xbusd->xbusd_path); | | 265 | printf("%s/\n", xa->xa_xbusd->xbusd_path); |
266 | for (i = 0; i < dir_n; i++) { | | 266 | for (i = 0; i < dir_n; i++) { |
267 | printf("\t/%s", dir[i]); | | 267 | printf("\t/%s", dir[i]); |
268 | err = xenbus_read(NULL, xa->xa_xbusd->xbusd_path, | | 268 | err = xenbus_read(NULL, xa->xa_xbusd->xbusd_path, |
269 | dir[i], NULL, &val); | | 269 | dir[i], NULL, &val); |
270 | if (err) { | | 270 | if (err) { |
271 | aprint_error_dev(self, "xenbus_read err %d\n", | | 271 | aprint_error_dev(self, "xenbus_read err %d\n", |
272 | err); | | 272 | err); |
273 | } else { | | 273 | } else { |
274 | printf(" = %s\n", val); | | 274 | printf(" = %s\n", val); |
275 | free(val, M_DEVBUF); | | 275 | free(val, M_DEVBUF); |
276 | } | | 276 | } |
277 | } | | 277 | } |
278 | } | | 278 | } |
279 | #endif /* XBD_DEBUG */ | | 279 | #endif /* XBD_DEBUG */ |
280 | sc->sc_xbusd = xa->xa_xbusd; | | 280 | sc->sc_xbusd = xa->xa_xbusd; |
281 | sc->sc_xbusd->xbusd_otherend_changed = xbd_backend_changed; | | 281 | sc->sc_xbusd->xbusd_otherend_changed = xbd_backend_changed; |
282 | | | 282 | |
283 | /* initialize free requests list */ | | 283 | /* initialize free requests list */ |
284 | SLIST_INIT(&sc->sc_xbdreq_head); | | 284 | SLIST_INIT(&sc->sc_xbdreq_head); |
285 | for (i = 0; i < XBD_RING_SIZE; i++) { | | 285 | for (i = 0; i < XBD_RING_SIZE; i++) { |
286 | sc->sc_reqs[i].req_id = i; | | 286 | sc->sc_reqs[i].req_id = i; |
287 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, &sc->sc_reqs[i], | | 287 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, &sc->sc_reqs[i], |
288 | req_next); | | 288 | req_next); |
289 | } | | 289 | } |
290 | | | 290 | |
291 | sc->sc_backend_status = BLKIF_STATE_DISCONNECTED; | | 291 | sc->sc_backend_status = BLKIF_STATE_DISCONNECTED; |
292 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; | | 292 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; |
293 | | | 293 | |
294 | ring = (void *)uvm_km_alloc(kernel_map, PAGE_SIZE, 0, UVM_KMF_WIRED); | | 294 | ring = (void *)uvm_km_alloc(kernel_map, PAGE_SIZE, 0, UVM_KMF_WIRED); |
295 | if (ring == NULL) | | 295 | if (ring == NULL) |
296 | panic("%s: can't alloc ring", device_xname(self)); | | 296 | panic("%s: can't alloc ring", device_xname(self)); |
297 | sc->sc_ring.sring = ring; | | 297 | sc->sc_ring.sring = ring; |
298 | | | 298 | |
299 | /* resume shared structures and tell backend that we are ready */ | | 299 | /* resume shared structures and tell backend that we are ready */ |
300 | if (xbd_xenbus_resume(self, PMF_Q_NONE) == false) { | | 300 | if (xbd_xenbus_resume(self, PMF_Q_NONE) == false) { |
301 | uvm_km_free(kernel_map, (vaddr_t)ring, PAGE_SIZE, | | 301 | uvm_km_free(kernel_map, (vaddr_t)ring, PAGE_SIZE, |
302 | UVM_KMF_WIRED); | | 302 | UVM_KMF_WIRED); |
303 | return; | | 303 | return; |
304 | } | | 304 | } |
305 | | | 305 | |
306 | if (!pmf_device_register(self, xbd_xenbus_suspend, xbd_xenbus_resume)) | | 306 | if (!pmf_device_register(self, xbd_xenbus_suspend, xbd_xenbus_resume)) |
307 | aprint_error_dev(self, "couldn't establish power handler\n"); | | 307 | aprint_error_dev(self, "couldn't establish power handler\n"); |
308 | | | 308 | |
309 | } | | 309 | } |
310 | | | 310 | |
311 | static int | | 311 | static int |
312 | xbd_xenbus_detach(device_t dev, int flags) | | 312 | xbd_xenbus_detach(device_t dev, int flags) |
313 | { | | 313 | { |
314 | struct xbd_xenbus_softc *sc = device_private(dev); | | 314 | struct xbd_xenbus_softc *sc = device_private(dev); |
315 | int bmaj, cmaj, i, mn, rc, s; | | 315 | int bmaj, cmaj, i, mn, rc, s; |
316 | | | 316 | |
317 | rc = disk_begindetach(&sc->sc_dksc.sc_dkdev, NULL, dev, flags); | | 317 | rc = disk_begindetach(&sc->sc_dksc.sc_dkdev, NULL, dev, flags); |
318 | if (rc != 0) | | 318 | if (rc != 0) |
319 | return rc; | | 319 | return rc; |
320 | | | 320 | |
321 | s = splbio(); /* XXXSMP */ | | 321 | s = splbio(); /* XXXSMP */ |
322 | DPRINTF(("%s: xbd_detach\n", device_xname(dev))); | | 322 | DPRINTF(("%s: xbd_detach\n", device_xname(dev))); |
323 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) { | | 323 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) { |
324 | sc->sc_shutdown = BLKIF_SHUTDOWN_LOCAL; | | 324 | sc->sc_shutdown = BLKIF_SHUTDOWN_LOCAL; |
325 | /* wait for requests to complete */ | | 325 | /* wait for requests to complete */ |
326 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && | | 326 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
327 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { | | 327 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { |
328 | /* XXXSMP */ | | 328 | /* XXXSMP */ |
329 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach", hz/2); | | 329 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach", hz/2); |
330 | } | | 330 | } |
331 | | | 331 | |
332 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosing); | | 332 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosing); |
333 | } | | 333 | } |
334 | if ((flags & DETACH_FORCE) == 0) { | | 334 | if ((flags & DETACH_FORCE) == 0) { |
335 | /* xbd_xenbus_detach already in progress */ | | 335 | /* xbd_xenbus_detach already in progress */ |
336 | wakeup(xbd_xenbus_detach); /* XXXSMP */ | | 336 | wakeup(xbd_xenbus_detach); /* XXXSMP */ |
337 | splx(s); | | 337 | splx(s); |
338 | return EALREADY; | | 338 | return EALREADY; |
339 | } | | 339 | } |
340 | while (xenbus_read_driver_state(sc->sc_xbusd->xbusd_otherend) | | 340 | while (xenbus_read_driver_state(sc->sc_xbusd->xbusd_otherend) |
341 | != XenbusStateClosed) { | | 341 | != XenbusStateClosed) { |
342 | /* XXXSMP */ | | 342 | /* XXXSMP */ |
343 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach2", hz/2); | | 343 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach2", hz/2); |
344 | } | | 344 | } |
345 | splx(s); | | 345 | splx(s); |
346 | | | 346 | |
347 | /* locate the major number */ | | 347 | /* locate the major number */ |
348 | bmaj = bdevsw_lookup_major(&xbd_bdevsw); | | 348 | bmaj = bdevsw_lookup_major(&xbd_bdevsw); |
349 | cmaj = cdevsw_lookup_major(&xbd_cdevsw); | | 349 | cmaj = cdevsw_lookup_major(&xbd_cdevsw); |
350 | | | 350 | |
351 | /* Nuke the vnodes for any open instances. */ | | 351 | /* Nuke the vnodes for any open instances. */ |
352 | for (i = 0; i < MAXPARTITIONS; i++) { | | 352 | for (i = 0; i < MAXPARTITIONS; i++) { |
