d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1484574843
freebsd-skq/sys/nlm/nlm_prot.h

449 lines
16 KiB
C
Raw Normal View History

Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
/*
* Please do not edit this file.
* It was generated using rpcgen.
*/
/* $FreeBSD$ */
#ifndef _NLM_PROT_H_RPCGEN
#define _NLM_PROT_H_RPCGEN
#include <rpc/rpc.h>
#ifdef __cplusplus
extern "C" {
#endif
#define LM_MAXSTRLEN 1024
#define MAXNAMELEN LM_MAXSTRLEN+1
enum nlm_stats {
nlm_granted = 0,
nlm_denied = 1,
nlm_denied_nolocks = 2,
nlm_blocked = 3,
nlm_denied_grace_period = 4,
nlm_deadlck = 5
};
typedef enum nlm_stats nlm_stats;
struct nlm_holder {
bool_t exclusive;
int svid;
netobj oh;
u_int l_offset;
u_int l_len;
};
typedef struct nlm_holder nlm_holder;
struct nlm_testrply {
nlm_stats stat;
union {
struct nlm_holder holder;
} nlm_testrply_u;
};
typedef struct nlm_testrply nlm_testrply;
struct nlm_stat {
nlm_stats stat;
};
typedef struct nlm_stat nlm_stat;
struct nlm_res {
netobj cookie;
nlm_stat stat;
};
typedef struct nlm_res nlm_res;
struct nlm_testres {
netobj cookie;
nlm_testrply stat;
};
typedef struct nlm_testres nlm_testres;
struct nlm_lock {
char *caller_name;
netobj fh;
netobj oh;
int svid;
u_int l_offset;
u_int l_len;
};
typedef struct nlm_lock nlm_lock;
struct nlm_lockargs {
netobj cookie;
bool_t block;
bool_t exclusive;
struct nlm_lock alock;
bool_t reclaim;
int state;
};
typedef struct nlm_lockargs nlm_lockargs;
struct nlm_cancargs {
netobj cookie;
bool_t block;
bool_t exclusive;
struct nlm_lock alock;
};
typedef struct nlm_cancargs nlm_cancargs;
struct nlm_testargs {
netobj cookie;
bool_t exclusive;
struct nlm_lock alock;
};
typedef struct nlm_testargs nlm_testargs;
struct nlm_unlockargs {
netobj cookie;
struct nlm_lock alock;
};
typedef struct nlm_unlockargs nlm_unlockargs;
/*
* The following enums are actually bit encoded for efficient
* boolean algebra.... DON'T change them.....
*/
enum fsh_mode {
fsm_DN = 0,
fsm_DR = 1,
fsm_DW = 2,
fsm_DRW = 3
};
typedef enum fsh_mode fsh_mode;
enum fsh_access {
fsa_NONE = 0,
fsa_R = 1,
fsa_W = 2,
fsa_RW = 3
};
typedef enum fsh_access fsh_access;
struct nlm_share {
char *caller_name;
netobj fh;
netobj oh;
fsh_mode mode;
fsh_access access;
};
typedef struct nlm_share nlm_share;
struct nlm_shareargs {
netobj cookie;
nlm_share share;
bool_t reclaim;
};
typedef struct nlm_shareargs nlm_shareargs;
struct nlm_shareres {
netobj cookie;
nlm_stats stat;
int sequence;
};
typedef struct nlm_shareres nlm_shareres;
struct nlm_notify {
char *name;
long state;
};
typedef struct nlm_notify nlm_notify;
/* definitions for NLM version 4 */
enum nlm4_stats {
nlm4_granted = 0,
nlm4_denied = 1,
nlm4_denied_nolocks = 2,
nlm4_blocked = 3,
nlm4_denied_grace_period = 4,
nlm4_deadlck = 5,
nlm4_rofs = 6,
nlm4_stale_fh = 7,
nlm4_fbig = 8,
nlm4_failed = 9
};
typedef enum nlm4_stats nlm4_stats;
struct nlm4_stat {
nlm4_stats stat;
};
typedef struct nlm4_stat nlm4_stat;
struct nlm4_holder {
bool_t exclusive;
u_int32_t svid;
netobj oh;
u_int64_t l_offset;
u_int64_t l_len;
};
typedef struct nlm4_holder nlm4_holder;
struct nlm4_lock {
char *caller_name;
netobj fh;
netobj oh;
u_int32_t svid;
u_int64_t l_offset;
u_int64_t l_len;
};
typedef struct nlm4_lock nlm4_lock;
struct nlm4_share {
char *caller_name;
netobj fh;
netobj oh;
fsh_mode mode;
fsh_access access;
};
typedef struct nlm4_share nlm4_share;
struct nlm4_testrply {
nlm4_stats stat;
union {
struct nlm4_holder holder;
} nlm4_testrply_u;
};
typedef struct nlm4_testrply nlm4_testrply;
struct nlm4_testres {
netobj cookie;
nlm4_testrply stat;
};
typedef struct nlm4_testres nlm4_testres;
struct nlm4_testargs {
netobj cookie;
bool_t exclusive;
struct nlm4_lock alock;
};
typedef struct nlm4_testargs nlm4_testargs;
struct nlm4_res {
netobj cookie;
nlm4_stat stat;
};
typedef struct nlm4_res nlm4_res;
struct nlm4_lockargs {
