10Sstevel@tonic-gate /* 20Sstevel@tonic-gate * CDDL HEADER START 30Sstevel@tonic-gate * 40Sstevel@tonic-gate * The contents of this file are subject to the terms of the 511507SVallish.Vaidyeshwara@Sun.COM * Common Development and Distribution License (the "License"). 611507SVallish.Vaidyeshwara@Sun.COM * You may not use this file except in compliance with the License. 70Sstevel@tonic-gate * 80Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 90Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 100Sstevel@tonic-gate * See the License for the specific language governing permissions 110Sstevel@tonic-gate * and limitations under the License. 120Sstevel@tonic-gate * 130Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 140Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 150Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 160Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 170Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 180Sstevel@tonic-gate * 190Sstevel@tonic-gate * CDDL HEADER END 200Sstevel@tonic-gate */ 210Sstevel@tonic-gate /* 22*13096SJordan.Vaughan@Sun.com * Copyright (c) 1986, 2010, Oracle and/or its affiliates. All rights reserved. 230Sstevel@tonic-gate */ 240Sstevel@tonic-gate 250Sstevel@tonic-gate /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ 260Sstevel@tonic-gate /* All Rights Reserved */ 270Sstevel@tonic-gate 280Sstevel@tonic-gate #ifndef _NFS_NFS_CLNT_H 290Sstevel@tonic-gate #define _NFS_NFS_CLNT_H 300Sstevel@tonic-gate 310Sstevel@tonic-gate #include <sys/utsname.h> 320Sstevel@tonic-gate #include <sys/kstat.h> 330Sstevel@tonic-gate #include <sys/time.h> 340Sstevel@tonic-gate #include <vm/page.h> 350Sstevel@tonic-gate #include <sys/thread.h> 360Sstevel@tonic-gate #include <nfs/rnode.h> 370Sstevel@tonic-gate #include <sys/list.h> 3811507SVallish.Vaidyeshwara@Sun.COM #include <sys/condvar_impl.h> 39*13096SJordan.Vaughan@Sun.com #include <sys/zone.h> 400Sstevel@tonic-gate 410Sstevel@tonic-gate #ifdef __cplusplus 420Sstevel@tonic-gate extern "C" { 430Sstevel@tonic-gate #endif 440Sstevel@tonic-gate 450Sstevel@tonic-gate #define HOSTNAMESZ 32 460Sstevel@tonic-gate #define ACREGMIN 3 /* min secs to hold cached file attr */ 470Sstevel@tonic-gate #define ACREGMAX 60 /* max secs to hold cached file attr */ 480Sstevel@tonic-gate #define ACDIRMIN 30 /* min secs to hold cached dir attr */ 490Sstevel@tonic-gate #define ACDIRMAX 60 /* max secs to hold cached dir attr */ 500Sstevel@tonic-gate #define ACMINMAX 3600 /* 1 hr is longest min timeout */ 510Sstevel@tonic-gate #define ACMAXMAX 36000 /* 10 hr is longest max timeout */ 520Sstevel@tonic-gate 530Sstevel@tonic-gate #define NFS_CALLTYPES 3 /* Lookups, Reads, Writes */ 540Sstevel@tonic-gate 550Sstevel@tonic-gate /* 560Sstevel@tonic-gate * rfscall() flags 570Sstevel@tonic-gate */ 580Sstevel@tonic-gate #define RFSCALL_SOFT 0x00000001 /* Do op as if fs was soft-mounted */ 590Sstevel@tonic-gate 600Sstevel@tonic-gate /* 610Sstevel@tonic-gate * Fake errno passed back from rfscall to indicate transfer size adjustment 620Sstevel@tonic-gate */ 630Sstevel@tonic-gate #define ENFS_TRYAGAIN 999 640Sstevel@tonic-gate 650Sstevel@tonic-gate /* 6611507SVallish.Vaidyeshwara@Sun.COM * The NFS specific async_reqs structure. iotype is grouped to support two 6711507SVallish.Vaidyeshwara@Sun.COM * types of async thread pools, please read comments section of mntinfo_t 6811507SVallish.Vaidyeshwara@Sun.COM * definition for more information. Care should be taken while adding new 6911507SVallish.Vaidyeshwara@Sun.COM * members to this group. 700Sstevel@tonic-gate */ 710Sstevel@tonic-gate 720Sstevel@tonic-gate enum iotype { 730Sstevel@tonic-gate NFS_PUTAPAGE, 740Sstevel@tonic-gate NFS_PAGEIO, 7511507SVallish.Vaidyeshwara@Sun.COM NFS_COMMIT, 7611507SVallish.Vaidyeshwara@Sun.COM NFS_READ_AHEAD, 770Sstevel@tonic-gate NFS_READDIR, 7811507SVallish.Vaidyeshwara@Sun.COM NFS_INACTIVE, 7911507SVallish.Vaidyeshwara@Sun.COM NFS_ASYNC_TYPES 800Sstevel@tonic-gate }; 8111507SVallish.Vaidyeshwara@Sun.COM #define NFS_ASYNC_PGOPS_TYPES (NFS_COMMIT + 1) 8211507SVallish.Vaidyeshwara@Sun.COM 8311507SVallish.Vaidyeshwara@Sun.COM /* 8411507SVallish.Vaidyeshwara@Sun.COM * NFS async requests queue type. 8511507SVallish.Vaidyeshwara@Sun.COM */ 8611507SVallish.Vaidyeshwara@Sun.COM 8711507SVallish.Vaidyeshwara@Sun.COM enum ioqtype { 8811507SVallish.Vaidyeshwara@Sun.COM NFS_ASYNC_QUEUE, 8911507SVallish.Vaidyeshwara@Sun.COM NFS_ASYNC_PGOPS_QUEUE, 9011507SVallish.Vaidyeshwara@Sun.COM NFS_MAX_ASYNC_QUEUES 9111507SVallish.Vaidyeshwara@Sun.COM }; 9211507SVallish.Vaidyeshwara@Sun.COM 9311507SVallish.Vaidyeshwara@Sun.COM /* 9411507SVallish.Vaidyeshwara@Sun.COM * Number of NFS async threads operating exclusively on page op requests. 