353 | mn = DISKMINOR(device_unit(dev), i); | | 353 | mn = DISKMINOR(device_unit(dev), i); |
354 | vdevgone(bmaj, mn, mn, VBLK); | | 354 | vdevgone(bmaj, mn, mn, VBLK); |
355 | vdevgone(cmaj, mn, mn, VCHR); | | 355 | vdevgone(cmaj, mn, mn, VCHR); |
356 | } | | 356 | } |
357 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED) { | | 357 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED) { |
358 | /* Delete all of our wedges. */ | | 358 | /* Delete all of our wedges. */ |
359 | dkwedge_delall(&sc->sc_dksc.sc_dkdev); | | 359 | dkwedge_delall(&sc->sc_dksc.sc_dkdev); |
360 | | | 360 | |
361 | /* Kill off any queued buffers. */ | | 361 | /* Kill off any queued buffers. */ |
362 | dk_drain(&sc->sc_dksc); | | 362 | dk_drain(&sc->sc_dksc); |
363 | bufq_free(sc->sc_dksc.sc_bufq); | | 363 | bufq_free(sc->sc_dksc.sc_bufq); |
364 | | | 364 | |
365 | /* detach disk */ | | 365 | /* detach disk */ |
366 | disk_detach(&sc->sc_dksc.sc_dkdev); | | 366 | disk_detach(&sc->sc_dksc.sc_dkdev); |
367 | disk_destroy(&sc->sc_dksc.sc_dkdev); | | 367 | disk_destroy(&sc->sc_dksc.sc_dkdev); |
368 | dk_detach(&sc->sc_dksc); | | 368 | dk_detach(&sc->sc_dksc); |
369 | } | | 369 | } |
370 | | | 370 | |
371 | hypervisor_mask_event(sc->sc_evtchn); | | 371 | hypervisor_mask_event(sc->sc_evtchn); |
372 | intr_disestablish(sc->sc_ih); | | 372 | intr_disestablish(sc->sc_ih); |
373 | | | 373 | |
374 | while (xengnt_status(sc->sc_ring_gntref)) { | | 374 | while (xengnt_status(sc->sc_ring_gntref)) { |
375 | /* XXXSMP */ | | 375 | /* XXXSMP */ |
376 | tsleep(xbd_xenbus_detach, PRIBIO, "xbd_ref", hz/2); | | 376 | tsleep(xbd_xenbus_detach, PRIBIO, "xbd_ref", hz/2); |
377 | } | | 377 | } |
378 | xengnt_revoke_access(sc->sc_ring_gntref); | | 378 | xengnt_revoke_access(sc->sc_ring_gntref); |
379 | uvm_km_free(kernel_map, (vaddr_t)sc->sc_ring.sring, | | 379 | uvm_km_free(kernel_map, (vaddr_t)sc->sc_ring.sring, |
380 | PAGE_SIZE, UVM_KMF_WIRED); | | 380 | PAGE_SIZE, UVM_KMF_WIRED); |
381 | | | 381 | |
382 | pmf_device_deregister(dev); | | 382 | pmf_device_deregister(dev); |
383 | | | 383 | |
384 | return 0; | | 384 | return 0; |
385 | } | | 385 | } |
386 | | | 386 | |
387 | static bool | | 387 | static bool |
388 | xbd_xenbus_suspend(device_t dev, const pmf_qual_t *qual) { | | 388 | xbd_xenbus_suspend(device_t dev, const pmf_qual_t *qual) { |
389 | | | 389 | |
390 | int s; | | 390 | int s; |
391 | struct xbd_xenbus_softc *sc; | | 391 | struct xbd_xenbus_softc *sc; |
392 | | | 392 | |
393 | sc = device_private(dev); | | 393 | sc = device_private(dev); |
394 | | | 394 | |
395 | s = splbio(); /* XXXSMP */ | | 395 | s = splbio(); /* XXXSMP */ |
396 | /* wait for requests to complete, then suspend device */ | | 396 | /* wait for requests to complete, then suspend device */ |
397 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && | | 397 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
398 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { | | 398 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { |
399 | /* XXXSMP */ | | 399 | /* XXXSMP */ |
400 | tsleep(xbd_xenbus_suspend, PRIBIO, "xbdsuspend", hz/2); | | 400 | tsleep(xbd_xenbus_suspend, PRIBIO, "xbdsuspend", hz/2); |
401 | } | | 401 | } |
402 | | | 402 | |
403 | hypervisor_mask_event(sc->sc_evtchn); | | 403 | hypervisor_mask_event(sc->sc_evtchn); |
404 | sc->sc_backend_status = BLKIF_STATE_SUSPENDED; | | 404 | sc->sc_backend_status = BLKIF_STATE_SUSPENDED; |
405 | intr_disestablish(sc->sc_ih); | | 405 | intr_disestablish(sc->sc_ih); |
406 | | | 406 | |
407 | splx(s); | | 407 | splx(s); |
408 | | | 408 | |
409 | xenbus_device_suspend(sc->sc_xbusd); | | 409 | xenbus_device_suspend(sc->sc_xbusd); |
410 | aprint_verbose_dev(dev, "removed event channel %d\n", sc->sc_evtchn); | | 410 | aprint_verbose_dev(dev, "removed event channel %d\n", sc->sc_evtchn); |
411 | | | 411 | |
412 | return true; | | 412 | return true; |
413 | } | | 413 | } |
414 | | | 414 | |
415 | static bool | | 415 | static bool |
416 | xbd_xenbus_resume(device_t dev, const pmf_qual_t *qual) | | 416 | xbd_xenbus_resume(device_t dev, const pmf_qual_t *qual) |
417 | { | | 417 | { |
418 | struct xbd_xenbus_softc *sc; | | 418 | struct xbd_xenbus_softc *sc; |
419 | struct xenbus_transaction *xbt; | | 419 | struct xenbus_transaction *xbt; |
420 | int error; | | 420 | int error; |
421 | blkif_sring_t *ring; | | 421 | blkif_sring_t *ring; |
422 | paddr_t ma; | | 422 | paddr_t ma; |
423 | const char *errmsg; | | 423 | const char *errmsg; |
424 | | | 424 | |
425 | sc = device_private(dev); | | 425 | sc = device_private(dev); |
426 | | | 426 | |
427 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { | | 427 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { |
428 | /* | | 428 | /* |
429 | * Device was suspended, so ensure that access associated to | | 429 | * Device was suspended, so ensure that access associated to |
430 | * the block I/O ring is revoked. | | 430 | * the block I/O ring is revoked. |
431 | */ | | 431 | */ |
432 | xengnt_revoke_access(sc->sc_ring_gntref); | | 432 | xengnt_revoke_access(sc->sc_ring_gntref); |
433 | } | | 433 | } |
434 | sc->sc_ring_gntref = GRANT_INVALID_REF; | | 434 | sc->sc_ring_gntref = GRANT_INVALID_REF; |
435 | | | 435 | |
436 | /* Initialize ring */ | | 436 | /* Initialize ring */ |
437 | ring = sc->sc_ring.sring; | | 437 | ring = sc->sc_ring.sring; |
438 | memset(ring, 0, PAGE_SIZE); | | 438 | memset(ring, 0, PAGE_SIZE); |
439 | SHARED_RING_INIT(ring); | | 439 | SHARED_RING_INIT(ring); |
440 | FRONT_RING_INIT(&sc->sc_ring, ring, PAGE_SIZE); | | 440 | FRONT_RING_INIT(&sc->sc_ring, ring, PAGE_SIZE); |
441 | | | 441 | |
442 | /* | | 442 | /* |
443 | * get MA address of the ring, and use it to set up the grant entry | | 443 | * get MA address of the ring, and use it to set up the grant entry |
444 | * for the block device | | 444 | * for the block device |
445 | */ | | 445 | */ |
446 | (void)pmap_extract_ma(pmap_kernel(), (vaddr_t)ring, &ma); | | 446 | (void)pmap_extract_ma(pmap_kernel(), (vaddr_t)ring, &ma); |
447 | error = xenbus_grant_ring(sc->sc_xbusd, ma, &sc->sc_ring_gntref); | | 447 | error = xenbus_grant_ring(sc->sc_xbusd, ma, &sc->sc_ring_gntref); |
448 | if (error) | | 448 | if (error) |
449 | goto abort_resume; | | 449 | goto abort_resume; |
450 | | | 450 | |
