2 * Device operations for the pnfs nfs4 file layout driver.
5 * The Regents of the University of Michigan
8 * Dean Hildebrand <dhildebz@umich.edu>
9 * Garth Goodson <Garth.Goodson@netapp.com>
11 * Permission is granted to use, copy, create derivative works, and
12 * redistribute this software and such derivative works for any purpose,
13 * so long as the name of the University of Michigan is not used in
14 * any advertising or publicity pertaining to the use or distribution
15 * of this software without specific, written prior authorization. If
16 * the above copyright notice or any other identification of the
17 * University of Michigan is included in any copy of any portion of
18 * this software, then the disclaimer below must also be included.
20 * This software is provided as is, without representation or warranty
21 * of any kind either express or implied, including without limitation
22 * the implied warranties of merchantability, fitness for a particular
23 * purpose, or noninfringement. The Regents of the University of
24 * Michigan shall not be liable for any damages, including special,
25 * indirect, incidental, or consequential damages, with respect to any
26 * claim arising out of or in connection with the use of the software,
27 * even if it has been or is hereafter advised of the possibility of
31 #include <linux/nfs_fs.h>
32 #include <linux/vmalloc.h>
35 #include "nfs4filelayout.h"
37 #define NFSDBG_FACILITY NFSDBG_PNFS_LD
42 * Data servers can be mapped to different device ids.
43 * nfs4_pnfs_ds reference counting
44 * - set to 1 on allocation
45 * - incremented when a device id maps a data server already in the cache.
46 * - decremented when deviceid is removed from the cache.
48 DEFINE_SPINLOCK(nfs4_ds_cache_lock);
49 static LIST_HEAD(nfs4_data_server_cache);
53 print_ds(struct nfs4_pnfs_ds *ds)
56 printk("%s NULL device\n", __func__);
59 printk(" ip_addr %x port %hu\n"
62 " cl_exchange_flags %x\n",
63 ntohl(ds->ds_ip_addr), ntohs(ds->ds_port),
64 atomic_read(&ds->ds_count), ds->ds_clp,
65 ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0);
69 print_ds_list(struct nfs4_file_layout_dsaddr *dsaddr)
74 printk("%s dsaddr->ds_num %d\n", __func__,
76 for (i = 0; i < dsaddr->ds_num; i++)
77 print_ds(dsaddr->ds_list[i]);
81 void print_deviceid(struct nfs4_deviceid *id)
85 dprintk("%s: device id= [%x%x%x%x]\n", __func__,
86 p[0], p[1], p[2], p[3]);
89 /* nfs4_ds_cache_lock is held */
90 static struct nfs4_pnfs_ds *
91 _data_server_lookup_locked(u32 ip_addr, u32 port)
93 struct nfs4_pnfs_ds *ds;
95 dprintk("_data_server_lookup: ip_addr=%x port=%hu\n",
96 ntohl(ip_addr), ntohs(port));
98 list_for_each_entry(ds, &nfs4_data_server_cache, ds_node) {
99 if (ds->ds_ip_addr == ip_addr &&
100 ds->ds_port == port) {
108 destroy_ds(struct nfs4_pnfs_ds *ds)
110 dprintk("--> %s\n", __func__);
115 nfs_put_client(ds->ds_clp);
120 nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr)
122 struct nfs4_pnfs_ds *ds;
125 print_deviceid(&dsaddr->deviceid.de_id);
127 for (i = 0; i < dsaddr->ds_num; i++) {
128 ds = dsaddr->ds_list[i];
130 if (atomic_dec_and_lock(&ds->ds_count,
131 &nfs4_ds_cache_lock)) {
132 list_del_init(&ds->ds_node);
133 spin_unlock(&nfs4_ds_cache_lock);
138 kfree(dsaddr->stripe_indices);
143 nfs4_fl_free_deviceid_callback(struct pnfs_deviceid_node *device)
145 struct nfs4_file_layout_dsaddr *dsaddr =
146 container_of(device, struct nfs4_file_layout_dsaddr, deviceid);
148 nfs4_fl_free_deviceid(dsaddr);
151 static struct nfs4_pnfs_ds *
152 nfs4_pnfs_ds_add(struct inode *inode, u32 ip_addr, u32 port)
154 struct nfs4_pnfs_ds *tmp_ds, *ds;
156 ds = kzalloc(sizeof(*tmp_ds), GFP_KERNEL);
160 spin_lock(&nfs4_ds_cache_lock);
161 tmp_ds = _data_server_lookup_locked(ip_addr, port);
162 if (tmp_ds == NULL) {
163 ds->ds_ip_addr = ip_addr;
165 atomic_set(&ds->ds_count, 1);
166 INIT_LIST_HEAD(&ds->ds_node);
168 list_add(&ds->ds_node, &nfs4_data_server_cache);
169 dprintk("%s add new data server ip 0x%x\n", __func__,
173 atomic_inc(&tmp_ds->ds_count);
174 dprintk("%s data server found ip 0x%x, inc'ed ds_count to %d\n",
175 __func__, tmp_ds->ds_ip_addr,
176 atomic_read(&tmp_ds->ds_count));
179 spin_unlock(&nfs4_ds_cache_lock);
185 * Currently only support ipv4, and one multi-path address.
187 static struct nfs4_pnfs_ds *
188 decode_and_add_ds(__be32 **pp, struct inode *inode)
190 struct nfs4_pnfs_ds *ds = NULL;
192 const char *ipend, *pstr;
196 __be32 *r_netid, *r_addr, *p = *pp;
199 nlen = be32_to_cpup(p++);
201 p += XDR_QUADLEN(nlen);
204 rlen = be32_to_cpup(p++);
206 p += XDR_QUADLEN(rlen);
209 /* Check that netid is "tcp" */
210 if (nlen != 3 || memcmp((char *)r_netid, "tcp", 3)) {
211 dprintk("%s: ERROR: non ipv4 TCP r_netid\n", __func__);
215 /* ipv6 length plus port is legal */
216 if (rlen > INET6_ADDRSTRLEN + 8) {
217 dprintk("%s: Invalid address, length %d\n", __func__,
221 buf = kmalloc(rlen + 1, GFP_KERNEL);
223 dprintk("%s: Not enough memory\n", __func__);
227 memcpy(buf, r_addr, rlen);
229 /* replace the port dots with dashes for the in4_pton() delimiter*/
230 for (i = 0; i < 2; i++) {
231 char *res = strrchr(buf, '.');
233 dprintk("%s: Failed finding expected dots in port\n",
240 /* Currently only support ipv4 address */
241 if (in4_pton(buf, rlen, (u8 *)&ip_addr, '-', &ipend) == 0) {
242 dprintk("%s: Only ipv4 addresses supported\n", __func__);
248 sscanf(pstr, "-%d-%d", &tmp[0], &tmp[1]);
249 port = htons((tmp[0] << 8) | (tmp[1]));
251 ds = nfs4_pnfs_ds_add(inode, ip_addr, port);
252 dprintk("%s: Decoded address and port %s\n", __func__, buf);
259 /* Decode opaque device data and return the result */
260 static struct nfs4_file_layout_dsaddr*
261 decode_device(struct inode *ino, struct pnfs_device *pdev)
266 __be32 *p = (__be32 *)pdev->area, *indicesp;
267 struct nfs4_file_layout_dsaddr *dsaddr;
269 /* Get the stripe count (number of stripe index) */
270 cnt = be32_to_cpup(p++);
271 dprintk("%s stripe count %d\n", __func__, cnt);
272 if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) {
273 printk(KERN_WARNING "%s: stripe count %d greater than "
274 "supported maximum %d\n", __func__,
275 cnt, NFS4_PNFS_MAX_STRIPE_CNT);
279 /* Check the multipath list count */
281 p += XDR_QUADLEN(cnt << 2);
282 num = be32_to_cpup(p++);
283 dprintk("%s ds_num %u\n", __func__, num);
284 if (num > NFS4_PNFS_MAX_MULTI_CNT) {
285 printk(KERN_WARNING "%s: multipath count %d greater than "
286 "supported maximum %d\n", __func__,
287 num, NFS4_PNFS_MAX_MULTI_CNT);
290 dsaddr = kzalloc(sizeof(*dsaddr) +
291 (sizeof(struct nfs4_pnfs_ds *) * (num - 1)),
296 dsaddr->stripe_indices = kzalloc(sizeof(u8) * cnt, GFP_KERNEL);
297 if (!dsaddr->stripe_indices)
300 dsaddr->stripe_count = cnt;
301 dsaddr->ds_num = num;
303 memcpy(&dsaddr->deviceid.de_id, &pdev->dev_id, sizeof(pdev->dev_id));
305 /* Go back an read stripe indices */
307 indexp = &dsaddr->stripe_indices[0];
308 for (i = 0; i < dsaddr->stripe_count; i++) {
309 *indexp = be32_to_cpup(p++);
314 /* Skip already read multipath list count */
317 for (i = 0; i < dsaddr->ds_num; i++) {
320 dummy = be32_to_cpup(p++); /* multipath count */
323 "%s: Multipath count %d not supported, "
324 "skipping all greater than 1\n", __func__,
327 for (j = 0; j < dummy; j++) {
329 dsaddr->ds_list[i] = decode_and_add_ds(&p, ino);
330 if (dsaddr->ds_list[i] == NULL)
334 /* skip extra multipath */
335 len = be32_to_cpup(p++);
336 p += XDR_QUADLEN(len);
337 len = be32_to_cpup(p++);
338 p += XDR_QUADLEN(len);
346 nfs4_fl_free_deviceid(dsaddr);
348 dprintk("%s ERROR: returning NULL\n", __func__);
353 * Decode the opaque device specified in 'dev'
354 * and add it to the list of available devices.
355 * If the deviceid is already cached, nfs4_add_deviceid will return
356 * a pointer to the cached struct and throw away the new.
358 static struct nfs4_file_layout_dsaddr*
359 decode_and_add_device(struct inode *inode, struct pnfs_device *dev)
361 struct nfs4_file_layout_dsaddr *dsaddr;
362 struct pnfs_deviceid_node *d;
364 dsaddr = decode_device(inode, dev);
366 printk(KERN_WARNING "%s: Could not decode or add device\n",
371 d = pnfs_add_deviceid(NFS_SERVER(inode)->nfs_client->cl_devid_cache,
374 return container_of(d, struct nfs4_file_layout_dsaddr, deviceid);
378 * Retrieve the information for dev_id, add it to the list
379 * of available devices, and return it.
381 struct nfs4_file_layout_dsaddr *
382 get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id)
384 struct pnfs_device *pdev = NULL;
387 struct page **pages = NULL;
388 struct nfs4_file_layout_dsaddr *dsaddr = NULL;
390 struct nfs_server *server = NFS_SERVER(inode);
393 * Use the session max response size as the basis for setting
394 * GETDEVICEINFO's maxcount
396 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
397 max_pages = max_resp_sz >> PAGE_SHIFT;
398 dprintk("%s inode %p max_resp_sz %u max_pages %d\n",
399 __func__, inode, max_resp_sz, max_pages);
401 pdev = kzalloc(sizeof(struct pnfs_device), GFP_KERNEL);
405 pages = kzalloc(max_pages * sizeof(struct page *), GFP_KERNEL);
410 for (i = 0; i < max_pages; i++) {
411 pages[i] = alloc_page(GFP_KERNEL);
417 pdev->area = vmap(pages, max_pages, VM_MAP, PAGE_KERNEL);
421 memcpy(&pdev->dev_id, dev_id, sizeof(*dev_id));
422 pdev->layout_type = LAYOUT_NFSV4_1_FILES;
425 pdev->pglen = PAGE_SIZE * max_pages;
428 rc = nfs4_proc_getdeviceinfo(server, pdev);
429 dprintk("%s getdevice info returns %d\n", __func__, rc);
434 * Found new device, need to decode it and then add it to the
435 * list of known devices for this mountpoint.
437 dsaddr = decode_and_add_device(inode, pdev);
439 if (pdev->area != NULL)
441 for (i = 0; i < max_pages; i++)
442 __free_page(pages[i]);
445 dprintk("<-- %s dsaddr %p\n", __func__, dsaddr);
449 struct nfs4_file_layout_dsaddr *
450 nfs4_fl_find_get_deviceid(struct nfs_client *clp, struct nfs4_deviceid *id)
452 struct pnfs_deviceid_node *d;
454 d = pnfs_find_get_deviceid(clp->cl_devid_cache, id);
455 return (d == NULL) ? NULL :
456 container_of(d, struct nfs4_file_layout_dsaddr, deviceid);