9511507SVallish.Vaidyeshwara@Sun.COM */ 9611507SVallish.Vaidyeshwara@Sun.COM #define NUM_ASYNC_PGOPS_THREADS 0x2 970Sstevel@tonic-gate 980Sstevel@tonic-gate struct nfs_async_read_req { 990Sstevel@tonic-gate void (*readahead)(); /* pointer to readahead function */ 1000Sstevel@tonic-gate u_offset_t blkoff; /* offset in file */ 1010Sstevel@tonic-gate struct seg *seg; /* segment to do i/o to */ 1020Sstevel@tonic-gate caddr_t addr; /* address to do i/o to */ 1030Sstevel@tonic-gate }; 1040Sstevel@tonic-gate 1050Sstevel@tonic-gate struct nfs_pageio_req { 1060Sstevel@tonic-gate int (*pageio)(); /* pointer to pageio function */ 1070Sstevel@tonic-gate page_t *pp; /* page list */ 1080Sstevel@tonic-gate u_offset_t io_off; /* offset in file */ 1090Sstevel@tonic-gate uint_t io_len; /* size of request */ 1100Sstevel@tonic-gate int flags; 1110Sstevel@tonic-gate }; 1120Sstevel@tonic-gate 1130Sstevel@tonic-gate struct nfs_readdir_req { 1140Sstevel@tonic-gate int (*readdir)(); /* pointer to readdir function */ 1150Sstevel@tonic-gate struct rddir_cache *rdc; /* pointer to cache entry to fill */ 1160Sstevel@tonic-gate }; 1170Sstevel@tonic-gate 1180Sstevel@tonic-gate struct nfs_commit_req { 1190Sstevel@tonic-gate void (*commit)(); /* pointer to commit function */ 1200Sstevel@tonic-gate page_t *plist; /* page list */ 1210Sstevel@tonic-gate offset3 offset; /* starting offset */ 1220Sstevel@tonic-gate count3 count; /* size of range to be commited */ 1230Sstevel@tonic-gate }; 1240Sstevel@tonic-gate 1250Sstevel@tonic-gate struct nfs_inactive_req { 1260Sstevel@tonic-gate void (*inactive)(); /* pointer to inactive function */ 1270Sstevel@tonic-gate }; 1280Sstevel@tonic-gate 1290Sstevel@tonic-gate struct nfs_async_reqs { 1300Sstevel@tonic-gate struct nfs_async_reqs *a_next; /* pointer to next arg struct */ 1310Sstevel@tonic-gate #ifdef DEBUG 1320Sstevel@tonic-gate kthread_t *a_queuer; /* thread id of queueing thread */ 1330Sstevel@tonic-gate #endif 1340Sstevel@tonic-gate struct vnode *a_vp; /* vnode pointer */ 1350Sstevel@tonic-gate struct cred *a_cred; /* cred pointer */ 1360Sstevel@tonic-gate enum iotype a_io; /* i/o type */ 1370Sstevel@tonic-gate union { 1380Sstevel@tonic-gate struct nfs_async_read_req a_read_args; 1390Sstevel@tonic-gate struct nfs_pageio_req a_pageio_args; 1400Sstevel@tonic-gate struct nfs_readdir_req a_readdir_args; 1410Sstevel@tonic-gate struct nfs_commit_req a_commit_args; 1420Sstevel@tonic-gate struct nfs_inactive_req a_inactive_args; 1430Sstevel@tonic-gate } a_args; 1440Sstevel@tonic-gate }; 1450Sstevel@tonic-gate 1460Sstevel@tonic-gate #define a_nfs_readahead a_args.a_read_args.readahead 1470Sstevel@tonic-gate #define a_nfs_blkoff a_args.a_read_args.blkoff 1480Sstevel@tonic-gate #define a_nfs_seg a_args.a_read_args.seg 1490Sstevel@tonic-gate #define a_nfs_addr a_args.a_read_args.addr 1500Sstevel@tonic-gate 1510Sstevel@tonic-gate #define a_nfs_putapage a_args.a_pageio_args.pageio 1520Sstevel@tonic-gate #define a_nfs_pageio a_args.a_pageio_args.pageio 1530Sstevel@tonic-gate #define a_nfs_pp a_args.a_pageio_args.pp 1540Sstevel@tonic-gate #define a_nfs_off a_args.a_pageio_args.io_off 1550Sstevel@tonic-gate #define a_nfs_len a_args.a_pageio_args.io_len 1560Sstevel@tonic-gate #define a_nfs_flags a_args.a_pageio_args.flags 1570Sstevel@tonic-gate 1580Sstevel@tonic-gate #define a_nfs_readdir a_args.a_readdir_args.readdir 1590Sstevel@tonic-gate #define a_nfs_rdc a_args.a_readdir_args.rdc 1600Sstevel@tonic-gate 1610Sstevel@tonic-gate #define a_nfs_commit a_args.a_commit_args.commit 1620Sstevel@tonic-gate #define a_nfs_plist a_args.a_commit_args.plist 1630Sstevel@tonic-gate #define a_nfs_offset a_args.a_commit_args.offset 1640Sstevel@tonic-gate #define a_nfs_count a_args.a_commit_args.count 1650Sstevel@tonic-gate 1660Sstevel@tonic-gate #define a_nfs_inactive a_args.a_inactive_args.inactive 1670Sstevel@tonic-gate 1680Sstevel@tonic-gate /* 1690Sstevel@tonic-gate * Due to the way the address space callbacks are used to execute a delmap, 1700Sstevel@tonic-gate * we must keep track of how many times the same thread has called 1710Sstevel@tonic-gate * VOP_DELMAP()->nfs_delmap()/nfs3_delmap(). This is done by having a list of 1720Sstevel@tonic-gate * nfs_delmapcall_t's associated with each rnode_t. This list is protected 1730Sstevel@tonic-gate * by the rnode_t's r_statelock. The individual elements do not need to be 1740Sstevel@tonic-gate * protected as they will only ever be created, modified and destroyed by 1750Sstevel@tonic-gate * one thread (the call_id). 