nvme_ns.c revision 240616 
1/*-
2 * Copyright (C) 2012 Intel Corporation
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: head/sys/dev/nvme/nvme_ns.c 240616 2012-09-17 19:23:01Z jimharris $");
29
30#include <sys/param.h>
31#include <sys/bio.h>
32#include <sys/bus.h>
33#include <sys/conf.h>
34#include <sys/disk.h>
35#include <sys/fcntl.h>
36#include <sys/ioccom.h>
37#include <sys/module.h>
38#include <sys/proc.h>
39
40#include <dev/pci/pcivar.h>
41
42#include "nvme_private.h"
43
44static void
45nvme_ns_cb(void *arg, const struct nvme_completion *status)
46{
47	struct nvme_completion	*cpl = arg;
48	struct mtx		*mtx;
49
50	/*
51	 * Copy status into the argument passed by the caller, so that
52	 *  the caller can check the status to determine if the
53	 *  the request passed or failed.
54	 */
55	memcpy(cpl, status, sizeof(*cpl));
56	mtx = mtx_pool_find(mtxpool_sleep, cpl);
57	mtx_lock(mtx);
58	wakeup(cpl);
59	mtx_unlock(mtx);
60}
61
62static int
63nvme_ns_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int flag,
64    struct thread *td)
65{
66	struct nvme_namespace	*ns;
67	struct nvme_controller	*ctrlr;
68	struct nvme_completion	cpl;
69	struct mtx		*mtx;
70
71	ns = cdev->si_drv1;
72	ctrlr = ns->ctrlr;
73
74	switch (cmd) {
75	case NVME_IDENTIFY_NAMESPACE:
76#ifdef CHATHAM2
77		/*
78		 * Don't refresh data on Chatham, since Chatham returns
79		 *  garbage on IDENTIFY anyways.
80		 */
81		if (pci_get_devid(ctrlr->dev) == CHATHAM_PCI_ID) {
82			memcpy(arg, &ns->data, sizeof(ns->data));
83			break;
84		}
85#endif
86		/* Refresh data before returning to user. */
87		mtx = mtx_pool_find(mtxpool_sleep, &cpl);
88		mtx_lock(mtx);
89		nvme_ctrlr_cmd_identify_namespace(ctrlr, ns->id, &ns->data,
90		    nvme_ns_cb, &cpl);
91		msleep(&cpl, mtx, PRIBIO, "nvme_ioctl", 0);
92		mtx_unlock(mtx);
93		if (cpl.sf_sc || cpl.sf_sct)
94			return (ENXIO);
95		memcpy(arg, &ns->data, sizeof(ns->data));
96		break;
97	case NVME_IO_TEST:
98	case NVME_BIO_TEST:
99		nvme_ns_test(ns, cmd, arg);
100		break;
101	case DIOCGMEDIASIZE:
102		*(off_t *)arg = (off_t)nvme_ns_get_size(ns);
103		break;
104	case DIOCGSECTORSIZE:
105		*(u_int *)arg = nvme_ns_get_sector_size(ns);
106		break;
107	default:
108		return (ENOTTY);
109	}
110
111	return (0);
112}
113
114static int
115nvme_ns_open(struct cdev *dev __unused, int flags, int fmt __unused,
116    struct thread *td)
117{
118	int error = 0;
119
120	if (flags & FWRITE)
121		error = securelevel_gt(td->td_ucred, 0);
122
123	return (error);
124}
125
126static int
127nvme_ns_close(struct cdev *dev __unused, int flags, int fmt __unused,
128    struct thread *td)
129{
130
131	return (0);
132}
133
134static void
135nvme_ns_strategy_done(void *arg, const struct nvme_completion *status)
136{
137	struct bio *bp = arg;
138
139	/*
140	 * TODO: add more extensive translation of NVMe status codes
141	 *  to different bio error codes (i.e. EIO, EINVAL, etc.)
