linux-module.c revision d7e09d0397e84eefbabfd9cb353221f3c6448d83
1/*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
19 *
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
22 * have any questions.
23 *
24 * GPL HEADER END
25 */
26/*
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2011, 2012, Intel Corporation.
31 */
32/*
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
35 *
36 * lustre/obdclass/linux/linux-module.c
37 *
38 * Object Devices Class Driver
39 * These are the only exported functions, they provide some generic
40 * infrastructure for managing object devices
41 */
42
43#define DEBUG_SUBSYSTEM S_CLASS
44
45#include <linux/module.h>
46#include <linux/errno.h>
47#include <linux/kernel.h>
48#include <linux/major.h>
49#include <linux/sched.h>
50#include <linux/lp.h>
51#include <linux/slab.h>
52#include <linux/ioport.h>
53#include <linux/fcntl.h>
54#include <linux/delay.h>
55#include <linux/skbuff.h>
56#include <linux/proc_fs.h>
57#include <linux/fs.h>
58#include <linux/poll.h>
59#include <linux/init.h>
60#include <linux/list.h>
61#include <linux/highmem.h>
62#include <asm/io.h>
63#include <asm/ioctls.h>
64#include <asm/poll.h>
65#include <asm/uaccess.h>
66#include <linux/miscdevice.h>
67#include <linux/seq_file.h>
68
69#include <linux/libcfs/libcfs.h>
70#include <obd_support.h>
71#include <obd_class.h>
72#include <linux/lnet/lnetctl.h>
73#include <lprocfs_status.h>
74#include <lustre_ver.h>
75#include <lustre/lustre_build_version.h>
76
77int proc_version;
78
79/* buffer MUST be at least the size of obd_ioctl_hdr */
80int obd_ioctl_getdata(char **buf, int *len, void *arg)
81{
82	struct obd_ioctl_hdr hdr;
83	struct obd_ioctl_data *data;
84	int err;
85	int offset = 0;
86	ENTRY;
87
88	err = copy_from_user(&hdr, (void *)arg, sizeof(hdr));
89	if ( err )
90		RETURN(err);
91
92	if (hdr.ioc_version != OBD_IOCTL_VERSION) {
93		CERROR("Version mismatch kernel (%x) vs application (%x)\n",
94		       OBD_IOCTL_VERSION, hdr.ioc_version);
95		RETURN(-EINVAL);
96	}
97
98	if (hdr.ioc_len > OBD_MAX_IOCTL_BUFFER) {
99		CERROR("User buffer len %d exceeds %d max buffer\n",
100		       hdr.ioc_len, OBD_MAX_IOCTL_BUFFER);
101		RETURN(-EINVAL);
102	}
103
104	if (hdr.ioc_len < sizeof(struct obd_ioctl_data)) {
105		CERROR("User buffer too small for ioctl (%d)\n", hdr.ioc_len);
106		RETURN(-EINVAL);
107	}
108
109	/* When there are lots of processes calling vmalloc on multi-core
110	 * system, the high lock contention will hurt performance badly,
111	 * obdfilter-survey is an example, which relies on ioctl. So we'd
112	 * better avoid vmalloc on ioctl path. LU-66 */
113	OBD_ALLOC_LARGE(*buf, hdr.ioc_len);
114	if (*buf == NULL) {
115		CERROR("Cannot allocate control buffer of len %d\n",
116		       hdr.ioc_len);
117		RETURN(-EINVAL);
118	}
119	*len = hdr.ioc_len;
120	data = (struct obd_ioctl_data *)*buf;
121
122	err = copy_from_user(*buf, (void *)arg, hdr.ioc_len);
123	if ( err ) {
124		OBD_FREE_LARGE(*buf, hdr.ioc_len);
125		RETURN(err);
126	}
127
128	if (obd_ioctl_is_invalid(data)) {
129		CERROR("ioctl not correctly formatted\n");
130		OBD_FREE_LARGE(*buf, hdr.ioc_len);
131		RETURN(-EINVAL);
132	}
133
134	if (data->ioc_inllen1) {
135		data->ioc_inlbuf1 = &data->ioc_bulk[0];
136		offset += cfs_size_round(data->ioc_inllen1);
137	}
138
139	if (data->ioc_inllen2) {
140		data->ioc_inlbuf2 = &data->ioc_bulk[0] + offset;
141		offset += cfs_size_round(data->ioc_inllen2);
142	}
143
144	if (data->ioc_inllen3) {
145		data->ioc_inlbuf3 = &data->ioc_bulk[0] + offset;
146		offset += cfs_size_round(data->ioc_inllen3);