1760Sstevel@tonic-gate * See nfs_delmap()/nfs3_delmap() for further explanation. 1770Sstevel@tonic-gate */ 1780Sstevel@tonic-gate typedef struct nfs_delmapcall { 1790Sstevel@tonic-gate kthread_t *call_id; 1800Sstevel@tonic-gate int error; /* error from delmap */ 1810Sstevel@tonic-gate list_node_t call_node; 1820Sstevel@tonic-gate } nfs_delmapcall_t; 1830Sstevel@tonic-gate 1840Sstevel@tonic-gate /* 1850Sstevel@tonic-gate * delmap address space callback args 1860Sstevel@tonic-gate */ 1870Sstevel@tonic-gate typedef struct nfs_delmap_args { 1880Sstevel@tonic-gate vnode_t *vp; 1890Sstevel@tonic-gate offset_t off; 1900Sstevel@tonic-gate caddr_t addr; 1910Sstevel@tonic-gate size_t len; 1920Sstevel@tonic-gate uint_t prot; 1930Sstevel@tonic-gate uint_t maxprot; 1940Sstevel@tonic-gate uint_t flags; 1950Sstevel@tonic-gate cred_t *cr; 1960Sstevel@tonic-gate nfs_delmapcall_t *caller; /* to retrieve errors from the cb */ 1970Sstevel@tonic-gate } nfs_delmap_args_t; 1980Sstevel@tonic-gate 1990Sstevel@tonic-gate #ifdef _KERNEL 2000Sstevel@tonic-gate extern nfs_delmapcall_t *nfs_init_delmapcall(void); 2010Sstevel@tonic-gate extern void nfs_free_delmapcall(nfs_delmapcall_t *); 2020Sstevel@tonic-gate extern int nfs_find_and_delete_delmapcall(rnode_t *, int *errp); 2030Sstevel@tonic-gate #endif /* _KERNEL */ 2040Sstevel@tonic-gate 2050Sstevel@tonic-gate /* 2060Sstevel@tonic-gate * The following structures, chhead and chtab, make up the client handle 2070Sstevel@tonic-gate * cache. chhead represents a quadruple(RPC program, RPC version, Protocol 2080Sstevel@tonic-gate * Family, and Transport). For example, a chhead entry could represent 2090Sstevel@tonic-gate * NFS/V3/IPv4/TCP requests. chhead nodes are linked together as a singly 2100Sstevel@tonic-gate * linked list and is referenced from chtable. 2110Sstevel@tonic-gate * 2120Sstevel@tonic-gate * chtab represents an allocated client handle bound to a particular 2130Sstevel@tonic-gate * quadruple. These nodes chain down from a chhead node. chtab 2140Sstevel@tonic-gate * entries which are on the chain are considered free, so a thread may simply 2150Sstevel@tonic-gate * unlink the first node without traversing the chain. When the thread is 2160Sstevel@tonic-gate * completed with its request, it puts the chtab node back on the chain. 2170Sstevel@tonic-gate */ 2180Sstevel@tonic-gate typedef struct chhead { 2190Sstevel@tonic-gate struct chhead *ch_next; /* next quadruple */ 2200Sstevel@tonic-gate struct chtab *ch_list; /* pointer to free client handle(s) */ 2210Sstevel@tonic-gate uint64_t ch_timesused; /* times this quadruple was requested */ 2220Sstevel@tonic-gate rpcprog_t ch_prog; /* RPC program number */ 2230Sstevel@tonic-gate rpcvers_t ch_vers; /* RPC version number */ 2240Sstevel@tonic-gate dev_t ch_dev; /* pseudo device number (i.e. /dev/udp) */ 2250Sstevel@tonic-gate char *ch_protofmly; /* protocol (i.e. NC_INET, NC_LOOPBACK) */ 2260Sstevel@tonic-gate } chhead_t; 2270Sstevel@tonic-gate 2280Sstevel@tonic-gate typedef struct chtab { 2290Sstevel@tonic-gate struct chtab *ch_list; /* next free client handle */ 2300Sstevel@tonic-gate struct chhead *ch_head; /* associated quadruple */ 2310Sstevel@tonic-gate time_t ch_freed; /* timestamp when freed */ 2320Sstevel@tonic-gate CLIENT *ch_client; /* pointer to client handle */ 2330Sstevel@tonic-gate } chtab_t; 2340Sstevel@tonic-gate 2350Sstevel@tonic-gate /* 2360Sstevel@tonic-gate * clinfo is a structure which encapsulates data that is needed to 2370Sstevel@tonic-gate * obtain a client handle from the cache 2380Sstevel@tonic-gate */ 2390Sstevel@tonic-gate typedef struct clinfo { 2400Sstevel@tonic-gate rpcprog_t cl_prog; /* RPC program number */ 2410Sstevel@tonic-gate rpcvers_t cl_vers; /* RPC version number */ 2420Sstevel@tonic-gate uint_t cl_readsize; /* transfer size */ 2430Sstevel@tonic-gate int cl_retrans; /* times to retry request */ 2440Sstevel@tonic-gate uint_t cl_flags; /* info flags */ 2450Sstevel@tonic-gate } clinfo_t; 2460Sstevel@tonic-gate 2470Sstevel@tonic-gate /* 2480Sstevel@tonic-gate * Failover information, passed opaquely through rfscall() 2490Sstevel@tonic-gate */ 2500Sstevel@tonic-gate typedef struct failinfo { 2510Sstevel@tonic-gate struct vnode *vp; 2520Sstevel@tonic-gate caddr_t fhp; 2530Sstevel@tonic-gate void (*copyproc)(caddr_t, vnode_t *); 