aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--fs/nfs/nfs4proc.c142
-rw-r--r--fs/nfs/nfs4xdr.c302
-rw-r--r--fs/nfs/pnfs.c385
-rw-r--r--fs/nfs/pnfs.h73
-rw-r--r--include/linux/nfs4.h2
-rw-r--r--include/linux/nfs_fs_sb.h1
-rw-r--r--include/linux/nfs_xdr.h49
7 files changed, 921 insertions, 33 deletions
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index a5f1edb45b47..7e14e991ddfa 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -55,6 +55,7 @@
55#include "internal.h" 55#include "internal.h"
56#include "iostat.h" 56#include "iostat.h"
57#include "callback.h" 57#include "callback.h"
58#include "pnfs.h"
58 59
59#define NFSDBG_FACILITY NFSDBG_PROC 60#define NFSDBG_FACILITY NFSDBG_PROC
60 61
@@ -5256,6 +5257,147 @@ out:
5256 dprintk("<-- %s status=%d\n", __func__, status); 5257 dprintk("<-- %s status=%d\n", __func__, status);
5257 return status; 5258 return status;
5258} 5259}
5260
5261static void
5262nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5263{
5264 struct nfs4_layoutget *lgp = calldata;
5265 struct inode *ino = lgp->args.inode;
5266 struct nfs_server *server = NFS_SERVER(ino);
5267
5268 dprintk("--> %s\n", __func__);
5269 if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5270 &lgp->res.seq_res, 0, task))
5271 return;
5272 rpc_call_start(task);
5273}
5274
5275static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5276{
5277 struct nfs4_layoutget *lgp = calldata;
5278 struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5279
5280 dprintk("--> %s\n", __func__);
5281
5282 if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5283 return;
5284
5285 switch (task->tk_status) {
5286 case 0:
5287 break;
5288 case -NFS4ERR_LAYOUTTRYLATER:
5289 case -NFS4ERR_RECALLCONFLICT:
5290 task->tk_status = -NFS4ERR_DELAY;
5291 /* Fall through */
5292 default:
5293 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5294 rpc_restart_call_prepare(task);
5295 return;
5296 }
5297 }
5298 lgp->status = task->tk_status;
5299 dprintk("<-- %s\n", __func__);
5300}
5301
5302static void nfs4_layoutget_release(void *calldata)
5303{
5304 struct nfs4_layoutget *lgp = calldata;
5305
5306 dprintk("--> %s\n", __func__);
5307 put_layout_hdr(lgp->args.inode);
5308 if (lgp->res.layout.buf != NULL)
5309 free_page((unsigned long) lgp->res.layout.buf);
5310 put_nfs_open_context(lgp->args.ctx);
5311 kfree(calldata);
5312 dprintk("<-- %s\n", __func__);
5313}
5314
5315static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5316 .rpc_call_prepare = nfs4_layoutget_prepare,
5317 .rpc_call_done = nfs4_layoutget_done,
5318 .rpc_release = nfs4_layoutget_release,
5319};
5320
5321int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5322{
5323 struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5324 struct rpc_task *task;
5325 struct rpc_message msg = {
5326 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5327 .rpc_argp = &lgp->args,
5328 .rpc_resp = &lgp->res,
5329 };
5330 struct rpc_task_setup task_setup_data = {
5331 .rpc_client = server->client,
5332 .rpc_message = &msg,
5333 .callback_ops = &nfs4_layoutget_call_ops,
5334 .callback_data = lgp,
5335 .flags = RPC_TASK_ASYNC,
5336 };
5337 int status = 0;
5338
5339 dprintk("--> %s\n", __func__);
5340
5341 lgp->res.layout.buf = (void *)__get_free_page(GFP_NOFS);
5342 if (lgp->res.layout.buf == NULL) {
5343 nfs4_layoutget_release(lgp);
5344 return -ENOMEM;
5345 }
5346
5347 lgp->res.seq_res.sr_slot = NULL;
5348 task = rpc_run_task(&task_setup_data);
5349 if (IS_ERR(task))
5350 return PTR_ERR(task);
5351 status = nfs4_wait_for_completion_rpc_task(task);
5352 if (status != 0)
5353 goto out;
5354 status = lgp->status;
5355 if (status != 0)
5356 goto out;
5357 status = pnfs_layout_process(lgp);
5358out:
5359 rpc_put_task(task);
5360 dprintk("<-- %s status=%d\n", __func__, status);
5361 return status;
5362}
5363
5364static int
5365_nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5366{
5367 struct nfs4_getdeviceinfo_args args = {
5368 .pdev = pdev,
5369 };
5370 struct nfs4_getdeviceinfo_res res = {
5371 .pdev = pdev,
5372 };
5373 struct rpc_message msg = {
5374 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5375 .rpc_argp = &args,
5376 .rpc_resp = &res,
5377 };
5378 int status;
5379
5380 dprintk("--> %s\n", __func__);
5381 status = nfs4_call_sync(server, &msg, &args, &res, 0);
5382 dprintk("<-- %s status=%d\n", __func__, status);
5383
5384 return status;
5385}
5386
5387int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5388{
5389 struct nfs4_exception exception = { };
5390 int err;
5391
5392 do {
5393 err = nfs4_handle_exception(server,
5394 _nfs4_proc_getdeviceinfo(server, pdev),
5395 &exception);
5396 } while (exception.retry);
5397 return err;
5398}
5399EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5400
5259#endif /* CONFIG_NFS_V4_1 */ 5401#endif /* CONFIG_NFS_V4_1 */