netobj cookie;
bool_t block;
bool_t exclusive;
struct nlm4_lock alock;
bool_t reclaim;
int state;
};
typedef struct nlm4_lockargs nlm4_lockargs;
struct nlm4_cancargs {
netobj cookie;
bool_t block;
bool_t exclusive;
struct nlm4_lock alock;
};
typedef struct nlm4_cancargs nlm4_cancargs;
struct nlm4_unlockargs {
netobj cookie;
struct nlm4_lock alock;
};
typedef struct nlm4_unlockargs nlm4_unlockargs;
struct nlm4_shareargs {
netobj cookie;
nlm4_share share;
bool_t reclaim;
};
typedef struct nlm4_shareargs nlm4_shareargs;
struct nlm4_shareres {
netobj cookie;
nlm4_stats stat;
int sequence;
};
typedef struct nlm4_shareres nlm4_shareres;
struct nlm_sm_status {
char *mon_name;
int state;
char priv[16];
};
typedef struct nlm_sm_status nlm_sm_status;
struct nlm4_notify {
char *name;
int32_t state;
};
typedef struct nlm4_notify nlm4_notify;
#define NLM_PROG ((unsigned long)(100021))
#define NLM_SM ((unsigned long)(0))
#define NLM_SM_NOTIFY ((unsigned long)(1))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_sm_notify_0(struct nlm_sm_status *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_sm_notify_0_svc(struct nlm_sm_status *, void *, struct svc_req *);
#define NLM_VERS ((unsigned long)(1))
#define NLM_TEST ((unsigned long)(1))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_test_1(struct nlm_testargs *, nlm_testres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_test_1_svc(struct nlm_testargs *, nlm_testres *, struct svc_req *);
#define NLM_LOCK ((unsigned long)(2))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_lock_1(struct nlm_lockargs *, nlm_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_lock_1_svc(struct nlm_lockargs *, nlm_res *, struct svc_req *);
#define NLM_CANCEL ((unsigned long)(3))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_cancel_1(struct nlm_cancargs *, nlm_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_cancel_1_svc(struct nlm_cancargs *, nlm_res *, struct svc_req *);
#define NLM_UNLOCK ((unsigned long)(4))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_unlock_1(struct nlm_unlockargs *, nlm_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_unlock_1_svc(struct nlm_unlockargs *, nlm_res *, struct svc_req *);
#define NLM_GRANTED ((unsigned long)(5))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_granted_1(struct nlm_testargs *, nlm_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_granted_1_svc(struct nlm_testargs *, nlm_res *, struct svc_req *);
#define NLM_TEST_MSG ((unsigned long)(6))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_test_msg_1(struct nlm_testargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_test_msg_1_svc(struct nlm_testargs *, void *, struct svc_req *);
#define NLM_LOCK_MSG ((unsigned long)(7))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_lock_msg_1(struct nlm_lockargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_lock_msg_1_svc(struct nlm_lockargs *, void *, struct svc_req *);
#define NLM_CANCEL_MSG ((unsigned long)(8))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_cancel_msg_1(struct nlm_cancargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_cancel_msg_1_svc(struct nlm_cancargs *, void *, struct svc_req *);
#define NLM_UNLOCK_MSG ((unsigned long)(9))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_unlock_msg_1(struct nlm_unlockargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_unlock_msg_1_svc(struct nlm_unlockargs *, void *, struct svc_req *);
#define NLM_GRANTED_MSG ((unsigned long)(10))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_granted_msg_1(struct nlm_testargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_granted_msg_1_svc(struct nlm_testargs *, void *, struct svc_req *);
#define NLM_TEST_RES ((unsigned long)(11))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_test_res_1(nlm_testres *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_test_res_1_svc(nlm_testres *, void *, struct svc_req *);
