Commit 597d0cae authored by David Teigland's avatar David Teigland Committed by Steven Whitehouse

[DLM] dlm: user locks

This changes the way the dlm handles user locks.  The core dlm is now
aware of user locks so they can be dealt with more efficiently.  There is
no more dlm_device module which previously managed its own duplicate copy
of every user lock.
Signed-off-by: default avatarPatrick Caulfield <pcaulfie@redhat.com>
Signed-off-by: default avatarDavid Teigland <teigland@redhat.com>
Signed-off-by: default avatarSteven Whitehouse <swhiteho@redhat.com>
parent 2eb168ca
......@@ -10,14 +10,6 @@ config DLM
A general purpose distributed lock manager for kernel or userspace
applications.
config DLM_DEVICE
tristate "DLM device for userspace access"
depends on DLM
help
This module creates a misc device through which the dlm lockspace
and locking functions become available to userspace applications
(usually through the libdlm library).
config DLM_DEBUG
bool "DLM debugging"
depends on DLM
......
obj-$(CONFIG_DLM) += dlm.o
obj-$(CONFIG_DLM_DEVICE) += dlm_device.o
dlm-y := ast.o \
config.o \
dir.o \
......@@ -15,7 +13,7 @@ dlm-y := ast.o \
recover.o \
recoverd.o \
requestqueue.o \
user.o \
util.o
dlm-$(CONFIG_DLM_DEBUG) += debug_fs.o
dlm_device-y := device.o
......@@ -13,7 +13,7 @@
#include "dlm_internal.h"
#include "lock.h"
#include "ast.h"
#include "user.h"
#define WAKE_ASTS 0
......@@ -34,6 +34,11 @@ void dlm_del_ast(struct dlm_lkb *lkb)
void dlm_add_ast(struct dlm_lkb *lkb, int type)
{
if (lkb->lkb_flags & DLM_IFL_USER) {
dlm_user_add_ast(lkb, type);
return;
}
spin_lock(&ast_queue_lock);
if (!(lkb->lkb_ast_type & (AST_COMP | AST_BAST))) {
kref_get(&lkb->lkb_ref);
......
This diff is collapsed.
......@@ -35,6 +35,7 @@
#include <linux/kref.h>
#include <linux/kernel.h>
#include <linux/jhash.h>
#include <linux/miscdevice.h>
#include <linux/mutex.h>
#include <asm/semaphore.h>
#include <asm/uaccess.h>
......@@ -68,6 +69,7 @@ struct dlm_mhandle;
#define log_error(ls, fmt, args...) \
printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args)
#define DLM_LOG_DEBUG
#ifdef DLM_LOG_DEBUG
#define log_debug(ls, fmt, args...) log_error(ls, fmt, ##args)
#else
......@@ -204,6 +206,9 @@ struct dlm_args {
#define DLM_IFL_MSTCPY 0x00010000
#define DLM_IFL_RESEND 0x00020000
#define DLM_IFL_DEAD 0x00040000
#define DLM_IFL_USER 0x00000001
#define DLM_IFL_ORPHAN 0x00000002
struct dlm_lkb {
struct dlm_rsb *lkb_resource; /* the rsb */
......@@ -231,6 +236,7 @@ struct dlm_lkb {
struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */
struct list_head lkb_wait_reply; /* waiting for remote reply */
struct list_head lkb_astqueue; /* need ast to be sent */
struct list_head lkb_ownqueue; /* list of locks for a process */
char *lkb_lvbptr;
struct dlm_lksb *lkb_lksb; /* caller's status block */
......@@ -409,6 +415,7 @@ struct rcom_lock {
struct dlm_ls {
struct list_head ls_list; /* list of lockspaces */
dlm_lockspace_t *ls_local_handle;
uint32_t ls_global_id; /* global unique lockspace ID */
uint32_t ls_exflags;
int ls_lvblen;