5260 5402
5261struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = { 5403struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 8b4dfa393f0f..f313c4cce7e4 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -52,6 +52,7 @@
52#include <linux/nfs_idmap.h> 52#include <linux/nfs_idmap.h>
53#include "nfs4_fs.h" 53#include "nfs4_fs.h"
54#include "internal.h" 54#include "internal.h"
55#include "pnfs.h"
55 56
56#define NFSDBG_FACILITY NFSDBG_XDR 57#define NFSDBG_FACILITY NFSDBG_XDR
57 58
@@ -310,6 +311,19 @@ static int nfs4_stat_to_errno(int);
310 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) 311 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5)
311#define encode_reclaim_complete_maxsz (op_encode_hdr_maxsz + 4) 312#define encode_reclaim_complete_maxsz (op_encode_hdr_maxsz + 4)
312#define decode_reclaim_complete_maxsz (op_decode_hdr_maxsz + 4) 313#define decode_reclaim_complete_maxsz (op_decode_hdr_maxsz + 4)
314#define encode_getdeviceinfo_maxsz (op_encode_hdr_maxsz + 4 + \
315 XDR_QUADLEN(NFS4_DEVICEID4_SIZE))
316#define decode_getdeviceinfo_maxsz (op_decode_hdr_maxsz + \
317 1 /* layout type */ + \
318 1 /* opaque devaddr4 length */ + \
319 /* devaddr4 payload is read into page */ \
320 1 /* notification bitmap length */ + \
321 1 /* notification bitmap */)
322#define encode_layoutget_maxsz (op_encode_hdr_maxsz + 10 + \
323 encode_stateid_maxsz)
324#define decode_layoutget_maxsz (op_decode_hdr_maxsz + 8 + \
325 decode_stateid_maxsz + \
326 XDR_QUADLEN(PNFS_LAYOUT_MAXSIZE))
313#else /* CONFIG_NFS_V4_1 */ 327#else /* CONFIG_NFS_V4_1 */
314#define encode_sequence_maxsz 0 328#define encode_sequence_maxsz 0
315#define decode_sequence_maxsz 0 329#define decode_sequence_maxsz 0
@@ -699,6 +713,20 @@ static int nfs4_stat_to_errno(int);
699#define NFS4_dec_reclaim_complete_sz (compound_decode_hdr_maxsz + \ 713#define NFS4_dec_reclaim_complete_sz (compound_decode_hdr_maxsz + \
700 decode_sequence_maxsz + \ 714 decode_sequence_maxsz + \
701 decode_reclaim_complete_maxsz) 715 decode_reclaim_complete_maxsz)
716#define NFS4_enc_getdeviceinfo_sz (compound_encode_hdr_maxsz + \
717 encode_sequence_maxsz +\
718 encode_getdeviceinfo_maxsz)
719#define NFS4_dec_getdeviceinfo_sz (compound_decode_hdr_maxsz + \
720 decode_sequence_maxsz + \
721 decode_getdeviceinfo_maxsz)
722#define NFS4_enc_layoutget_sz (compound_encode_hdr_maxsz + \
723 encode_sequence_maxsz + \
724 encode_putfh_maxsz + \
725 encode_layoutget_maxsz)
726#define NFS4_dec_layoutget_sz (compound_decode_hdr_maxsz + \
727 decode_sequence_maxsz + \
728 decode_putfh_maxsz + \
729 decode_layoutget_maxsz)
702 730
703const u32 nfs41_maxwrite_overhead = ((RPC_MAX_HEADER_WITH_AUTH + 731const u32 nfs41_maxwrite_overhead = ((RPC_MAX_HEADER_WITH_AUTH +
704 compound_encode_hdr_maxsz + 732 compound_encode_hdr_maxsz +
@@ -1737,6 +1765,58 @@ static void encode_sequence(struct xdr_stream *xdr,
1737#endif /* CONFIG_NFS_V4_1 */ 1765#endif /* CONFIG_NFS_V4_1 */
1738} 1766}
1739 1767
1768#ifdef CONFIG_NFS_V4_1
1769static void
1770encode_getdeviceinfo(struct xdr_stream *xdr,
1771 const struct nfs4_getdeviceinfo_args *args,
1772 struct compound_hdr *hdr)
1773{
1774 __be32 *p;
1775
1776 p = reserve_space(xdr, 16 + NFS4_DEVICEID4_SIZE);
1777 *p++ = cpu_to_be32(OP_GETDEVICEINFO);
1778 p = xdr_encode_opaque_fixed(p, args->pdev->dev_id.data,
1779 NFS4_DEVICEID4_SIZE);
1780 *p++ = cpu_to_be32(args->pdev->layout_type);
1781 *p++ = cpu_to_be32(args->pdev->pglen); /* gdia_maxcount */
1782 *p++ = cpu_to_be32(0); /* bitmap length 0 */
1783 hdr->nops++;
1784 hdr->replen += decode_getdeviceinfo_maxsz;
1785}
1786
1787static void
1788encode_layoutget(struct xdr_stream *xdr,
1789 const struct nfs4_layoutget_args *args,
1790 struct compound_hdr *hdr)
1791{
1792 nfs4_stateid stateid;
1793 __be32 *p;
1794
1795 p = reserve_space(xdr, 44 + NFS4_STATEID_SIZE);
1796 *p++ = cpu_to_be32(OP_LAYOUTGET);
1797 *p++ = cpu_to_be32(0); /* Signal layout available */
1798 *p++ = cpu_to_be32(args->type);
1799 *p++ = cpu_to_be32(args->range.iomode);
1800 p = xdr_encode_hyper(p, args->range.offset);
1801 p = xdr_encode_hyper(p, args->range.length);
1802 p = xdr_encode_hyper(p, args->minlength);
1803 pnfs_get_layout_stateid(&stateid, NFS_I(args->inode)->layout,
1804 args->ctx->state);
1805 p = xdr_encode_opaque_fixed(p, &stateid.data, NFS4_STATEID_SIZE);
1806 *p = cpu_to_be32(args->maxcount);
1807
1808 dprintk("%s: 1st type:0x%x iomode:%d off:%lu len:%lu mc:%d\n",
1809 __func__,
1810 args->type,
1811 args->range.iomode,
1812 (unsigned long)args->range.offset,
1813 (unsigned long)args->range.length,
1814 args->maxcount);
1815 hdr->nops++;
1816 hdr->replen += decode_layoutget_maxsz;
1817}
1818#endif /* CONFIG_NFS_V4_1 */
1819
1740/* 1820/*
1741 * END OF "GENERIC" ENCODE ROUTINES. 1821 * END OF "GENERIC" ENCODE ROUTINES.