#define NLM_LOCK_RES ((unsigned long)(12))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_lock_res_1(nlm_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_lock_res_1_svc(nlm_res *, void *, struct svc_req *);
#define NLM_CANCEL_RES ((unsigned long)(13))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_cancel_res_1(nlm_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_cancel_res_1_svc(nlm_res *, void *, struct svc_req *);
#define NLM_UNLOCK_RES ((unsigned long)(14))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_unlock_res_1(nlm_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_unlock_res_1_svc(nlm_res *, void *, struct svc_req *);
#define NLM_GRANTED_RES ((unsigned long)(15))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_granted_res_1(nlm_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_granted_res_1_svc(nlm_res *, void *, struct svc_req *);
extern int nlm_prog_1_freeresult(SVCXPRT *, xdrproc_t, caddr_t);
#define NLM_VERSX ((unsigned long)(3))
#define NLM_SHARE ((unsigned long)(20))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_share_3(nlm_shareargs *, nlm_shareres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_share_3_svc(nlm_shareargs *, nlm_shareres *, struct svc_req *);
#define NLM_UNSHARE ((unsigned long)(21))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_unshare_3(nlm_shareargs *, nlm_shareres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_unshare_3_svc(nlm_shareargs *, nlm_shareres *, struct svc_req *);
#define NLM_NM_LOCK ((unsigned long)(22))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_nm_lock_3(nlm_lockargs *, nlm_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_nm_lock_3_svc(nlm_lockargs *, nlm_res *, struct svc_req *);
#define NLM_FREE_ALL ((unsigned long)(23))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm_free_all_3(nlm_notify *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm_free_all_3_svc(nlm_notify *, void *, struct svc_req *);
extern int nlm_prog_3_freeresult(SVCXPRT *, xdrproc_t, caddr_t);
#define NLM_VERS4 ((unsigned long)(4))
#define NLM4_TEST ((unsigned long)(1))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_test_4(nlm4_testargs *, nlm4_testres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_test_4_svc(nlm4_testargs *, nlm4_testres *, struct svc_req *);
#define NLM4_LOCK ((unsigned long)(2))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_lock_4(nlm4_lockargs *, nlm4_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_lock_4_svc(nlm4_lockargs *, nlm4_res *, struct svc_req *);
#define NLM4_CANCEL ((unsigned long)(3))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_cancel_4(nlm4_cancargs *, nlm4_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_cancel_4_svc(nlm4_cancargs *, nlm4_res *, struct svc_req *);
#define NLM4_UNLOCK ((unsigned long)(4))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_unlock_4(nlm4_unlockargs *, nlm4_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_unlock_4_svc(nlm4_unlockargs *, nlm4_res *, struct svc_req *);
#define NLM4_GRANTED ((unsigned long)(5))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_granted_4(nlm4_testargs *, nlm4_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_granted_4_svc(nlm4_testargs *, nlm4_res *, struct svc_req *);
#define NLM4_TEST_MSG ((unsigned long)(6))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_test_msg_4(nlm4_testargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_test_msg_4_svc(nlm4_testargs *, void *, struct svc_req *);
#define NLM4_LOCK_MSG ((unsigned long)(7))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_lock_msg_4(nlm4_lockargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_lock_msg_4_svc(nlm4_lockargs *, void *, struct svc_req *);