......@@ -444,6 +451,8 @@ struct dlm_ls {
wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
int ls_uevent_result;
struct miscdevice ls_device;
/* recovery related */
struct timer_list ls_timer;
......@@ -461,6 +470,7 @@ struct dlm_ls {
spinlock_t ls_recover_list_lock;
int ls_recover_list_count;
wait_queue_head_t ls_wait_general;
struct mutex ls_clear_proc_locks;
struct list_head ls_root_list; /* root resources */
struct rw_semaphore ls_root_sem; /* protect root_list */
......@@ -475,6 +485,40 @@ struct dlm_ls {
#define LSFL_RCOM_READY 3
#define LSFL_UEVENT_WAIT 4
/* much of this is just saving user space pointers associated with the
lock that we pass back to the user lib with an ast */
struct dlm_user_args {
struct dlm_user_proc *proc; /* each process that opens the lockspace
device has private data
(dlm_user_proc) on the struct file,
the process's locks point back to it*/
struct dlm_lksb lksb;
int old_mode;
int update_user_lvb;
struct dlm_lksb __user *user_lksb;
void __user *castparam;
void __user *castaddr;
void __user *bastparam;
void __user *bastaddr;
};
#define DLM_PROC_FLAGS_CLOSING 1
#define DLM_PROC_FLAGS_COMPAT 2
/* locks list is kept so we can remove all a process's locks when it
exits (or orphan those that are persistent) */
struct dlm_user_proc {
dlm_lockspace_t *lockspace;
unsigned long flags; /* DLM_PROC_FLAGS */
struct list_head asts;
spinlock_t asts_spin;
struct list_head locks;
spinlock_t locks_spin;
wait_queue_head_t wait;
};
static inline int dlm_locking_stopped(struct dlm_ls *ls)
{
return !test_bit(LSFL_RUNNING, &ls->ls_flags);
......
......@@ -55,8 +55,9 @@
R: do_xxxx()
L: receive_xxxx_reply() <- R: send_xxxx_reply()
*/
#include <linux/types.h>
#include "dlm_internal.h"
#include <linux/dlm_device.h>
#include "memory.h"
#include "lowcomms.h"
#include "requestqueue.h"
......@@ -69,6 +70,7 @@
#include "rcom.h"
#include "recover.h"
#include "lvb_table.h"
#include "user.h"
#include "config.h"
static int send_request(struct dlm_rsb *r, struct dlm_lkb *lkb);
......@@ -84,6 +86,8 @@ static void __receive_convert_reply(struct dlm_rsb *r, struct dlm_lkb *lkb,
struct dlm_message *ms);
static int receive_extralen(struct dlm_message *ms);
#define FAKE_USER_AST (void*)0xff00ff00
/*
* Lock compatibilty matrix - thanks Steve
* UN = Unlocked state. Not really a state, used as a flag
......@@ -152,7 +156,7 @@ static const int __quecvt_compat_matrix[8][8] = {
{0, 0, 0, 0, 0, 0, 0, 0} /* PD */
};
static void dlm_print_lkb(struct dlm_lkb *lkb)
void dlm_print_lkb(struct dlm_lkb *lkb)
{
printk(KERN_ERR "lkb: nodeid %d id %x remid %x exflags %x flags %x\n"
" status %d rqmode %d grmode %d wait_type %d ast_type %d\n",
......@@ -291,7 +295,7 @@ static int search_rsb_list(struct list_head *head, char *name, int len,
if (len == r->res_length && !memcmp(name, r->res_name, len))
goto found;
}
return -ENOENT;
return -EBADR;
found:
if (r->res_nodeid && (flags & R_MASTER))
......@@ -376,7 +380,7 @@ static int find_rsb(struct dlm_ls *ls, char *name, int namelen,
if (!error)
goto out;
if (error == -ENOENT && !(flags & R_CREATE))