2540Sstevel@tonic-gate int (*lookupproc)(vnode_t *, char *, vnode_t **, struct pathname *, 2550Sstevel@tonic-gate int, vnode_t *, struct cred *, int); 2560Sstevel@tonic-gate int (*xattrdirproc)(vnode_t *, vnode_t **, bool_t, cred_t *, int); 2570Sstevel@tonic-gate } failinfo_t; 2580Sstevel@tonic-gate 2590Sstevel@tonic-gate /* 2600Sstevel@tonic-gate * Static server information 2610Sstevel@tonic-gate * 2620Sstevel@tonic-gate * These fields are protected by sv_lock: 2630Sstevel@tonic-gate * sv_flags 2640Sstevel@tonic-gate */ 2650Sstevel@tonic-gate typedef struct servinfo { 2660Sstevel@tonic-gate struct knetconfig *sv_knconf; /* bound TLI fd */ 2670Sstevel@tonic-gate struct knetconfig *sv_origknconf; /* For RDMA save orig knconf */ 2680Sstevel@tonic-gate struct netbuf sv_addr; /* server's address */ 2690Sstevel@tonic-gate nfs_fhandle sv_fhandle; /* this server's filehandle */ 2700Sstevel@tonic-gate struct sec_data *sv_secdata; /* security data for rpcsec module */ 2710Sstevel@tonic-gate char *sv_hostname; /* server's hostname */ 2720Sstevel@tonic-gate int sv_hostnamelen; /* server's hostname length */ 2730Sstevel@tonic-gate uint_t sv_flags; /* see below */ 2740Sstevel@tonic-gate struct servinfo *sv_next; /* next in list */ 2750Sstevel@tonic-gate kmutex_t sv_lock; 2760Sstevel@tonic-gate } servinfo_t; 2770Sstevel@tonic-gate 2780Sstevel@tonic-gate /* 2790Sstevel@tonic-gate * The values for sv_flags. 2800Sstevel@tonic-gate */ 2810Sstevel@tonic-gate #define SV_ROOT_STALE 0x1 /* root vnode got ESTALE */ 2820Sstevel@tonic-gate 2830Sstevel@tonic-gate /* 2840Sstevel@tonic-gate * Switch from RDMA knconf to original mount knconf 2850Sstevel@tonic-gate */ 2860Sstevel@tonic-gate 2870Sstevel@tonic-gate #define ORIG_KNCONF(mi) (mi->mi_curr_serv->sv_origknconf ? \ 2880Sstevel@tonic-gate mi->mi_curr_serv->sv_origknconf : mi->mi_curr_serv->sv_knconf) 2890Sstevel@tonic-gate 290*13096SJordan.Vaughan@Sun.com #if defined(_KERNEL) 2910Sstevel@tonic-gate /* 2920Sstevel@tonic-gate * NFS private data per mounted file system 2930Sstevel@tonic-gate * The mi_lock mutex protects the following fields: 2940Sstevel@tonic-gate * mi_flags 2950Sstevel@tonic-gate * mi_printed 2960Sstevel@tonic-gate * mi_down 2970Sstevel@tonic-gate * mi_tsize 2980Sstevel@tonic-gate * mi_stsize 2990Sstevel@tonic-gate * mi_curread 3000Sstevel@tonic-gate * mi_curwrite 3010Sstevel@tonic-gate * mi_timers 3020Sstevel@tonic-gate * mi_curr_serv 3030Sstevel@tonic-gate * mi_readers 3040Sstevel@tonic-gate * mi_klmconfig 3050Sstevel@tonic-gate * 3060Sstevel@tonic-gate * The mi_async_lock mutex protects the following fields: 3070Sstevel@tonic-gate * mi_async_reqs 3080Sstevel@tonic-gate * mi_async_req_count 3090Sstevel@tonic-gate * mi_async_tail 31011507SVallish.Vaidyeshwara@Sun.COM * mi_async_curr[NFS_MAX_ASYNC_QUEUES] 3110Sstevel@tonic-gate * mi_async_clusters 3120Sstevel@tonic-gate * mi_async_init_clusters 31311507SVallish.Vaidyeshwara@Sun.COM * mi_threads[NFS_MAX_ASYNC_QUEUES] 3140Sstevel@tonic-gate * mi_manager_thread 3150Sstevel@tonic-gate * 3160Sstevel@tonic-gate * Normally the netconfig information for the mount comes from 3170Sstevel@tonic-gate * mi_curr_serv and mi_klmconfig is NULL. If NLM calls need to use a 3180Sstevel@tonic-gate * different transport, mi_klmconfig contains the necessary netconfig 3190Sstevel@tonic-gate * information. 3200Sstevel@tonic-gate * 3210Sstevel@tonic-gate * 'mi_zone' is initialized at structure creation time, and never 3220Sstevel@tonic-gate * changes; it may be read without a lock. 3230Sstevel@tonic-gate * 3240Sstevel@tonic-gate * mi_zone_node is linkage into the mi4_globals.mig_list, and is 3250Sstevel@tonic-gate * protected by mi4_globals.mig_list_lock. 3260Sstevel@tonic-gate * 3270Sstevel@tonic-gate * Locking order: 3280Sstevel@tonic-gate * mi_globals::mig_lock > mi_async_lock > mi_lock 3290Sstevel@tonic-gate */ 3300Sstevel@tonic-gate typedef struct mntinfo { 3310Sstevel@tonic-gate kmutex_t mi_lock; /* protects mntinfo fields */ 3320Sstevel@tonic-gate struct servinfo *mi_servers; /* server list */ 3330Sstevel@tonic-gate struct servinfo *mi_curr_serv; /* current server */ 3340Sstevel@tonic-gate kcondvar_t mi_failover_cv; /* failover synchronization */ 3350Sstevel@tonic-gate int mi_readers; /* failover - users of mi_curr_serv */ 3360Sstevel@tonic-gate struct vfs *mi_vfsp; /* back pointer to vfs */ 3370Sstevel@tonic-gate enum vtype mi_type; /* file type of the root vnode */ 3380Sstevel@tonic-gate uint_t mi_flags; /* see below */ 3390Sstevel@tonic-gate uint_t mi_tsize; /* max read transfer size (bytes) */ 3400Sstevel@tonic-gate uint_t mi_stsize; /* max write transfer size (bytes) */ 3410Sstevel@tonic-gate int mi_timeo; /* inital timeout in 10th sec */ 3420Sstevel@tonic-gate int mi_retrans; /* times to retry request */ 3430Sstevel@tonic-gate hrtime_t mi_acregmin; /* min time to hold cached file attr */ 3440Sstevel@tonic-gate hrtime_t mi_acregmax; /* max time to hold cached file attr */ 3450Sstevel@tonic-gate hrtime_t mi_acdirmin; /* min time to hold cached dir attr */ 3460Sstevel@tonic-gate hrtime_t mi_acdirmax; /* max time to hold cached dir attr */ 3470Sstevel@tonic-gate len_t mi_maxfilesize; /* for pathconf _PC_FILESIZEBITS */ 3480Sstevel@tonic-gate /* 3490Sstevel@tonic-gate * Extra fields for congestion control, one per NFS call type, 3500Sstevel@tonic-gate * plus one global one. 3510Sstevel@tonic-gate */ 3520Sstevel@tonic-gate struct rpc_timers mi_timers[NFS_CALLTYPES+1]; 3530Sstevel@tonic-gate int mi_curread; /* current read size */ 3540Sstevel@tonic-gate int mi_curwrite; /* current write size */ 3550Sstevel@tonic-gate /* 35611507SVallish.Vaidyeshwara@Sun.COM * Async I/O management 35711507SVallish.Vaidyeshwara@Sun.COM * We have 2 pools of threads working on async I/O: 35811507SVallish.Vaidyeshwara@Sun.COM * (i) Threads which work on all async queues. Default number of 35911507SVallish.Vaidyeshwara@Sun.COM * threads in this queue is 8. Threads in this pool work on async 36011507SVallish.Vaidyeshwara@Sun.COM * queue pointed by mi_async_curr[NFS_ASYNC_QUEUE]. Number of 36111507SVallish.Vaidyeshwara@Sun.COM * active threads in this pool is tracked by 36211507SVallish.Vaidyeshwara@Sun.COM * mi_threads[NFS_ASYNC_QUEUE]. 36311507SVallish.Vaidyeshwara@Sun.COM * (ii)Threads which work only on page op async queues. 36411507SVallish.Vaidyeshwara@Sun.COM * Page ops queue comprises of NFS_PUTAPAGE, NFS_PAGEIO & 36511507SVallish.Vaidyeshwara@Sun.COM * NFS_COMMIT. Default number of threads in this queue is 2 36611507SVallish.Vaidyeshwara@Sun.COM * (NUM_ASYNC_PGOPS_THREADS). Threads in this pool work on async 36711507SVallish.Vaidyeshwara@Sun.COM * queue pointed by mi_async_curr[NFS_ASYNC_PGOPS_QUEUE]. Number 36811507SVallish.Vaidyeshwara@Sun.COM * of active threads in this pool is tracked by 36911507SVallish.Vaidyeshwara@Sun.COM * mi_threads[NFS_ASYNC_PGOPS_QUEUE]. 3700Sstevel@tonic-gate */ 3710Sstevel@tonic-gate struct nfs_async_reqs *mi_async_reqs[NFS_ASYNC_TYPES]; 3720Sstevel@tonic-gate struct nfs_async_reqs *mi_async_tail[NFS_ASYNC_TYPES]; 37311507SVallish.Vaidyeshwara@Sun.COM struct nfs_async_reqs **mi_async_curr[NFS_MAX_ASYNC_QUEUES]; 37411507SVallish.Vaidyeshwara@Sun.COM /* current async queue */ 3750Sstevel@tonic-gate uint_t mi_async_clusters[NFS_ASYNC_TYPES]; 3760Sstevel@tonic-gate uint_t mi_async_init_clusters; 3770Sstevel@tonic-gate uint_t mi_async_req_count; /* # outstanding work requests */ 3780Sstevel@tonic-gate kcondvar_t mi_async_reqs_cv; /* signaled when there's work */ 37911507SVallish.Vaidyeshwara@Sun.COM ushort_t mi_threads[NFS_MAX_ASYNC_QUEUES]; 38011507SVallish.Vaidyeshwara@Sun.COM /* number of active async threads */ 3810Sstevel@tonic-gate ushort_t mi_max_threads; /* max number of async worker threads */ 3820Sstevel@tonic-gate kthread_t *mi_manager_thread; /* async manager thread */ 3830Sstevel@tonic-gate kcondvar_t mi_async_cv; /* signaled when the last worker dies */ 38411507SVallish.Vaidyeshwara@Sun.COM kcondvar_t mi_async_work_cv[NFS_MAX_ASYNC_QUEUES]; 38511507SVallish.Vaidyeshwara@Sun.COM /* tell workers to work */ 3860Sstevel@tonic-gate kmutex_t mi_async_lock; /* lock to protect async list */ 3870Sstevel@tonic-gate /* 3880Sstevel@tonic-gate * Other stuff 3890Sstevel@tonic-gate */ 3900Sstevel@tonic-gate struct pathcnf *mi_pathconf; /* static pathconf kludge */ 3910Sstevel@tonic-gate rpcprog_t mi_prog; /* RPC program number */ 3920Sstevel@tonic-gate rpcvers_t mi_vers; /* RPC program version number */ 3930Sstevel@tonic-gate char **mi_rfsnames; /* mapping to proc names */ 3940Sstevel@tonic-gate kstat_named_t *mi_reqs; /* count of requests */ 3950Sstevel@tonic-gate uchar_t *mi_call_type; /* dynamic retrans call types */ 3960Sstevel@tonic-gate uchar_t *mi_ss_call_type; /* semisoft call type */ 3970Sstevel@tonic-gate uchar_t *mi_timer_type; /* dynamic retrans timer types */ 3980Sstevel@tonic-gate clock_t mi_printftime; /* last error printf time */ 3990Sstevel@tonic-gate /* 4000Sstevel@tonic-gate * ACL entries 4010Sstevel@tonic-gate */ 4020Sstevel@tonic-gate char **mi_aclnames; /* mapping to proc names */ 4030Sstevel@tonic-gate