451 | error = xenbus_alloc_evtchn(sc->sc_xbusd, &sc->sc_evtchn); | | 451 | error = xenbus_alloc_evtchn(sc->sc_xbusd, &sc->sc_evtchn); |
452 | if (error) | | 452 | if (error) |
453 | goto abort_resume; | | 453 | goto abort_resume; |
454 | | | 454 | |
455 | aprint_verbose_dev(dev, "using event channel %d\n", | | 455 | aprint_verbose_dev(dev, "using event channel %d\n", |
456 | sc->sc_evtchn); | | 456 | sc->sc_evtchn); |
457 | sc->sc_ih = intr_establish_xname(0, &xen_pic, sc->sc_evtchn, IST_LEVEL, | | 457 | sc->sc_ih = intr_establish_xname(0, &xen_pic, sc->sc_evtchn, IST_LEVEL, |
458 | IPL_BIO, &xbd_handler, sc, false, device_xname(dev)); | | 458 | IPL_BIO, &xbd_handler, sc, false, device_xname(dev)); |
459 | KASSERT(sc->sc_ih != NULL); | | 459 | KASSERT(sc->sc_ih != NULL); |
460 | | | 460 | |
461 | again: | | 461 | again: |
462 | xbt = xenbus_transaction_start(); | | 462 | xbt = xenbus_transaction_start(); |
463 | if (xbt == NULL) | | 463 | if (xbt == NULL) |
464 | return false; | | 464 | return false; |
465 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, | | 465 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
466 | "ring-ref","%u", sc->sc_ring_gntref); | | 466 | "ring-ref","%u", sc->sc_ring_gntref); |
467 | if (error) { | | 467 | if (error) { |
468 | errmsg = "writing ring-ref"; | | 468 | errmsg = "writing ring-ref"; |
469 | goto abort_transaction; | | 469 | goto abort_transaction; |
470 | } | | 470 | } |
471 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, | | 471 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
472 | "event-channel", "%u", sc->sc_evtchn); | | 472 | "event-channel", "%u", sc->sc_evtchn); |
473 | if (error) { | | 473 | if (error) { |
474 | errmsg = "writing event channel"; | | 474 | errmsg = "writing event channel"; |
475 | goto abort_transaction; | | 475 | goto abort_transaction; |
476 | } | | 476 | } |
477 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, | | 477 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
478 | "protocol", "%s", XEN_IO_PROTO_ABI_NATIVE); | | 478 | "protocol", "%s", XEN_IO_PROTO_ABI_NATIVE); |
479 | if (error) { | | 479 | if (error) { |
480 | errmsg = "writing protocol"; | | 480 | errmsg = "writing protocol"; |
481 | goto abort_transaction; | | 481 | goto abort_transaction; |
482 | } | | 482 | } |
483 | error = xenbus_transaction_end(xbt, 0); | | 483 | error = xenbus_transaction_end(xbt, 0); |
484 | if (error == EAGAIN) | | 484 | if (error == EAGAIN) |
485 | goto again; | | 485 | goto again; |
486 | if (error != 0) { | | 486 | if (error != 0) { |
487 | xenbus_dev_fatal(sc->sc_xbusd, error, | | 487 | xenbus_dev_fatal(sc->sc_xbusd, error, |
488 | "completing transaction"); | | 488 | "completing transaction"); |
489 | return false; | | 489 | return false; |
490 | } | | 490 | } |
491 | | | 491 | |
492 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateInitialised); | | 492 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateInitialised); |
493 | | | 493 | |
494 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { | | 494 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { |
495 | /* | | 495 | /* |
496 | * device was suspended, softc structures are | | 496 | * device was suspended, softc structures are |
497 | * already initialized - we use a shortcut | | 497 | * already initialized - we use a shortcut |
498 | */ | | 498 | */ |
499 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; | | 499 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; |
500 | xenbus_device_resume(sc->sc_xbusd); | | 500 | xenbus_device_resume(sc->sc_xbusd); |
501 | hypervisor_enable_event(sc->sc_evtchn); | | 501 | hypervisor_enable_event(sc->sc_evtchn); |
502 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); | | 502 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); |
503 | } | | 503 | } |
504 | | | 504 | |
505 | return true; | | 505 | return true; |
506 | | | 506 | |
507 | abort_resume: | | 507 | abort_resume: |
508 | xenbus_dev_fatal(sc->sc_xbusd, error, "resuming device"); | | 508 | xenbus_dev_fatal(sc->sc_xbusd, error, "resuming device"); |
509 | return false; | | 509 | return false; |
510 | | | 510 | |
511 | abort_transaction: | | 511 | abort_transaction: |
512 | xenbus_transaction_end(xbt, 1); | | 512 | xenbus_transaction_end(xbt, 1); |
513 | xenbus_dev_fatal(sc->sc_xbusd, error, "%s", errmsg); | | 513 | xenbus_dev_fatal(sc->sc_xbusd, error, "%s", errmsg); |
514 | return false; | | 514 | return false; |
515 | } | | 515 | } |
516 | | | 516 | |
517 | static void | | 517 | static void |
518 | xbd_backend_changed(void *arg, XenbusState new_state) | | 518 | xbd_backend_changed(void *arg, XenbusState new_state) |
519 | { | | 519 | { |
520 | struct xbd_xenbus_softc *sc = device_private((device_t)arg); | | 520 | struct xbd_xenbus_softc *sc = device_private((device_t)arg); |
521 | struct disk_geom *dg; | | 521 | struct disk_geom *dg; |
522 | | | 522 | |
523 | char buf[9]; | | 523 | char buf[9]; |
524 | int s; | | 524 | int s; |
525 | DPRINTF(("%s: new backend state %d\n", | | 525 | DPRINTF(("%s: new backend state %d\n", |
526 | device_xname(sc->sc_dksc.sc_dev), new_state)); | | 526 | device_xname(sc->sc_dksc.sc_dev), new_state)); |
527 | | | 527 | |
528 | switch (new_state) { | | 528 | switch (new_state) { |
529 | case XenbusStateUnknown: | | 529 | case XenbusStateUnknown: |
530 | case XenbusStateInitialising: | | 530 | case XenbusStateInitialising: |
531 | case XenbusStateInitWait: | | 531 | case XenbusStateInitWait: |
532 | case XenbusStateInitialised: | | 532 | case XenbusStateInitialised: |
533 | break; | | 533 | break; |
534 | case XenbusStateClosing: | | 534 | case XenbusStateClosing: |
535 | s = splbio(); /* XXXSMP */ | | 535 | s = splbio(); /* XXXSMP */ |
536 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) | | 536 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) |
537 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; | | 537 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; |
538 | /* wait for requests to complete */ | | 538 | /* wait for requests to complete */ |
539 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && | | 539 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
540 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { | | 540 | disk_isbusy(&sc->sc_dksc.sc_dkdev)) { |
541 | /* XXXSMP */ | | 541 | /* XXXSMP */ |
542 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach", hz/2); | | 542 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach", hz/2); |
543 | } | | 543 | } |
544 | splx(s); | | 544 | splx(s); |
545 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosed); | | 545 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosed); |
546 | break; | | 546 | break; |
547 | case XenbusStateConnected: | | 547 | case XenbusStateConnected: |
548 | /* | | 548 | /* |
549 | * note that xbd_backend_changed() can only be called by | | 549 | * note that xbd_backend_changed() can only be called by |
550 | * the xenbus thread. | | 550 | * the xenbus thread. |
551 | */ | | 551 | */ |
552 | | | 552 | |
553 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED || | | 553 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED || |
554 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED) | | 554 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED) |
555 | /* already connected */ | | 555 | /* already connected */ |
556 | return; | | 556 | return; |
557 | | | 557 | |
558 | xbd_connect(sc); | | 558 | xbd_connect(sc); |
559 | sc->sc_shutdown = BLKIF_SHUTDOWN_RUN; | | 559 | sc->sc_shutdown = BLKIF_SHUTDOWN_RUN; |
560 | hypervisor_enable_event(sc->sc_evtchn); | | 560 | hypervisor_enable_event(sc->sc_evtchn); |
561 | | | 561 | |
562 | sc->sc_xbdsize = | | 562 | sc->sc_xbdsize = |
563 | sc->sc_sectors * (uint64_t)sc->sc_secsize / DEV_BSIZE; | | 563 | sc->sc_sectors * (uint64_t)sc->sc_secsize / DEV_BSIZE; |
564 | dg = &sc->sc_dksc.sc_dkdev.dk_geom; | | 564 | dg = &sc->sc_dksc.sc_dkdev.dk_geom; |
565 | memset(dg, 0, sizeof(*dg)); | | 565 | memset(dg, 0, sizeof(*dg)); |
566 | | | 566 | |
567 | dg->dg_secperunit = sc->sc_xbdsize; | | 567 | dg->dg_secperunit = sc->sc_xbdsize; |
568 | dg->dg_secsize = DEV_BSIZE; | | 568 | dg->dg_secsize = DEV_BSIZE; |
569 | dg->dg_ntracks = 1; | | 569 | dg->dg_ntracks = 1; |
570 | // XXX: Ok to hard-code DEV_BSIZE? | | 570 | // XXX: Ok to hard-code DEV_BSIZE? |
571 | dg->dg_nsectors = 1024 * (1024 / dg->dg_secsize); | | 571 | dg->dg_nsectors = 1024 * (1024 / dg->dg_secsize); |
572 | dg->dg_ncylinders = dg->dg_secperunit / dg->dg_nsectors; | | 572 | dg->dg_ncylinders = dg->dg_secperunit / dg->dg_nsectors; |
573 | | | 573 | |
574 | bufq_alloc(&sc->sc_dksc.sc_bufq, "fcfs", 0); | | 574 | bufq_alloc(&sc->sc_dksc.sc_bufq, "fcfs", 0); |
575 | dk_attach(&sc->sc_dksc); | | 575 | dk_attach(&sc->sc_dksc); |
576 | disk_attach(&sc->sc_dksc.sc_dkdev); | | 576 | disk_attach(&sc->sc_dksc.sc_dkdev); |
577 | | | 577 | |
578 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; | | 578 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; |
579 | | | 579 | |
580 | /* try to read the disklabel */ | | 580 | /* try to read the disklabel */ |
581 | dk_getdisklabel(&sc->sc_dksc, 0 /* XXX ? */); | | 581 | dk_getdisklabel(&sc->sc_dksc, 0 /* XXX ? */); |
582 | format_bytes(buf, sizeof(buf), sc->sc_sectors * sc->sc_secsize); | | 582 | format_bytes(buf, sizeof(buf), sc->sc_sectors * sc->sc_secsize); |
583 | aprint_verbose_dev(sc->sc_dksc.sc_dev, | | 583 | aprint_verbose_dev(sc->sc_dksc.sc_dev, |
584 | "%s, %d bytes/sect x %" PRIu64 " sectors\n", | | 584 | "%s, %d bytes/sect x %" PRIu64 " sectors\n", |
585 | buf, (int)dg->dg_secsize, sc->sc_xbdsize); | | 585 | buf, (int)dg->dg_secsize, sc->sc_xbdsize); |
586 | /* Discover wedges on this disk. */ | | 586 | /* Discover wedges on this disk. */ |
587 | dkwedge_discover(&sc->sc_dksc.sc_dkdev); | | 587 | dkwedge_discover(&sc->sc_dksc.sc_dkdev); |
588 | | | 588 | |
589 | disk_set_info(sc->sc_dksc.sc_dev, &sc->sc_dksc.sc_dkdev, NULL); | | 589 | disk_set_info(sc->sc_dksc.sc_dev, &sc->sc_dksc.sc_dkdev, NULL); |
590 | | | 590 | |
591 | /* the disk should be working now */ | | 591 | /* the disk should be working now */ |
592 | config_pending_decr(sc->sc_dksc.sc_dev); | | 592 | config_pending_decr(sc->sc_dksc.sc_dev); |
593 | break; | | 593 | break; |
594 | default: | | 594 | default: |
595 | panic("bad backend state %d", new_state); | | 595 | panic("bad backend state %d", new_state); |
596 | } | | 596 | } |
597 | } | | 597 | } |
598 | | | 598 | |
599 | static void | | 599 | static void |
600 | xbd_connect(struct xbd_xenbus_softc *sc) | | 600 | xbd_connect(struct xbd_xenbus_softc *sc) |
601 | { | | 601 | { |
602 | int err; | | 602 | int err; |
603 | unsigned long long sectors; | | 603 | unsigned long long sectors; |
604 | u_long cache_flush; | | 604 | u_long cache_flush; |
605 | | | 605 | |
606 | err = xenbus_read_ul(NULL, | | 606 | err = xenbus_read_ul(NULL, |
607 | sc->sc_xbusd->xbusd_path, "virtual-device", &sc->sc_handle, 10); | | 607 | sc->sc_xbusd->xbusd_path, "virtual-device", &sc->sc_handle, 10); |
608 | if (err) | | 608 | if (err) |
609 | panic("%s: can't read number from %s/virtual-device\n", | | 609 | panic("%s: can't read number from %s/virtual-device\n", |
610 | device_xname(sc->sc_dksc.sc_dev), | | 610 | device_xname(sc->sc_dksc.sc_dev), |
611 | sc->sc_xbusd->xbusd_otherend); | | 611 | sc->sc_xbusd->xbusd_otherend); |
612 | err = xenbus_read_ull(NULL, | | 612 | err = xenbus_read_ull(NULL, |
613 | sc->sc_xbusd->xbusd_otherend, "sectors", §ors, 10); | | 613 | sc->sc_xbusd->xbusd_otherend, "sectors", §ors, 10); |
614 | if (err) | | 614 | if (err) |
615 | panic("%s: can't read number from %s/sectors\n", | | 615 | panic("%s: can't read number from %s/sectors\n", |
616 | device_xname(sc->sc_dksc.sc_dev), | | 616 | device_xname(sc->sc_dksc.sc_dev), |
617 | sc->sc_xbusd->xbusd_otherend); | | 617 | sc->sc_xbusd->xbusd_otherend); |
618 | sc->sc_sectors = sectors; | | 618 | sc->sc_sectors = sectors; |
619 | | | 619 | |
620 | err = xenbus_read_ul(NULL, | | 620 | err = xenbus_read_ul(NULL, |
621 | sc->sc_xbusd->xbusd_otherend, "info", &sc->sc_info, 10); | | 621 | sc->sc_xbusd->xbusd_otherend, "info", &sc->sc_info, 10); |
622 | if (err) | | 622 | if (err) |
623 | panic("%s: can't read number from %s/info\n", | | 623 | panic("%s: can't read number from %s/info\n", |
624 | device_xname(sc->sc_dksc.sc_dev), | | 624 | device_xname(sc->sc_dksc.sc_dev), |
625 | sc->sc_xbusd->xbusd_otherend); | | 625 | sc->sc_xbusd->xbusd_otherend); |
626 | err = xenbus_read_ul(NULL, | | 626 | err = xenbus_read_ul(NULL, |
627 | sc->sc_xbusd->xbusd_otherend, "sector-size", &sc->sc_secsize, 10); | | 627 | sc->sc_xbusd->xbusd_otherend, "sector-size", &sc->sc_secsize, 10); |