#define NLM4_CANCEL_MSG ((unsigned long)(8))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_cancel_msg_4(nlm4_cancargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_cancel_msg_4_svc(nlm4_cancargs *, void *, struct svc_req *);
#define NLM4_UNLOCK_MSG ((unsigned long)(9))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_unlock_msg_4(nlm4_unlockargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_unlock_msg_4_svc(nlm4_unlockargs *, void *, struct svc_req *);
#define NLM4_GRANTED_MSG ((unsigned long)(10))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_granted_msg_4(nlm4_testargs *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_granted_msg_4_svc(nlm4_testargs *, void *, struct svc_req *);
#define NLM4_TEST_RES ((unsigned long)(11))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_test_res_4(nlm4_testres *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_test_res_4_svc(nlm4_testres *, void *, struct svc_req *);
#define NLM4_LOCK_RES ((unsigned long)(12))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_lock_res_4(nlm4_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_lock_res_4_svc(nlm4_res *, void *, struct svc_req *);
#define NLM4_CANCEL_RES ((unsigned long)(13))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_cancel_res_4(nlm4_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_cancel_res_4_svc(nlm4_res *, void *, struct svc_req *);
#define NLM4_UNLOCK_RES ((unsigned long)(14))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_unlock_res_4(nlm4_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_unlock_res_4_svc(nlm4_res *, void *, struct svc_req *);
#define NLM4_GRANTED_RES ((unsigned long)(15))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_granted_res_4(nlm4_res *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_granted_res_4_svc(nlm4_res *, void *, struct svc_req *);
#define NLM4_SHARE ((unsigned long)(20))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_share_4(nlm4_shareargs *, nlm4_shareres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_share_4_svc(nlm4_shareargs *, nlm4_shareres *, struct svc_req *);
#define NLM4_UNSHARE ((unsigned long)(21))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_unshare_4(nlm4_shareargs *, nlm4_shareres *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_unshare_4_svc(nlm4_shareargs *, nlm4_shareres *, struct svc_req *);
#define NLM4_NM_LOCK ((unsigned long)(22))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_nm_lock_4(nlm4_lockargs *, nlm4_res *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_nm_lock_4_svc(nlm4_lockargs *, nlm4_res *, struct svc_req *);
#define NLM4_FREE_ALL ((unsigned long)(23))
Re-implement the client side of rpc.lockd in the kernel. This implementation provides the correct semantics for flock(2) style locks which are used by the lockf(1) command line tool and the pidfile(3) library. It also implements recovery from server restarts and ensures that dirty cache blocks are written to the server before obtaining locks (allowing multiple clients to use file locking to safely share data). Sponsored by: Isilon Systems PR: 94256 MFC after: 2 weeks
2008-06-26 10:21:54 +00:00
extern enum clnt_stat nlm4_free_all_4(nlm4_notify *, void *, CLIENT *, struct rpc_callextra *, struct timeval);
Add the new kernel-mode NFS Lock Manager. To use it instead of the user-mode lock manager, build a kernel with the NFSLOCKD option and add '-k' to 'rpc_lockd_flags' in rc.conf. Highlights include: * Thread-safe kernel RPC client - many threads can use the same RPC client handle safely with replies being de-multiplexed at the socket upcall (typically driven directly by the NIC interrupt) and handed off to whichever thread matches the reply. For UDP sockets, many RPC clients can share the same socket. This allows the use of a single privileged UDP port number to talk to an arbitrary number of remote hosts. * Single-threaded kernel RPC server. Adding support for multi-threaded server would be relatively straightforward and would follow approximately the Solaris KPI. A single thread should be sufficient for the NLM since it should rarely block in normal operation. * Kernel mode NLM server supporting cancel requests and granted callbacks. I've tested the NLM server reasonably extensively - it passes both my own tests and the NFS Connectathon locking tests running on Solaris, Mac OS X and Ubuntu Linux. * Userland NLM client supported. While the NLM server doesn't have support for the local NFS client's locking needs, it does have to field async replies and granted callbacks from remote NLMs that the local client has contacted. We relay these replies to the userland rpc.lockd over a local domain RPC socket. * Robust deadlock detection for the local lock manager. In particular it will detect deadlocks caused by a lock request that covers more than one blocking request. As required by the NLM protocol, all deadlock detection happens synchronously - a user is guaranteed that if a lock request isn't rejected immediately, the lock will eventually be granted. The old system allowed for a 'deferred deadlock' condition where a blocked lock request could wake up and find that some other deadlock-causing lock owner had beaten them to the lock. * Since both local and remote locks are managed by the same kernel locking code, local and remote processes can safely use file locks for mutual exclusion. Local processes have no fairness advantage compared to remote processes when contending to lock a region that has just been unlocked - the local lock manager enforces a strict first-come first-served model for both local and remote lockers. Sponsored by: Isilon Systems PR: 95247 107555 115524 116679 MFC after: 2 weeks
2008-03-26 15:23:12 +00:00
extern bool_t nlm4_free_all_4_svc(nlm4_notify *, void *, struct svc_req *);
extern int nlm_prog_4_freeresult(SVCXPRT *, xdrproc_t, caddr_t);
/* the xdr functions */
extern bool_t xdr_nlm_stats(XDR *, nlm_stats*);
extern bool_t xdr_nlm_holder(XDR *, nlm_holder*);
extern bool_t xdr_nlm_testrply(XDR *, nlm_testrply*);
extern bool_t xdr_nlm_stat(XDR *, nlm_stat*);
extern bool_t xdr_nlm_res(XDR *, nlm_res*);
extern bool_t xdr_nlm_testres(XDR *, nlm_testres*);
extern bool_t xdr_nlm_lock(XDR *, nlm_lock*);
extern bool_t xdr_nlm_lockargs(XDR *, nlm_lockargs*);
extern bool_t xdr_nlm_cancargs(XDR *, nlm_cancargs*);
extern bool_t xdr_nlm_testargs(XDR *, nlm_testargs*);
extern bool_t xdr_nlm_unlockargs(XDR *, nlm_unlockargs*);
extern bool_t xdr_fsh_mode(XDR *, fsh_mode*);
extern bool_t xdr_fsh_access(XDR *, fsh_access*);
extern bool_t xdr_nlm_share(XDR *, nlm_share*);
extern bool_t xdr_nlm_shareargs(XDR *, nlm_shareargs*);
extern bool_t xdr_nlm_shareres(XDR *, nlm_shareres*);
extern bool_t xdr_nlm_notify(XDR *, nlm_notify*);
extern bool_t xdr_nlm4_stats(XDR *, nlm4_stats*);
extern bool_t xdr_nlm4_stat(XDR *, nlm4_stat*);
extern bool_t xdr_nlm4_holder(XDR *, nlm4_holder*);
extern bool_t xdr_nlm4_lock(XDR *, nlm4_lock*);
extern bool_t xdr_nlm4_share(XDR *, nlm4_share*);
extern bool_t xdr_nlm4_testrply(XDR *, nlm4_testrply*);
extern bool_t xdr_nlm4_testres(XDR *, nlm4_testres*);
extern bool_t xdr_nlm4_testargs(XDR *, nlm4_testargs*);
extern bool_t xdr_nlm4_res(XDR *, nlm4_res*);
extern bool_t xdr_nlm4_lockargs(XDR *, nlm4_lockargs*);
extern bool_t xdr_nlm4_cancargs(XDR *, nlm4_cancargs*);
extern bool_t xdr_nlm4_unlockargs(XDR *, nlm4_unlockargs*);
extern bool_t xdr_nlm4_shareargs(XDR *, nlm4_shareargs*);
extern bool_t xdr_nlm4_shareres(XDR *, nlm4_shareres*);
extern bool_t xdr_nlm_sm_status(XDR *, nlm_sm_status*);
extern bool_t xdr_nlm4_notify(XDR *, nlm4_notify*);
#ifdef __cplusplus
}
#endif
#endif /* !_NLM_PROT_H_RPCGEN */
Reference in New Issue Copy Permalink