kstat_named_t *mi_aclreqs; /* count of acl requests */ 4040Sstevel@tonic-gate uchar_t *mi_acl_call_type; /* dynamic retrans call types */ 4050Sstevel@tonic-gate uchar_t *mi_acl_ss_call_type; /* semisoft call types */ 4060Sstevel@tonic-gate uchar_t *mi_acl_timer_type; /* dynamic retrans timer types */ 4070Sstevel@tonic-gate /* 4080Sstevel@tonic-gate * Client Side Failover stats 4090Sstevel@tonic-gate */ 4100Sstevel@tonic-gate uint_t mi_noresponse; /* server not responding count */ 4110Sstevel@tonic-gate uint_t mi_failover; /* failover to new server count */ 4120Sstevel@tonic-gate uint_t mi_remap; /* remap to new server count */ 4130Sstevel@tonic-gate /* 4140Sstevel@tonic-gate * Kstat statistics 4150Sstevel@tonic-gate */ 4160Sstevel@tonic-gate struct kstat *mi_io_kstats; 4170Sstevel@tonic-gate struct kstat *mi_ro_kstats; 4180Sstevel@tonic-gate struct knetconfig *mi_klmconfig; 4190Sstevel@tonic-gate /* 4200Sstevel@tonic-gate * Zones support. 4210Sstevel@tonic-gate */ 422*13096SJordan.Vaughan@Sun.com struct zone *mi_zone; /* Zone in which FS is mounted */ 423*13096SJordan.Vaughan@Sun.com zone_ref_t mi_zone_ref; /* Reference to aforementioned zone */ 4240Sstevel@tonic-gate list_node_t mi_zone_node; /* Linkage into per-zone mi list */ 4251068Svv149972 /* 4261068Svv149972 * Serializes threads in failover_remap. 4271068Svv149972 * Need to acquire this lock first in failover_remap() function 4281068Svv149972 * before acquiring any other rnode lock. 4291068Svv149972 */ 4301068Svv149972 kmutex_t mi_remap_lock; 4310Sstevel@tonic-gate } mntinfo_t; 432*13096SJordan.Vaughan@Sun.com #endif /* _KERNEL */ 4330Sstevel@tonic-gate 4340Sstevel@tonic-gate /* 4350Sstevel@tonic-gate * vfs pointer to mount info 4360Sstevel@tonic-gate */ 4370Sstevel@tonic-gate #define VFTOMI(vfsp) ((mntinfo_t *)((vfsp)->vfs_data)) 4380Sstevel@tonic-gate 4390Sstevel@tonic-gate /* 4400Sstevel@tonic-gate * vnode pointer to mount info 4410Sstevel@tonic-gate */ 4420Sstevel@tonic-gate #define VTOMI(vp) ((mntinfo_t *)(((vp)->v_vfsp)->vfs_data)) 4430Sstevel@tonic-gate 4440Sstevel@tonic-gate /* 4450Sstevel@tonic-gate * The values for mi_flags. 4460Sstevel@tonic-gate */ 4470Sstevel@tonic-gate #define MI_HARD 0x1 /* hard or soft mount */ 4480Sstevel@tonic-gate #define MI_PRINTED 0x2 /* not responding message printed */ 4490Sstevel@tonic-gate #define MI_INT 0x4 /* interrupts allowed on hard mount */ 4500Sstevel@tonic-gate #define MI_DOWN 0x8 /* server is down */ 4510Sstevel@tonic-gate #define MI_NOAC 0x10 /* don't cache attributes */ 4520Sstevel@tonic-gate #define MI_NOCTO 0x20 /* no close-to-open consistency */ 4530Sstevel@tonic-gate #define MI_DYNAMIC 0x40 /* dynamic transfer size adjustment */ 4540Sstevel@tonic-gate #define MI_LLOCK 0x80 /* local locking only (no lockmgr) */ 4550Sstevel@tonic-gate #define MI_GRPID 0x100 /* System V group id inheritance */ 4560Sstevel@tonic-gate #define MI_RPCTIMESYNC 0x200 /* RPC time sync */ 4570Sstevel@tonic-gate #define MI_LINK 0x400 /* server supports link */ 4580Sstevel@tonic-gate #define MI_SYMLINK 0x800 /* server supports symlink */ 4590Sstevel@tonic-gate #define MI_READDIRONLY 0x1000 /* use readdir instead of readdirplus */ 4600Sstevel@tonic-gate #define MI_ACL 0x2000 /* server supports NFS_ACL */ 4610Sstevel@tonic-gate #define MI_BINDINPROG 0x4000 /* binding to server is changing */ 4620Sstevel@tonic-gate #define MI_LOOPBACK 0x8000 /* Set if this is a loopback mount */ 4630Sstevel@tonic-gate #define MI_SEMISOFT 0x10000 /* soft reads, hard modify */ 4640Sstevel@tonic-gate #define MI_NOPRINT 0x20000 /* don't print messages */ 4650Sstevel@tonic-gate #define MI_DIRECTIO 0x40000 /* do direct I/O */ 4660Sstevel@tonic-gate #define MI_EXTATTR 0x80000 /* server supports extended attrs */ 4670Sstevel@tonic-gate #define MI_ASYNC_MGR_STOP 0x100000 /* tell async mgr to die */ 468264Sthurlow #define MI_DEAD 0x200000 /* mount has been terminated */ 4690Sstevel@tonic-gate 4700Sstevel@tonic-gate /* 4710Sstevel@tonic-gate * Read-only mntinfo statistics 4720Sstevel@tonic-gate */ 4730Sstevel@tonic-gate struct mntinfo_kstat { 4740Sstevel@tonic-gate char mik_proto[KNC_STRSIZE]; 4750Sstevel@tonic-gate uint32_t mik_vers; 4760Sstevel@tonic-gate uint_t mik_flags; 4770Sstevel@tonic-gate uint_t mik_secmod; 4780Sstevel@tonic-gate uint32_t mik_curread; 4790Sstevel@tonic-gate uint32_t mik_curwrite; 4800Sstevel@tonic-gate int mik_timeo; 4810Sstevel@tonic-gate int mik_retrans; 4820Sstevel@tonic-gate uint_t mik_acregmin; 4830Sstevel@tonic-gate uint_t mik_acregmax; 4840Sstevel@tonic-gate uint_t mik_acdirmin; 4850Sstevel@tonic-gate uint_t mik_acdirmax; 4860Sstevel@tonic-gate struct { 4870Sstevel@tonic-gate uint32_t srtt; 4880Sstevel@tonic-gate uint32_t deviate; 4890Sstevel@tonic-gate uint32_t rtxcur; 4900Sstevel@tonic-gate } mik_timers[NFS_CALLTYPES+1]; 4910Sstevel@tonic-gate uint32_t mik_noresponse; 4920Sstevel@tonic-gate uint32_t mik_failover; 4930Sstevel@tonic-gate uint32_t mik_remap; 4940Sstevel@tonic-gate char mik_curserver[SYS_NMLN]; 4950Sstevel@tonic-gate }; 4960Sstevel@tonic-gate 4970Sstevel@tonic-gate /* 49811507SVallish.Vaidyeshwara@Sun.COM * Macro to wakeup sleeping async worker threads. 49911507SVallish.Vaidyeshwara@Sun.COM */ 50011507SVallish.Vaidyeshwara@Sun.COM #define NFS_WAKE_ASYNC_WORKER(work_cv) { \ 50111507SVallish.Vaidyeshwara@Sun.COM if (CV_HAS_WAITERS(&work_cv[NFS_ASYNC_QUEUE])) \ 50211507SVallish.Vaidyeshwara@Sun.COM cv_signal(&work_cv[NFS_ASYNC_QUEUE]); \ 50311507SVallish.Vaidyeshwara@Sun.COM else if (CV_HAS_WAITERS(&work_cv[NFS_ASYNC_PGOPS_QUEUE])) \ 50411507SVallish.Vaidyeshwara@Sun.COM cv_signal(&work_cv[NFS_ASYNC_PGOPS_QUEUE]); \ 50511507SVallish.Vaidyeshwara@Sun.COM } 50611507SVallish.Vaidyeshwara@Sun.COM 50711507SVallish.Vaidyeshwara@Sun.COM #define NFS_WAKEALL_ASYNC_WORKERS(work_cv) { \ 50811507SVallish.Vaidyeshwara@Sun.COM cv_broadcast(&work_cv[NFS_ASYNC_QUEUE]); \ 50911507SVallish.Vaidyeshwara@Sun.COM cv_broadcast(&work_cv[NFS_ASYNC_PGOPS_QUEUE]); \ 51011507SVallish.Vaidyeshwara@Sun.COM } 51111507SVallish.Vaidyeshwara@Sun.COM 51211507SVallish.Vaidyeshwara@Sun.COM /* 5130Sstevel@tonic-gate * Mark cached attributes as timed out 5140Sstevel@tonic-gate * 5150Sstevel@tonic-gate * The caller must not be holding the rnode r_statelock mutex. 5160Sstevel@tonic-gate */ 5170Sstevel@tonic-gate #define PURGE_ATTRCACHE(vp) { \ 5180Sstevel@tonic-gate rnode_t *rp = VTOR(vp); \ 5190Sstevel@tonic-gate mutex_enter(&rp->r_statelock); \ 5200Sstevel@tonic-gate PURGE_ATTRCACHE_LOCKED(rp); \ 5210Sstevel@tonic-gate mutex_exit(&rp->r_statelock); \ 5220Sstevel@tonic-gate } 5230Sstevel@tonic-gate 5240Sstevel@tonic-gate #define PURGE_ATTRCACHE_LOCKED(rp) { \ 5250Sstevel@tonic-gate ASSERT(MUTEX_HELD(&rp->r_statelock)); \ 5260Sstevel@tonic-gate rp->r_attrtime = gethrtime(); \ 5270Sstevel@tonic-gate rp->r_mtime = rp->r_attrtime; \ 5280Sstevel@tonic-gate } 5290Sstevel@tonic-gate 5300Sstevel@tonic-gate /* 5310Sstevel@tonic-gate * Is the attribute cache valid? 5320Sstevel@tonic-gate */ 5330Sstevel@tonic-gate #define ATTRCACHE_VALID(vp) (gethrtime() < VTOR(vp)->r_attrtime) 5340Sstevel@tonic-gate 5350Sstevel@tonic-gate /* 5360Sstevel@tonic-gate * Flags to indicate whether to purge the DNLC for non-directory vnodes 5370Sstevel@tonic-gate * in a call to nfs_purge_caches. 5380Sstevel@tonic-gate */ 5390Sstevel@tonic-gate #define NFS_NOPURGE_DNLC 0 5400Sstevel@tonic-gate #define NFS_PURGE_DNLC 1 5410Sstevel@tonic-gate 5420Sstevel@tonic-gate /* 5430Sstevel@tonic-gate * If returned error is ESTALE flush all caches. 5440Sstevel@tonic-gate */ 5450Sstevel@tonic-gate #define PURGE_STALE_FH(error, vp, cr) \ 5460Sstevel@tonic-gate if ((error) == ESTALE) { \ 5470Sstevel@tonic-gate struct rnode *rp = VTOR(vp); \ 5480Sstevel@tonic-gate if (vp->v_flag & VROOT) { \ 5490Sstevel@tonic-gate servinfo_t *svp = rp->r_server; \ 5500Sstevel@tonic-gate mutex_enter(&svp->sv_lock); \ 5510Sstevel@tonic-gate svp->sv_flags |= SV_ROOT_STALE; \ 5520Sstevel@tonic-gate mutex_exit(&svp->sv_lock); \ 5530Sstevel@tonic-gate } \ 5540Sstevel@tonic-gate mutex_enter(&rp->r_statelock); \ 5550Sstevel@tonic-gate rp->r_flags |= RSTALE; \ 5560Sstevel@tonic-gate if (!rp->r_error) \ 5570Sstevel@tonic-gate rp->r_error = (error); \ 5580Sstevel@tonic-gate mutex_exit(&rp->r_statelock); \ 5590Sstevel@tonic-gate if (vn_has_cached_data(vp)) \ 5600Sstevel@tonic-gate nfs_invalidate_pages((vp), (u_offset_t)0, (cr)); \ 5610Sstevel@tonic-gate nfs_purge_caches((vp), NFS_PURGE_DNLC, (cr)); \ 5620Sstevel@tonic-gate } 5630Sstevel@tonic-gate 5640Sstevel@tonic-gate /* 5650Sstevel@tonic-gate * Is cache valid? 5660Sstevel@tonic-gate * Swap is always valid, if no attributes (attrtime == 0) or 5670Sstevel@tonic-gate * if mtime matches cached mtime it is valid 5680Sstevel@tonic-gate * NOTE: mtime is now a timestruc_t. 5690Sstevel@tonic-gate * Caller should be holding the rnode r_statelock mutex. 5700Sstevel@tonic-gate */ 5710Sstevel@tonic-gate #define CACHE_VALID(rp, mtime, fsize) \ 5720Sstevel@tonic-gate ((RTOV(rp)->v_flag & VISSWAP) == VISSWAP || \ 5730Sstevel@tonic-gate (((mtime).tv_sec == (rp)->r_attr.va_mtime.tv_sec && \ 5740Sstevel@tonic-gate (mtime).tv_nsec == (rp)->r_attr.va_mtime.tv_nsec) && \ 5750Sstevel@tonic-gate ((fsize) == (rp)->r_attr.va_size))) 5760Sstevel@tonic-gate 5770Sstevel@tonic-gate /* 5780Sstevel@tonic-gate * Macro to detect forced unmount or a zone shutdown. 5790Sstevel@tonic-gate */ 5800Sstevel@tonic-gate #define FS_OR_ZONE_GONE(vfsp) \ 5810Sstevel@tonic-gate (((vfsp)->vfs_flag & VFS_UNMOUNTED) || \ 5820Sstevel@tonic-gate zone_status_get(curproc->p_zone) >= ZONE_IS_SHUTTING_DOWN) 5830Sstevel@tonic-gate 5840Sstevel@tonic-gate /* 5850Sstevel@tonic-gate * Convert NFS tunables to hrtime_t units, seconds to nanoseconds. 5860Sstevel@tonic-gate */ 5870Sstevel@tonic-gate #define SEC2HR(sec) ((sec) * (long long)NANOSEC) 5880Sstevel@tonic-gate #define HR2SEC(hr) ((hr) / (long long)NANOSEC) 5890Sstevel@tonic-gate 5900Sstevel@tonic-gate /* 5910Sstevel@tonic-gate * Structure to identify owner of a PC file share reservation. 5920Sstevel@tonic-gate */ 5930Sstevel@tonic-gate struct nfs_owner { 5940Sstevel@tonic-gate int magic; /* magic uniquifying number */ 5950Sstevel@tonic-gate char hname[16]; /* first 16 bytes of hostname */ 5960Sstevel@tonic-gate char lowner[8]; /* local owner from fcntl */ 5970Sstevel@tonic-gate }; 5980Sstevel@tonic-gate 5990Sstevel@tonic-gate /* 6000Sstevel@tonic-gate * Values for magic. 6010Sstevel@tonic-gate */ 6020Sstevel@tonic-gate #define NFS_OWNER_MAGIC 0x1D81E 6030Sstevel@tonic-gate 6040Sstevel@tonic-gate /* 6050Sstevel@tonic-gate * Support for extended attributes 6060Sstevel@tonic-gate */ 6070Sstevel@tonic-gate #define XATTR_DIR_NAME "/@/" /* used for DNLC entries */ 6080Sstevel@tonic-gate #define XATTR_RPATH "ExTaTtR" /* used for r_path for failover */ 6090Sstevel@tonic-gate 6100Sstevel@tonic-gate /* 6110Sstevel@tonic-gate * Short hand for checking to see whether the file system was mounted 6120Sstevel@tonic-gate * interruptible or not. 6130Sstevel@tonic-gate */ 6140Sstevel@tonic-gate #define INTR(vp) (VTOMI(vp)->mi_flags & MI_INT) 6150Sstevel@tonic-gate 6160Sstevel@tonic-gate /* 6170Sstevel@tonic-gate * Short hand for checking whether failover is enabled or not 6180Sstevel@tonic-gate */ 6190Sstevel@tonic-gate #define FAILOVER_MOUNT(mi) (mi->mi_servers->sv_next) 6200Sstevel@tonic-gate 6210Sstevel@tonic-gate /* 6220Sstevel@tonic-gate * How long will async threads wait for additional work. 6230Sstevel@tonic-gate */ 6240Sstevel@tonic-gate #define NFS_ASYNC_TIMEOUT (60 * 1 * hz) /* 1 minute */ 6250Sstevel@tonic-gate 6260Sstevel@tonic-gate #ifdef _KERNEL 6270Sstevel@tonic-gate extern int clget(clinfo_t *, servinfo_t *, cred_t *, CLIENT **, 6280Sstevel@tonic-gate struct chtab **); 6290Sstevel@tonic-gate extern void clfree(CLIENT *, struct chtab *); 6300Sstevel@tonic-gate extern void nfs_mi_zonelist_add(mntinfo_t *); 6310Sstevel@tonic-gate extern void nfs_free_mi(mntinfo_t *); 6320Sstevel@tonic-gate extern void nfs_mnt_kstat_init(struct vfs *); 6330Sstevel@tonic-gate #endif 6340Sstevel@tonic-gate 6350Sstevel@tonic-gate /* 6360Sstevel@tonic-gate * Per-zone data for managing client handles. Included here solely for the 6370Sstevel@tonic-gate * benefit of MDB. 6380Sstevel@tonic-gate */ 6390Sstevel@tonic-gate /* 6400Sstevel@tonic-gate * client side statistics 6410Sstevel@tonic-gate */ 6420Sstevel@tonic-gate struct clstat { 6430Sstevel@tonic-gate kstat_named_t calls; /* client requests */ 6440Sstevel@tonic-gate kstat_named_t badcalls; /* rpc failures */ 6450Sstevel@tonic-gate kstat_named_t clgets; /* client handle gets */ 6460Sstevel@tonic-gate kstat_named_t cltoomany; /* client handle cache misses */ 6470Sstevel@tonic-gate #ifdef DEBUG 6480Sstevel@tonic-gate kstat_named_t clalloc; /* number of client handles */ 6490Sstevel@tonic-gate kstat_named_t noresponse; /* server not responding cnt */ 6500Sstevel@tonic-gate kstat_named_t failover; /* server failover count */ 6510Sstevel@tonic-gate kstat_named_t remap; /* server remap count */ 6520Sstevel@tonic-gate #endif 6530Sstevel@tonic-gate }; 6540Sstevel@tonic-gate 6550Sstevel@tonic-gate struct nfs_clnt { 6560Sstevel@tonic-gate struct chhead *nfscl_chtable; 6570Sstevel@tonic-gate kmutex_t nfscl_chtable_lock; 6580Sstevel@tonic-gate zoneid_t nfscl_zoneid; 6590Sstevel@tonic-gate list_node_t nfscl_node; 6600Sstevel@tonic-gate struct clstat nfscl_stat; 6610Sstevel@tonic-gate }; 6620Sstevel@tonic-gate 6630Sstevel@tonic-gate #ifdef __cplusplus 6640Sstevel@tonic-gate } 6650Sstevel@tonic-gate #endif 6660Sstevel@tonic-gate 6670Sstevel@tonic-gate #endif /* _NFS_NFS_CLNT_H */ 668