if (error == -EBADR && !(flags & R_CREATE))
goto out;
/* the rsb was found but wasn't a master copy */
......@@ -920,7 +924,7 @@ static void set_lvb_lock_pc(struct dlm_rsb *r, struct dlm_lkb *lkb,
if (!(lkb->lkb_exflags & DLM_LKF_VALBLK))
return;
b = dlm_lvb_operations[lkb->lkb_grmode + 1][lkb->lkb_rqmode + 1];
b = dlm_lvb_operations[lkb->lkb_grmode + 1][lkb->lkb_rqmode + 1];
if (b == 1) {
int len = receive_extralen(ms);
memcpy(lkb->lkb_lvbptr, ms->m_extra, len);
......@@ -963,6 +967,8 @@ static void revert_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
lkb->lkb_rqmode = DLM_LOCK_IV;
switch (lkb->lkb_status) {
case DLM_LKSTS_GRANTED:
break;
case DLM_LKSTS_CONVERT:
move_lkb(r, lkb, DLM_LKSTS_GRANTED);
break;
......@@ -1727,6 +1733,11 @@ static int do_unlock(struct dlm_rsb *r, struct dlm_lkb *lkb)
return -DLM_EUNLOCK;
}
/* FIXME: if revert_lock() finds that the lkb is granted, we should
skip the queue_cast(ECANCEL). It indicates that the request/convert
completed (and queued a normal ast) just before the cancel; we don't
want to clobber the sb_result for the normal ast with ECANCEL. */
static int do_cancel(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
revert_lock(r, lkb);
......@@ -2739,7 +2750,7 @@ static void receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
confirm_master(r, error);
break;
case -ENOENT:
case -EBADR:
case -ENOTBLK:
/* find_rsb failed to find rsb or rsb wasn't master */
r->res_nodeid = -1;
......@@ -3545,3 +3556,284 @@ int dlm_recover_process_copy(struct dlm_ls *ls, struct dlm_rcom *rc)
return 0;
}
int dlm_user_request(struct dlm_ls *ls, struct dlm_user_args *ua,
int mode, uint32_t flags, void *name, unsigned int namelen,
uint32_t parent_lkid)
{
struct dlm_lkb *lkb;
struct dlm_args args;
int error;
lock_recovery(ls);
error = create_lkb(ls, &lkb);
if (error) {
kfree(ua);
goto out;
}
if (flags & DLM_LKF_VALBLK) {
ua->lksb.sb_lvbptr = kmalloc(DLM_USER_LVB_LEN, GFP_KERNEL);
if (!ua->lksb.sb_lvbptr) {
kfree(ua);
__put_lkb(ls, lkb);
error = -ENOMEM;
goto out;
}
}
/* After ua is attached to lkb it will be freed by free_lkb().
When DLM_IFL_USER is set, the dlm knows that this is a userspace
lock and that lkb_astparam is the dlm_user_args structure. */
error = set_lock_args(mode, &ua->lksb, flags, namelen, parent_lkid,
FAKE_USER_AST, ua, FAKE_USER_AST, &args);
lkb->lkb_flags |= DLM_IFL_USER;
ua->old_mode = DLM_LOCK_IV;
if (error) {
__put_lkb(ls, lkb);
goto out;
}
error = request_lock(ls, lkb, name, namelen, &args);
switch (error) {
case 0:
break;
case -EINPROGRESS:
error = 0;
break;
case -EAGAIN:
error = 0;
/* fall through */
default:
__put_lkb(ls, lkb);
goto out;
}
/* add this new lkb to the per-process list of locks */
spin_lock(&ua->proc->locks_spin);
kref_get(&lkb->lkb_ref);
list_add_tail(&lkb->lkb_ownqueue, &ua->proc->locks);
spin_unlock(&ua->proc->locks_spin);
out:
unlock_recovery(ls);
return error;
}
int dlm_user_convert(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
int mode, uint32_t flags, uint32_t lkid, char *lvb_in)
{
struct dlm_lkb *lkb;