147	}
148
149	if (data->ioc_inllen4) {
150		data->ioc_inlbuf4 = &data->ioc_bulk[0] + offset;
151	}
152
153	EXIT;
154	return 0;
155}
156EXPORT_SYMBOL(obd_ioctl_getdata);
157
158int obd_ioctl_popdata(void *arg, void *data, int len)
159{
160	int err;
161
162	err = copy_to_user(arg, data, len);
163	if (err)
164		err = -EFAULT;
165	return err;
166}
167EXPORT_SYMBOL(obd_ioctl_popdata);
168
169/*  opening /dev/obd */
170static int obd_class_open(struct inode * inode, struct file * file)
171{
172	ENTRY;
173
174	try_module_get(THIS_MODULE);
175	RETURN(0);
176}
177
178/*  closing /dev/obd */
179static int obd_class_release(struct inode * inode, struct file * file)
180{
181	ENTRY;
182
183	module_put(THIS_MODULE);
184	RETURN(0);
185}
186
187/* to control /dev/obd */
188static long obd_class_ioctl(struct file *filp, unsigned int cmd,
189			    unsigned long arg)
190{
191	int err = 0;
192	ENTRY;
193
194	/* Allow non-root access for OBD_IOC_PING_TARGET - used by lfs check */
195	if (!cfs_capable(CFS_CAP_SYS_ADMIN) && (cmd != OBD_IOC_PING_TARGET))
196		RETURN(err = -EACCES);
197	if ((cmd & 0xffffff00) == ((int)'T') << 8) /* ignore all tty ioctls */
198		RETURN(err = -ENOTTY);
199
200	err = class_handle_ioctl(cmd, (unsigned long)arg);
201
202	RETURN(err);
203}
204
205/* declare character device */
206static struct file_operations obd_psdev_fops = {
207	.owner	  = THIS_MODULE,
208	.unlocked_ioctl = obd_class_ioctl, /* unlocked_ioctl */
209	.open	   = obd_class_open,      /* open */
210	.release	= obd_class_release,   /* release */
211};
212
213/* modules setup */
214psdev_t obd_psdev = {
215	.minor = OBD_DEV_MINOR,
216	.name  = OBD_DEV_NAME,
217	.fops  = &obd_psdev_fops,
218};
219
220
221#ifdef LPROCFS
222int obd_proc_read_version(char *page, char **start, off_t off, int count,
223			  int *eof, void *data)
224{
225	*eof = 1;
226	return snprintf(page, count, "lustre: %s\nkernel: %s\nbuild:  %s\n",
227			LUSTRE_VERSION_STRING, "patchless_client",
228			BUILD_VERSION);
229}
230
231int obd_proc_read_pinger(char *page, char **start, off_t off, int count,
232			 int *eof, void *data)
233{
234	*eof = 1;
235	return snprintf(page, count, "%s\n",
236			"on"
237		       );
238}
239
240/**
241 * Check all obd devices health
242 *
243 * \param page
244 * \param start
245 * \param off
246 * \param count
247 * \param eof
248 * \param data
249 *		  proc read function parameters, please refer to kernel
250 *		  code fs/proc/generic.c proc_file_read()
251 * \param data [in] unused
252 *
253 * \retval number of characters printed
254 */
255static int obd_proc_read_health(char *page, char **start, off_t off,
256				int count, int *eof, void *data)
257{
258	int rc = 0, i;
259	*eof = 1;
260
261	if (libcfs_catastrophe)
262		rc += snprintf(page + rc, count - rc, "LBUG\n");
263
264	read_lock(&obd_dev_lock);
265	for (i = 0; i < class_devno_max(); i++) {
266		struct obd_device *obd;
267
268		obd = class_num2obd(i);
269		if (obd == NULL || !obd->obd_attached || !obd->obd_set_up)
270			continue;
271
272		LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC);
273		if (obd->obd_stopping)
274			continue;
275
276		class_incref(obd, __FUNCTION__, current);
277		read_unlock(&obd_dev_lock);
278
279		if (obd_health_check(NULL, obd)) {
280			rc += snprintf(page + rc, count - rc,
281				       "device %s reported unhealthy\n",
282				       obd->obd_name);
283		}
284		class_decref(obd, __FUNCTION__, current);
285		read_lock(&obd_dev_lock);
286	}
287	read_unlock(&obd_dev_lock);
288
289	if (rc == 0)
290		return snprintf(page, count, "healthy\n");
291