142	 */
143	if (status->sf_sc || status->sf_sct) {
144		bp->bio_error = EIO;
145		bp->bio_flags |= BIO_ERROR;
146		bp->bio_resid = bp->bio_bcount;
147	} else
148		bp->bio_resid = 0;
149
150	biodone(bp);
151}
152
153static void
154nvme_ns_strategy(struct bio *bp)
155{
156	struct nvme_namespace	*ns;
157	int			err;
158
159	ns = bp->bio_dev->si_drv1;
160	err = nvme_ns_bio_process(ns, bp, nvme_ns_strategy_done);
161
162	if (err) {
163		bp->bio_error = EIO;
164		bp->bio_flags |= BIO_ERROR;
165		bp->bio_resid = bp->bio_bcount;
166		biodone(bp);
167	}
168
169}
170
171static struct cdevsw nvme_ns_cdevsw = {
172	.d_version =	D_VERSION,
173	.d_flags =	D_DISK,
174	.d_open =	nvme_ns_open,
175	.d_close =	nvme_ns_close,
176	.d_read =	nvme_ns_physio,
177	.d_write =	nvme_ns_physio,
178	.d_strategy =	nvme_ns_strategy,
179	.d_ioctl =	nvme_ns_ioctl
180};
181
182uint32_t
183nvme_ns_get_max_io_xfer_size(struct nvme_namespace *ns)
184{
185	return ns->ctrlr->max_xfer_size;
186}
187
188uint32_t
189nvme_ns_get_sector_size(struct nvme_namespace *ns)
190{
191	return (1 << ns->data.lbaf[0].lbads);
192}
193
194uint64_t
195nvme_ns_get_num_sectors(struct nvme_namespace *ns)
196{
197	return (ns->data.nsze);
198}
199
200uint64_t
201nvme_ns_get_size(struct nvme_namespace *ns)
202{
203	return (nvme_ns_get_num_sectors(ns) * nvme_ns_get_sector_size(ns));
204}
205
206uint32_t
207nvme_ns_get_flags(struct nvme_namespace *ns)
208{
209	return (ns->flags);
210}
211
212const char *
213nvme_ns_get_serial_number(struct nvme_namespace *ns)
214{
215	return ((const char *)ns->ctrlr->cdata.sn);
216}
217
218const char *
219nvme_ns_get_model_number(struct nvme_namespace *ns)
220{
221	return ((const char *)ns->ctrlr->cdata.mn);
222}
223
224static void
225nvme_ns_bio_done(void *arg, const struct nvme_completion *status)
226{
227	struct bio	*bp = arg;
228	nvme_cb_fn_t	bp_cb_fn;
229
230	bp_cb_fn = bp->bio_driver1;
231
232	if (bp->bio_driver2)
233		free(bp->bio_driver2, M_NVME);
234
235	bp_cb_fn(bp, status);
236}
237
238int
239nvme_ns_bio_process(struct nvme_namespace *ns, struct bio *bp,
240	nvme_cb_fn_t cb_fn)
241{
242	struct nvme_dsm_range *dsm_range;
243
244	bp->bio_driver1 = cb_fn;
245
246	switch (bp->bio_cmd) {
247	case BIO_READ:
248		nvme_ns_cmd_read(ns, bp->bio_data,
249		    bp->bio_offset/nvme_ns_get_sector_size(ns),
250		    bp->bio_bcount/nvme_ns_get_sector_size(ns),
251		    nvme_ns_bio_done, bp);
252		break;
253	case BIO_WRITE:
254		nvme_ns_cmd_write(ns, bp->bio_data,
255		    bp->bio_offset/nvme_ns_get_sector_size(ns),
256		    bp->bio_bcount/nvme_ns_get_sector_size(ns),
257		    nvme_ns_bio_done, bp);
258		break;
259	case BIO_FLUSH:
260		nvme_ns_cmd_flush(ns, nvme_ns_bio_done, bp);
261		break;
262	case BIO_DELETE:
263		/*
264		 * Note: Chatham2 doesn't support DSM, so this code
265		 *  can't be fully tested yet.