1742 */ 1822 */
@@ -2554,6 +2634,51 @@ static int nfs4_xdr_enc_reclaim_complete(struct rpc_rqst *req, uint32_t *p,
2554 return 0; 2634 return 0;
2555} 2635}
2556 2636
2637/*
2638 * Encode GETDEVICEINFO request
2639 */
2640static int nfs4_xdr_enc_getdeviceinfo(struct rpc_rqst *req, uint32_t *p,
2641 struct nfs4_getdeviceinfo_args *args)
2642{
2643 struct xdr_stream xdr;
2644 struct compound_hdr hdr = {
2645 .minorversion = nfs4_xdr_minorversion(&args->seq_args),
2646 };
2647
2648 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
2649 encode_compound_hdr(&xdr, req, &hdr);
2650 encode_sequence(&xdr, &args->seq_args, &hdr);
2651 encode_getdeviceinfo(&xdr, args, &hdr);
2652
2653 /* set up reply kvec. Subtract notification bitmap max size (2)
2654 * so that notification bitmap is put in xdr_buf tail */
2655 xdr_inline_pages(&req->rq_rcv_buf, (hdr.replen - 2) << 2,
2656 args->pdev->pages, args->pdev->pgbase,
2657 args->pdev->pglen);
2658
2659 encode_nops(&hdr);
2660 return 0;
2661}
2662
2663/*
2664 * Encode LAYOUTGET request
2665 */
2666static int nfs4_xdr_enc_layoutget(struct rpc_rqst *req, uint32_t *p,
2667 struct nfs4_layoutget_args *args)
2668{
2669 struct xdr_stream xdr;
2670 struct compound_hdr hdr = {
2671 .minorversion = nfs4_xdr_minorversion(&args->seq_args),
2672 };
2673
2674 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
2675 encode_compound_hdr(&xdr, req, &hdr);
2676 encode_sequence(&xdr, &args->seq_args, &hdr);
2677 encode_putfh(&xdr, NFS_FH(args->inode), &hdr);
2678 encode_layoutget(&xdr, args, &hdr);
2679 encode_nops(&hdr);
2680 return 0;
2681}
2557#endif /* CONFIG_NFS_V4_1 */ 2682#endif /* CONFIG_NFS_V4_1 */
2558 2683
2559static void print_overflow_msg(const char *func, const struct xdr_stream *xdr) 2684static void print_overflow_msg(const char *func, const struct xdr_stream *xdr)
@@ -4830,6 +4955,134 @@ out_overflow:
4830#endif /* CONFIG_NFS_V4_1 */ 4955#endif /* CONFIG_NFS_V4_1 */
4831} 4956}
4832 4957
4958#if defined(CONFIG_NFS_V4_1)
4959
4960static int decode_getdeviceinfo(struct xdr_stream *xdr,
4961 struct pnfs_device *pdev)
4962{
4963 __be32 *p;
4964 uint32_t len, type;
4965 int status;
4966
4967 status = decode_op_hdr(xdr, OP_GETDEVICEINFO);
4968 if (status) {
4969 if (status == -ETOOSMALL) {
4970 p = xdr_inline_decode(xdr, 4);
4971 if (unlikely(!p))
4972 goto out_overflow;
4973 pdev->mincount = be32_to_cpup(p);
4974 dprintk("%s: Min count too small. mincnt = %u\n",
4975 __func__, pdev->mincount);
4976 }
4977 return status;
4978 }
4979
4980 p = xdr_inline_decode(xdr, 8);
4981 if (unlikely(!p))
4982 goto out_overflow;
4983 type = be32_to_cpup(p++);
4984 if (type != pdev->layout_type) {
4985 dprintk("%s: layout mismatch req: %u pdev: %u\n",
4986 __func__, pdev->layout_type, type);
4987 return -EINVAL;
4988 }
4989 /*
4990 * Get the length of the opaque device_addr4. xdr_read_pages places
4991 * the opaque device_addr4 in the xdr_buf->pages (pnfs_device->pages)
4992 * and places the remaining xdr data in xdr_buf->tail
4993 */
4994 pdev->mincount = be32_to_cpup(p);
4995 xdr_read_pages(xdr, pdev->mincount); /* include space for the length */
4996
4997 /* Parse notification bitmap, verifying that it is zero. */
4998 p = xdr_inline_decode(xdr, 4);
4999 if (unlikely(!p))
5000 goto out_overflow;
5001 len = be32_to_cpup(p);
5002 if (len) {
5003 int i;
5004
5005 p = xdr_inline_decode(xdr, 4 * len);
5006 if (unlikely(!p))
5007 goto out_overflow;
5008 for (i = 0; i < len; i++, p++) {
5009 if (be32_to_cpup(p)) {
5010 dprintk("%s: notifications not supported\n",
5011 __func__);
5012 return -EIO;
5013 }
5014 }
5015 }
5016 return 0;
5017out_overflow:
5018 print_overflow_msg(__func__, xdr);
5019 return -EIO;
5020}
5021
5022static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req,
5023 struct nfs4_layoutget_res *res)
5024{
5025 __be32 *p;
5026 int status;
5027 u32 layout_count;
5028
5029 status = decode_op_hdr(xdr, OP_LAYOUTGET);
5030 if (status)
5031 return status;
5032 p = xdr_inline_decode(xdr, 8 + NFS4_STATEID_SIZE);
5033 if (unlikely(!p))
5034 goto out_overflow;
5035 res->return_on_close = be32_to_cpup(p++);
5036 p = xdr_decode_opaque_fixed(p, res->stateid.data, NFS4_STATEID_SIZE);
5037 layout_count = be32_to_cpup(p);
5038 if (!layout_count) {
5039 dprintk("%s: server responded with empty layout array\n",
5040 __func__);
5041 return -EINVAL;
5042 }
5043
5044 p = xdr_inline_decode(xdr, 24);
5045 if (unlikely(!p))
5046 goto out_overflow;
5047 p = xdr_decode_hyper(p, &res->range.offset);
5048 p = xdr_decode_hyper(p, &res->range.length);
5049 res->range.iomode = be32_to_cpup(p++);
5050 res->type = be32_to_cpup(p++);
5051
5052 status = decode_opaque_inline(xdr, &res->layout.len, (char **)&p);
5053 if (unlikely(status))
5054 return status;
5055
5056 dprintk("%s roff:%lu rlen:%lu riomode:%d, lo_type:0x%x, lo.len:%d\n",
5057 __func__,
5058 (unsigned long)res->range.offset,
5059 (unsigned long)res->range.length,
5060 res->range.iomode,
5061 res->type,
5062 res->layout.len);
5063
5064 /* nfs4_proc_layoutget allocated a single page */
5065 if (res->layout.len > PAGE_SIZE)
5066 return -ENOMEM;
5067 memcpy(res->layout.buf, p, res->layout.len);
5068
5069 if (layout_count > 1) {
5070 /* We only handle a length one array at the moment. Any
5071 * further entries are just ignored. Note that this means
5072 * the client may see a response that is less than the
5073 * minimum it requested.
5074 */
5075 dprintk("%s: server responded with %d layouts, dropping tail\n",
5076 __func__, layout_count);
5077 }
5078
5079 return 0;
5080out_overflow:
5081 print_overflow_msg(__func__, xdr);
5082 return -EIO;
5083}
5084#endif /* CONFIG_NFS_V4_1 */
5085
4833/* 5086/*
4834 * END OF "GENERIC" DECODE ROUTINES. 5087 * END OF "GENERIC" DECODE ROUTINES.
4835 */ 5088 */
@@ -5857,6 +6110,53 @@ static int nfs4_xdr_dec_reclaim_complete(struct rpc_rqst *rqstp, uint32_t *p,
5857 status = decode_reclaim_complete(&xdr, (void *)NULL); 6110 status = decode_reclaim_complete(&xdr, (void *)NULL);
5858 return status; 6111 return status;
5859} 6112}
6113
6114/*
6115 * Decode GETDEVINFO response
6116 */
6117static int nfs4_xdr_dec_getdeviceinfo(struct rpc_rqst *rqstp, uint32_t *p,
6118 struct nfs4_getdeviceinfo_res *res)
6119{
6120 struct xdr_stream xdr;
6121 struct compound_hdr hdr;
6122 int status;
6123
6124 xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p);
6125 status = decode_compound_hdr(&xdr, &hdr);
6126 if (status != 0)
6127 goto out;
6128 status = decode_sequence(&xdr, &res->seq_res, rqstp);
6129 if (status != 0)
6130 goto out;
6131 status = decode_getdeviceinfo(&xdr, res->pdev);
6132out:
6133 return status;
6134}
6135
6136/*
6137 * Decode LAYOUTGET response
6138 */
6139static int nfs4_xdr_dec_layoutget(struct rpc_rqst *rqstp, uint32_t *p,
6140 struct nfs4_layoutget_res *res)
6141{
6142 struct xdr_stream xdr;
6143 struct compound_hdr hdr;
6144 int status;
6145
6146 xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p);
6147 status = decode_compound_hdr(&xdr, &hdr);
6148 if (status)
6149 goto out;
6150 status = decode_sequence(&xdr, &res->seq_res, rqstp);
6151 if (status)
6152 goto out;
6153 status = decode_putfh(&xdr);
6154 if (status)
6155 goto out;
6156 status = decode_layoutget(&xdr, rqstp, res);
6157out:
6158 return status;
6159}
5860#endif /* CONFIG_NFS_V4_1 */ 6160#endif /* CONFIG_NFS_V4_1 */
5861 6161
5862__be32 *nfs4_decode_dirent(struct xdr_stream *xdr, struct nfs_entry *entry, 6162__be32 *nfs4_decode_dirent(struct xdr_stream *xdr, struct nfs_entry *entry,
@@ -6048,6 +6348,8 @@ struct rpc_procinfo nfs4_procedures[] = {
6048 PROC(SEQUENCE, enc_sequence, dec_sequence), 6348 PROC(SEQUENCE, enc_sequence, dec_sequence),
6049 PROC(GET_LEASE_TIME, enc_get_lease_time, dec_get_lease_time), 6349 PROC(GET_LEASE_TIME, enc_get_lease_time, dec_get_lease_time),
6050 PROC(RECLAIM_COMPLETE, enc_reclaim_complete, dec_reclaim_complete), 6350 PROC(RECLAIM_COMPLETE, enc_reclaim_complete, dec_reclaim_complete),
6351 PROC(GETDEVICEINFO, enc_getdeviceinfo, dec_getdeviceinfo),
6352 PROC(LAYOUTGET, enc_layoutget, dec_layoutget),
6051#endif /* CONFIG_NFS_V4_1 */ 6353#endif /* CONFIG_NFS_V4_1 */
6052}; 6354};
6053 6355
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 891a0c36f992..d1ad7df3479e 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -140,6 +140,11 @@ pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *ld_type)
140 printk(KERN_ERR "%s id 0 is reserved\n", __func__); 140 printk(KERN_ERR "%s id 0 is reserved\n", __func__);
141 return status; 141 return status;
142 } 142 }
143 if (!ld_type->alloc_lseg || !ld_type->free_lseg) {
144 printk(KERN_ERR "%s Layout driver must provide "
145 "alloc_lseg and free_lseg.\n", __func__);
146 return status;
147 }
143 148
144 spin_lock(&pnfs_spinlock); 149 spin_lock(&pnfs_spinlock);
145 tmp = find_pnfs_driver_locked(ld_type->id); 150 tmp = find_pnfs_driver_locked(ld_type->id);
@@ -168,6 +173,10 @@ pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *ld_type)
168} 173}
169EXPORT_SYMBOL_GPL(pnfs_unregister_layoutdriver); 174EXPORT_SYMBOL_GPL(pnfs_unregister_layoutdriver);
170 175
176/*
177 * pNFS client layout cache
178 */
179
171static void 180static void
172get_layout_hdr_locked(struct pnfs_layout_hdr *lo) 181get_layout_hdr_locked(struct pnfs_layout_hdr *lo)
173{ 182{
@@ -190,7 +199,7 @@ put_layout_hdr_locked(struct pnfs_layout_hdr *lo)
190 } 199 }
191} 200}
192 201
193static void 202void
194put_layout_hdr(struct inode *inode) 203put_layout_hdr(struct inode *inode)
195{ 204{
196 spin_lock(&inode->i_lock); 205 spin_lock(&inode->i_lock);
@@ -215,7 +224,7 @@ destroy_lseg(struct kref *kref)
215 struct inode *ino = lseg->layout->inode; 224 struct inode *ino = lseg->layout->inode;
216 225
217 dprintk("--> %s\n", __func__); 226 dprintk("--> %s\n", __func__);
218 kfree(lseg); 227 NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg);
219 /* Matched by get_layout_hdr_locked in pnfs_insert_layout */ 228 /* Matched by get_layout_hdr_locked in pnfs_insert_layout */
220 put_layout_hdr(ino); 229 put_layout_hdr(ino);
221} 230}
@@ -249,6 +258,9 @@ pnfs_clear_lseg_list(struct pnfs_layout_hdr *lo, struct list_head *tmp_list)
249 /* List does not take a reference, so no need for put here */ 258 /* List does not take a reference, so no need for put here */
250 list_del_init(&lo->layouts); 259 list_del_init(&lo->layouts);
251 spin_unlock(&clp->cl_lock); 260 spin_unlock(&clp->cl_lock);
261 write_seqlock(&lo->seqlock);
262 clear_bit(NFS_LAYOUT_STATEID_SET, &lo->state);
263 write_sequnlock(&lo->seqlock);
252 264
253 dprintk("%s:Return\n", __func__); 265 dprintk("%s:Return\n", __func__);
254} 266}
@@ -307,40 +319,135 @@ pnfs_destroy_all_layouts(struct nfs_client *clp)
307 } 319 }
308} 320}
309 321
310static void pnfs_insert_layout(struct pnfs_layout_hdr *lo, 322/* update lo->stateid with new if is more recent
311 struct pnfs_layout_segment *lseg); 323 *
324 * lo->stateid could be the open stateid, in which case we just use what given.
325 */
326static void
327pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo,
328 const nfs4_stateid *new)
329{
330 nfs4_stateid *old = &lo->stateid;
331 bool overwrite = false;
332
333 write_seqlock(&lo->seqlock);
334 if (!test_bit(NFS_LAYOUT_STATEID_SET, &lo->state) ||
335 memcmp(old->stateid.other, new->stateid.other, sizeof(new->stateid.other)))
336 overwrite = true;
337 else {
338 u32 oldseq, newseq;
339
340 oldseq = be32_to_cpu(old->stateid.seqid);
341 newseq = be32_to_cpu(new->stateid.seqid);
342 if ((int)(newseq - oldseq) > 0)
343 overwrite = true;
344 }
345 if (overwrite)
346 memcpy(&old->stateid, &new->stateid, sizeof(new->stateid));
347 write_sequnlock(&lo->seqlock);
348}
349
350static void
351pnfs_layout_from_open_stateid(struct pnfs_layout_hdr *lo,
352 struct nfs4_state *state)
353{
354 int seq;
355
356 dprintk("--> %s\n", __func__);
357 write_seqlock(&lo->seqlock);
358 do {
359 seq = read_seqbegin(&state->seqlock);
360 memcpy(lo->stateid.data, state->stateid.data,
361 sizeof(state->stateid.data));
362 } while (read_seqretry(&state->seqlock, seq));
363 set_bit(NFS_LAYOUT_STATEID_SET, &lo->state);
364 write_sequnlock(&lo->seqlock);
365 dprintk("<-- %s\n", __func__);
366}
367
368void
369pnfs_get_layout_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo,
370 struct nfs4_state *open_state)
371{
372 int seq;
312 373
313/* Get layout from server. */ 374 dprintk("--> %s\n", __func__);
375 do {
376 seq = read_seqbegin(&lo->seqlock);
377 if (!test_bit(NFS_LAYOUT_STATEID_SET, &lo->state)) {
378 /* This will trigger retry of the read */
379 pnfs_layout_from_open_stateid(lo, open_state);
380 } else
381 memcpy(dst->data, lo->stateid.data,
382 sizeof(lo->stateid.data));
383 } while (read_seqretry(&lo->seqlock, seq));
384 dprintk("<-- %s\n", __func__);
385}
386
387/*
388* Get layout from server.
389* for now, assume that whole file layouts are requested.
390* arg->offset: 0
391* arg->length: all ones
392*/
314static struct pnfs_layout_segment * 393static struct pnfs_layout_segment *
315send_layoutget(struct pnfs_layout_hdr *lo, 394send_layoutget(struct pnfs_layout_hdr *lo,
316 struct nfs_open_context *ctx, 395 struct nfs_open_context *ctx,
317 u32 iomode) 396 u32 iomode)
318{ 397{
319 struct inode *ino = lo->inode; 398 struct inode *ino = lo->inode;
320 struct pnfs_layout_segment *lseg; 399 struct nfs_server *server = NFS_SERVER(ino);
400 struct nfs4_layoutget *lgp;
401 struct pnfs_layout_segment *lseg = NULL;
402
403 dprintk("--> %s\n", __func__);
321 404
322 /* Lets pretend we sent LAYOUTGET and got a response */ 405 BUG_ON(ctx == NULL);
323 lseg = kzalloc(sizeof(*lseg), GFP_KERNEL); 406 lgp = kzalloc(sizeof(*lgp), GFP_KERNEL);
407 if (lgp == NULL) {
408 put_layout_hdr(lo->inode);
409 return NULL;
410 }
411 lgp->args.minlength = NFS4_MAX_UINT64;
412 lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE;
413 lgp->args.range.iomode = iomode;
414 lgp->args.range.offset = 0;
415 lgp->args.range.length = NFS4_MAX_UINT64;
416 lgp->args.type = server->pnfs_curr_ld->id;
417 lgp->args.inode = ino;
418 lgp->args.ctx = get_nfs_open_context(ctx);
419 lgp->lsegpp = &lseg;
420
421 /* Synchronously retrieve layout information from server and
422 * store in lseg.
423 */
424 nfs4_proc_layoutget(lgp);
324 if (!lseg) { 425 if (!lseg) {
426 /* remember that LAYOUTGET failed and suspend trying */
325 set_bit(lo_fail_bit(iomode), &lo->state); 427 set_bit(lo_fail_bit(iomode), &lo->state);
326 spin_lock(&ino->i_lock);
327 put_layout_hdr_locked(lo);
328 spin_unlock(&ino->i_lock);
329 return NULL;
330 } 428 }
331 init_lseg(lo, lseg);
332 lseg->iomode = IOMODE_RW;
333 spin_lock(&ino->i_lock);
334 pnfs_insert_layout(lo, lseg);
335 put_layout_hdr_locked(lo);
336 spin_unlock(&ino->i_lock);
337 return lseg; 429 return lseg;
338} 430}
339 431
432/*
433 * Compare two layout segments for sorting into layout cache.
434 * We want to preferentially return RW over RO layouts, so ensure those
435 * are seen first.
436 */
437static s64
438cmp_layout(u32 iomode1, u32 iomode2)
439{
440 /* read > read/write */
441 return (int)(iomode2 == IOMODE_READ) - (int)(iomode1 == IOMODE_READ);
442}
443
340static void 444static void
341pnfs_insert_layout(struct pnfs_layout_hdr *lo, 445pnfs_insert_layout(struct pnfs_layout_hdr *lo,
342 struct pnfs_layout_segment *lseg) 446 struct pnfs_layout_segment *lseg)
343{ 447{
448 struct pnfs_layout_segment *lp;
449 int found = 0;
450
344 dprintk("%s:Begin\n", __func__); 451 dprintk("%s:Begin\n", __func__);
345 452
346 assert_spin_locked(&lo->inode->i_lock); 453 assert_spin_locked(&lo->inode->i_lock);
@@ -352,19 +459,28 @@ pnfs_insert_layout(struct pnfs_layout_hdr *lo,
352 list_add_tail(&lo->layouts, &clp->cl_layouts); 459 list_add_tail(&lo->layouts, &clp->cl_layouts);
353 spin_unlock(&clp->cl_lock); 460 spin_unlock(&clp->cl_lock);
354 } 461 }
355 get_layout_hdr_locked(lo); 462 list_for_each_entry(lp, &lo->segs, fi_list) {
356 /* STUB - add the constructed lseg if necessary */ 463 if (cmp_layout(lp->range.iomode, lseg->range.iomode) > 0)
357 if (list_empty(&lo->segs)) { 464 continue;
465 list_add_tail(&lseg->fi_list, &lp->fi_list);
466 dprintk("%s: inserted lseg %p "
467 "iomode %d offset %llu length %llu before "
468 "lp %p iomode %d offset %llu length %llu\n",
469 __func__, lseg, lseg->range.iomode,
470 lseg->range.offset, lseg->range.length,
471 lp, lp->range.iomode, lp->range.offset,
472 lp->range.length);
473 found = 1;
474 break;
475 }
476 if (!found) {
358 list_add_tail(&lseg->fi_list, &lo->segs); 477 list_add_tail(&lseg->fi_list, &lo->segs);
359 dprintk("%s: inserted lseg %p iomode %d at tail\n", 478 dprintk("%s: inserted lseg %p "
360 __func__, lseg, lseg->iomode); 479 "iomode %d offset %llu length %llu at tail\n",
361 } else { 480 __func__, lseg, lseg->range.iomode,
362 /* There is no harm for the moment in calling this 481 lseg->range.offset, lseg->range.length);
363 * with the lock held, and the call will be removed
364 * with the STUB.
365 */
366 put_lseg(lseg);
367 } 482 }
483 get_layout_hdr_locked(lo);
368 484
369 dprintk("%s:Return\n", __func__); 485 dprintk("%s:Return\n", __func__);
370} 486}
@@ -380,6 +496,7 @@ alloc_init_layout_hdr(struct inode *ino)
380 lo->refcount = 1; 496 lo->refcount = 1;
381 INIT_LIST_HEAD(&lo->layouts); 497 INIT_LIST_HEAD(&lo->layouts);
382 INIT_LIST_HEAD(&lo->segs); 498 INIT_LIST_HEAD(&lo->segs);
499 seqlock_init(&lo->seqlock);
383 lo->inode = ino; 500 lo->inode = ino;
384 return lo; 501 return lo;
385} 502}
@@ -407,11 +524,46 @@ pnfs_find_alloc_layout(struct inode *ino)
407 return nfsi->layout; 524 return nfsi->layout;
408} 525}
409 526
410/* STUB - LAYOUTGET never succeeds, so cache is empty */ 527/*
528 * iomode matching rules:
529 * iomode lseg match
530 * ----- ----- -----
531 * ANY READ true
532 * ANY RW true
533 * RW READ false
534 * RW RW true
535 * READ READ true
536 * READ RW true
537 */
538static int
539is_matching_lseg(struct pnfs_layout_segment *lseg, u32 iomode)
540{
541 return (iomode != IOMODE_RW || lseg->range.iomode == IOMODE_RW);
542}
543
544/*
545 * lookup range in layout
546 */
411static struct pnfs_layout_segment * 547static struct pnfs_layout_segment *
412pnfs_has_layout(struct pnfs_layout_hdr *lo, u32 iomode) 548pnfs_has_layout(struct pnfs_layout_hdr *lo, u32 iomode)
413{ 549{
414 return NULL; 550 struct pnfs_layout_segment *lseg, *ret = NULL;
551
552 dprintk("%s:Begin\n", __func__);
553
554 assert_spin_locked(&lo->inode->i_lock);
555 list_for_each_entry(lseg, &lo->segs, fi_list) {
556 if (is_matching_lseg(lseg, iomode)) {
557 ret = lseg;
558 break;
559 }
560 if (cmp_layout(iomode, lseg->range.iomode) > 0)
561 break;
562 }
563
564 dprintk("%s:Return lseg %p ref %d\n",
565 __func__, ret, ret ? atomic_read(&ret->kref.refcount) : 0);
566 return ret;
415} 567}
416 568
417/* 569/*
@@ -448,7 +600,7 @@ pnfs_update_layout(struct inode *ino,
448 if (test_bit(lo_fail_bit(iomode), &nfsi->layout->state)) 600 if (test_bit(lo_fail_bit(iomode), &nfsi->layout->state))
449 goto out_unlock; 601 goto out_unlock;
450 602
451 get_layout_hdr_locked(lo); 603 get_layout_hdr_locked(lo); /* Matched in nfs4_layoutget_release */
452 spin_unlock(&ino->i_lock); 604 spin_unlock(&ino->i_lock);
453 605
454 lseg = send_layoutget(lo, ctx, iomode); 606 lseg = send_layoutget(lo, ctx, iomode);
@@ -460,3 +612,172 @@ out_unlock:
460 spin_unlock(&ino->i_lock); 612 spin_unlock(&ino->i_lock);
461 goto out; 613 goto out;
462} 614}
615
616int
617pnfs_layout_process(struct nfs4_layoutget *lgp)
618{
619 struct pnfs_layout_hdr *lo = NFS_I(lgp->args.inode)->layout;
620 struct nfs4_layoutget_res *res = &lgp->res;
621 struct pnfs_layout_segment *lseg;
622 struct inode *ino = lo->inode;
623 int status = 0;
624
625 /* Inject layout blob into I/O device driver */
626 lseg = NFS_SERVER(ino)->pnfs_curr_ld->alloc_lseg(lo, res);
627 if (!lseg || IS_ERR(lseg)) {
628 if (!lseg)
629 status = -ENOMEM;
630 else
631 status = PTR_ERR(lseg);
632 dprintk("%s: Could not allocate layout: error %d\n",
633 __func__, status);
634 goto out;
635 }
636
637 spin_lock(&ino->i_lock);
638 init_lseg(lo, lseg);
639 lseg->range = res->range;
640 *lgp->lsegpp = lseg;
641 pnfs_insert_layout(lo, lseg);
642
643 /* Done processing layoutget. Set the layout stateid */
644 pnfs_set_layout_stateid(lo, &res->stateid);
645 spin_unlock(&ino->i_lock);
646out:
647 return status;
648}
649
650/*
651 * Device ID cache. Currently supports one layout type per struct nfs_client.
652 * Add layout type to the lookup key to expand to support multiple types.
653 */
654int
655pnfs_alloc_init_deviceid_cache(struct nfs_client *clp,
656 void (*free_callback)(struct pnfs_deviceid_node *))
657{
658 struct pnfs_deviceid_cache *c;
659
660 c = kzalloc(sizeof(struct pnfs_deviceid_cache), GFP_KERNEL);
661 if (!c)
662 return -ENOMEM;
663 spin_lock(&clp->cl_lock);
664 if (clp->cl_devid_cache != NULL) {
665 atomic_inc(&clp->cl_devid_cache->dc_ref);
666 dprintk("%s [kref [%d]]\n", __func__,
667 atomic_read(&clp->cl_devid_cache->dc_ref));
668 kfree(c);
669 } else {
670 /* kzalloc initializes hlists */
671 spin_lock_init(&c->dc_lock);
672 atomic_set(&c->dc_ref, 1);
673 c->dc_free_callback = free_callback;
674 clp->cl_devid_cache = c;
675 dprintk("%s [new]\n", __func__);
676 }
677 spin_unlock(&clp->cl_lock);
678 return 0;
679}
680EXPORT_SYMBOL_GPL(pnfs_alloc_init_deviceid_cache);
681
682/*
683 * Called from pnfs_layoutdriver_type->free_lseg
684 * last layout segment reference frees deviceid
685 */
686void
687pnfs_put_deviceid(struct pnfs_deviceid_cache *c,
688 struct pnfs_deviceid_node *devid)
689{
690 struct nfs4_deviceid *id = &devid->de_id;
691 struct pnfs_deviceid_node *d;
692 struct hlist_node *n;
693 long h = nfs4_deviceid_hash(id);
694
695 dprintk("%s [%d]\n", __func__, atomic_read(&devid->de_ref));
696 if (!atomic_dec_and_lock(&devid->de_ref, &c->dc_lock))
697 return;
698
699 hlist_for_each_entry_rcu(d, n, &c->dc_deviceids[h], de_node)
700 if (!memcmp(&d->de_id, id, sizeof(*id))) {
701 hlist_del_rcu(&d->de_node);
702 spin_unlock(&c->dc_lock);
703 synchronize_rcu();
704 c->dc_free_callback(devid);
705 return;
706 }
707 spin_unlock(&c->dc_lock);
708 /* Why wasn't it found in the list? */
709 BUG();
710}
711EXPORT_SYMBOL_GPL(pnfs_put_deviceid);
712
713/* Find and reference a deviceid */
714struct pnfs_deviceid_node *
715pnfs_find_get_deviceid(struct pnfs_deviceid_cache *c, struct nfs4_deviceid *id)
716{
717 struct pnfs_deviceid_node *d;
718 struct hlist_node *n;
719 long hash = nfs4_deviceid_hash(id);
720
721 dprintk("--> %s hash %ld\n", __func__, hash);
722 rcu_read_lock();
723 hlist_for_each_entry_rcu(d, n, &c->dc_deviceids[hash], de_node) {
724 if (!memcmp(&d->de_id, id, sizeof(*id))) {
725 if (!atomic_inc_not_zero(&d->de_ref)) {
726 goto fail;
727 } else {
728 rcu_read_unlock();
729 return d;
730 }
731 }
732 }
733fail:
734 rcu_read_unlock();
735 return NULL;
736}
737EXPORT_SYMBOL_GPL(pnfs_find_get_deviceid);
738
739/*
740 * Add a deviceid to the cache.
741 * GETDEVICEINFOs for same deviceid can race. If deviceid is found, discard new
742 */
743struct pnfs_deviceid_node *
744pnfs_add_deviceid(struct pnfs_deviceid_cache *c, struct pnfs_deviceid_node *new)
745{
746 struct pnfs_deviceid_node *d;
747 long hash = nfs4_deviceid_hash(&new->de_id);
748
749 dprintk("--> %s hash %ld\n", __func__, hash);
750 spin_lock(&c->dc_lock);
751 d = pnfs_find_get_deviceid(c, &new->de_id);
752 if (d) {
753 spin_unlock(&c->dc_lock);
754 dprintk("%s [discard]\n", __func__);
755 c->dc_free_callback(new);
756 return d;
757 }
758 INIT_HLIST_NODE(&new->de_node);
759 atomic_set(&new->de_ref, 1);
760 hlist_add_head_rcu(&new->de_node, &c->dc_deviceids[hash]);
761 spin_unlock(&c->dc_lock);
762 dprintk("%s [new]\n", __func__);
763 return new;
764}
765EXPORT_SYMBOL_GPL(pnfs_add_deviceid);
766
767void
768pnfs_put_deviceid_cache(struct nfs_client *clp)
769{
770 struct pnfs_deviceid_cache *local = clp->cl_devid_cache;
771
772 dprintk("--> %s cl_devid_cache %p\n", __func__, clp->cl_devid_cache);
773 if (atomic_dec_and_lock(&local->dc_ref, &clp->cl_lock)) {
774 int i;
775 /* Verify cache is empty */
776 for (i = 0; i < NFS4_DEVICE_ID_HASH_SIZE; i++)
777 BUG_ON(!hlist_empty(&local->dc_deviceids[i]));
778 clp->cl_devid_cache = NULL;
779 spin_unlock(&clp->cl_lock);
780 kfree(local);
781 }
782}
783EXPORT_SYMBOL_GPL(pnfs_put_deviceid_cache);
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 1c3eb02f4944..cbba28cb02a7 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -32,7 +32,7 @@
32 32
33struct pnfs_layout_segment { 33struct pnfs_layout_segment {
34 struct list_head fi_list; 34 struct list_head fi_list;
35 u32 iomode; 35 struct pnfs_layout_range range;
36 struct kref kref; 36 struct kref kref;
37 struct pnfs_layout_hdr *layout; 37 struct pnfs_layout_hdr *layout;
38}; 38};
@@ -44,6 +44,7 @@ struct pnfs_layout_segment {
44enum { 44enum {
45 NFS_LAYOUT_RO_FAILED = 0, /* get ro layout failed stop trying */ 45 NFS_LAYOUT_RO_FAILED = 0, /* get ro layout failed stop trying */
46 NFS_LAYOUT_RW_FAILED, /* get rw layout failed stop trying */ 46 NFS_LAYOUT_RW_FAILED, /* get rw layout failed stop trying */
47 NFS_LAYOUT_STATEID_SET, /* have a valid layout stateid */
47}; 48};
48 49
49/* Per-layout driver specific registration structure */ 50/* Per-layout driver specific registration structure */
@@ -54,26 +55,96 @@ struct pnfs_layoutdriver_type {
54 struct module *owner; 55 struct module *owner;
55 int (*initialize_mountpoint) (struct nfs_server *); 56 int (*initialize_mountpoint) (struct nfs_server *);
56 int (*uninitialize_mountpoint) (struct nfs_server *); 57 int (*uninitialize_mountpoint) (struct nfs_server *);
58 struct pnfs_layout_segment * (*alloc_lseg) (struct pnfs_layout_hdr *layoutid, struct nfs4_layoutget_res *lgr);
59 void (*free_lseg) (struct pnfs_layout_segment *lseg);
57}; 60};
58 61
59struct pnfs_layout_hdr { 62struct pnfs_layout_hdr {
60 unsigned long refcount; 63 unsigned long refcount;
61 struct list_head layouts; /* other client layouts */ 64 struct list_head layouts; /* other client layouts */
62 struct list_head segs; /* layout segments list */ 65 struct list_head segs; /* layout segments list */
66 seqlock_t seqlock; /* Protects the stateid */
67 nfs4_stateid stateid;
63 unsigned long state; 68 unsigned long state;
64 struct inode *inode; 69 struct inode *inode;
65}; 70};
66 71
72struct pnfs_device {
73 struct nfs4_deviceid dev_id;
74 unsigned int layout_type;
75 unsigned int mincount;
76 struct page **pages;
77 void *area;
78 unsigned int pgbase;
79 unsigned int pglen;
80};
81
82/*
83 * Device ID RCU cache. A device ID is unique per client ID and layout type.