628 | if (err) | | 628 | if (err) |
629 | panic("%s: can't read number from %s/sector-size\n", | | 629 | panic("%s: can't read number from %s/sector-size\n", |
630 | device_xname(sc->sc_dksc.sc_dev), | | 630 | device_xname(sc->sc_dksc.sc_dev), |
631 | sc->sc_xbusd->xbusd_otherend); | | 631 | sc->sc_xbusd->xbusd_otherend); |
632 | | | 632 | |
633 | err = xenbus_read_ul(NULL, sc->sc_xbusd->xbusd_otherend, | | 633 | err = xenbus_read_ul(NULL, sc->sc_xbusd->xbusd_otherend, |
634 | "feature-flush-cache", &cache_flush, 10); | | 634 | "feature-flush-cache", &cache_flush, 10); |
635 | if (err) | | 635 | if (err) |
636 | cache_flush = 0; | | 636 | cache_flush = 0; |
637 | if (cache_flush > 0) | | 637 | if (cache_flush > 0) |
638 | sc->sc_cache_flush = 1; | | 638 | sc->sc_cache_flush = 1; |
639 | else | | 639 | else |
640 | sc->sc_cache_flush = 0; | | 640 | sc->sc_cache_flush = 0; |
641 | | | 641 | |
642 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); | | 642 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); |
643 | } | | 643 | } |
644 | | | 644 | |
645 | static int | | 645 | static int |
646 | xbd_handler(void *arg) | | 646 | xbd_handler(void *arg) |
647 | { | | 647 | { |
648 | struct xbd_xenbus_softc *sc = arg; | | 648 | struct xbd_xenbus_softc *sc = arg; |
649 | struct buf *bp; | | 649 | struct buf *bp; |
650 | RING_IDX resp_prod, i; | | 650 | RING_IDX resp_prod, i; |
651 | int more_to_do; | | 651 | int more_to_do; |
652 | int seg; | | 652 | int seg; |
653 | | | 653 | |
654 | DPRINTF(("xbd_handler(%s)\n", device_xname(sc->sc_dksc.sc_dev))); | | 654 | DPRINTF(("xbd_handler(%s)\n", device_xname(sc->sc_dksc.sc_dev))); |
655 | | | 655 | |
656 | if (__predict_false(sc->sc_backend_status != BLKIF_STATE_CONNECTED)) | | 656 | if (__predict_false(sc->sc_backend_status != BLKIF_STATE_CONNECTED)) |
657 | return 0; | | 657 | return 0; |
658 | again: | | 658 | again: |
659 | resp_prod = sc->sc_ring.sring->rsp_prod; | | 659 | resp_prod = sc->sc_ring.sring->rsp_prod; |
660 | xen_rmb(); /* ensure we see replies up to resp_prod */ | | 660 | xen_rmb(); /* ensure we see replies up to resp_prod */ |
661 | for (i = sc->sc_ring.rsp_cons; i != resp_prod; i++) { | | 661 | for (i = sc->sc_ring.rsp_cons; i != resp_prod; i++) { |
662 | blkif_response_t *rep = RING_GET_RESPONSE(&sc->sc_ring, i); | | 662 | blkif_response_t *rep = RING_GET_RESPONSE(&sc->sc_ring, i); |
663 | struct xbd_req *xbdreq = &sc->sc_reqs[rep->id]; | | 663 | struct xbd_req *xbdreq = &sc->sc_reqs[rep->id]; |
664 | bp = xbdreq->req_bp; | | 664 | |
665 | DPRINTF(("xbd_handler(%p): b_bcount = %ld\n", | | | |
666 | xbdreq->req_bp, (long)bp->b_bcount)); | | | |
667 | if (rep->operation == BLKIF_OP_FLUSH_DISKCACHE) { | | 665 | if (rep->operation == BLKIF_OP_FLUSH_DISKCACHE) { |
| | | 666 | KASSERT(xbdreq->req_bp == NULL); |
668 | xbdreq->req_sync.s_error = rep->status; | | 667 | xbdreq->req_sync.s_error = rep->status; |
669 | xbdreq->req_sync.s_done = 1; | | 668 | xbdreq->req_sync.s_done = 1; |
670 | wakeup(xbdreq); /* XXXSMP */ | | 669 | wakeup(xbdreq); /* XXXSMP */ |
671 | /* caller will free the req */ | | 670 | /* caller will free the req */ |
672 | continue; | | 671 | continue; |
673 | } | | 672 | } |
| | | 673 | |
| | | 674 | if (rep->operation != BLKIF_OP_READ && |
| | | 675 | rep->operation != BLKIF_OP_WRITE) { |
| | | 676 | aprint_error_dev(sc->sc_dksc.sc_dev, |
| | | 677 | "bad operation %d from backend\n", rep->operation); |
| | | 678 | continue; |
| | | 679 | } |
| | | 680 | |
674 | for (seg = xbdreq->req_nr_segments - 1; seg >= 0; seg--) { | | 681 | for (seg = xbdreq->req_nr_segments - 1; seg >= 0; seg--) { |
675 | if (__predict_false( | | 682 | if (__predict_false( |
676 | xengnt_status(xbdreq->req_gntref[seg]))) { | | 683 | xengnt_status(xbdreq->req_gntref[seg]))) { |
677 | aprint_verbose_dev(sc->sc_dksc.sc_dev, | | 684 | aprint_verbose_dev(sc->sc_dksc.sc_dev, |
678 | "grant still used by backend\n"); | | 685 | "grant still used by backend\n"); |
679 | sc->sc_ring.rsp_cons = i; | | 686 | sc->sc_ring.rsp_cons = i; |
680 | xbdreq->req_nr_segments = seg + 1; | | 687 | xbdreq->req_nr_segments = seg + 1; |
681 | goto done; | | 688 | goto done; |
682 | } | | 689 | } |
683 | xengnt_revoke_access(xbdreq->req_gntref[seg]); | | 690 | xengnt_revoke_access(xbdreq->req_gntref[seg]); |
684 | xbdreq->req_nr_segments--; | | 691 | xbdreq->req_nr_segments--; |
685 | } | | 692 | } |
686 | if (rep->operation != BLKIF_OP_READ && | | 693 | KASSERT(xbdreq->req_nr_segments == 0); |
687 | rep->operation != BLKIF_OP_WRITE) { | | 694 | |
688 | aprint_error_dev(sc->sc_dksc.sc_dev, | | 695 | bp = xbdreq->req_bp; |
689 | "bad operation %d from backend\n", rep->operation); | | 696 | xbdreq->req_bp = NULL; |
690 | bp->b_error = EIO; | | 697 | DPRINTF(("%s(%p): b_bcount = %ld\n", __func__, |
691 | bp->b_resid = bp->b_bcount; | | 698 | bp, (long)bp->b_bcount)); |
692 | goto next; | | 699 | |
693 | } | | | |
694 | if (rep->status != BLKIF_RSP_OKAY) { | | 700 | if (rep->status != BLKIF_RSP_OKAY) { |
695 | bp->b_error = EIO; | | 701 | bp->b_error = EIO; |
696 | bp->b_resid = bp->b_bcount; | | 702 | bp->b_resid = bp->b_bcount; |
697 | goto next; | | 703 | goto next; |
698 | } | | 704 | } |
699 | /* b_resid was set in dk_start */ | | 705 | /* b_resid was set in dk_start */ |
700 | next: | | 706 | next: |
701 | dk_done(&sc->sc_dksc, bp); | | 707 | dk_done(&sc->sc_dksc, bp); |
702 | | | 708 | |
703 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, req_next); | | 709 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, req_next); |
704 | } | | 710 | } |
705 | done: | | 711 | done: |
706 | xen_rmb(); | | 712 | xen_rmb(); |
707 | sc->sc_ring.rsp_cons = i; | | 713 | sc->sc_ring.rsp_cons = i; |
708 | | | 714 | |
709 | RING_FINAL_CHECK_FOR_RESPONSES(&sc->sc_ring, more_to_do); | | 715 | RING_FINAL_CHECK_FOR_RESPONSES(&sc->sc_ring, more_to_do); |
710 | if (more_to_do) | | 716 | if (more_to_do) |
711 | goto again; | | 717 | goto again; |
712 | | | 718 | |
713 | if (sc->sc_xbdreq_wait) | | 719 | if (sc->sc_xbdreq_wait) |
714 | wakeup(&sc->sc_xbdreq_wait); /* XXXSMP */ | | 720 | wakeup(&sc->sc_xbdreq_wait); /* XXXSMP */ |
715 | else | | 721 | else |
716 | dk_start(&sc->sc_dksc, NULL); | | 722 | dk_start(&sc->sc_dksc, NULL); |
717 | return 1; | | 723 | return 1; |