struct dlm_args args;
struct dlm_user_args *ua;
int error;
lock_recovery(ls);
error = find_lkb(ls, lkid, &lkb);
if (error)
goto out;
/* user can change the params on its lock when it converts it, or
add an lvb that didn't exist before */
ua = (struct dlm_user_args *)lkb->lkb_astparam;
if (flags & DLM_LKF_VALBLK && !ua->lksb.sb_lvbptr) {
ua->lksb.sb_lvbptr = kmalloc(DLM_USER_LVB_LEN, GFP_KERNEL);
if (!ua->lksb.sb_lvbptr) {
error = -ENOMEM;
goto out_put;
}
}
if (lvb_in && ua->lksb.sb_lvbptr)
memcpy(ua->lksb.sb_lvbptr, lvb_in, DLM_USER_LVB_LEN);
ua->castparam = ua_tmp->castparam;
ua->castaddr = ua_tmp->castaddr;
ua->bastparam = ua_tmp->bastparam;
ua->bastaddr = ua_tmp->bastaddr;
ua->old_mode = lkb->lkb_grmode;
error = set_lock_args(mode, &ua->lksb, flags, 0, 0, FAKE_USER_AST, ua,
FAKE_USER_AST, &args);
if (error)
goto out_put;
error = convert_lock(ls, lkb, &args);
if (error == -EINPROGRESS || error == -EAGAIN)
error = 0;
out_put:
dlm_put_lkb(lkb);
out:
unlock_recovery(ls);
kfree(ua_tmp);
return error;
}
int dlm_user_unlock(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
uint32_t flags, uint32_t lkid, char *lvb_in)
{
struct dlm_lkb *lkb;
struct dlm_args args;
struct dlm_user_args *ua;
int error;
lock_recovery(ls);
error = find_lkb(ls, lkid, &lkb);
if (error)
goto out;
ua = (struct dlm_user_args *)lkb->lkb_astparam;
if (lvb_in && ua->lksb.sb_lvbptr)
memcpy(ua->lksb.sb_lvbptr, lvb_in, DLM_USER_LVB_LEN);
ua->castparam = ua_tmp->castparam;
error = set_unlock_args(flags, ua, &args);
if (error)
goto out_put;
error = unlock_lock(ls, lkb, &args);
if (error == -DLM_EUNLOCK)
error = 0;
if (error)
goto out_put;
spin_lock(&ua->proc->locks_spin);
list_del(&lkb->lkb_ownqueue);
spin_unlock(&ua->proc->locks_spin);
/* this removes the reference for the proc->locks list added by
dlm_user_request */
unhold_lkb(lkb);
out_put:
dlm_put_lkb(lkb);
out:
unlock_recovery(ls);
return error;
}
int dlm_user_cancel(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
uint32_t flags, uint32_t lkid)
{
struct dlm_lkb *lkb;
struct dlm_args args;
struct dlm_user_args *ua;
int error;
lock_recovery(ls);
error = find_lkb(ls, lkid, &lkb);
if (error)
goto out;
ua = (struct dlm_user_args *)lkb->lkb_astparam;
ua->castparam = ua_tmp->castparam;
error = set_unlock_args(flags, ua, &args);
if (error)
goto out_put;
error = cancel_lock(ls, lkb, &args);
if (error == -DLM_ECANCEL)
error = 0;
if (error)
goto out_put;
/* this lkb was removed from the WAITING queue */
if (lkb->lkb_grmode == DLM_LOCK_IV) {
spin_lock(&ua->proc->locks_spin);
list_del(&lkb->lkb_ownqueue);
spin_unlock(&ua->proc->locks_spin);
unhold_lkb(lkb);
}
out_put:
dlm_put_lkb(lkb);
out:
unlock_recovery(ls);
return error;
}
static int orphan_proc_lock(struct dlm_ls *ls, struct dlm_lkb *lkb)
{
struct dlm_user_args *ua = (struct dlm_user_args *)lkb->lkb_astparam;
if (ua->lksb.sb_lvbptr)
kfree(ua->lksb.sb_lvbptr);
kfree(ua);
lkb->lkb_astparam = (long)NULL;
/* TODO: propogate to master if needed */
return 0;
}
/* The force flag allows the unlock to go ahead even if the lkb isn't granted.