84 */
85#define NFS4_DEVICE_ID_HASH_BITS 5
86#define NFS4_DEVICE_ID_HASH_SIZE (1 << NFS4_DEVICE_ID_HASH_BITS)
87#define NFS4_DEVICE_ID_HASH_MASK (NFS4_DEVICE_ID_HASH_SIZE - 1)
88
89static inline u32
90nfs4_deviceid_hash(struct nfs4_deviceid *id)
91{
92 unsigned char *cptr = (unsigned char *)id->data;
93 unsigned int nbytes = NFS4_DEVICEID4_SIZE;
94 u32 x = 0;
95
96 while (nbytes--) {
97 x *= 37;
98 x += *cptr++;
99 }
100 return x & NFS4_DEVICE_ID_HASH_MASK;
101}
102
103struct pnfs_deviceid_node {
104 struct hlist_node de_node;
105 struct nfs4_deviceid de_id;
106 atomic_t de_ref;
107};
108
109struct pnfs_deviceid_cache {
110 spinlock_t dc_lock;
111 atomic_t dc_ref;
112 void (*dc_free_callback)(struct pnfs_deviceid_node *);
113 struct hlist_head dc_deviceids[NFS4_DEVICE_ID_HASH_SIZE];
114};
115
116extern int pnfs_alloc_init_deviceid_cache(struct nfs_client *,
117 void (*free_callback)(struct pnfs_deviceid_node *));
118extern void pnfs_put_deviceid_cache(struct nfs_client *);
119extern struct pnfs_deviceid_node *pnfs_find_get_deviceid(
120 struct pnfs_deviceid_cache *,
121 struct nfs4_deviceid *);
122extern struct pnfs_deviceid_node *pnfs_add_deviceid(
123 struct pnfs_deviceid_cache *,
124 struct pnfs_deviceid_node *);
125extern void pnfs_put_deviceid(struct pnfs_deviceid_cache *c,
126 struct pnfs_deviceid_node *devid);
127
67extern int pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *); 128extern int pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *);
68extern void pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *); 129extern void pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *);
69 130
131/* nfs4proc.c */
132extern int nfs4_proc_getdeviceinfo(struct nfs_server *server,
133 struct pnfs_device *dev);
134extern int nfs4_proc_layoutget(struct nfs4_layoutget *lgp);
135
136/* pnfs.c */
70struct pnfs_layout_segment * 137struct pnfs_layout_segment *
71pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx, 138pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx,
72 enum pnfs_iomode access_type); 139 enum pnfs_iomode access_type);
73void set_pnfs_layoutdriver(struct nfs_server *, u32 id); 140void set_pnfs_layoutdriver(struct nfs_server *, u32 id);
74void unset_pnfs_layoutdriver(struct nfs_server *); 141void unset_pnfs_layoutdriver(struct nfs_server *);
142int pnfs_layout_process(struct nfs4_layoutget *lgp);
75void pnfs_destroy_layout(struct nfs_inode *); 143void pnfs_destroy_layout(struct nfs_inode *);
76void pnfs_destroy_all_layouts(struct nfs_client *); 144void pnfs_destroy_all_layouts(struct nfs_client *);
145void put_layout_hdr(struct inode *inode);
146void pnfs_get_layout_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo,
147 struct nfs4_state *open_state);
77 148
78 149
79static inline int lo_fail_bit(u32 iomode) 150static inline int lo_fail_bit(u32 iomode)
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index 34da32436ac0..a9683d6acaa4 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -545,6 +545,8 @@ enum {
545 NFSPROC4_CLNT_SEQUENCE, 545 NFSPROC4_CLNT_SEQUENCE,
546 NFSPROC4_CLNT_GET_LEASE_TIME, 546 NFSPROC4_CLNT_GET_LEASE_TIME,
547 NFSPROC4_CLNT_RECLAIM_COMPLETE, 547 NFSPROC4_CLNT_RECLAIM_COMPLETE,
548 NFSPROC4_CLNT_LAYOUTGET,
549 NFSPROC4_CLNT_GETDEVICEINFO,
548}; 550};
549 551
550/* nfs41 types */ 552/* nfs41 types */
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 4d62f1581ed1..452d96436d26 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -83,6 +83,7 @@ struct nfs_client {
83 u32 cl_exchange_flags; 83 u32 cl_exchange_flags;
84 struct nfs4_session *cl_session; /* sharred session */ 84 struct nfs4_session *cl_session; /* sharred session */
85 struct list_head cl_layouts; 85 struct list_head cl_layouts;
86 struct pnfs_deviceid_cache *cl_devid_cache; /* pNFS deviceid cache */
86#endif /* CONFIG_NFS_V4_1 */ 87#endif /* CONFIG_NFS_V4_1 */
87 88
88#ifdef CONFIG_NFS_FSCACHE 89#ifdef CONFIG_NFS_FSCACHE
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 065f9d105d05..ba6cc8f223c9 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -187,6 +187,55 @@ struct nfs4_get_lease_time_res {
187 struct nfs4_sequence_res lr_seq_res; 187 struct nfs4_sequence_res lr_seq_res;
188}; 188};
189 189
190#define PNFS_LAYOUT_MAXSIZE 4096
191
192struct nfs4_layoutdriver_data {
193 __u32 len;
194 void *buf;
195};
196
197struct pnfs_layout_range {
198 u32 iomode;
199 u64 offset;
200 u64 length;
201};
202
203struct nfs4_layoutget_args {
204 __u32 type;
205 struct pnfs_layout_range range;
206 __u64 minlength;
207 __u32 maxcount;
208 struct inode *inode;
209 struct nfs_open_context *ctx;
210 struct nfs4_sequence_args seq_args;
211};
212
213struct nfs4_layoutget_res {
214 __u32 return_on_close;
215 struct pnfs_layout_range range;
216 __u32 type;
217 nfs4_stateid stateid;
218 struct nfs4_layoutdriver_data layout;
219 struct nfs4_sequence_res seq_res;
220};
221
222struct nfs4_layoutget {
223 struct nfs4_layoutget_args args;
224 struct nfs4_layoutget_res res;
225 struct pnfs_layout_segment **lsegpp;
226 int status;
227};
228
229struct nfs4_getdeviceinfo_args {
230 struct pnfs_device *pdev;
231 struct nfs4_sequence_args seq_args;
232};
233
234struct nfs4_getdeviceinfo_res {
235 struct pnfs_device *pdev;
236 struct nfs4_sequence_res seq_res;
237};
238
190/* 239/*
191 * Arguments to the open call. 240 * Arguments to the open call.
192 */ 241 */
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-09 18:11:41 -0500