718 | } | | 724 | } |
719 | | | 725 | |
720 | static void | | 726 | static void |
721 | xbdminphys(struct buf *bp) | | 727 | xbdminphys(struct buf *bp) |
722 | { | | 728 | { |
723 | if (bp->b_bcount > XBD_MAX_XFER) { | | 729 | if (bp->b_bcount > XBD_MAX_XFER) { |
724 | bp->b_bcount = XBD_MAX_XFER; | | 730 | bp->b_bcount = XBD_MAX_XFER; |
725 | } | | 731 | } |
726 | minphys(bp); | | 732 | minphys(bp); |
727 | } | | 733 | } |
728 | | | 734 | |
729 | int | | 735 | int |
730 | xbdopen(dev_t dev, int flags, int fmt, struct lwp *l) | | 736 | xbdopen(dev_t dev, int flags, int fmt, struct lwp *l) |
731 | { | | 737 | { |
732 | struct xbd_xenbus_softc *sc; | | 738 | struct xbd_xenbus_softc *sc; |
733 | | | 739 | |
734 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 740 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
735 | if (sc == NULL) | | 741 | if (sc == NULL) |
736 | return (ENXIO); | | 742 | return (ENXIO); |
737 | if ((flags & FWRITE) && (sc->sc_info & VDISK_READONLY)) | | 743 | if ((flags & FWRITE) && (sc->sc_info & VDISK_READONLY)) |
738 | return EROFS; | | 744 | return EROFS; |
739 | | | 745 | |
740 | DPRINTF(("xbdopen(0x%04x, %d)\n", dev, flags)); | | 746 | DPRINTF(("xbdopen(0x%04x, %d)\n", dev, flags)); |
741 | return dk_open(&sc->sc_dksc, dev, flags, fmt, l); | | 747 | return dk_open(&sc->sc_dksc, dev, flags, fmt, l); |
742 | } | | 748 | } |
743 | | | 749 | |
744 | int | | 750 | int |
745 | xbdclose(dev_t dev, int flags, int fmt, struct lwp *l) | | 751 | xbdclose(dev_t dev, int flags, int fmt, struct lwp *l) |
746 | { | | 752 | { |
747 | struct xbd_xenbus_softc *sc; | | 753 | struct xbd_xenbus_softc *sc; |
748 | | | 754 | |
749 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 755 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
750 | | | 756 | |
751 | DPRINTF(("xbdclose(%d, %d)\n", dev, flags)); | | 757 | DPRINTF(("xbdclose(%d, %d)\n", dev, flags)); |
752 | return dk_close(&sc->sc_dksc, dev, flags, fmt, l); | | 758 | return dk_close(&sc->sc_dksc, dev, flags, fmt, l); |
753 | } | | 759 | } |
754 | | | 760 | |
755 | void | | 761 | void |
756 | xbdstrategy(struct buf *bp) | | 762 | xbdstrategy(struct buf *bp) |
757 | { | | 763 | { |
758 | struct xbd_xenbus_softc *sc; | | 764 | struct xbd_xenbus_softc *sc; |
759 | | | 765 | |
760 | sc = device_lookup_private(&xbd_cd, DISKUNIT(bp->b_dev)); | | 766 | sc = device_lookup_private(&xbd_cd, DISKUNIT(bp->b_dev)); |
761 | | | 767 | |
762 | DPRINTF(("xbdstrategy(%p): b_bcount = %ld\n", bp, | | 768 | DPRINTF(("xbdstrategy(%p): b_bcount = %ld\n", bp, |
763 | (long)bp->b_bcount)); | | 769 | (long)bp->b_bcount)); |
764 | | | 770 | |
765 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { | | 771 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { |
766 | bp->b_error = EIO; | | 772 | bp->b_error = EIO; |
767 | biodone(bp); | | 773 | biodone(bp); |
768 | return; | | 774 | return; |
769 | } | | 775 | } |
770 | if (__predict_false((sc->sc_info & VDISK_READONLY) && | | 776 | if (__predict_false((sc->sc_info & VDISK_READONLY) && |
771 | (bp->b_flags & B_READ) == 0)) { | | 777 | (bp->b_flags & B_READ) == 0)) { |
772 | bp->b_error = EROFS; | | 778 | bp->b_error = EROFS; |
773 | biodone(bp); | | 779 | biodone(bp); |
774 | return; | | 780 | return; |
775 | } | | 781 | } |
776 | | | 782 | |
777 | dk_strategy(&sc->sc_dksc, bp); | | 783 | dk_strategy(&sc->sc_dksc, bp); |
778 | return; | | 784 | return; |
779 | } | | 785 | } |
780 | | | 786 | |
781 | int | | 787 | int |
782 | xbdsize(dev_t dev) | | 788 | xbdsize(dev_t dev) |
783 | { | | 789 | { |
784 | struct xbd_xenbus_softc *sc; | | 790 | struct xbd_xenbus_softc *sc; |
785 | | | 791 | |
786 | DPRINTF(("xbdsize(%d)\n", dev)); | | 792 | DPRINTF(("xbdsize(%d)\n", dev)); |
787 | | | 793 | |
788 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 794 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
789 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) | | 795 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) |
790 | return -1; | | 796 | return -1; |
791 | return dk_size(&sc->sc_dksc, dev); | | 797 | return dk_size(&sc->sc_dksc, dev); |
792 | } | | 798 | } |
793 | | | 799 | |
794 | int | | 800 | int |
795 | xbdread(dev_t dev, struct uio *uio, int flags) | | 801 | xbdread(dev_t dev, struct uio *uio, int flags) |
796 | { | | 802 | { |
797 | struct xbd_xenbus_softc *sc = | | 803 | struct xbd_xenbus_softc *sc = |
798 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 804 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
799 | struct dk_softc *dksc = &sc->sc_dksc; | | 805 | struct dk_softc *dksc = &sc->sc_dksc; |
800 | | | 806 | |
801 | if (!DK_ATTACHED(dksc)) | | 807 | if (!DK_ATTACHED(dksc)) |
802 | return ENXIO; | | 808 | return ENXIO; |
803 | return physio(xbdstrategy, NULL, dev, B_READ, xbdminphys, uio); | | 809 | return physio(xbdstrategy, NULL, dev, B_READ, xbdminphys, uio); |
804 | } | | 810 | } |
805 | | | 811 | |
806 | int | | 812 | int |
807 | xbdwrite(dev_t dev, struct uio *uio, int flags) | | 813 | xbdwrite(dev_t dev, struct uio *uio, int flags) |
808 | { | | 814 | { |
809 | struct xbd_xenbus_softc *sc = | | 815 | struct xbd_xenbus_softc *sc = |
810 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 816 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
811 | struct dk_softc *dksc = &sc->sc_dksc; | | 817 | struct dk_softc *dksc = &sc->sc_dksc; |
812 | | | 818 | |
813 | if (!DK_ATTACHED(dksc)) | | 819 | if (!DK_ATTACHED(dksc)) |
814 | return ENXIO; | | 820 | return ENXIO; |
815 | if (__predict_false(sc->sc_info & VDISK_READONLY)) | | 821 | if (__predict_false(sc->sc_info & VDISK_READONLY)) |
816 | return EROFS; | | 822 | return EROFS; |
817 | return physio(xbdstrategy, NULL, dev, B_WRITE, xbdminphys, uio); | | 823 | return physio(xbdstrategy, NULL, dev, B_WRITE, xbdminphys, uio); |
818 | } | | 824 | } |
819 | | | 825 | |
820 | int | | 826 | int |
821 | xbdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) | | 827 | xbdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) |
822 | { | | 828 | { |
823 | struct xbd_xenbus_softc *sc = | | 829 | struct xbd_xenbus_softc *sc = |
824 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 830 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
825 | struct dk_softc *dksc; | | 831 | struct dk_softc *dksc; |
826 | int error; | | 832 | int error; |
827 | int s; | | 833 | int s; |
828 | struct xbd_req *xbdreq; | | 834 | struct xbd_req *xbdreq; |
829 | blkif_request_t *req; | | 835 | blkif_request_t *req; |
830 | int notify; | | 836 | int notify; |
831 | | | 837 | |
832 | DPRINTF(("xbdioctl(%d, %08lx, %p, %d, %p)\n", | | 838 | DPRINTF(("xbdioctl(%d, %08lx, %p, %d, %p)\n", |
833 | dev, cmd, data, flag, l)); | | 839 | dev, cmd, data, flag, l)); |
834 | dksc = &sc->sc_dksc; | | 840 | dksc = &sc->sc_dksc; |
835 | | | 841 | |
836 | switch (cmd) { | | 842 | switch (cmd) { |
837 | case DIOCSSTRATEGY: | | 843 | case DIOCSSTRATEGY: |
838 | error = EOPNOTSUPP; | | 844 | error = EOPNOTSUPP; |
839 | break; | | 845 | break; |
840 | case DIOCCACHESYNC: | | 846 | case DIOCCACHESYNC: |
841 | if (sc->sc_cache_flush <= 0) { | | 847 | if (sc->sc_cache_flush <= 0) { |
842 | if (sc->sc_cache_flush == 0) { | | 848 | if (sc->sc_cache_flush == 0) { |
843 | aprint_error_dev(sc->sc_dksc.sc_dev, | | 849 | aprint_error_dev(sc->sc_dksc.sc_dev, |
844 | "WARNING: cache flush not supported " | | 850 | "WARNING: cache flush not supported " |
845 | "by backend\n"); | | 851 | "by backend\n"); |
846 | sc->sc_cache_flush = -1; | | 852 | sc->sc_cache_flush = -1; |
847 | } | | 853 | } |
848 | return EOPNOTSUPP; | | 854 | return EOPNOTSUPP; |
849 | } | | 855 | } |
850 | | | 856 | |
851 | s = splbio(); /* XXXSMP */ | | 857 | s = splbio(); /* XXXSMP */ |
852 | | | 858 | |
853 | while (RING_FULL(&sc->sc_ring)) { | | 859 | while (RING_FULL(&sc->sc_ring)) { |
854 | sc->sc_xbdreq_wait = 1; | | 860 | sc->sc_xbdreq_wait = 1; |
855 | /* XXXSMP */ | | 861 | /* XXXSMP */ |
856 | tsleep(&sc->sc_xbdreq_wait, PRIBIO, "xbdreq", 0); | | 862 | tsleep(&sc->sc_xbdreq_wait, PRIBIO, "xbdreq", 0); |
857 | } | | 863 | } |
858 | sc->sc_xbdreq_wait = 0; | | 864 | sc->sc_xbdreq_wait = 0; |
859 | | | 865 | |
860 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); | | 866 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); |
861 | if (__predict_false(xbdreq == NULL)) { | | 867 | if (__predict_false(xbdreq == NULL)) { |
862 | DPRINTF(("xbdioctl: no req\n")); | | 868 | DPRINTF(("xbdioctl: no req\n")); |
863 | error = ENOMEM; | | 869 | error = ENOMEM; |
864 | } else { | | 870 | } else { |
865 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); | | 871 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); |
866 | req = RING_GET_REQUEST(&sc->sc_ring, | | 872 | req = RING_GET_REQUEST(&sc->sc_ring, |
867 | sc->sc_ring.req_prod_pvt); | | 873 | sc->sc_ring.req_prod_pvt); |
868 | req->id = xbdreq->req_id; | | 874 | req->id = xbdreq->req_id; |
869 | req->operation = BLKIF_OP_FLUSH_DISKCACHE; | | 875 | req->operation = BLKIF_OP_FLUSH_DISKCACHE; |
870 | req->handle = sc->sc_handle; | | 876 | req->handle = sc->sc_handle; |
871 | xbdreq->req_sync.s_done = 0; | | 877 | xbdreq->req_sync.s_done = 0; |
872 | sc->sc_ring.req_prod_pvt++; | | 878 | sc->sc_ring.req_prod_pvt++; |
873 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, | | 879 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, |
874 | notify); | | 880 | notify); |
875 | if (notify) | | 881 | if (notify) |
876 | hypervisor_notify_via_evtchn(sc->sc_evtchn); | | 882 | hypervisor_notify_via_evtchn(sc->sc_evtchn); |
877 | /* request sent, now wait for completion */ | | 883 | /* request sent, now wait for completion */ |
878 | while (xbdreq->req_sync.s_done == 0) { | | 884 | while (xbdreq->req_sync.s_done == 0) { |
879 | /* XXXSMP */ | | 885 | /* XXXSMP */ |
880 | tsleep(xbdreq, PRIBIO, "xbdsync", 0); | | 886 | tsleep(xbdreq, PRIBIO, "xbdsync", 0); |
881 | } | | 887 | } |
882 | if (xbdreq->req_sync.s_error == BLKIF_RSP_EOPNOTSUPP) | | 888 | if (xbdreq->req_sync.s_error == BLKIF_RSP_EOPNOTSUPP) |
883 | error = EOPNOTSUPP; | | 889 | error = EOPNOTSUPP; |
884 | else if (xbdreq->req_sync.s_error == BLKIF_RSP_OKAY) | | 890 | else if (xbdreq->req_sync.s_error == BLKIF_RSP_OKAY) |
885 | error = 0; | | 891 | error = 0; |
886 | else | | 892 | else |
887 | error = EIO; | | 893 | error = EIO; |
888 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, | | 894 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, |
889 | req_next); | | 895 | req_next); |
890 | } | | 896 | } |
891 | splx(s); | | 897 | splx(s); |
892 | break; | | 898 | break; |
893 | | | 899 | |
894 | default: | | 900 | default: |
895 | error = dk_ioctl(dksc, dev, cmd, data, flag, l); | | 901 | error = dk_ioctl(dksc, dev, cmd, data, flag, l); |
896 | break; | | 902 | break; |
897 | } | | 903 | } |
898 | | | 904 | |
899 | return error; | | 905 | return error; |
900 | } | | 906 | } |
901 | | | 907 | |
902 | int | | 908 | int |
903 | xbddump(dev_t dev, daddr_t blkno, void *va, size_t size) | | 909 | xbddump(dev_t dev, daddr_t blkno, void *va, size_t size) |
904 | { | | 910 | { |
905 | struct xbd_xenbus_softc *sc; | | 911 | struct xbd_xenbus_softc *sc; |
906 | | | 912 | |
907 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); | | 913 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
908 | if (sc == NULL) | | 914 | if (sc == NULL) |
909 | return (ENXIO); | | 915 | return (ENXIO); |
910 | | | 916 | |
911 | DPRINTF(("xbddump(%d, %" PRId64 ", %p, %lu)\n", dev, blkno, va, | | 917 | DPRINTF(("xbddump(%d, %" PRId64 ", %p, %lu)\n", dev, blkno, va, |
912 | (unsigned long)size)); | | 918 | (unsigned long)size)); |
913 | return dk_dump(&sc->sc_dksc, dev, blkno, va, size); | | 919 | return dk_dump(&sc->sc_dksc, dev, blkno, va, size); |
914 | } | | 920 | } |
915 | | | 921 | |
916 | static int | | 922 | static int |
917 | xbd_diskstart(device_t self, struct buf *bp) | | 923 | xbd_diskstart(device_t self, struct buf *bp) |
918 | { | | 924 | { |
919 | struct xbd_xenbus_softc *sc = device_private(self); | | 925 | struct xbd_xenbus_softc *sc = device_private(self); |
920 | struct xbd_req *xbdreq; | | 926 | struct xbd_req *xbdreq; |
921 | blkif_request_t *req; | | 927 | blkif_request_t *req; |
922 | size_t bcount, off; | | 928 | size_t bcount, off; |
923 | paddr_t ma; | | 929 | paddr_t ma; |
924 | vaddr_t va; | | 930 | vaddr_t va; |
925 | int nsects, nbytes, seg; | | 931 | int nsects, nbytes, seg; |
926 | int notify, error = 0; | | 932 | int notify, error = 0; |
927 | int s; | | 933 | int s; |
928 | | | 934 | |
929 | DPRINTF(("xbd_diskstart(%p): b_bcount = %ld\n", | | 935 | DPRINTF(("xbd_diskstart(%p): b_bcount = %ld\n", |
930 | bp, (long)bp->b_bcount)); | | 936 | bp, (long)bp->b_bcount)); |
931 | | | 937 | |
932 | if (sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { | | 938 | if (sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { |
933 | error = EIO; | | 939 | error = EIO; |
934 | goto err; | | 940 | goto err; |
935 | } | | 941 | } |
936 | | | 942 | |
937 | if (bp->b_rawblkno < 0 || bp->b_rawblkno > sc->sc_xbdsize) { | | 943 | if (bp->b_rawblkno < 0 || bp->b_rawblkno > sc->sc_xbdsize) { |
938 | /* invalid block number */ | | 944 | /* invalid block number */ |
939 | error = EINVAL; | | 945 | error = EINVAL; |
940 | goto err; | | 946 | goto err; |
941 | } | | 947 | } |
942 | | | 948 | |
943 | s = splbio(); /* XXX SMP */ | | 949 | s = splbio(); /* XXX SMP */ |
944 | | | 950 | |
945 | if (__predict_false( | | 951 | if (__predict_false( |
946 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED)) { | | 952 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED)) { |
947 | /* device is suspended, do not consume buffer */ | | 953 | /* device is suspended, do not consume buffer */ |
948 | DPRINTF(("%s: (xbd_diskstart) device suspended\n", | | 954 | DPRINTF(("%s: (xbd_diskstart) device suspended\n", |
949 | sc->sc_dksc.sc_xname)); | | 955 | sc->sc_dksc.sc_xname)); |
950 | error = EAGAIN; | | 956 | error = EAGAIN; |
951 | goto out; | | 957 | goto out; |
952 | } | | 958 | } |
953 | | | 959 | |
954 | if (RING_FULL(&sc->sc_ring) || sc->sc_xbdreq_wait) { | | 960 | if (RING_FULL(&sc->sc_ring) || sc->sc_xbdreq_wait) { |
955 | DPRINTF(("xbd_diskstart: ring_full\n")); | | 961 | DPRINTF(("xbd_diskstart: ring_full\n")); |
956 | error = EAGAIN; | | 962 | error = EAGAIN; |
957 | goto out; | | 963 | goto out; |
958 | } | | 964 | } |
959 | | | 965 | |
960 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); | | 966 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); |
961 | if (__predict_false(xbdreq == NULL)) { | | 967 | if (__predict_false(xbdreq == NULL)) { |
962 | DPRINTF(("xbd_diskstart: no req\n")); | | 968 | DPRINTF(("xbd_diskstart: no req\n")); |
963 | error = EAGAIN; | | 969 | error = EAGAIN; |
964 | goto out; | | 970 | goto out; |
965 | } | | 971 | } |
966 | | | 972 | |
967 | if ((vaddr_t)bp->b_data & (XEN_BSIZE - 1)) { | | 973 | if ((vaddr_t)bp->b_data & (XEN_BSIZE - 1)) { |
968 | DPRINTF(("xbd_diskstart: no align\n")); | | 974 | DPRINTF(("xbd_diskstart: no align\n")); |
969 | error = EINVAL; | | 975 | error = EINVAL; |
970 | goto out; | | 976 | goto out; |
971 | } | | 977 | } |
972 | | | 978 | |
973 | xbdreq->req_bp = bp; | | 979 | xbdreq->req_bp = bp; |
974 | | | 980 | |
975 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); | | 981 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); |
976 | req = RING_GET_REQUEST(&sc->sc_ring, sc->sc_ring.req_prod_pvt); | | 982 | req = RING_GET_REQUEST(&sc->sc_ring, sc->sc_ring.req_prod_pvt); |
977 | req->id = xbdreq->req_id; | | 983 | req->id = xbdreq->req_id; |
978 | req->operation = | | 984 | req->operation = |
979 | bp->b_flags & B_READ ? BLKIF_OP_READ : BLKIF_OP_WRITE; | | 985 | bp->b_flags & B_READ ? BLKIF_OP_READ : BLKIF_OP_WRITE; |
980 | req->sector_number = bp->b_rawblkno; | | 986 | req->sector_number = bp->b_rawblkno; |
981 | req->handle = sc->sc_handle; | | 987 | req->handle = sc->sc_handle; |
982 | | | 988 | |
983 | va = (vaddr_t)bp->b_data & ~PAGE_MASK; | | 989 | va = (vaddr_t)bp->b_data & ~PAGE_MASK; |
984 | off = (vaddr_t)bp->b_data & PAGE_MASK; | | 990 | off = (vaddr_t)bp->b_data & PAGE_MASK; |
985 | bcount = bp->b_bcount; | | 991 | bcount = bp->b_bcount; |
986 | bp->b_resid = 0; | | 992 | bp->b_resid = 0; |
987 | for (seg = 0; bcount > 0;) { | | 993 | for (seg = 0; bcount > 0;) { |
988 | pmap_extract_ma(pmap_kernel(), va, &ma); | | 994 | pmap_extract_ma(pmap_kernel(), va, &ma); |
989 | KASSERT((ma & (XEN_BSIZE - 1)) == 0); | | 995 | KASSERT((ma & (XEN_BSIZE - 1)) == 0); |
990 | if (bcount > PAGE_SIZE - off) | | 996 | if (bcount > PAGE_SIZE - off) |
991 | nbytes = PAGE_SIZE - off; | | 997 | nbytes = PAGE_SIZE - off; |
992 | else | | 998 | else |
993 | nbytes = bcount; | | 999 | nbytes = bcount; |
994 | nsects = nbytes >> XEN_BSHIFT; | | 1000 | nsects = nbytes >> XEN_BSHIFT; |
995 | req->seg[seg].first_sect = off >> XEN_BSHIFT; | | 1001 | req->seg[seg].first_sect = off >> XEN_BSHIFT; |
996 | req->seg[seg].last_sect = | | 1002 | req->seg[seg].last_sect = |
997 | (off >> XEN_BSHIFT) + nsects - 1; | | 1003 | (off >> XEN_BSHIFT) + nsects - 1; |
998 | KASSERT(req->seg[seg].first_sect <= | | 1004 | KASSERT(req->seg[seg].first_sect <= |
999 | req->seg[seg].last_sect); | | 1005 | req->seg[seg].last_sect); |
1000 | KASSERT(req->seg[seg].last_sect < 8); | | 1006 | KASSERT(req->seg[seg].last_sect < 8); |
1001 | if (__predict_false(xengnt_grant_access( | | 1007 | if (__predict_false(xengnt_grant_access( |
1002 | sc->sc_xbusd->xbusd_otherend_id, ma, | | 1008 | sc->sc_xbusd->xbusd_otherend_id, ma, |
1003 | (bp->b_flags & B_READ) == 0, | | 1009 | (bp->b_flags & B_READ) == 0, |
1004 | &xbdreq->req_gntref[seg]))) | | 1010 | &xbdreq->req_gntref[seg]))) |
1005 | panic("xbd_diskstart: xengnt_grant_access"); /* XXX XXX !!! */ | | 1011 | panic("xbd_diskstart: xengnt_grant_access"); /* XXX XXX !!! */ |
1006 | req->seg[seg].gref = xbdreq->req_gntref[seg]; | | 1012 | req->seg[seg].gref = xbdreq->req_gntref[seg]; |
1007 | seg++; | | 1013 | seg++; |
1008 | KASSERT(seg <= BLKIF_MAX_SEGMENTS_PER_REQUEST); | | 1014 | KASSERT(seg <= BLKIF_MAX_SEGMENTS_PER_REQUEST); |
1009 | va += PAGE_SIZE; | | 1015 | va += PAGE_SIZE; |
1010 | off = 0; | | 1016 | off = 0; |
1011 | bcount -= nbytes; | | 1017 | bcount -= nbytes; |
1012 | } | | 1018 | } |
1013 | xbdreq->req_nr_segments = req->nr_segments = seg; | | 1019 | xbdreq->req_nr_segments = req->nr_segments = seg; |
1014 | sc->sc_ring.req_prod_pvt++; | | 1020 | sc->sc_ring.req_prod_pvt++; |
1015 | | | 1021 | |
1016 | out: | | 1022 | out: |
1017 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, notify); | | 1023 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, notify); |
1018 | if (notify) | | 1024 | if (notify) |
1019 | hypervisor_notify_via_evtchn(sc->sc_evtchn); | | 1025 | hypervisor_notify_via_evtchn(sc->sc_evtchn); |
1020 | splx(s); /* XXXSMP */ | | 1026 | splx(s); /* XXXSMP */ |
1021 | err: | | 1027 | err: |
1022 | return error; | | 1028 | return error; |
1023 | } | | 1029 | } |