266		 */
267		dsm_range =
268		    malloc(sizeof(struct nvme_dsm_range), M_NVME,
269		    M_ZERO | M_NOWAIT);
270		dsm_range->length =
271		    bp->bio_bcount/nvme_ns_get_sector_size(ns);
272		dsm_range->starting_lba =
273		    bp->bio_offset/nvme_ns_get_sector_size(ns);
274		bp->bio_driver2 = dsm_range;
275		nvme_ns_cmd_deallocate(ns, dsm_range, 1, nvme_ns_bio_done, bp);
276		break;
277	default:
278		return (EIO);
279	}
280
281	return (0);
282}
283
284#ifdef CHATHAM2
285static void
286nvme_ns_populate_chatham_data(struct nvme_namespace *ns)
287{
288	struct nvme_controller		*ctrlr;
289	struct nvme_namespace_data	*nsdata;
290
291	ctrlr = ns->ctrlr;
292	nsdata = &ns->data;
293
294	nsdata->nsze = ctrlr->chatham_lbas;
295	nsdata->ncap = ctrlr->chatham_lbas;
296	nsdata->nuse = ctrlr->chatham_lbas;
297
298	/* Chatham2 doesn't support thin provisioning. */
299	nsdata->nsfeat.thin_prov = 0;
300
301	/* Set LBA size to 512 bytes. */
302	nsdata->lbaf[0].lbads = 9;
303}
304#endif /* CHATHAM2 */
305
306int
307nvme_ns_construct(struct nvme_namespace *ns, uint16_t id,
308    struct nvme_controller *ctrlr)
309{
310	struct nvme_completion	cpl;
311	struct mtx		*mtx;
312	int			status;
313
314	ns->ctrlr = ctrlr;
315	ns->id = id;
316
317#ifdef CHATHAM2
318	if (pci_get_devid(ctrlr->dev) == CHATHAM_PCI_ID)
319		nvme_ns_populate_chatham_data(ns);
320	else {
321#endif
322		mtx = mtx_pool_find(mtxpool_sleep, &cpl);
323
324		mtx_lock(mtx);
325		nvme_ctrlr_cmd_identify_namespace(ctrlr, id, &ns->data,
326		    nvme_ns_cb, &cpl);
327		status = msleep(&cpl, mtx, PRIBIO, "nvme_start", hz*5);
328		mtx_unlock(mtx);
329		if ((status != 0) || cpl.sf_sc || cpl.sf_sct) {
330			printf("nvme_identify_namespace failed!\n");
331			return (ENXIO);
332		}
333#ifdef CHATHAM2
334	}
335#endif
336
337	if (ctrlr->cdata.oncs.dsm && ns->data.nsfeat.thin_prov)
338		ns->flags |= NVME_NS_DEALLOCATE_SUPPORTED;
339
340	if (ctrlr->cdata.vwc.present)
341		ns->flags |= NVME_NS_FLUSH_SUPPORTED;
342
343/*
344 * MAKEDEV_ETERNAL was added in r210923, for cdevs that will never
345 *  be destroyed.  This avoids refcounting on the cdev object.
346 *  That should be OK case here, as long as we're not supporting PCIe
347 *  surprise removal nor namespace deletion.
348 */
349#ifdef MAKEDEV_ETERNAL_KLD
350	ns->cdev = make_dev_credf(MAKEDEV_ETERNAL_KLD, &nvme_ns_cdevsw, 0,
351	    NULL, UID_ROOT, GID_WHEEL, 0600, "nvme%dns%d",
352	    device_get_unit(ctrlr->dev), ns->id);
353#else
354	ns->cdev = make_dev_credf(0, &nvme_ns_cdevsw, 0,
355	    NULL, UID_ROOT, GID_WHEEL, 0600, "nvme%dns%d",
356	    device_get_unit(ctrlr->dev), ns->id);
357#endif
358
359	if (ns->cdev) {
360		ns->cdev->si_drv1 = ns;
361	}
362
363	return (0);
364}
365