292	rc += snprintf(page + rc, count - rc, "NOT HEALTHY\n");
293	return rc;
294}
295
296static int obd_proc_rd_jobid_var(char *page, char **start, off_t off,
297				int count, int *eof, void *data)
298{
299	return snprintf(page, count, "%s\n", obd_jobid_var);
300}
301
302static int obd_proc_wr_jobid_var(struct file *file, const char *buffer,
303				unsigned long count, void *data)
304{
305	if (!count || count > JOBSTATS_JOBID_VAR_MAX_LEN)
306		return -EINVAL;
307
308	memset(obd_jobid_var, 0, JOBSTATS_JOBID_VAR_MAX_LEN + 1);
309	/* Trim the trailing '\n' if any */
310	memcpy(obd_jobid_var, buffer, count - (buffer[count - 1] == '\n'));
311	return count;
312}
313
314/* Root for /proc/fs/lustre */
315struct proc_dir_entry *proc_lustre_root = NULL;
316EXPORT_SYMBOL(proc_lustre_root);
317
318struct lprocfs_vars lprocfs_base[] = {
319	{ "version", obd_proc_read_version, NULL, NULL },
320	{ "pinger", obd_proc_read_pinger, NULL, NULL },
321	{ "health_check", obd_proc_read_health, NULL, NULL },
322	{ "jobid_var", obd_proc_rd_jobid_var,
323		       obd_proc_wr_jobid_var, NULL },
324	{ 0 }
325};
326#else
327#define lprocfs_base NULL
328#endif /* LPROCFS */
329
330static void *obd_device_list_seq_start(struct seq_file *p, loff_t *pos)
331{
332	if (*pos >= class_devno_max())
333		return NULL;
334
335	return pos;
336}
337
338static void obd_device_list_seq_stop(struct seq_file *p, void *v)
339{
340}
341
342static void *obd_device_list_seq_next(struct seq_file *p, void *v, loff_t *pos)
343{
344	++*pos;
345	if (*pos >= class_devno_max())
346		return NULL;
347
348	return pos;
349}
350
351static int obd_device_list_seq_show(struct seq_file *p, void *v)
352{
353	loff_t index = *(loff_t *)v;
354	struct obd_device *obd = class_num2obd((int)index);
355	char *status;
356
357	if (obd == NULL)
358		return 0;
359
360	LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC);
361	if (obd->obd_stopping)
362		status = "ST";
363	else if (obd->obd_inactive)
364		status = "IN";
365	else if (obd->obd_set_up)
366		status = "UP";
367	else if (obd->obd_attached)
368		status = "AT";
369	else
370		status = "--";
371
372	return seq_printf(p, "%3d %s %s %s %s %d\n",
373			  (int)index, status, obd->obd_type->typ_name,
374			  obd->obd_name, obd->obd_uuid.uuid,
375			  atomic_read(&obd->obd_refcount));
376}
377
378struct seq_operations obd_device_list_sops = {
379	.start = obd_device_list_seq_start,
380	.stop = obd_device_list_seq_stop,
381	.next = obd_device_list_seq_next,
382	.show = obd_device_list_seq_show,
383};
384
385static int obd_device_list_open(struct inode *inode, struct file *file)
386{
387	struct proc_dir_entry *dp = PDE(inode);
388	struct seq_file *seq;
389	int rc = seq_open(file, &obd_device_list_sops);
390
391	if (rc)
392		return rc;
393
394	seq = file->private_data;
395	seq->private = dp->data;
396
397	return 0;
398}
399
400struct file_operations obd_device_list_fops = {
401	.owner   = THIS_MODULE,
402	.open    = obd_device_list_open,
403	.read    = seq_read,
404	.llseek  = seq_lseek,
405	.release = seq_release,
406};
407
408int class_procfs_init(void)
409{
410	int rc;
411	ENTRY;
412
413	obd_sysctl_init();
414	proc_lustre_root = lprocfs_register("fs/lustre", NULL,
415					    lprocfs_base, NULL);
416	rc = lprocfs_seq_create(proc_lustre_root, "devices", 0444,
417				&obd_device_list_fops, NULL);
418	if (rc)
419		CERROR("error adding /proc/fs/lustre/devices file\n");
420	RETURN(0);
421}
422
423int class_procfs_clean(void)
424{
425	ENTRY;
426	if (proc_lustre_root) {
427		lprocfs_remove(&proc_lustre_root);
428	}
429	RETURN(0);
430}
431