Regardless of what rsb queue the lock is on, it's removed and freed. */
static int unlock_proc_lock(struct dlm_ls *ls, struct dlm_lkb *lkb)
{
struct dlm_user_args *ua = (struct dlm_user_args *)lkb->lkb_astparam;
struct dlm_args args;
int error;
/* FIXME: we need to handle the case where the lkb is in limbo
while the rsb is being looked up, currently we assert in
_unlock_lock/is_remote because rsb nodeid is -1. */
set_unlock_args(DLM_LKF_FORCEUNLOCK, ua, &args);
error = unlock_lock(ls, lkb, &args);
if (error == -DLM_EUNLOCK)
error = 0;
return error;
}
/* The ls_clear_proc_locks mutex protects against dlm_user_add_asts() which
1) references lkb->ua which we free here and 2) adds lkbs to proc->asts,
which we clear here. */
/* proc CLOSING flag is set so no more device_reads should look at proc->asts
list, and no more device_writes should add lkb's to proc->locks list; so we
shouldn't need to take asts_spin or locks_spin here. this assumes that
device reads/writes/closes are serialized -- FIXME: we may need to serialize
them ourself. */
void dlm_clear_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc)
{
struct dlm_lkb *lkb, *safe;
lock_recovery(ls);
mutex_lock(&ls->ls_clear_proc_locks);
list_for_each_entry_safe(lkb, safe, &proc->locks, lkb_ownqueue) {
if (lkb->lkb_ast_type) {
list_del(&lkb->lkb_astqueue);
unhold_lkb(lkb);
}
list_del(&lkb->lkb_ownqueue);
if (lkb->lkb_exflags & DLM_LKF_PERSISTENT) {
lkb->lkb_flags |= DLM_IFL_ORPHAN;
orphan_proc_lock(ls, lkb);
} else {
lkb->lkb_flags |= DLM_IFL_DEAD;
unlock_proc_lock(ls, lkb);
}
/* this removes the reference for the proc->locks list
added by dlm_user_request, it may result in the lkb
being freed */
dlm_put_lkb(lkb);
}
mutex_unlock(&ls->ls_clear_proc_locks);
unlock_recovery(ls);
}
......@@ -14,6 +14,7 @@
#define __LOCK_DOT_H__
void dlm_print_rsb(struct dlm_rsb *r);
void dlm_print_lkb(struct dlm_lkb *lkb);
int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery);
int dlm_modes_compat(int mode1, int mode2);
int dlm_find_rsb(struct dlm_ls *ls, char *name, int namelen,
......@@ -31,6 +32,16 @@ void dlm_recover_waiters_pre(struct dlm_ls *ls);
int dlm_recover_master_copy(struct dlm_ls *ls, struct dlm_rcom *rc);
int dlm_recover_process_copy(struct dlm_ls *ls, struct dlm_rcom *rc);
int dlm_user_request(struct dlm_ls *ls, struct dlm_user_args *ua, int mode,
uint32_t flags, void *name, unsigned int namelen, uint32_t parent_lkid);
int dlm_user_convert(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
int mode, uint32_t flags, uint32_t lkid, char *lvb_in);
int dlm_user_unlock(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
uint32_t flags, uint32_t lkid, char *lvb_in);
int dlm_user_cancel(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
uint32_t flags, uint32_t lkid);
void dlm_clear_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc);
static inline int is_master(struct dlm_rsb *r)
{
return !r->res_nodeid;
......
......@@ -270,12 +270,36 @@ struct dlm_ls *dlm_find_lockspace_global(uint32_t id)
return ls;
}
struct dlm_ls *dlm_find_lockspace_local(void *id)
struct dlm_ls *dlm_find_lockspace_local(dlm_lockspace_t *lockspace)
{
struct dlm_ls *ls = id;
struct dlm_ls *ls;
spin_lock(&lslist_lock);
ls->ls_count++;
list_for_each_entry(ls, &lslist, ls_list) {
if (ls->ls_local_handle == lockspace) {
ls->ls_count++;
goto out;
}
}
ls = NULL;
out:
spin_unlock(&lslist_lock);
return ls;
}
struct dlm_ls *dlm_find_lockspace_device(int minor)
{
struct dlm_ls *ls;
spin_lock(&lslist_lock);
list_for_each_entry(ls, &lslist, ls_list) {
if (ls->ls_device.minor == minor) {
ls->ls_count++;
goto out;
}
}
ls = NULL;
out:
spin_unlock(&lslist_lock);
return ls;
}
......@@ -436,6 +460,7 @@ static int new_lockspace(char *name, int namelen, void **lockspace,
init_rwsem(&ls->ls_in_recovery);
INIT_LIST_HEAD(&ls->ls_requestqueue);
mutex_init(&ls->ls_requestqueue_mutex);
mutex_init(&ls->ls_clear_proc_locks);
ls->ls_recover_buf = kmalloc(dlm_config.buffer_size, GFP_KERNEL);
if (!ls->ls_recover_buf)
......@@ -444,6 +469,7 @@ static int new_lockspace(char *name, int namelen, void **lockspace,
INIT_LIST_HEAD(&ls->ls_recover_list);
spin_lock_init(&ls->ls_recover_list_lock);
ls->ls_recover_list_count = 0;
ls->ls_local_handle = ls;
init_waitqueue_head(&ls->ls_wait_general);
INIT_LIST_HEAD(&ls->ls_root_list);
init_rwsem(&ls->ls_root_sem);
......
......@@ -18,6 +18,7 @@ int dlm_lockspace_init(void);
void dlm_lockspace_exit(void);
struct dlm_ls *dlm_find_lockspace_global(uint32_t id);
struct dlm_ls *dlm_find_lockspace_local(void *id);
struct dlm_ls *dlm_find_lockspace_device(int minor);
void dlm_put_lockspace(struct dlm_ls *ls);
#endif /* __LOCKSPACE_DOT_H__ */
......
......@@ -14,6 +14,7 @@
#include "dlm_internal.h"
#include "lockspace.h"
#include "lock.h"
#include "user.h"
#include "memory.h"
#include "lowcomms.h"
#include "config.h"
......@@ -50,10 +51,16 @@ static int __init init_dlm(void)
if (error)
goto out_debug;
error = dlm_user_init();
if (error)
goto out_lowcomms;
printk("DLM (built %s %s) installed\n", __DATE__, __TIME__);
return 0;
out_lowcomms:
dlm_lowcomms_exit();
out_debug:
dlm_unregister_debugfs();
out_config:
......@@ -68,6 +75,7 @@ static int __init init_dlm(void)
static void __exit exit_dlm(void)
{
dlm_user_exit();
dlm_lowcomms_exit();
dlm_config_exit();
dlm_memory_exit();
......
......@@ -84,6 +84,15 @@ struct dlm_lkb *allocate_lkb(struct dlm_ls *ls)
void free_lkb(struct dlm_lkb *lkb)
{
if (lkb->lkb_flags & DLM_IFL_USER) {
struct dlm_user_args *ua;
ua = (struct dlm_user_args *)lkb->lkb_astparam;
if (ua) {
if (ua->lksb.sb_lvbptr)
kfree(ua->lksb.sb_lvbptr);
kfree(ua);
}
}
kmem_cache_free(lkb_cache, lkb);
}
......
This diff is collapsed.
/*
* Copyright (C) 2006 Red Hat, Inc. All rights reserved.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU General Public License v.2.
*/
#ifndef __USER_DOT_H__
#define __USER_DOT_H__
void dlm_user_add_ast(struct dlm_lkb *lkb, int type);
int dlm_user_init(void);
void dlm_user_exit(void);
#endif
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment